Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
3D
Phase Theory
Phase Theory is the latest empirical and conceptual innovation in syntactic theory
within the Chomskyan generative tradition. Adopting a crosslinguistic perspective, this book provides an introduction to Phase Theory, tracing the development
of phases in minimalist syntax. It reviews both empirical and theoretical arguments in favor of phases, and examines the role phases play at the interface
with semantics and phonology. Analyzing current phasehood diagnostics,
it applies them in a systematic fashion to a broad range of syntactic categories,
both phases and non-phases. It concludes with a discussion of some of the more
contentious issues in Phase Theory, involving crosslinguistic variation with
respect to phasehood and the dynamic versus static nature of phases.
Barbara Citko is Associate Professor of Linguistics at the University of
Washington.
{DESKTOP}NEW FOLDER//9781107040847SER.3D
ii
{DESKTOP}NEW FOLDER//9781107040847TTL.3D
Phase Theory
An Introduction
BARBARA CITKO
University of Washington
{DESKTOP}NEW FOLDER//9781107040847IMP.3D
iv
{DESKTOP}NEW FOLDER//9781107040847TOC.3D
Contents
Acknowledgments
List of abbreviations
page vii
viii
1
Introduction
1
1.1
1.2
1.3
1.4
2
General architecture
External and Internal Merge
Features
Agree
7
10
14
20
Introducing phases
23
2.1
2.2
2.3
2.4
2.5
2.6
23
27
29
31
41
46
Phasehood diagnostics
58
3.1
3.2
3.3
3.4
58
60
66
67
71
Classic phases
71
91
108
Other ph(r)ases
124
124
137
151
v
{DESKTOP}NEW FOLDER//9781107040847TOC.3D
vi
vi
Contents
Variation in phasehood
161
161
168
177
185
185
199
Summary
205
References
Index
207
223
{DESKTOP}NEW FOLDER//9781107040847ACK.3D
Acknowledgments
First, I would like to thank the many cohorts of students who took my graduate
syntax classes at the University of Washington. The idea for this book came from
you and all the questions you asked (that I may or may not have had answers to
at the time). I hope next time I will! I also want to thank all my colleagues in
the Department of Linguistics for creating an intellectually stimulating work
environment, and especially my fellow syntacticians, Edith Aldridge, Karen
Zagona and Julia Herschensohn, for encouragement and support, especially in
the nal stages of the project; and Toshi Ogihara for very kindly answering all my
last-minute semantic questions.
Portions of this book were written at the Whiteley Writing Center, a wonderful writing oasis (on an island!), which I thank for the good writing vibes. Some
parts were written when I was a visiting scholar at the Department of Linguistics
and Philosophy at MIT in the Fall of 2011. I thank the department for the
welcoming atmosphere it creates for all its visitors, and for all the inspiring
classes, talks and conversations I had while I was there. In particular, I would
like to thank David Pesetsky, Sabine Iatridou and Shigeru Miyagawa for their
openness and interest in my research. I also want to thank Andrew Winnard
and the whole editorial team at Cambridge University Press for their patience
and guidance throughout the project, Alexander Sugar for help proong the
manuscript and Brent Woo for help with the index.
Given that the literature on phases in syntactic theory is vast and evergrowing, it is impossible to do justice here to all the phenomena that bear on all
the issues surrounding phases, to all the literature on phases and to all the
researchers working on phases. I alone remain responsible for any omissions or
misinterpretations.
Last but not least, I thank my husband Randy for his support and perspective.
I dedicate this book to the memory of Icarus.
vii
{DESKTOP}NEW FOLDER//9781107040847LOA.3D
viii
Abbreviations
The following list includes not only abbreviations used in this book but also the
abbreviations often used in the primary literature on phases.
ACC
ACD
ATB
AUGB
AUX
BPS
C
CHL
CAUS
C-I
CP
DNS
DP
ECM
EM
EPP
ERG
FEM
FI
FIFA
FIFB
FL
FP
G
H
phi-features
phonological component
semantic component
Accusative
Antecedent-Contained Deletion
across the board
approaching UG from below
Auxiliary
Bare Phrase Structure
Case
Computation of Human Language
Causative
conceptualintentional
Complementizer Phrase
Derivation of Narrow Syntax
Determiner Phrase
Exceptional Case Marking
External Merge
Extended Projection Principle
Ergative
Feminine
Feature Inheritance
Feature Inheritance from Above
Feature Inheritance from Below
language faculty
Functional Projection
Goal
Head
viii
{DESKTOP}NEW FOLDER//9781107040847LOA.3D
List of abbreviations
HTA
IC
iF
IM
INTER
INDEF
iwh
LA
LBE
LCA
LDA
Lex
LF
LI
MASC
MOM
MSO
N
NEUT
NMLZ
NOM
NPI
NS
NSL
NSR
NUM
P
PART
PF
P-feature
PH
PIC
PL
PLA
PP
PrP
PST
Q
QP
QR
RNR
RP
SA
ix
{DESKTOP}NEW FOLDER//9781107040847LOA.3D
SG
SM
SMT
S-O
SOR
SOT
T
TNP
TP
uF
uQ
val
VIR
vP
VPE
List of abbreviations
Singular
Sensorimotor
Strong Minimalist Thesis
Spell-Out
Subject to Object Raising
Sequence of Tense
Tense
Traditional Noun Phrase
Tense Phrase
uninterpretable or unvalued feature
uninterpretable Q feature
Feature Value
Virile
verb Phrase
VP Ellipsis
{DESKTOP}NEW FOLDER//9781107040847INT.3D
Introduction
This research survey combines an introduction to Phase Theory with an assessment of the state of the art in Phase Theory. The term Phase Theory refers to a set
of theoretical innovations in post-2000 minimalism (Chomsky 2000, 2001, 2004,
2005, 2008).1 One of the core ideas in minimalism is the idea that the language
faculty is an optimal solution to the constraints imposed on it by the two cognitive
systems with which it interacts: the system of thought and the articulatory
perceptual system. What Phase Theory adds to this picture is the idea that the
language faculty interacts with these two cognitive systems at very specic points
during the syntactic derivation, and, consequently, that syntactic derivations are
constructed in chunks referred to as phases.2 In most general terms, phases
cannot be accessed by the narrow syntax once they are transferred to the
interfaces.
My goal in this survey is to combine an introduction to a given issue within
Phase Theory with an overview of the existing research on this issue (and an
assessment thereof), giving the reader a sense of what is fairly settled upon and
what is still under debate.3 The fact that there is lot of research that relies on
phases shows a need for a survey that situates phases in current syntactic theory,
introduces the technical details of Phase Theory, synthesizes the existing research
Chomsky in his writings is very careful about distinguishing a program from a theory, emphasizing
the programmatic nature of minimalism (see in particular Boeckx 2006 for a more detailed
discussion of this distinction). Chomsky does not use the term Phase Theory in his early writings
on phases but does so more recently: One goal of Phase Theory is to provide the mechanisms to
distinguish copies from repetitions, as straightforwardly as possible (Chomsky 2012: 3).
The resulting model is also sometimes referred to as a Multiple Spell-Out (MSO) model. The idea of
Multiple Spell-Out goes back to Uriagereka (1999) (see also Uriagereka 2012 for a fuller, booklength exposition).
There are a number of monographs, volumes and collections that focus on various aspects of Phase
Theory, which I build on and am intellectually indebted to (see, among others, Frascarelli 2006,
Gallego 2010, 2012, Grohmann 2009a, 2009b, 2009c, McGinnis & Richards 2005, and the individual
contributions in the two issues of Linguistic Analysis 33, guest-edited by Kleanthes Grohmann).
{DESKTOP}NEW FOLDER//9781107040847INT.3D
Introduction
on phases (pointing out issues that might be still contentious), outlines directions
for future research, and, last but not least, standardizes the notation.
Even though many (though not all) syntacticians (explicitly or implicitly)
assume the concept of a phase, there seems to be less of a consensus regarding
many of the most fundamental properties of phases, such as those listed in (1).
(1) a. How do we dene phases?
b. What categories count as phases?
c. Do the same categories count as phases with respect to semantic and phonological considerations?
d. Are phases dynamic or static?
e. Is there any crosslinguistic variation with respect to phasehood?
f. How do phases interact with the interfaces?
Following Chomsky (2004), I take beyond explanatory adequacy to refer to the why-questions about
language, captured by the following quote: In principle, then, we can seek a level of explanation
deeper than explanatory adequacy, asking not only what the properties of language are but also why
they are that way. (Chomsky 2004: 105).
{DESKTOP}NEW FOLDER//9781107040847INT.3D
Introduction
See also Boeckx (2006), Boeckx & Grohmann (2007), Den Dikken (2007) and Gallego (2010) for a
discussion of these different views of phases, and of the problems some of them raise.
{DESKTOP}NEW FOLDER//9781107040847INT.3D
Introduction
The discussion of phasehood diagnostics also raises the question of whether there
is any crosslinguistic variation with respect to phasehood. This is the issue which
I come back to in Chapter 6. While variation with respect to whether a given
language has phases or not seems highly unlikely and implausible, given the
conceptual arguments in favor of having phases to begin with (such as reducing
computational load and being independently motivated by the interfaces), it is
certainly possible for languages to differ with respect to what categories count as
phases.
Chapter 4 Classic phases discusses in detail three categories that are commonly assumed to be phases CPs, vPs and DPs and applies the diagnostics
established in Chapter 3 to these categories.6 The phasehood status of CP is
relatively easy to establish: the evidence in favor of successive cyclicity from the
literature on A-bar dependencies is typically taken as evidence for CPs being phases
(see, for example, Lahne 2008 for an illuminating overview). The evidence includes
phenomena like wh-copying (Felser 2004, Manetta 2010, McDaniel 1989, among
many others), scope marking (Dayal 1996, Lutz Mller & von Stechow 2000,
Stepanov 2000, among others), complementizer agreement (Carstens 2003,
Carstens & Diercks 2011, Haegeman 1992, Haegeman & Van Koppen 2011,
Zwart 1993, 1997), wh-quantier stranding (McCloskey 2000, 2001), reconstruction (Barss 1986, 2001) and left branch extraction (Wiland 2010). The evidence in
favor of vPs being phases is similar in spirit. This chapter also reviews the debate on
whether unaccusative and passive vPs constitute phases, as argued for by Legate
(2003), and against by Den Dikken (2006a). While many of the facts that are
typically deemed to bear on the issue of C or v being a phase head might be given
6
The discussion in Chapters 4 and 5 is a sequence of case studies. There are other categories that are
conspicuously absent from the discussion (APs, AdvPs, various functional projections in the left
periphery of a clause) whose phasehood we might wonder about. I thank Kleanthes Grohmann for
raising this issue.
{DESKTOP}NEW FOLDER//9781107040847INT.3D
Introduction
{DESKTOP}NEW FOLDER//9781107040847INT.3D
Introduction
covert movement and by Felser (2004) to handle wh-copying. This chapter also
addresses crosslinguistic variation with respect to phasehood: if phases are
dynamic and head movement can extend phasehood, a certain amount of variation will come from independent considerations (such as the presence or absence
of certain types of head movement). Variation in phasehood can also follow from
variation in lexical inventories.
Chapter 7 Phases at the interfaces examines the roles phases play at the
interfaces, putting them in the more general context of the syntaxphonology and
syntaxsemantics interface. With respect to the PF interface, it focuses on the
questions of whether phases (or Spell-Out domains) are relevant and substantive
phonological units, and how these phasal or Spell-Out units are manipulated by
phonology. This chapter examines the role phases play in determining linear
order (see Fox & Pesetskys (2005) Cyclic Linearization) and nuclear stress (see
Adger 2007, Kahnemuyipour 2003, 2004, 2005, Kratzer & Selkirk 2007, among
others).
The potential evidence for the signicance of phases at the syntaxsemantics
interface comes not only from phenomena like scope ambiguities (on the assumption that Quantier Raising is constrained by phasehood) and the propositional
status of phases both of which feature prominently as phasehood diagnostics
but also from the idea that the boundary between vP and CP phasal domains
corresponds to the distinction between nuclear scope and restrictive domain in the
tripartite quanticational structure, as proposed explicitly by Biskup (2009a).
Chapter 8 Summary provides a brief summary and conclusion.
{DESKTOP}NEW FOLDER//9781107040847C01.3D
1
The Minimalist Program
{DESKTOP}NEW FOLDER//9781107040847C01.3D
The minimum that the language faculty (FL) has to accomplish is to interface
with language-external systems. The two external systems in question are
the sensorimotor (SM) system and the conceptual-intentional (C-I) system.
