|AUTHOR: Müller, Gereon
TITLE: Constraints on Displacement
SUBTITLE: A phase-based approach
SERIES TITLE: Language Faculty and Beyond: Internal and External Variation in
PUBLISHER: John Benjamins Publishing Company
Atakan İnce, Columbia, MD, USA
Gereon Müller’s “Constraints on Displacement” reduces locality constraints on
movement of the (Generalized) Minimal Link Condition ((G)MLC) and the Condition
on Extraction Domain (CED) to the more basic principles of the Phase
Impenetrability Condition (PIC) and Edge Feature Condition (EFC) under a
strictly derivational system where every phrase counts as a phase. Such a
reduction provides us with a more minimalistic grammar, a new analysis of
syntactic islands, accounting for data that the above-mentioned constraints
could not account for, as well as showing the power and importance of the basic
principles given above.
The book contains an introduction and seven chapters. The seven chapters are:
Locality Constraints, (G)MLC and CED in Minimalist Syntax, On Deriving (G)MLC
Effects from the PIC, On Deriving CED Effects from the PIC, Operator Island
Effects, Movement from Verb-Second Clauses, Island Repair by Ellipsis.
The Introduction provides an overview of the book and introduces the PIC from
Chomsky (2000, 2001) and his version of EFC:
1. Phase Impenetrability Condition (PIC): The domain of a head X of a phase XP
is not accessible to operations outside XP; only X and its edge are accessible
to such operations.
2. Edge Feature Condition: The head X of phase XP may be assigned an edge
feature before the phase XP is otherwise complete, but only if there is no other
way to produce a balanced phase.
Here, “phase balance” refers to the existence of a matching feature either in
the workspace or at the left edge of a phase for every structure-building feature.
Chapter 1, “Locality Constraints,” gives an overview of all the locality
constraints on movement with their revised versions and how they evolved
throughout the history of generative syntax. These constraints are categorized
according to whether they are local vs. non-local, derivational vs.
representational, etc. Ultimately, these constraints are categorized into two
groups depending on whether they restrict movement to rigid vs. relativized
Müller also introduces three meta-constraints for ‘good’ constraints:
3. Constraints should be as simple and general as possible (p. 28)
4. Constraints should not be complex (p. 30)
5. Constraints are of type (i) or (ii):
(i) principles of efficient computation
(ii) interface conditions
According to Müller, movement constraints other than (G)MLC and CED are either
problematic or can be reduced to (G)MLC and CED.
Chapter 2, “(G)MLC and CED in Minimalist Syntax,” presents a more detailed
overview of both (G)MLC and CED and provides conceptual and empirical arguments
against these conditions. The conclusion is that neither of these constraints is
compatible with a strictly derivational, phase-based minimalist grammar. It also
reviews the literature on these conditions (both the Minimalist Program and
other frameworks such as GPSG and HPSG) and points out their problematic
aspects. The definitions of these conditions are given below:
6. The (Generalized) Minimal Link Condition (p. 53)
In a structure α[•F•] . . . [ . . . β[F] . . . γ[F] . . . ] . . ., movement to
[•F•] can only affect the category bearing the [F] feature that is closer to [•F•]
[NB: “Structure-building features for external Merge and movement are
accompanied by bullets,” p. 4.]
7. Condition on Extraction Domain (CED): (p. 59)
a. Movement must not cross a barrier.
b. An XP is a barrier iff it is not a complement.
Chapter 3, “On Deriving (G)MLC Effects from the PIC,” argues that (G)MLC effects
follows from the PIC. Thus, it can be dispensed with. This is based on three
assumptions in the chapter. The first assumption is that features on lexical
items drive syntactic operations. The second is that all phrases are phases;
and, the last one is that edge feature insertion is based on ‘phase balance’: “A
phase is balanced iff, for every movement-inducing feature [•F•] in the
numeration, there is a distinct potentially available feature [F]” (p. 128).
In a sense, phase balance enables successive-cyclic movement via edge feature
insertion. These assumptions account for superiority effects in (8b) as follows:
a. (I wonder) who(1) bought what(2).
b. *(I wonder) what(2) bought who(1).
To be able to move to Spec, CP, what(2) first needs to move to Spec, VP, which
requires edge feature insertion. However, since VP is balanced because there is
another wh-phrase (i.e. who(1)) in the workspace, no edge feature can be
inserted, which disables movement of what(2).