The conditions imposed by these two external systems are referred to as
Legibility Conditions, Bare Output Conditions or Interface Conditions
(IC).1 The Strong Minimalist Thesis (SMT), given in (3), states that language
is designed to interface with the external systems in an optimum way.2
(3) The substantive thesis is that language design may really be optimal in some
respects, approaching a perfect solution to minimal design specications.
(Chomsky 2000: 93)
The general architecture of the language faculty is as follows. Language has three
components: Narrow Syntax (NS), the phonological component and the
semantic component .3 For the most part, we will be concerned here with
Narrow Syntax and its computational processes.
Each derivation starts with a set of lexical items which are manipulated in the
course of the derivation by the syntactic operations Merge and Agree. I will discuss
these two operations in more detail in Sections 1.2 and 1.4, respectively. The
lexical item is, strictly speaking, a bundle of features, not a primitive syntactic
object (hence the quotes).4 This set of lexical items is called a Lexical Array (LA)
and is represented as an unordered set. A Lexical Array augmented by information
1
2
S0 refers to the genetically determined initial state in the process of language acquisition, which is
what UG provides.
There is no PF or LF cycle and thus there are no PF or LF operations. The terms are used to refer to
PF or LF representations. The terms PF interface and LF interface are used to refer to the interface
with SM or C-I systems, respectively.
This assumes something like the Late Vocabulary Insertion model of Distributed Morphology of
Halle & Marantz (1993) and much later work.
{DESKTOP}NEW FOLDER//9781107040847C01.3D
General architecture
on how many times each lexical item is selected from the lexicon is called a
Numeration (N). This information is represented by the subscripts. In simple
cases, Numerations and Lexical Arrays are equivalent, as shown in (4bc).
(4) a. Icarus likes nuts.
b. LA = {Icarus, likes, nuts, v, T, C}
c. N = {Icarus1, likes1, nuts1, v1, T1, C1}
The two diverge when a single item is used more than once in a given derivation,
as in the infamous example given in (5a).5
(5) a. Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo.
(Pinker 1994: 2006, crediting Annie Senghas)
b. LA = {Buffalo, buffalo, buffalo, C, v, T}
c. N = {Buffalo3, buffalo2, buffalo2, C2, v2, T2}
However, one does not need to resort to exotic examples of this sort to show how
Lexical Arrays and Numerations differ; relatively simple sentences with any level
of embedding make the same point:
(6) a. Icarus thinks he likes nuts.
b. LA = {Icarus, thinks, he, likes, nuts, v, T, C}
c. N = {Icarus1, thinks1, he1, likes1, nuts1, v2, T2, C2}
Its notoriety comes from lexical ambiguity, of which it provides a very extreme illustration, not from
the distinction between a lexical array, which in this case contains distinct lexical items buffalo (i.e.
the proper name Buffalo, the common noun buffalo, and the less commonly used transitive verb to
buffalo meaning to bully) and the Numeration that includes multiple occurrences of each of them.
The following paraphrase helps distinguish the different meanings of buffalo:
(i) The buffalo from Buffalo that (another) buffalo from Buffalo bullies himself bullies (yet another)
buffalo from Buffalo.
See, however, Frampton & Gutmann (2002) for a proposal that syntax only generates convergent
derivations, and Preminger (2011) for a proposal that having unvalued features does not always lead
to a crash.
{DESKTOP}NEW FOLDER//9781107040847C01.3D
10
10
the SM interface can interpret features having to do with linear order (if such
exist), syllable structure, prosodic structure or intonation. The C-I interface, on
the other hand, should be able to interpret features having to do with scope,
quantication, referentiality, specicity, propositional status etc. Neither interface
can interpret formal features, such as structural case features or categorial features. Since features play such a major role in minimalism and there are quite a
few contentious issues surrounding them, we will devote an entire section to them
(Section 1.3).
The three basic operations that manipulate lexical items selected from the
lexicon are External Merge, Internal Merge and Agree. These three, in conjunction with a more detailed discussion of features, are the focus of the next three
sections.
Elsewhere, I have argued that Merge can also create structures in which a single object can end up
shared between two objects, referring to this type of Merge as Parallel Merge (see Citko 2011b and
the references therein), illustrated in (iii). Parallel Merge combines the properties of External
Merge and Internal Merge. Before Merge takes place, Z and YP are disjoint (as in External Merge)
but Z merges with a subpart of YP (as in Internal Merge).
(i) Merge X and Y, Project Y
YP
Y
XP
YP
Y
ZP
XP
Chomsky (2007) excludes such a possibility on the grounds that it requires new operations and
conditions on what counts as a copy, hence additional properties of UG (Chomsky 2007: 8, note 10).
{DESKTOP}NEW FOLDER//9781107040847C01.3D
11
11
itself be the output of a prior Merge operation, as shown in (7d), where the
complex object that is the output of (7b) merges with Z.8
(7) a. Select X and Y from the Numeration
b. Merge X and Y
X
X
Z
Z
X
Y
Following Chomsky (1994), I assume a bare-bones approach to phrase structure, referred to as Bare
Phrase Structure (BPS), in which the X-bar status of a given projection (in particular, whether it is a
head (X0), a phrase (XP) or an intermediate projection (X0 )) is not a primitive of the theory, as it can
always be derived from the structure in which it appears. Thus a head is an element that is not a
projection of another element of the same type. It can (but it does not have to) project further. A
phrase is an element that does not project any further. Again, it can, but it does not have to, be a
projection of another element of the same type. The representations in (7ad) are BPS representations; the ones in (i) and (ii) are their more traditional X-bar theoretical counterparts, with the ones in
(ib) and (iib) simplied in that they lack vacuous intermediate projections.
(i)
a. XP
b.
X
X
(ii)
a.
XP
X
YP
YP
ZP
b.
Z
ZP
Z
XP
XP
X
YP
X
X
YP
Even though I will be using representations of the kind given in (ib)(iib) throughout this book,
mostly for familiaritys sake, I do assume a BPS approach throughout. The advantages of such an
approach are twofold. First, it allows a given element to be simultaneously a head and a phrase,
something that the behavior of clitics suggests has to be allowed in the grammar. Y in (7b) and (7d) is
such an element; it is a head because it is not a projection of another Y and it is a phrase because it
does not project any further. Second, BPS dispenses with vacuous intermediate projections, a
welcome move from the perspective of the Strong Minimalist Thesis (SMT).
{DESKTOP}NEW FOLDER//9781107040847C01.3D
12
12
The output of External Merge could also be represented in set notation. (8a) is
equivalent to (7b) and (8b) to (7d).
(8) a. {X, {X, Y}}
b. {Z, {Z, {X, {X, Y}}}}
It is standard to assume that the result of Merge also has to have a label, bolded in
(8ab) above.9 The label is determined by one of the two merged elements; other
options would violate the Inclusiveness Condition (IC), which prohibits the introduction of new elements in the course of the derivation. Any label other than X or Y
in (8a) (and other than Z and X in (8b)) would be a violation of this condition.
Chomsky (1995) discusses two other logical possibilities. In a set consisting of two
elements (such as X and Y in (8a)), the label could also be the intersection of X and Y
or the union of X and Y. However, he excludes the rst option on the grounds that the
intersection in many cases is a null set. The second option is excluded because
the union of X and Y is contradictory if X and Y differ in features, as is the case when
one is N and the other V, for example. This leaves us with only one option: either for
X or Y to project as the label. The question of which of the two it is, or the even more
fundamental question of whether labels are necessary at all, is far from being settled.
A common and fairly intuitive idea is that the object which has its selectional feature
checked as a result of Merge is the object that determines the label. For example,
when a verb or a preposition merges with a noun, this verb (or preposition) has its
selectional feature satised via the Merge operation. This means that the label is
going to be determined by this verb or preposition.10,11 Many existing labeling
algorithms make reference to selection; the one from Cecchetto & Donati (2010),
given in (9) below, is a representative example:12
10
11
An interesting issue is whether all the features of X determine the label, or only a subset thereof.
While it seems clear (and commonly assumed) that categorial features (N, V, D etc.) are part of the
label, what other features need to project is somewhat less clear (see Citko 2008a, 2011b,
Cecchetto & Donati 2010, Donati & Cecchetto 2011, for a discussion of this and related issues).
Things get a little more complicated if c-selection is removed from our syntactic toolkit, as
Chomsky suggests in Beyond Explanatory Adequacy (2004).
Chomsky (2008) proposes a different labeling algorithm, given in (i) below, which does not make
reference to selection.
(i) In {H, }, H a lexical item (LI), H is the label.
If is internally merged to forming {, }then the label of is the label of {, }
(Chomsky 2008: 145)
12
Other examples similar in spirit, as also noted by Cecchetto & Donati (2010) and Gallego (2010)
include those proposed by Adger (2003), Pesetsky & Torrego (2006) and Boeckx (2008b):
(i) The head is the syntactic object which selects in any Merge operation. (Adger 2003: 91)
(ii) If and merge, some feature F of must probe F on . (Pesetsky & Torrego 2006)
(ii) The label of {, } is whichever of or probes the other, where the Probe = Lexical Item whose uF
gets valued. (Boeckx 2008b: 96)
{DESKTOP}NEW FOLDER//9781107040847C01.3D
13
13
(9) The label of a syntactic object {, } is the feature(s) that act(s) as a Probe of the
merging operation creating {, }. (Cecchetto & Donati 2010: 245)
Chomsky (2004) considers the possibility that the grammar requires no labels
whatsoever (see also Collins 2002, Gallego 2010, Seely 2006, among others).13
While intuitively appealing from the perspective of the Strong Minimalist
Thesis, eliminating labels raises some non-trivial issues, such as how
selection or relativized minimality works. I will not discuss these issues here
(as they strike me as tangential to the issue of phases. Instead, I refer the
interested reader to Citko (2008a, 2011b) and the references therein for
relevant discussion, pointing to the conclusion that labels are not dispensable.
One argument often adduced against the existence of labels comes from the
Inclusiveness Condition. However, if a label is simply a copy of the features of
one of the two merged elements (or more likely, of a subset of its features),
inclusiveness is not violated.
Internal Merge, which as Chomsky suggests in Beyond Explanatory
Adequacy (2004) is a very natural way to capture movement, is similar to
External Merge in that it takes two objects and combines them to form one larger
object. The only difference is that one of these two objects is already part of the
other one (hence the term Internal Merge). (10) below provides an illustration; Y
undergoes Internal Merge with X or, to use more traditional terminology, moves
to the specier of X.
(10) Internal Merge of X and Y14
a.
X
Y
b.
X
X EPP Y
X
X
X EPP
The label in Internal Merge structures is determined by the element that drives
movement. In (10ab), for example, it is some feature of X that drives movement
of Y to its specier. I assume that the feature in question is the EPP feature, which
is checked by this movement. Checking is marked by a strikethrough. In what
13
14
However, Chomsky (2013) seems to depart from this conclusion, and points out a number of cases
in which the lack of labels forces movement. This suggests that in other (non-movement) cases,
labels have to be in principle available.
(10a) represents movement (Internal Merge in our terms) in a familiar Copy and Merge notation,
whereas (10b) does so in a way that reects the term Internal Merge perhaps a little more
directly. Both are BPS-style representations. Strictly speaking, the movement illustrated here
(i.e. movement of a complement to a specier of the same head) might be banned by independent
anti-locality principles, which I discuss in more detail in Chapters 4 and 5.
{DESKTOP}NEW FOLDER//9781107040847C01.3D
14
14
follows, I use the term valuation to refer what used to be known as checking and
reserve the term checking for Extended Projection Principle (EPP) feature
checking, the only remnant of checking in the current system.
The third operation that plays a crucial role in minimalist syntax is Agree.
Since it is an operation that affects features, let me digress briey and discuss
features rst. This is the focus of the next section, which dicusses syntactic
features and the role they play in Narrow Syntax.
1.3 Features
Examples of syntactic features that populate syntactic literature include categorial features (such as N features or V features), case features, -features
(gender, number, person), EPP features, wh-features, and Q features, among
others. Features play a crucial role in minimalism in that they drive syntactic
operations.
Features come in two guises: Interpretable Features (iF) and
Uninterpretable Features (uF). Chomsky (2001, 2004) reduces the distinction
between these two types to feature valuation. Uninterpretable features are simply
features that enter the derivation unvalued, whereas interpretable ones are the
ones that come with inherently specied feature values:
(11) An uninterpretable feature F must be distinguished somehow in LEX from
interpretable features. The simplest way, introducing no new devices, is to
enter F without value: for example, [uNumber]. That is particularly natural
because the value is redundant, determined by Agree. (Chomsky 2004: 116)
I will use the following notation to distinguish these two types, where val stands
for feature value.
(12) a. Interpretable feature: iF[val] (e.g. iT[past], i[3sg, masc])
b. Uninterpretable feature: uF[ ] (e.g. uCase[ ], u[ ])
{DESKTOP}NEW FOLDER//9781107040847C01.3D
15
Features
15
it seems plausible to assume the C-I interface can interpret features that have
some semantic import; examples of such features might be features involving
tense (past, present, future (and null, on some accounts)), aspect, quantication or
-features (person, number and gender). The SM interface, on the other hand,
can interpret features that are relevant for articulation. This allows us to distinguish features that are interpretable at a given interface from features which the
interface cannot interpret: a feature is interpretable at a given interface if it makes
a distinction between two representations at this interface.15,16 Svenonius
(2007a) (see also Adger & Svenonius 2011 for relevant discussion) puts it
succinctly as follows:
(13) A feature F is an X feature iff F can constitute a distinction between two different
X representations.(Svenonius 2007a: 375)
15
16
This view of interpretable features bears some resemblance to distinctive features, not an unwelcome result. Distinctive features in phonology are features that distinguish between two phonological representations. By analogy, we could think of features that are interpretable at the C-I
interface as distinctive semantic features.
Svenonius also makes the following distinction between module-internal features (for example,
syntax-internal features) and interface features (for example, syntaxsemantics features). Features
that have no semantic import at all, such as class features, or arguably EPP features, provide good
examples of syntax-internal features.
(i) For any F, and any modules X and Y,
a. F is an X-internal feature iff F is an X feature and not a feature of any other module.
b. F is an X-Y interface feature iff F is an X feature and a Y feature. (Svenonius 2007a: 375)
17
{DESKTOP}NEW FOLDER//9781107040847C01.3D
16
16
third-person singular feminine noun; the presence of third-person singular feminine agreement on the verb, however, is a result of run-of-the-mill subjectverb
agreement.18 In (15), on the other hand, the noun ptak bird is specied as a thirdperson singular masculine noun and the verb agrees with it.
(14) a. Papuga
siedziaa
parrot.3SG.FEM sit.PST.3SG.FEM
A parrot sat on a tree.
na
on
drzewie.
tree
na
on
drzewie.
tree
[Polish]
b. N = {PARROTi[3SG.FEM], . . ., SITu[ ], }
(15) a. Ptak
siedzia
bird.3SG.MASC sit.PST.3SG.MASC
A bird sat on a tree.
b. N = {BIRDi[3SG.MASC], . . ., SITu[ ], }
The idea that -features are always interpretable on nouns has sometimes been
questioned in the relevant literature. We have seen above (note 18) that number
might be different from other features in this respect. Legate (2002, 2012) also
raises some issues for the way Chomsky characterizes the distinction between
interpretable and uninterpretable features, pointing out that it is not clear in what
sense gender features in languages with non-natural gender are semantically
18
Number is different from person and gender. For most nouns, gender specication is inherent and
invariant; the Polish noun ksika book is always third person feminine but it can be either singular or
plural. This suggests that number might not be inherent to the noun itself in the same way person and
gender are. One way to implement it is to think of Number as being uninterpretable on the noun but
interpretable on a higher Number head, as shown in (i) (see Danon 2011 for a more concrete proposal).
DP
(i)
NumP
NP
NumiNum[val]
NiPers[val]
iGen[val]
uNum[ ]
However, there do exist nouns that can be either feminine or masculine (such as natural gender
marked nouns in (ii)), or nouns that are always singular or plural (such as so-called pluralia tantum
nouns in (iii)).
(ii) aktor/aktorka
actor/actress
lekarz/lekarka
pilot/pilotka
(iii) spodnie
nozyczki
pants
scissors
Since little hinges on this distinction as far as the specics of Phase Theory are concerned, I will
abstract away from it in what follows.
{DESKTOP}NEW FOLDER//9781107040847C01.3D
17
Features
17
kniga
book3SG.FEM
livre
book.3SG.MASC
Buch
book.3SG.NEUT
[Russian]
[French]
[German]
walkeduT[PST]
A slightly different explanation (but not one that is incompatible with tense
feature on the verb being uninterpretable) comes from Pesetsky and Torregos
(2001) work on case. In their view, Nominative case is an instance of an uninterpretable T feature on the subject:
(18) TiT[PST]
DPuT[NOM]
{DESKTOP}NEW FOLDER//9781107040847C01.3D
18
18
2000, among others), in which Accusative case marks aspectual notions such as
delimitedness or specicity:
(19) a. Ali
[Turkish]
(20) a. Matti
[Finnish]
bir
kitab
ald.
Ali one book.ACC bought
A book is such that Ali bought it.
b. Ali
bir
kitap
ald.
Ali
one
book
bought
Ali bought some book or other.
luk-i
kirja-t
(tunni-ssa).
Matti-SG/NOM read-PST/3SG book-PL-ACC (hour-INESS)
Matti read the books (in an hour).
b. Matti
luk-i
kirjo-j-a
(tunni-n).
Matti-SG/NOM read-PST/3SG book-PL-PART (hour-ACC)
Matti read books (for an hour). (Megerdoomian 2000: 31617)
{DESKTOP}NEW FOLDER//9781107040847C01.3D
19
Features
19
interaction between feature valuation, Transfer and deletion till the next
chapter. To preview, we will see there that feature valuation and deletion
have to happen simultaneously at the point of Transfer to the interfaces. There
is thus no need for the derivation to keep track of which features entered the
derivation valued and which ones did not.
As pointed out by Pesetsky & Torrego, Chomskys reasoning for equating
interpretability with valuation has to do with the fact that Narrow Syntax
lacks the look-ahead property to determine whether a given feature is going
to be interpretable at the interface or not. It is important to note that the timing
issue alluded to here only arises if we equate uninterpretability with lack
of value; if uninterpretable features remain uninterpretable throughout the derivation, the issue does not arise. Indeed, this equivalence has also been questioned
in the literature. Pesetsky & Torrego (2007) and Bokovi (2011), for example,
argue that valuation and interpretability should be taken as independ ent of
each other. This allows feature combinations that were contradictory if
feature interpretability amounts to valuation. The new possibilities are bolded
in (22).
(22) Types of features
uF[val]
iF[val]
uF[ ]
iF[ ]
For Pesetsky & Torrego (2007), tense on the verb is valued in the lexicon
(which, they argue, is plausible given the fact that quite often tense surfaces on
the verb).19 It is nevertheless uninterpretable and remains so throughout the
derivation. The tense feature on the T head, on the other hand, is interpretable
but unvalued:
(23) a. TiT[ ]
b. TiT[PST]
walkeduT[PST]
walkeduT[PST]
There are thus two plausible candidates for what an uninterpretable T feature
might be: a tense feature on the verb or a Nominative case feature on the
noun. Furthermore, Pesetsky & Torrego argue that such a dissociation of interpretability and valuation also removes the need for two types of features
on interrogative complementizer heads and wh-phrases. Complementizer
heads in wh-questions are typically thought to require two types of features: the
feature that marks the clause as interrogative (essentially Chengs (1991) typing
19
I am simplifying Pesetsky and Torregos system somewhat. For them, feature valuation is feature
sharing of the kind proposed by Frampton and Gutmann (2000), for example. Instead of (23), they
suggest (i), in which [2] is the value. The number 2 has no signicance, it just marks the same
feature value in both locations:
(i) TenseiT[2]
walkeduT+past[2]
{DESKTOP}NEW FOLDER//9781107040847C01.3D
20
20
WHiwh,uQ[ ]
WHiwh,uQ[val]
WHuQ[wh]
WHuQ[wh]
1.4 Agree
The discussion in the previous section relied on the assumption that Agree is an
operation responsible for valuing unvalued features.21 In this section, we turn to
the details of the Agree mechanism. Following the general consensus in the eld,
I will capitalize Agree when I use it in this technical sense and distinguish it from
the descriptive terms agreement or agrees with, used quite broadly in the
literature.
For Agree to be possible, the following conditions have to hold (cf. Chomsky
2000: 1223).
(26) a. The Probe and the Goal have to be active, where being active means having
uninterpretable/unvalued features.
THE ACTIVITY CONDITION
20
21
Some sources take C to have an interpretable wh-feature and an uninterpretable Q feature, and the
wh-phrase to have an uninterpretable wh-feature and an uninterpretable Q feature. This seems to be
a terminological rather than a substantive difference.
Chomsky describes Agree as an operation responsible for deleting uninterpretable features under
matching:
(i) The erasure of uninterpretable features of probe and goal is the operation we called Agree.
(Chomsky 2000: 122)
Feature deletion is distinct from feature valuation, which raises the question of whether we have been
mistaken in taking Agree to be responsible for feature valuation. I do not believe so. As we will see in
Section 2.6, if valuation and deletion happen simultaneously (at the phase level), there is no incompatibility here.
{DESKTOP}NEW FOLDER//9781107040847C01.3D
21
Agree
21
b. The features of the Probe and Goal have to match, where matching refers to
feature identity.
THE MATCHING CONDITION
c. The Goal has to be inside the domain of the Probe, where the domain of the
Probe is its sister.
THE DOMAIN CONDITION
d. The Goal has to be in a local relationship, where locality is closest ccommand.
THE LOCALITY CONDITION
When all these conditions are met, the unvalued features get valued (and deleted).
A typical conguration satisfying all four conditions is given in (27). Agree
is indicated by a dotted line (a convention I am going to employ throughout
the book). Agree is typically assumed to be impossible if both the Probe and
the Goal have only unvalued features.22
(27)
a.
PuF[
b.
AGREE
GiF[val], uF[
PuF[val]
GiF[val], uF[val]
DPi[3sg.fem], uC[ ]
DPi[3sg.fem], uC[Nom]
22
Some researchers allow such non-standard Agree (see Adger 2003: 169, for example).
{DESKTOP}NEW FOLDER//9781107040847C01.3D
22
22
In my own work on Parallel Merge structures (see Citko 2011b for the most
complete exposition), I argued in favor of the other possibility: the possibility of a
single Goal agreeing simultaneously with multiple Probes:
(30) a. Tu[ ], iT[val]
Tu[ ], iT[val]
b. Tu[val], iT[val] Tu[val], iT[val]
DPi[val], uC[ ]
DPi[val], uC[val]
Since the issue of whether Agree is a binary operation or not does not bear
directly on any aspect of Phase Theory, I will not elaborate on it further here,
and I refer the interested reader to the works cited above (and the references
therein) for data, evidence and further discussion. The one thing about Agree
that is relevant to Phase Theory concerns the locality conditions on Agree; in
particular the issue of whether Agree is possible across phase boundaries. This
is the issue we come back to in Section 2.4, which focuses on the Phase
Impenetrability Condition.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
23
2
Introducing phases
2.1
This chapter introduces the concept of a phase, providing both theoretical and
empirical motivation for it. To the best of my knowledge, the term rst appeared
in Chomskys (2000) Minimalist Inquiries, where phases (to be more specic,
lexical subarrays associated with phases) were introduced as a solution to a
problem arising from the Merge over Move (MOM) principle. The concept of
a phase, however, builds on many prior principles involving locality domains:
cycles, barriers, islands, to name a few.
Let us start by looking at the Merge over Move Principle in more detail. What
it says is that, all things being equal, Merge is preferred over Move.2 Consider the
following contrast involving a raising predicate be likely, modeled upon
Chomskys (2000) (non-parrot) examples:
(1) a. There are likely to be many parrots at the clay lick right now.
b. *There are likely many parrots to be at the clay lick right now.
See also Hornstein Nunes & Grohmann (2005) for a very lucid overview of the rationale behind the
Merge over Move principle. The logic of the presentation here mirrors theirs. See, however,
Castillo, Drury & Gohmann (2004) for a discussion of some issues MOM raises and ways to
resolve these issues.
Chomsky points out that MOM could be derived from more general economy principles, if Move is
a more complex operation than Merge, consisting of Copy and Merge. It is not clear, however, if the
logic survives when Move is treated as a variant of Merge, i.e. Internal Merge. There is no sense in
which Internal Merge is more complex than External Merge.
23
{DESKTOP}NEW FOLDER//9781107040847C02.3D
24
24
Introducing phases
We have a choice here: the EPP feature of T can be checked either by merging
the expletive there in [Spec,TP] or by moving the DP many parrots there.