The rest of the chapter suggests an account for (lack of) Superiority effects in
certain constructions and languages. Such a formulation of (G)MLC, Müller
notices, also accounts for wh-intervention effects where there is no
c-command/dominance relation between the relevant wh-phrases, that the standard
(G)MLC could not account for.
Chapter 4, “On Deriving CED Effects from the PIC,” argues that CED effects also
follow from the PIC, and that therefore CED can also be dispensed with. For
this, two additional assumptions are introduced to the three assumptions made in
Chapter 3: first, features are ordered on lexical items. The second assumption
is that edge features are inserted as long as the phase head is active. A phase
is active as long as it has features to discharge. Under these assumptions, α is
a barrier if it is merged in a phase as the last operation in that phase.
Therefore, CED effects are not a matter of position but rather timing of
insertion. ‘Structure-building’ features (that trigger internal and external
merge) and ‘probe’ features (that trigger Agree operations) occur in different
stacks. Features on lexical heads also obey pushdown automata: whichever feature
is introduced last in the stack is checked first. Here, Müller adds two more
conditions for edge feature insertion to EFC in (2): an edge feature can be
inserted to a phase head if (a) the relevant phase head has not discharged all
its structure-building or probe features yet, and (b) the relevant edge feature
ends up on top of the phase head’s list of structure building features.
In the following example,
9. *[DP2 Who(m) ] has [DP1 a comment about t2 ] annoyed you?
Insertion of an edge feature to the relevant phase is required for 'who(m)' to
be extracted out of DP1. However, no further edge feature can be inserted
because after insertion of DP2 the phase is inactive/inert lacking any
structure-building or probe feature to be checked. Therefore, CED effects follow
from the PIC.
The rest of the chapter focuses on freezing effects and melting effects. Under
Müller’s formulization of CED effects, a phrase YP in Spec, XP is supposed to
cease being a barrier if another phrase moves to an outer specifier of the phase
XP. In other words, YP melts. Müller gives data from German and Czech that
confirms melting effects. The chapter ends with an appendix where Müller
contends that movement-related morphology (in Irish, for instance) makes
successive-cyclic movement possible.
Chapter 5, “Operator Island Effects,” focuses on wh-island and topic island
effects, and starts with the observation that (G)MLC (as developed in Chapter 3)
cannot account for these islands. The reason is that a [+topic]/wh-phrase is
expected to be able to move through the edge of a Topic/interrogative
Complementizer Phrase to satisfy phase balance if (a) there is another
Topic/complementizer head in the search space, or if (b) the lower
Topic/interrogative Complementizer Phrase through which a [+topic]/wh-phrase
needs to move through checks its relevant topic/wh-features with another
For this reason, Müller introduces (a) the Intermediate Step Corollary and (b)
feature ‘maraudage’ (Georgi et al. 2009). The Intermediate Step Corollary
requires that “[i]ntermediate movement steps to specifiers of X . . . must take
place before a final specifier is merged in XP” (p. 176). In maraudage, the
phrase to be moved long-distance marauds the A-bar features of the lower
complementizer head before any phrase is attracted by the same complementizer
head. Since all the relevant A-bar features of the complementizer are discharged
by the lower A-bar element, the higher A-bar element cannot move to the same CP
to discharge any feature.
Chapter 6, “Movement from Verb-Second Clauses,” offers an account for why
verb-second complement clauses in German only allow movement into a higher
verb-second clause but not into a higher verb-final clause, whereas verb-final
complement clauses headed by ‘dass’ (‘that’) allow movement into higher clauses
of both types. Müller argues for a PIC-based account for this dichotomy.
Chapter 7, “Island Repair by Ellipsis,” develops a model of ellipsis (actually,
sluicing) phase-by-phase: ellipsis-by-phase. In this chapter, Müller introduces
a third type of feature: ellipsis features. They occur on phase heads and
trigger ellipsis of the complement of a phase head. In sluicing cases, not only
the relevant complementizer head hosting the wh-phrase but also all the lower
phase heads bear the ellipsis feature. Discharge of the ellipsis feature is the
last operation on a phase head, and an edge feature can be inserted on this
phase head, enabling a wh-phrase to move without violating the PIC, which also
accounts for the island-insensitive nature of sluicing.
Müller’s book is very successful in reducing certain locality effects on
movement to the PIC. This makes the Minimalist Program more minimalistic, doing
away with constraints such as (G)MLC and CED.