However, only one of these choices leads to a convergent derivation. If
we move many parrots to [Spec,TP], the derivation proceeds as follows:
(3) a. Move [DP many parrots] to [Spec,TP], checking the EPP feature of T
TP
DPi[3pl],uC[ ] T
many parrots T
to EPP
VP
AdjP
Adj
likely
TP
DP i[3pl],uC[
many parrots
T
to EPP
VP
be many parrots at the clay lick right now
c. Merge [AP likely many parrots to be at the clay lick right now] with the verb are
VP
V
are
AdjP
Adj
likely
TP
DPi[3pl],uC[
many parrots
T
to EPP
T
VP
d. Merge [VP are likely many parrots to be at the clay lick right now] with T
Move are to T
Agree between T and [DP many parrots]
{DESKTOP}NEW FOLDER//9781107040847C02.3D
25
25
TP
T EPP,u[3pl] VP
are
V
AdjP
are
Adj
TP
likely
DPi[3pl],uC[Nom] T
many parrots T
to EPP
VP
TP
T
there
TEPP,u[3pl] VP
are
V
AdjP
are
Adj
TP
likely
DPi[3sg],uC[Nom] T
VP
many parrots T
to EPP
be many parrots at the clay lick right now
f. There are likely many parrots to be at the clay lick right now.
Interestingly, nothing seems to go wrong with this derivation, yet the result
(example (3f)) is ungrammatical. The uninterpretable features are all checked
and/or valued, the Numeration is exhausted, and locality is respected. This is
where the Merge over Move (MOM) principle comes in: the step in (3a) violates
MOM, as we are moving the DP many parrots to [Spec,TP] rather than merging
the expletive there. If we merge the expletive there rst, the derivation proceeds
as schematized in (4) instead, leading to the grammatical (4f).
(4) a. Merge there with [TP to be many parrots at the clay lick right now], checking the
EPP feature of T
TP
there
T
toEPP
VP
{DESKTOP}NEW FOLDER//9781107040847C02.3D
26
26
Introducing phases
b. Merge [TP there to be many parrots at the clay lick right now] with the adjective
likely
AdjP
Adj
likely
TP
there
T
toEPP
VP
VP
V
are
AdjP
Adj
likely
TP
T
there
T
toEPP
VP
be many parrots[3pl],uC[ ] at the clay lick right now
d. Merge the [VP are likely many parrots to be at the clay lick right now] with the
matrix T
Agree between T and many parrots
Move are to T
TP
Tu[pl],EPP VP
are
V
are
AdjP
Adj
TP
likely
T
there
T
to EPP
VP
{DESKTOP}NEW FOLDER//9781107040847C02.3D
27
Motivating phases
27
TP
there
Tuj[3pl],EPP VP
are
V
AdjP
are
Adj
TP
likely
there
T
T
to EPP
VP
Since its Numeration, given in (6a) below, contains an expletive, the prediction is
that (5) should be ungrammatical, since at the stage given in (6b), according to
MOM, the Merge of there should win out over the Move of many parrots.
(6) a. N = {there, is, a, strong, likelihood, that, many, parrots, will, be, at, the, clay lick,
right, now}
b.
TP
T
T
VP
willu[3sg]EPP
be many parrots i[3pl],uC[Nom]t be at the clay lick right now
c.
TP
there
T
willu[3sg]EPP
VP
be many parrotsi[3pl],uC[Nom] be at the clay lick right now
{DESKTOP}NEW FOLDER//9781107040847C02.3D
28
28
Introducing phases
If that happens, however, the derivation is going to run into problems when the
matrix T is merged. Raising of there from the subject position of the embedded
nite clause to the matrix subject position, which results in (7), is unmotivated;
hence ruled out by economy.
(7) * Therei is a strong likelihood that ti will be many parrots at the clay lick right now.
The solution to this problem relies on the intuition that at the point in the
derivation when the EPP feature of the embedded T needs to get checked
(which is immediately upon the Merge of T), the expletive there is not available.
In Minimalist Inquiries, Chomsky implements it in terms of so-called subarrays.
(8) Suppose we select LA as before . . . Suppose further that at each stage of the
derivation a subset LAi is extracted, placed in active memory (the work space),
and submitted to the procedure L. When LAi is exhausted, the computation may
proceed if possible. Or it may return to LA and extract LAj, proceeding as before.
(Chomsky 2000: 106)
Crucially, only when a given subarray has been exhausted can the next subarray
be accessed, which can be thought of as a form of cyclicity in lexical access. With
lexical arrays in place, the Numeration becomes a more structured object, which
instead of being a set (as in (9a)), is a set of sets (as in (9b)). These subsets are the
subarrays we have just learned about.
(9) a. N = { a, b, c, d, e, f }
b. N = { {a, b},{c, d},{e, f} }
Now, the choice between Merge and Move (and the preference for Merge) arises
only if the relevant subarray contains both the expletive and the potentially
movable DP. What this means for our problematic example (5) above is that the
Merge over Move principle does not apply since the expletive there and the DP
people are in two different subarrays, as shown in (10).
(10) N = { {there, is, a, strong, likelihood}, {that, many, parrots, will, be, at, the, clay
lick, right, now } }
The only way to check the EPP feature of T is thus to move many people to [Spec,
TP], which will eventually lead to the grammatical string (11b).
(11) a.
TP
DP i[3pl],uC[Nom] T
many parrots
T EPP u[3pl] VP
will
be many parrots at the clay lick right now
b. There is a strong likelihood that many parrots will be at the clay lick right now.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
29
Phasehood properties
29
Given the focus of this book on phases, the question is: what is the relationship between subarrays (the need for which arises from problems relating to the
MOM principle) and phases? In Minimalist Inquiries, this relationship is fairly
straightforward. Chomskys characterization of phases is simply in terms of
subarrays:
(12) A phase of a derivation is a syntactic object derived . . . by choice of LAi.
(Chomsky 2000: 106)
Such a characterization of phases only raises the question of how lexical subarrays (LAi) are chosen. Otherwise, the argument is circular: phases are determined relative to subarrays, and subarrays are determined relative to phases. To
avoid such circularity, we need an independent way to characterize phases (or
subarrays). This is the focus of the next section.
Chomsky actually describes subarrays in those terms in Minimalist Inquiries, but this seems to be a
better characterization of phases. A subarray is a set, so it cannot be a syntactic object that can have
the semantic type of a proposition or be movable, for example. A phase, on the other hand, being a
hierarchically organized object, can have these properties.
Even though Chomsky does not explicitly address the phase status of verb phrases involving
ditransitive verbs, it seems reasonable to group them with transitive ones. I will come back to
them in Chapter 5.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
30
30
Introducing phases
By analogy, one could also make a case that unergative verbs are not complete in
terms of their argument structure (because they lack an internal argument which
can sometimes be realized (as in (15b)), or that many transitive verbs are not
complete in terms of their argument structure because they have double object
counterparts, as shown in (16b).
(15) a. John ran.
b. John ran a race.
(16) a. John baked a cake.
b. John baked Mary a cake.
This is different from the issue of whether the phase head itself can be a target of operations from
outside the phase (as opposed to the issue of whether the phase head can trigger any operations). The
Phase Impenetrability Condition, the focus of the next section, determines how much of a given
phase is accessible from the outside.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
31
31
(18) The head of a phase is inert after the phase is completed, triggering no further
operations. (Chomsky 2000: 107)
The idea of phase heads being the loci of uninterpretable features is often viewed
as a denitional property of phase heads. This is the view articulated explicitly by
Gallego (2010), Legate (2012) and Miyagawa (2011), among others, whose
formulations are given in (19ac).
(19) a. Uninterpretable features signal phase boundaries. (Gallego 2010: 151)
b. A C/v possessing an unvalued -feature in the numeration is a phase head.
(Legate 2012: 239)
c. Case identies phase heads. (Miyagawa 2011: 1273)
Let us start with the original version of PIC, from Chomskys (2000) Minimalist
Inquiries.
(21) Phase Impenetrability Condition
In phase with head H, the domain of H is not accessible to operations
outside , only H and its edge are accessible to such operations.
(Chomsky 2000: 108)
{DESKTOP}NEW FOLDER//9781107040847C02.3D
32
32
Introducing phases
means if there are any features that are unvalued (or unchecked), they are going to
remain so. The general conguration PIC refers to is shown in (22a), and applied
to a more concrete clausal conguration in (22b); HP/vP is the phase (with H/v
being a phase head), YP/VP is the Spell-Out domain, and H/v and its specier
constitute the phase edge. In (22ab), as well as in the tree diagrams that follow,
the spelled-out portions are faded.6
(22) a.
PHASE
XP
X
HP
H
H
YP
SPELL-OUT DOMAIN
EDGE
b.
PHASE
TP
T
vP
DP
VP
SPELL-OUT DOMAIN
EDGE
Thus the only way something can move out of the Spell-Out domain is if it gets to
the phase edge rst. This is how the Phase Impenetrability Condition, in conjunction with (23) which states that phase heads have the requisite features to
trigger movement forces movement out of the phase to proceed through the
phase edge.7
(23) The head H of phase Ph may be assigned an EPP-feature.
(Chomsky 2000: 109)
I borrow the convention from Hornstein, Nunes Grohmann (2005). They, however, use it to mark
checked features rather than spelled-out domains.
The EPP feature is also sometimes referred to as the P-feature (Periphery feature) (see Chomsky
2000: 144, note 5).
{DESKTOP}NEW FOLDER//9781107040847C02.3D
33
33
oat, to name a few). If, for example, a wh-phrase in a long-distance question passes
successive-cyclically through the edge of vP or CP (the edge being there because of
the extra EPP (or P) feature, as per (23)), we should be able to see evidence of it
having passed through that edge. This is illustrated schematically in (24).
(24) [CP WH [C C . . . [vP WH [v v . . . [CP WH [C C . . . [vP WH v [VP V WH ] ] ] ] ] ] ] ]
As briey mentioned above, there are at least two major versions of the Phase
Impenetrability Condition: one from Chomskys (2000) Minimalist Inquiries
and the other one from Chomskys (2001) Derivation by Phase. The two are
juxtaposed in (25ab). I will refer to the two versions as a Strong PIC and a Weak
PIC, or PIC1 and PIC2, respectively (following Chomsky 2001, Mller 2004,
Richards 2011, among others, who also provide a more detailed discussion of the
differences between them).8
(25) a. The domain of H is not accessible to operations outside HP
[emphasis mine, B.C.]; only H and its edge are accessible
to such operations.
STRONG PIC/PIC1
b. The domain of H is not accessible to operations at ZP
[emphasis mine, B.C.]; only H and its edge are accessible to
such operations.
WEAK PIC/PIC2
(Chomsky 2001: 1314)
The two denitions differ with respect to when the domain of the phase head H
becomes inaccessible: as soon as HP is complete versus at the point the next
phase head (Z) is merged. Thus the two versions impose different restrictions
on how long (in derivational terms) the elements inside the complement
domain (complement of H) can be probed. If the phase HP is directly dominated by Z (another phase head), the two versions of PIC are empirically
equivalent. Both state that at the level of ZP, only the edge of H (i.e. H itself, its
speciers and adjuncts) are accessible to Z. However, the two versions of PIC
diverge in their empirical consequences in congurations of the kind given in
(26a) below, where there is a non-phase head between the two phase heads Z
and H (and corresponding phrases), given in bold.
8
Mller (2004), however, argues in favor of a yet different type of PIC, given in (i) below, which he
dubs PIC3, in which movement proceeds through the edge of every phrase (and in which every
phrase is thus a phase). I take here the more restrictive (and the more standardly assumed)
approach to phasehood, in which only some phrases are phases. I also refrain from digressing
into a detailed comparison of PIC3 with its predecessors, I refer the interested reader to Mllers
work instead.
(i) Phrase Impenetrability Condition (PIC3)
The domain of a head X of a phrase XP is not accessible to operations outside XP; only X and its
edge are accessible to such operations.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
34
34
Introducing phases
(26) a.
XP
X
HP
PHASE 1
H
H
YP
SPELL-OUT DOMAIN
EDGE
b.
ZP
Z
PHASE 2
XP
X
HP
PHASE 1
H
H
SPELL-OUT DOMAIN
YP
The two versions of PIC also make different predictions with respect to what X
can agree with in (26ab). Under PIC1, X cannot agree with YP (or anything
inside YP), since YP is spelled out as soon as X is merged, as shown in (26a).
Under PIC2, however, X can agree with YP since YP is not spelled out until Z is
merged, as shown in (26b).
A concrete illustration comes from the relationship between T and a direct
object. According to PIC1, T cannot agree with DP, as shown in (27). According
to PIC2, T can undergo such an Agree relationship, since the complement of v
becomes inaccessible only at the next phase level, the CP level, as shown in
(28b). The two versions of PIC do not differ in their predictions regarding Agree
between T and the elements at the edge of vP (the external argument, shifted
object if present, and the v head itself ).
(27)
TP
Tu[ ]
v
Agree between T
and DP blocked
vP
VP
SPELL-OUT DOMAIN
...DPi[val],uC[
]...
{DESKTOP}NEW FOLDER//9781107040847C02.3D
35
(28) a.
35
TP
Tu[val]
vP
VP
Agree between T
and DP possible
b.
DPi[val],uC[val]
CP
C
TP
Tu[val]
vP
VP
SPELL-OUT DOMAIN
...DPi[val],uC[val]...
The reason to believe that this type of Agree should be possible comes from
languages that allow so-called quirky Nominative objects. On the assumption
that Nominative case is a result of Agree between a nite T and the object in
question, VP cannot be spelled out before Agree takes place. This is thus an
argument in favor of PIC2 over PIC1. Illustrative examples from Icelandic and
Polish are given in (29ab), respectively.9, 10
(29) a. Henni hfu
leist
eir.
her.DAT had.3PL bored.at they.NOM
She had found them boring. (Sigursson 2002: 692)
b. Marii
podobaa sie ta
ksiazka.
Maria.DAT please
REFL
Maria liked this book.
9
this
[Icelandic]
[Polish]
book.NOM
It is clear that T, rather than some unorthodox form of v, values Nominative case on the object,
given that in Polish it remains Nominative under negation. If v were involved in valuing the case of
the (Nominative) object, we would expect Genitive case instead (so-called Genitive of Negation)
under negation. This is what happens obligatorily with Accusative objects:
(i) Marii
Maria.DAT
nie
podobaja
sie
kwiaty/*kwiatw.
not
please
REFL
owers.NOM/*GEN
[Polish]
nie
lubi
kwiatw/*kwiaty.
not
likes
owers.GEN/ACC
Many of the examples with Dative subjects and Nominative objects involve so-called psych-verbs,
which is interesting in itself but does not change the fundamental insight about case valuation (see
Belletti & Rizzi 1988 for a classic early account of psych-verbs, and Pesetsky 1995 and Landau
2010, among many others, for more recent ones).
{DESKTOP}NEW FOLDER//9781107040847C02.3D
36
36
Introducing phases
Abstracting away from the non-trivial issue of how Agree can take place
across a Dative experiencer in such a conguration, Agree between T and DP is
only possible under PIC2; under PIC1,VP is spelled out as soon as T is merged. A
similar issue arises in expletive constructions; if the associate of the expletive
bears Nominative case and if v in this case is also a phase head (a somewhat
controversial assumption, as we will see in Chapter 5), the example in (30) also
requires Agree between T and the VP internal argument, schematized in (30).11,12
(30) a. There arrived a train.
b. [CP C [TP there T u[3sg] [vP [v [VP arrived [DP a train]i[3sg],uC[Nom] ] ] ] ] ]
Again, if the VPs are spelled out as soon as T is merged, the case feature on the
associate of the expletive cannot be valued.
Richards (2011) argues that the differences between the two versions of PIC
are reducible to which subarray contains T. He contrasts two general approaches
to phasehood, which he dubs the subarrays approach versus all-powerful head
approach, and argues in favor of the subarray-based one (see, however, the
discussion in Section 2.2 above). Since in principle T could belong either to the
subarray headed by C or to the one headed by v, depending on which of
the two it is, the effect is either PIC1 or PIC2.
(31) a. N= { {C, T},{v, V} }
b. N = { {C},{T,v},{V} }
STRONG PIC/PIC1
WEAK PIC/PIC2
However, given the empirical problems associated with PIC1 (as well as the
inherent circularity in dening phases in terms of subarrays and subarrays in
terms of phases), it is not clear why we would ever need both versions.
Richards (2011) points out that PIC1 might be preferable conceptually, as
PIC2 increases the search space, thus increasing computational complexity
11
The assumption that the associate is Nominative seems problematic in light of the examples
involving pronouns, such as the one in (i), in which the case of the associate is Accusative rather
than Nominative. This led Belletti (1988) and Lasnik (1992) to a different proposal, relying on socalled partitive case. Interestingly though, even in such cases the associate determines agreement,
as shown by the contrast between (ii) and (iii).
(i) There arrived him/*he.
(ii) There arrives him.
(iii) There arrive them.
12
The so-called transitive expletive constructions, such as the one in (i) from Icelandic, are not a
problem, since the Nominative argument is at the phase edge, as shown in (ii).
(i) a
there
boruu
sennilega
margir
jlasveinar
bjgun.
ate
probably
many
Christmas.trolls
the.sausages
[Icelandic]
Many Christmas trolls probably ate the sausages.(Bobalijik & Jonas1996: 230)
(ii) [CP C [TP there T u[3pl] [vP [DP many Christmas trolls]i[3pl]uC[Nom] [v [VP eaten [DP the sausages] i[3pl]
uC[Acc]]
]]]]
{DESKTOP}NEW FOLDER//9781107040847C02.3D
37
37
(see also Mller 2004). Richards also notes, together with Bokovi (2002),
Abels (2003), Matushansky (2005), Boeckx & Grohmann (2007) and Epstein
(2007) among others, that there is a mismatch between what is a phase and what
is transferred to the interfaces. Phases are identied by their heads, but it is the
complement of the phase head that gets transferred to the interfaces. Richards
system avoids this problem. For him, the phase conceived in terms of subarrays
is the Spell-Out unit.
An important question about the Phase Impenetrability Condition, which
has been given a fair amount of attention in the literature on phases, is
whether all syntactic mechanisms are subject to it or constrained by it. The
two most relevant syntactic mechanisms (arguably, the only two) are Move/
Internal Merge and Agree. The discussion so far implies that they both
have to obey PIC. However, this conclusion is not uncontroversial (see, for
example, Bokovi 2003, 2007, Bhatt 2005 for claims to the contrary).
Bokovi, based on data of the following sort, argues that Agree into nite
clauses is possible, and, consequently, that Agree cannot be subject to PIC.
In what follows, we will refer to this type of Agree as Long Distance Agree
(LDA).
(32) nan qlilu lnrk -nin-et [iqun -rtm v-nen-at qora-t]. [Chukchee]
he-INSTR regrets-3-PL
that 3SG-lost-3-PL
reindeer-PL(NOM)
He regrets that he lost the reindeers.
(Bokovic 2003: 57, citing Inenlikj & Nedjalkov 1973)
In (32), the matrix verb is plural. Since the only plural element in the clause is
the embedded object, Bokovi takes this to indicate the presence of an Agree
relationship between the matrix v and the embedded object. This leaves us with
the question of whether such LDA is universally possible, or subject to parametric variation. Bokovi takes the latter view, attributing the difference
between languages like Chukchee and ones like English to the presence or
absence of -features on CP. If CPs have -features, which he takes to be the
case in English, LDA is impossible (presumably, due to intervention effects, as
a violation of Attract Closest).
The phenomenon of LDA crossing a nite CP boundary is also attested in
Tsez, another language Bokovi discusses. His analysis, however, departs
from Polinsky & Potsdam (2001), from whom he draws his data. (33a) illustrates what Polinsky and Potsdam dub Properly Local Agreement (PLA);
(33b) illustrates the LDA pattern we are interested in here. Polinsky and
Potsdam, arguing against specierhead-based accounts of agreement in general, reanalyze LDA as involving covert topic movement. The element triggering agreement moves to the edge of the embedded clause, from which it is
accessible to Agree with the matrix V-v complex. In (33a), the verb agrees with
the embedded clause. In (33b), which is an example of LDA, the verb agrees
with the embedded object.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
38
38
Introducing phases
(33) a. enir
[ua magalu
bacrui] r-iyxo.
[Tsez]
mother [boy bread.III.ABS ate].IV
IV-know
The mother knows [the boy ate the bread].
The mother knows the boy ate the bread.
b. enir
[ua magalu
bacrui] b-iyxo.
III-know
Mother [boy bread.III.ABS ate]
The mother knows [the boy ate the bread].
The mother knows the boy ate the bread. (Polinsky & Potsdam 2001: 584)
[HindiUrdu]
HindiUrdu is slightly different from the Chukchee and Tzez data we have just
seen in that LDA is only possible into innitival clauses, which leads Bhatt to
correlate LDA with restructuring. It also leads him to dissociate case from
agreement; in LDA examples, the embedded object agrees with the matrix verb
but does not get case from it.13
A slightly different version of the Phase Impenetrability Condition, which also
allows LDA into embedded CPs, comes from Landaus work on control (see also
witkos 2010a, b).14 In a series of works, Landau develops a theory of control, often
referred to as Agree-based Calculus of Control (see Landau 1999, 2000, 2003,
2008), in which the interpretation of PRO is determined by what it undergoes Agree
13
14
In order to account for Hindi LDA, Bhatt (2005) also reformulates Agree (dubbing it AGREE). His
version of AGREE differs from the standard Agree in that it allows the Goal to have no unvalued
features. In standard Agree cases, having unvalued features (uCase features in most cases we
have seen in this section) was a denitional characteristic of a Goal; this was the property that
made it active.
This is slightly different from the issue of Agree being exempt from PIC. Here we are dealing with
PIC being reformulated in a way that allows certain elements to agree with outside others their
phase (without being at the phase edge).
{DESKTOP}NEW FOLDER//9781107040847C02.3D
39
39
with. In Landaus system, the relationship between PRO and its controller is
mediated via Agree, but, crucially, not necessarily Agree between PRO and its
controller. Rather, he proposes two mechanisms of control, which correlate with the
distinction between Partial and Exhaustive control.15 The difference between
the two types of control, which Landau convincingly argues are both subspecies
of Obligatory Control, is illustrated by the contrast between the example in
(35a), illustrating Partial Control, and the one in (36a), illustrating Exhaustive
Control. The former allows a mismatch in plurality between a (singular) controller
(i.e. matrix subject) and a plural embedded PRO. The plurality of PRO, which
Landau indicates by the subscript 1+, is evidenced by the fact that it is possible with
distributive predicates like gather, meet or do something together. Both Exhaustive
and Partial Control involve a number of Agree relationships. Interestingly though,
neither involves a direct Agree relationship between the controller and PRO. In
Partial Control cases, such as the one in (35a), none of the three Agree relationships
violates PIC.
(35) a. Maria1 preferred [PRO1+ to go for a walk together].
b.
TP
DP
T
vP
Maria
v
VP
preferred
CP
C
TP
T C PRO
T
T
vP
However, the Agree Landau posits for Exhaustive Control is different and it does
involve Agree across a CP boundary, in violation of the Phase Impenetrability
Condition. This is the Agree between the matrix T and the PRO in the embedded
[Spec,TP] position.
15
The distinction between Exhaustive and Partial Control is a matter of lexical selection. Some verbs
are inherently specied as selecting Partial Control complements, whereas others are specied as
selecting Exhaustive Control complements.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
40
40
Introducing phases
b.
TP
T
DP
T1
vP
v
Maria
VP
v
managed
CP
C
TP
T
PRO
T
vP
PRO go for a walk
(37)
VP
managed
CP
TP
T
PRO
T
vP
PRO go for walk
Thus, in order to allow Agree between T and PRO in such cases, Landau modies
the Phase Impenetrability Condition. He does not take the route Bokovi took
(essentially exempting Agree from PIC). Instead, he allows Agree into the
complement domain under certain circumstances, namely when the Goal in this
16
17
The only difference between subject and object control in both Exhaustive and Partial Control
cases is that the rst Agree relationship is between v and the object (rather than T and the subject).
As far as PIC is concerned, the same problem arises: the Agree between v and PRO in Partial
Control cases also violates it.
Or even earlier if we assume the earlier version of PIC, on which the complement of a phase head
becomes inaccessible as soon as the next head is merged, which in this case is the matrix V.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
41
Multiple Spell-Out
41
The possibility of Agree not being subject to the PIC raises a very fundamental question concerning the relationship between Transfer and PIC.