The first chapter gives a very good review of different locality constraints on
movement and how they evolved throughout the history of generative syntax.
Chapter 2 gives a very detailed review of (G)MLC and CED and of various works
on deriving the (G)MLC and CED effects.
Therefore, the book is must read for linguists working on (locality of)
movement, islands, and phases. The first two chapters can also be used in
introductory courses on syntax and seminars on (locality of) movement, islands,
and phases at the graduate level.
Müller’s work is a phase-based version of Takahashi (1994) in that it assumes
that a phrase must move through every other phrase (which is also a phase) on
the way to its ultimate landing site. The rationale for that in this book is the
PIC. It also suggests an analysis for previously puzzling data from German and
introduces novel data.
However, the last chapter, “Island Repair by Ellipsis,” has some problematic
aspects. The first problem is related to the spreading of the ellipsis feature.
To avoid any violation of Inclusiveness, Müller contends that the ellipsis
feature is “optionally inserted on any lexical item in the numeration” (p. 312).
He also contends that a head with the ellipsis feature other than the
Complementizer head can merge with a phrase only if the head of the second
phrase also bears the ellipsis feature. Any lexical head above the [+wh]
complementizer with the ellipsis feature does not bear the ellipsis feature.
One problem is that if the ellipsis feature can be inserted in a lexical head in
sluicing, it can be inserted in the same lexical head any time. Then, ellipsis
of any phrase across the clausal spine is predicted: VP, vP, AspP, DP, NP. For
instance, a Verb head might be expected to bear the ellipsis feature in the
a. Did you see the boy?
b. No, I called *(the boy).
However, (10b) is ungrammatical. Therefore, Müller’s theory needs to be able to
block such elliptical structures.
In a similar way, sluicing is common in many languages (see, for example,
Merchant & Simpson (2012)); however, VP ellipsis is not observed in every
language that allows sluicing. Under Müller’s theory (as well as İnce (2012)),
VP ellipsis is expected to occur in every language in which sluicing occurs
because the ellipsis feature is expected to be inserted in v in non-sluicing
cases as well as in sluicing cases. In sum, Müller’s theory of ellipsis
Another problem is related to island-sensitivity of VP ellipsis:
11. *They want to hire someone who speaks a Balkan language, but I don’t know
which (Balkan language) they do [VP want to hire someone who speaks t].
(Fox & Lasnik, 2003: 147 (ex. 18))
[Editor’s note: “[VP want to hire someone who speaks t]” should contain
strikethrough, but it may not appear in this format]
To explain the ungrammaticality of VP ellipsis cases as in (11) where the island
is inside the elision site, Müller speculates that there are two types of
ellipsis feature. One is inserted in sluicing cases, and the other is inserted
in VP ellipsis cases: “Assuming that there are two types of deletion features,
one might simply stipulate that one type is in effect too weak to keep the head
active at the decisive stage of the derivation, thereby making a circumvention
of island effects impossible in the case of VP ellipsis even if the island
itself also undergoes deletion” (p. 310).
This is only a rewording of the puzzle. Müller refers to insertion of an edge
feature in a phase head for a DP to escape the elision site (to avoid any PIC
violation) with “keep[ing] the head active,” whether there is an island in the
elision site or not. If the ellipsis feature is too weak to keep the head
active, then no edge feature insertion occurs, and no element can escape the
elision site whether it is extracted out of an island within the elision site or
not. In sum, such a stipulation is too restrictive.
In conclusion, although Müller’s book is in general successful, it requires some
refinement in the last chapter.
Fox, Danny & Howard Lasnik. 2003. Successive cyclic movement and island repair:
the difference between sluicing and VP ellipsis. Linguistic Inquiry 34:143-154.
Georgi, Doreen, Fabian Heck & Gereon Müller. 2009. Maraudage. Manuscript,
İnce, Atakan. 2012. ‘Sluicing in Turkish,’ in Merchant, J. & A. Simpson (eds.),
Sluicing: Cross-Linguistic Perspectives, 248-269. Oxford: Oxford University Press.
Merchant, Jason & Andrew Simpson (eds.). 2012. Sluicing: Cross-Linguistic
Perspectives. Oxford: Oxford University Press.
Takahashi, Daiko. 1994. Minimality of Movement. Ph.D. thesis, University of
ABOUT THE REVIEWER