Implicit in the discussion so far has been the assumption that the reason the
complement of a phase head becomes inaccessible at a certain point in the
derivation (i.e. when the next phase head is merged) is the fact that it has been
transferred to the interfaces. From this perspective, Agree that violates PIC,
which would be Agree after Transfer, should never be possible. But is the
conclusion that PIC determines Transfer a logical necessity? Could it be that
PIC determines locality but not Transfer? For the proposals we discussed in
this section, PIC determines both Transfer and locality. This, however, is only
one possibility. Another possibility would be to assume that in certain cases
(such as the cases of LDA considered here), Transfer to the interfaces is simply
delayed. This would keep the concept of Transfer intact, in that once an object
is transferred, it is no longer visible to syntactic processes such as Merge,
Move or Agree, but would allow delayed Transfer, the possibility we will
discuss in more detail in Chapter 6.18
18
This is essentially what Richards (2012) does by taking Spell-Out to occur when the complement of
a phase head contains only interpretable features.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
42
42
Introducing phases
fact that (most) derivations contain more than one phase, Spell-Out is going to
occur more than once per derivation. In this respect, phase-theoretical syntax
constitutes a major departure from its predecessors (minimalist and preminimalist alike), where there was only one point at which syntax could
communicate with phonology and semantics. The difference is illustrated
schematically in (39) and (40).
(39)
SINGLE SPELL-OUT
C-I
SM
(40)
MULTIPLE SPELL-OUT
C-I
SM
C-I
SM
C-I
SM
Logically speaking, the Transfer to each interface could happen at different times
(the Non-Simultaneous Spell-Out of Marui 2005, for example), as shown
below.19
(41)
SM
C-I
SM
C-I
SM
C-I
SM
SM
SM
C-I
19
Since there is only one point of Transfer under the so-called inverted T (or Y) model illustrated in
(39), non-simultaneous Transfer was not an option.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
43
Multiple Spell-Out
43
b.
C-I
C-I
C-I
SM
(43)
YP
Y
X
ZP X
WP
With the concept of Multiple Spell-Out in mind, let us look at how the
theoretical pieces we have seen so far come together. The example we will look
at is given in (44a); its Numeration, given in (44b), contains two phase heads,
indicated in bold. This means that that there are two subarrays, each containing
one phase head. If DP is a phase a possibility which I abstract away from until
Chapter 4 the Numeration will be (44c) instead.
(44) a. Parrots like nuts.
b. N = { {C1, T1}, {Parrots1, v1, like1, D2, nuts1} }
c. N = { {C1, T1}, {D1, parrots1}, {v1, like1}, {D1, nuts1} }
20
It also helps reconcile Bare Phrase Structure with Kaynes (1994) Linear Correspondence Axiom.
If the specier becomes one big lexical item, it is not an issue that its subparts are not ordered with
respect to the elements in the rest of the structure.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
44
44
Introducing phases
The relevant features are given in (45), and the derivation proceeds along the lines
given in (46):
(45) CiForce[Decl]
TiT[pres], u[ ],EPP
DuC[ ]
parrotsi[3pl]
vu[ ]
nutsi[3pl]
(46) a. Merge D with nuts
DPi[3pl], uC[
D uC[
nutsi[3pl]
VP
like
DPi[3pl], uC[
DuC[ ]
nutsi[3pl]
c. Merge VP with v
Agree between v and DP21
vP
vu[3pl]
VP
like
DPi[3pl]uC[Acc]
DuC[Acc]
nutsi[3pl]
v
v
like vu[3pl] like
V
DPi[3pl], uC[Acc]
DuC[Acc]
21
nutsi[3pl]
The Agree between v and the direct object results in v getting its -features valued by the
object. This Agree is realized overtly in languages in which verbs agree with both subjects and
objects.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
45
Multiple Spell-Out
45
DPi[3pl], uC[
DuC[
parrotsi[3pl]
vP
DPuC[
],i[3pl]
VP
D parrots v
like vu[3pl] like
DPi[3pl], uC[Acc]
DuC[Acc]
nutsi[3pl]
g. Merge T with vP
Agree between T and [DP parrots]
TP
TEPP,u[3pl]
vP
DPuC[Nom],i[3pl]
D
parrots v
VP
DPi[3pl],uC[Acc]
DuC[Acc]
nutsi[3pl]
TP
T
DPuC[Nom],i[3pl]
parrots v
like vu[3pl] like
VP
DPi[3pl],uC[Acc]
DuC[Acc]
nutsi[3pl]
{DESKTOP}NEW FOLDER//9781107040847C02.3D
46
46
Introducing phases
CP
C
TP
DPuC[Nom],i[3pl]
vP
v
DP
D
parrots v
VP
DPi[3pl], uC[Acc]
nuts
{DESKTOP}NEW FOLDER//9781107040847C02.3D
47
Feature Inheritance
47
to assume that it is not the same kind of feature as all other features, and is more
of a second-order feature in the Adger & Svenonius (2011) sense. If we treat
EPP as a second-order feature, it becomes less of a mystery that it lacks an
interpretable counterpart.22
Taking T to be a phase head instead would also cause both conceptual and
empirical problems. Conceptually, TPs are not complete or independent at the
interfaces in the same sense that CPs are. Empirically, if TP were a phase, we
would encounter similar problems to the ones we encountered in the previous
section with the earlier version of PIC, so-called PIC1. If T were a phase head,
however, the same problems would arise even with PIC2 in place. In particular,
Agree between T and a VP-internal object (which is what is needed in (47a))
would never be possible:
(47) a. Henni hfu
leist
eir.
her.DAT had.3PL bored.at they.NOM
She had found them boring. (Sigursson 2002: 692)
b.
[Icelandic]
TP
Tu[
Agree between T
and DP blocked
vP
VP
...DPi[val], uC[ ]...
If T inherits its uninterpretable -features from C, we can reconcile the idea that
phase heads are hosts of uninterpretable features with the idea that T undergoes
Agree with the subject (and sometimes with the object), thus valuing the unvalued/uninterpretable features of both.
A corollary of the claim that phase heads have uninterpretable features is the
claim that derivations are evaluated at phase levels. Independent evidence in
favor of this claim comes from locality considerations. The idea that locality is
evaluated at phase levels eliminates the need for equidistance, which arose from a
locality violation in run-of-the-mill wh-questions. A simple formulation of equidistance is given in (48).23
22
23
Adger (2003) treats EPP as a D-feature. This is a bit of an oversimplication given that other
elements besides DPs can check it.
The original formulation of the equidistance principle (from The Minimalist Program) is given
below:
(i) If , are in the same minimal domain, they are equidistant from . (Chomsky 1995: 184)
See Chomsky (1995: 17780) for a detailed explanation of what counts as a minimal domain.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
48
48
Introducing phases
At the point in the derivation illustrated in (49b), Agree between T and the subject
is blocked by the intervening wh-object. The movement of the subject over the
intervening wh-object is also blocked by its presence. Equidistance, which made
speciers of the same head equidistant from any higher Goal, was essentially a
stipulation to allow such movement (or Agree) to take place.
(49) a. What do parrots like?
b.
CP
TP
TEPP,u[ ]
vP
whati[3sg], uC[Acc]
Agree between T
and subject
blocked
parrotsuC[
vP
],i[3pl]
vEPP
VP
like
what
24
The equidistance principle was set against a different theoretical backdrop, with Agreement
projections still in place, and before the emergence of concepts like Probe, Goal and Agree. The
basic logic behind it has not changed, though.
The idea that all operations take place simultaneously at the phase level also removes the
countercyclicity in Holmbergs Generalization; the verb had to move to T before the object
could undergo object shift. Chomsky (2001) addresses the countercyclic nature of head movement, which together with other problematic aspects of head movement (such as the fact that
the moved head does not c-command its trace and is subject to different locality conditions than
other types of movement) led him to suggest that head movement is phonological.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
49
Feature Inheritance
49
wh-phrases cannot remain in [Spec,vP] and that the landing site of overt object
shift cannot be [Spec,vP], either.25
(50)
CP
C
CiForce[Q], EPP TP
T
Ti[3sg],
uC[Acc] vP
x
whati[3sg], uC[Acc] vP
parrotsuC[Nom ],i[3pl]
vEPP
VP
like
what
Feature Inheritance thus provides a concrete way of implementing the idea that
operations happen simultaneously at the phase level. If uninterpretable -features
of T start out on C and get inherited by T from C, there is simply no way for T to
have these features before C is merged.
This is what Chomsky (2008) proposes. Richards (2008) argues that Feature
Inheritance is obligatory and deduces its obligatoriness (as stated in (51)) from
the interplay of the two independently motivated principles, Value-Transfer
Simultaneity (given in (52a)) and the formulation of the Phase Impenetrability
Condition (given in (52b)).26
(51) Feature Inheritance
uF must spread from edge to nonedge (i.e. from C to T, v* to V, etc.)
(Richards 2008: 567)
(52) a. Value-Transfer Simultaneity
Value and Transfer of uFs must happen together.
b. Phase Impenetrability Condition
The edge and nonedge (complement) of a phase are transferred separately.
(Richards 2008: 5668)
25
26
Chomsky states explicitly that no clear case of stranding in Spec-v is known (Chomsky 2004:
123). This conclusion might be too strong, in light of the data to be discussed in Section 4.2 of
Chapter 4, which does seem to suggest that stranding sometimes is possible in the intermediate
[Spec,vP] position.
Strictly speaking, (52b) is a consequence of PIC not PIC itself.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
50
50
Introducing phases
First, Richards argues that feature valuation and Transfer have to happen at the
same time. The relative timing of these two operations is something that we
have not been explicit about so far. Uninterpretable features (being uninterpretable, thus by denition not readable by the interfaces) have to be deleted by
the time of Transfer. However, if feature interpretability reduces to feature
valuation, with uninterpretable features being simply unvalued (and interpretable features valued), unvalued features that were valued during the derivation
via Agree become indistinguishable from features that were valued to begin
with. In essence, uninterpretable features become indistinguishable from
interpretable features. Thus, once they are valued there is no way to know
that they have to be deleted (unless they come with I acquired my value
versus I was born with my value tags, a somewhat speculative solution, to say
the least). The problem disappears if feature valuation and feature deletion
happen simultaneously.27 This is precisely what Feature Inheritance allows.
There are nevertheless some nontrivial questions that Feature Inheritance
raises, listed in (53).
(53) a. Do all uninterpretable features have to be inherited?
b. Can other features (besides -features) be inherited?
c. Can a non-phase head inherit its uninterpretable features from a phase head
below it?
27
This is also a point Epstein & Seely (2002) make. The conclusion they draw is different. In order to
ensure that feature valuation and Spell-Out happen simultaneously, they propose Spell-Out
happens whenever feature valuation happens.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
51
Feature Inheritance
51
complementizer agrees with the embedded subject, in (54b) it agrees with the
matrix subject, and in (54c) the verb moved to C agrees with the wh-phrase in
the specier of CP.
(54) a. Kpeinzen dan-k (ik) morgen
goan.
[West Flemish]
I.think
that-I (I)
tomorrow go
I think that Ill go tomorrow. (Carstens 2005: 222, citing Haegeman 1992)
b. Alfredi ka-bol-el-a
baba-ndu a-li
ba-kha-khil-e.
[Lubukusu]
1Alfred 1S-said-AP-FV 2-person 1-that 2S-FUT-conquer
Alfred told the people that they will win. (Diercks 2011: 1)
c. Bik
bi--ks-l-
bbo bkulu
mwm mu-mwlo? [Kilega]
8what 8CA-A-give-PERF-FV 2that 2woman 1chief 183village
What did those women give the chief in the village? (Carstens 2005: 220)
(55)
28
The complementizer agreement is different from the so-called oating inection (discussed
by Booij & Rubach 1987, Embick 1995, Franks & Baski 1999, among many others),
illustrated in (i).
(i) My-smy
we-AUX.1PL
wyjechali
Wrocawia
rano.
leave.PART.VIR.PL
from
Wrocaw
morning
wyjechali-s
my
Wrocawia
leave. PART.VIR.PL-AUX.1PL
from
Wrocaw
that
nish.PST.PL.MASC
to
widziaes
you
it
saw
(v) Tys
to
widzia.
(iii) Ty
tos
widzia
this
saw
you
{DESKTOP}NEW FOLDER//9781107040847C02.3D
52
52
Introducing phases
DONATE:
SHARE:
KEEP
C
Cu
Cu
Tu
Tu
T
The latter two options are the ones that are problematic if Feature Inheritance is
obligatory.29 Ouali attributes the so-called anti-agreement effect in Tamazight
Berber, illustrated in (57) below, to the ability of C to not pass the uninterpretable -features to T (the KEEP option given in (56c)). The anti-agreement effect
refers to the fact that in examples involving subject extraction, subjectverb
agreement is blocked. The data below provide a concrete illustration. (57a) is
the base example showing that verbs typically agree with subjects, (57b) shows
that this is impossible if the subject is wh-moved, and (57c) shows that in such
cases the non-agreeing participial form surfaces (hence the term antiagreement).
(57) a. thamttut thla
araw.
woman
3SF.see.PERF boys
The woman saw the boys.
b. *mani thamttut ag
thla
araw.
which woman
COMP
3SF.see.PERF boys
Which woman saw the boys?
c. mani thamttut ag
lan
araw.
which woman
COMP
see.PERF.PART boys
Which woman saw the boys? (Ouali 2008: 164)
[Tamazight Berber]
Ouali argues that in (57b) T does not inherit any -features from C, and, as a
result, subjectverb agreement is impossible.
Another plausible case of C keeping its uninterpretable -features comes
from Legates (2011) analysis of subject-initial orders in V2 languages.
Legate refers to it as under-inheritance, but it is equivalent to Oualis KEEP option.
What is interesting about subject-initial V2 clauses is that the subject behaves as
if it were in an A (rather than A-bar) position, in spite of the fact that it occupies
the [Spec,CP] position. The evidence that the [Spec,CP] subject position is an Aposition comes from reconstruction effects. Legate presents the following contrast from Dutch, due to Craenenbroeck & Haegeman (2007), to show that the
29
Ouali also argues that the three options are ordered with respect to each other, but nothing here
hinges on this argument.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
53
Feature Inheritance
53
Interestingly, as pointed out by Richards (2011), PIC2 (the weaker version) is informulable with
Feature Inheritance in place. If the Merge of C is what triggers the Spell-Out of VP, the
Nominative object becomes inaccessible when C is merged. However, T does not inherit its
uninterpretable -features, which are the features that allow it to undergo Agree with the object,
until C is merged.
{DESKTOP}NEW FOLDER//9781107040847C02.3D
54
54
Introducing phases
a.
CP
C
TP
FIFA
Tu
b.
vP
VP
Vu
(60)
a.
VP
FIFB
CP
Vu
b.
TP
Tu
vP
v
The question now is whether the two FIFB options in (60) are theoretically
possible and empirically justied. There does not seem to be anything that
excludes them on theoretical grounds; if anything, they seem conceptually
more plausible if we pursue the analogy between Feature Inheritance and movement. The FIFA options in (59a) and (59b) would be the options corresponding to
lowering movement operations and the FIFB options would correspond to standard (upwards) movement operations. Nevertheless, let me end with a piece of
speculation concerning the potential empirical support for (60a). Perhaps this
is what happens in Subject-to-Object Raising (SOR) / Exceptional Case
Marking (ECM) congurations, where the matrix V inherits uninterpretable
features from the embedded C, in addition to inheriting them from the matrix v.
The latter option was suggested by Chomsky (2008). On either account, SOR
targets [Spec,VP]; however, FIFB might provide a new take on the old puzzle of
why some languages allow SOR and others, such as Polish (as shown in (61ab)),
do not (see Bokovi 1997, Dziwirek 2000, Kayne 1981, Lasnik 1998, Rooryck
1997, among others, for a discussion of crosslinguistic variation with respect to
SOR).
{DESKTOP}NEW FOLDER//9781107040847C02.3D
55
Feature Inheritance
(61) a. *Maria
uwaza
Jana
byc
madrym.
Maria.NOM considers Jan.ACC be.INF smart.INSTR
Maria considers Jan to be smart.
b. *Maria spodziewa sie Jana
byc
na czas.
Maria
expects
REFL
Jan.ACC be.INF on time
Maria expects Jan to be on time.
55
[Polish]
This could only work on an independent assumption that control and ECM
clauses are categorically non-distinct and both are CPs (see, for example,
Bokovi 1997 and Pesetsky 1991). This seems problematic given the defective
tense of ECM clauses (see Stowell 1981 and Martin 1996, among others). But
maybe this defective nature of CP is what motivates FIFB. This is not what
happens in control clauses, whose CPs bear irrealis tense. Perhaps more generally, this is impossible in Slavic languages, whose CPs are more temporally
independent than their English counterparts, as evidenced, for example, by the
lack of Sequence of Tense (SOT) (see Khomitsevich 2007 for a recent discussion
of SOT in Russian, and Lasnik 1998 on the connection between tense and the lack
of ECM, also in Russian).
Let me conclude this chapter with a concrete illustration of how exactly
Feature Inheritance affects Narrow Syntax. In particular, let us look at how the
derivation of the example (44) above, repeated below as (62a), changes with
Feature Inheritance in place. There are two cases of Feature Inheritance to
consider: T inherits its uninterpretable -features from C, and V inherits its
uninterpretable -features from v. What is signicant is that a number of
processes happen simultaneously at phase levels (the steps (63c) and (63h)).
These are: Feature Inheritance, Agree (and feature valuation), movement and
Transfer.
(62) a. Parrots like nuts.
b. N = { {C1, T1}, {parrots1, v1, like1, D2, nuts1} }
(63) a. Merge D with nuts
DPi[3pl], uC[
DuC[
nutsi[3pl]
DPi[3pl], uC[
DuC[ ]
nutsi[3pl]
{DESKTOP}NEW FOLDER//9781107040847C02.3D
56
56
Introducing phases
vP
vu[
VP
likeu[3pl]
DPij[3pl], uC[Acc]
DuC[Acc]
nutsij[3pl]
v
v
v likeu[3pl]
like
DPi[3pl], uC[Acc]
DuC[Acc]
nutsi[3pl]
DPi[3pl], uC[
DuC[
parrotsi[3pl]
],i[3pl]
D parrots
v
VP
likeu,[3pl] v like
DPi[3pl], uC[Acc]
DuC[Acc] nutsi[3pl]
g. Merge T with vP
TP
T
vP
DPuC[
D
],i[3pl]
parrots v
like v
VP
like
DPi[3pl], uC[Acc]
D
nuts
{DESKTOP}NEW FOLDER//9781107040847C02.3D
57
Feature Inheritance
57
CP
Cu[3pl]
TP
DPuC[Nom],i[3pl] T
D parrots
TEPP,u[3pl] vP
DPuC[Nom],i[3pl] v
parrots v
VP
DPi[3pl], uC[Acc]
nuts
j. Transfer CP to PF and LF
In sum, we have seen the motivation for phases stemming from the Merge over
Move principle, as well as reviewed the existing characterizations of phases (e.g.
phases dened in terms of subarrays versus phases dened in terms of the properties of phase heads). We have also seen how the introduction of phases changes the
overall architecture of the grammar in that it results in Multiple (and potentially
Non-Simultaneous) Spell-Out and how the property of phase heads as the loci of
uninterpretable features leads to Feature Inheritance. Throughout this chapter, we
have assumed that certain types of vPs and CPs (and perhaps DPs) are phases,
without providing much evidence for this assumption, and without applying concrete
diagnostics to these categories. This is what we turn to in the next two chapters. In
Chapter 3, we discuss various diagnostics that have been suggested in the literature to
determine the phasehood of a given category, with an eye towards determining
which of these diagnostics are real, and which ones are not. The diagnostics
established in this chapter will serve as the basis for the discussion of specic phases
in the chapters that follow.
31
I assume that the EPP feature of T is also inherited from C. This is not uncontroversial, but it seems
innocuous in the case at hand, where TP is dominated by C and where C does not need any EPP
features of its own. More interesting are cases in which TP is not dominated by C but nevertheless
has an EPP feature, as in the successive cyclic raising example in (i), and cases in which both C and
T have an EPP feature, as in the object wh-question in (ii).
(i) Parrots seem to be likely to like nuts.
(ii) What do parrots like?
{DESKTOP}NEW FOLDER//9781107040847C03.3D
58
3
Phasehood diagnostics
I borrow the title of this section from the title of chapter 1 of Levin & Hovavs (1995) book
Unaccusativity: At the SyntaxLexical Semantics Interface, which is The anatomy of a diagnostic:
the resultative construction.
58
{DESKTOP}NEW FOLDER//9781107040847C03.3D
59
59
I do not think anyone would question the idea that all phases have to be maximal projections, since
maximal projections have the level of independence and completeness that non-maximal projections do not have. However, the question of whether something would go wrong at the interfaces if
phases were not maximal projections, or if phases were not always maximal projections, is
ultimately an empirical question. But allowing non-maximal projections to be phases would devoid
phases of completeness associated with phasehood.
{DESKTOP}NEW FOLDER//9781107040847C03.3D
60
60
Phasehood diagnostics
diagnostics than for others. For example, being a domain for nuclear stress
assignment would be a fairly straightforward example of a PF diagnostic,
whereas being propositional would count as a solid case of an LF diagnostic.
But some are harder to classify. For example, is being a target for Quantier
Raising (QR) or reconstruction a semantic or a syntactic diagnostic? This
depends on whether we think of QR (or reconstruction) as a semantic or a
syntactic process, and we know that there are different views on this matter.
Similarly, if ellipsis is a diagnostic, should it count as a PF, LF or a syntactic
diagnostic? Surely, ellipsis has an effect on both pronunciation and interpretation, and depending on whom we ask, it is treated as either a PF process (literal
deletion) or an LF process (reconstruction of missing material at LF).
Difculties in classication notwithstanding, I will proceed on the assumption that phasehood diagnostics can indeed be subdivided into three types:
syntactic, semantic and phonological, with the caveat that sometimes the
distinctions between the three types might be less than perfect. I will thus
not attach too much importance to this classication and leave open the
possibility that the same diagnostic can belong to more than one group, and,
for example, be both a syntactic and a semantic diagnostic. It is not even
clear whether the issue of classication is a substantive, as opposed to a
terminological issue. After all, does it really matter whether a given diagnostic
is a PF, LF or a syntactic phasehood diagnostic, as long as it is a reliable
phasehood diagnostic?
The two more concrete questions we can ask about phases concern the
properties of phase heads. If phases are syntactic engines, to borrow
Richards (2011) term, the following questions come to mind:
(2) a. Does X trigger Spell-Out?
b. Is X the source of uninterpretable features?
The answer to (2a) will help us diagnose the PF and LF status of phase; if phase
heads determine Spell-Out, the domain that is spelled out is a domain that will
feed phonology and semantics. We might therefore expect both phonological and
semantic processes to apply to this domain. Sections 3.2 and 3.3 will thus be
devoted to PF and LF phasehood diagnostics, respectively. The afrmative
answer to (2b), on the other hand, will help us uncover syntactic diagnostics; if
only phase heads are inherently endowed with uninterpretable features, then they
are going to trigger operations that rely on these features. This will lead us to
concrete syntactic diagnostics in Section 3.4.
3.2 PF diagnostics
This section focuses on the questions of what it means for a phase to be
independent at PF, what it means for a phase head to trigger Spell-Out, and
more importantly how we know that a phase is PF independent or has
{DESKTOP}NEW FOLDER//9781107040847C03.3D
61
PF diagnostics
61
Perhaps the most natural potential PF diagnostic is the one in (3a), involving
phonetic isolability. Simply put, according to this diagnostic, if XP can stand
alone (and function as a response to a question, for example), XP is a phase. The
diagnostic in (3b) says that if a given XP can be moved (by processes like
fronting of any kind, extraposition, clefting etc.), it should count as a phase.
However, these are standard constituency tests (as any introductory syntax
textbook can conrm), so a much larger number of categories pass as phases
according to this diagnostic. Matushansky (2005) also notes that the ability to
be isolated or undergo movement does not seem to be a very reliable phasehood
diagnostic (although she nevertheless applies it to DPs).3 The reason why we
might be tempted to think of it as such is that taking phonetic independence to
be a property of phases has an intuitive appeal. However, both isolability and
movement are broader characteristics of phrases, or perhaps of constituents
even more generally. Thus, example (4a) shows that DPs can stand alone as
sentence fragments. However, the examples in (4bf) show that many other
phrases can do so as well.
(4) a. What did Mary study?
a. [DP Western philosophy]
b. How many tests did Mary take to become a doctor?
b. [NumP Two]
c. Where did Mary go?
c. [PP To Cambridge]
d. What does everyone believe?
d. [CP That the earth is at]
e. What did Tom try?
e. [CP To be a good syntactician]
f. What color is this page?
f. [AP White]
Matushansky also considers extraposition, clefting and though-movement as PF phasehood diagnostics. She starts of by comparing TPs, CPs and vPs, and then applies the same diagnostics to test
DPs for phasehood.
{DESKTOP}NEW FOLDER//9781107040847C03.3D
62
62
Phasehood diagnostics
Since different types of verb phrases have been claimed to differ in terms of their
phasehood status (something we will discuss in detail in Section 4.2 of
Chapter 4), I am going to keep them distinct for the purposes of the (putative)
diagnostics considered here.
(6) a. What Mary did was [vP read Chomskys Syntactic Structures]
b. What John did was [vP/VP arrive late].
c. *What John did/was [vP/VP given a book].
UNERGATIVES
UNACCUSATIVES
PASSIVES
The examples in (7ag) show that quite a few types of phrases can undergo
movement:
(7) a. [DP This new movie], we all admired.
b. [PP To this store], John went after the meeting.
c. [TP To become a top syntactician], Mary is very likely.
d. [CP To become a top syntactician], John really wanted.
e. [CP That the earth is not at], nobody doubts.
f. *[AdjP Interesting], we read books.
g. *[NumP Five], we read books.
All three verb phrase types under consideration here behave alike in this respect:
(8) a. [vP Read Chomskys new manuscript], all linguists should.
b. [vP/VP Arrive at the station], the train should have by now.
c. [vP/VP Eaten by sharks], nobody wants to be.
UNERGATIVES
UNACCUSATIVES
PASSIVES
And the examples below make a similar point with respect to Right Node Raising
(RNR):5
(9) a. John read and Bill reviewed [DP a new article on RNR].
b. John knows and Bill believes [CP that syntactic phases exist].
A reasonable question to ask is what type of derivation pseudoclefts involve; in particular whether
they involve movement of the pivot. Derivational details notwithstanding, it is clear that there is a
pause before the pivot of the pseudocleft, which is suggestive of some degree of phonetic
independence. Similar considerations apply to Right Node Raising.
See also Bokovi (2002) for a discussion of RNR (of TPs in particular) as an argument against
treating isolability as a diagnostic.
{DESKTOP}NEW FOLDER//9781107040847C03.3D
63
PF diagnostics
63
c. John should and Bill wont [vP read a new article on RNR].
d. John should but Bill wont [vP/VP arrive on time].
However, RNR seems to be a very permissive diagnostic, given that RNR can
even apply to subwords:6
(10) a. My analysis over and your analysis under-generates.
b. Ex and current smokers have high blood pressure.
c. Sheila likes pre- and post-colonial architecture. (Barros & Vicente 2010: 9)
This shows that isolability (the ability to stand alone and the ability to undergo
movement) is not a very reliable diagnostic. It is also not clear that there are any
reasons to expect that movement or the ability to stand alone should count as
phasehood diagnostics to begin with. There is nothing in the way syntactic
derivations proceed under phase-theoretical assumptions that should make us
believe that only phases should be able to undergo syntactic movement, or that
only phases should be able to stand alone. For example, we could conclude from
the grammaticality of sentence fragments of the kind given in (11b) that AdjP are
phases. But we could also conclude that they involve a more complex (phasal)
structure with some portion deleted (namely (11c)). Needless to say, this makes
the diagnostic vacuous, unless we have a clear sense of what it means for a given
string to stand alone.
(11) a. What color is the sky today?
b. [AdjP Grey].
c. [CP it is grey]
Phase Theory also does not seem to make any predictions about the movability of a given constituent. It only makes predictions about the movement
path (namely, the Phase Impenetrability Condition states that movement out of
a phase is possible only if it proceeds through the phase edge). Thus, even
though both the ability to stand alone and the ability to function as a unit for the
purposes of movement seem like reasonable and intuitive ways to understand
phonetic independence, there is nothing in the theory that leads to such a
prediction; the conditions imposed by the interfaces do not lead to a prediction
that only phases should be able to move or stand alone. However, the ability of
phase heads to determine Spell-Out does lead to two concrete diagnostics: one
having to do with determining phonological domains and the other one having
to do with ellipsis.
If phase heads determine Spell-Out, they should also determine domains for
indisputably PF-related processes, such as nuclear stress assignment (and, arguably, linear order). We will discuss this in more detail in Chapter 7, where we
focus on the syntaxphonology interface and the role phases play at this interface.
For now, sufce it to note that given the role phase heads play in determining
6
{DESKTOP}NEW FOLDER//9781107040847C03.3D
64
64
Phasehood diagnostics
This is not to say that being a complement of a phase head is sufcient to license ellipsis. Rouveret
(2012), for example, proposes the licensing condition in (i) to capture crosslinguistic variation with
respect to VPE.
(i) Licensing Condition on VPE
VPE is available in a given structure if, and only if, vs uninterpretable [tense] feature is valued at
the v-level.
He argues that in some languages, the tense feature is valued at the v level, whereas for others it is
valued at the T level.
This view of ellipsis does raise questions about other ellipsis types, such as gapping, pseudogapping, Right Node Raising or argument ellipsis, which do not readily lend themselves to being treated
as involving deletion of a complement of a phase head. For some of them, there are alternative
treatments that do not involve ellipsis. For example, Right Node Raising has been reanalyzed in a
multidominant fashion (see Citko 2011a, b and the references therein), and gapping has been argued
to involve ATB movement (see Johnson 2000, 2009). Logically speaking, it could also be the case
that ellipsis is not directly linked to Spell-Out, and, consequently, the grammar has to include some
process directly responsible for ellipsis that goes beyond spelling out the complement of a phase
head as null. Aelbrecht (2009), for example, argues explicitly against the claim that only phase
heads can license ellipsis.
{DESKTOP}NEW FOLDER//9781107040847C03.3D
65
PF diagnostics
65
This is not the only view of ellipsis that has been proposed within the connes
of Phase Theory. Bokovi (to appear) argues that either complete phases or
complements of phase heads can be elided. To illustrate with a concrete example,
he argues, following ener & Takahashi (2009), that (13b) involves ellipsis of a
complete noun phrase, since it allows for both a strict and sloppy reading, the
latter being characteristic of ellipsis.9
(13) a. Taroo-wa sannin-no sensei-o
Taro-TOP
three-GEN teacher-ACC
Taro respects three teachers.
b. Hanako-mo e sonkeisiteiru.
sonkeisiteiru.
respects
[Japanese]
Hanako-also
respects
(Lit.) Hanako respects e, too.
(Bokovic to appear: 22, citing Sener & Takahashi 2009: 3)
Possessors, particles and quantiers can escape ellipsis, which Bokovi takes to
mean that complements of phase heads can also be elided. Interestingly, he notes
that when both a quantier and a possessor is present, the quantier can escape
ellipsis but the possessor cannot:
(14) *B sensei-wa hotondo-no Ziroo-no tikoku-o
yurus-anakat-ta.
Prof. B-TOP
most-GEN
Ziro-GEN tardiness-ACC forgive-NEG-PST
Prof. B didnt forgive most of Ziros tardiness.
(Bokovic to appear 28, citing Takahashi 2001)
This is related to an independent claim that phases are dynamic in that the highest
projection in a given extended projection constitutes a phase. In (14), the highest
projection is the QP hosting the quantier most. This is the projection that
constitutes a nominal phase. The possessor occupies a specier of a lower
(non-phasal) head. In Bokovis theory, (14) is ungrammatical because it
involves ellipsis of a complement of a complement of a phase head. We will
discuss the possibilities of phases being dynamic in more detail in Chapter 6. For
now, let me conclude the discussion of PF diagnostics by listing the questions that
survived the scrutiny of this section; positive answers to these questions classify a
given constituent as a PF phase. I do so in (15); the answers to (15b) and (15c)
establish the answer to (15a).
(15) a. Does X trigger Spell-Out?
b. Does XP constitute (or determine) a prosodic domain?
c. Can the complement of X be elided?
See, however, Merchant (2013) for a discussion of some issues with considering sloppy identity a
reliable diagnostic of ellipsis.
{DESKTOP}NEW FOLDER//9781107040847C03.3D
66
66
Phasehood diagnostics
3.3 LF diagnostics
A common characterization of phases is that, in addition to being independent at
PF, they should also be independent at LF. What this means is that they should be
interpretable and complete in some semantically relevant sense. However, as
often pointed out, it is not very clear what it means to be complete in a semantic
sense. Furthermore, what counts as complete can mean different things depending on the category. One possibility, explored by Matushansky (2005), would be
to equate semantic completeness with being saturated; in other words, being of
type e or t. From this perspective, only elements of type e (individuals) or type t
(truth values) would be phases. Thus only certain types of DPs would be phases:
proper names (being of type e), but not quantied DPs (being of type <<e,t >,t>).
On the Generalized Quantier theory, no DP would ever be a phase, since all DPs
are of the unsaturated type <<e,t >,t>, as also noted by Matushansky (2005).
Furthermore, even if we take semantic saturation to be a viable diagnostic, it is not
going to be a reliable one, given that the semantic type of a given constituent can
change if something moves out of it.
Chomsky often refers to a slightly different kind of completeness: the
kind associated with a complete argument structure. However, it is not clear
that an unaccusative verb is any less complete than an unergative one; both
require only one argument, yet Chomsky only treats unergative (and transitive)
vPs as phases. Treating unergatives as hidden transitives with a hidden object,
along the lines of Hale & Keyser (1993), provides a possible solution to this
problem.
There are other (perhaps more concrete) diagnostics that are arguably also
semantic, which I list in (18). One has to do with reconstruction, and the other one
with QR. Classifying them as semantic is somewhat controversial, since they
both have to do with overt movement out of XP, which is undeniably a syntactic
process. However, since they deal with the interpretation of elements moving out
of phases, it seems innocuous enough to treat them as semantic (or both syntactic
and semantic).
(16) a. Can an element moving through the edge of XP be interpreted at the edge of
XP?
b. Is XP a target for Quantier Raising?
10
{DESKTOP}NEW FOLDER//9781107040847C03.3D
67
Syntactic diagnostics
67
not clear that this is a general diagnostic that can be applied to test the phasehood
of other categories, I will not rely on it here as a diagnostic, but will come back to
it in Section 7.2, which deals with the syntaxsemantics interface and the role
phases play at that interface.
Quantier Raising might seem odd as a phasehood diagnostic, given the
classic view that QR targets TPs, non-phases on standard assumptions (see May
1977, 1985, for example). There are, however, many independent factors
involved (such as semantic type compatibility, economy or locality considerations), which make it impossible for QR to target many phrases, but the availability of QR in cases where these factors do not intervene does help us establish
that movement can target the edge of a given a category, a property associated
with phases. Bruening (2001), for example, argues that QR is subject to economy
in that it targets the closest constituent of the right semantic type (type t). Thus, vP
is always going to be the closest semantically appropriate QR landing site for vPinternal quantiers. This can explain clause-boundedness of QR; since vP is
always closer than a CP, CP is never a potential landing site, and successive
cyclic movement from the edge of the embedded vP to the edge of matrix vP
would be banned by the Phase Impenetrability Condition. Thus PIC, as shown by
Cecchetto (2004), can explain clause-bounded nature of QR. Simply put, if QR
could target CPs, there would be no way to explain it. QR should be able to take
place successive-cyclically, the way wh-movement takes place.
The question is whether movement out of XP has to proceed through the edge of
XP. I will refer to this property of phases as the edge property.11 This general
question can be broken up into more specic questions about the effects of such
movement, listed in (18).12
11
12
This is slightly different from perhaps the more typical usage of the term edge property, which
takes it to be the feature that triggers movement to the edge.
These are familiar from the vast literature on successive cyclic movement (see Boeckx 2003,
2008a, Felser 2004 and the references therein, for more detailed discussion of successive cyclic
movement, and Lahne 2008 in particular for a discussion of these (and other) diagnostics in the
context of Phase Theory).
{DESKTOP}NEW FOLDER//9781107040847C03.3D
68
68
Phasehood diagnostics
It may well be the case that some of these diagnostics are phonological and
some are semantic, or that they cannot be classied as falling neatly into one
(or the other) class. As we saw above, deciding whether a given diagnostic is a
PF, LF or a syntactic diagnostic might not matter for all practical intents and
purposes as long as we assume phase integrity, take the same categories to
count as both PF and LF phases, and take Transfer to PF and LF to occur at the
same time. This is not uncontroversial; we alluded to the possibility of nonsimultaneous Transfer/Spell-Out in Section 2.5, and we will discuss it in more
detail in Chapter 6.
There are two other diagnostics that do not seem to be easily classiable
as either PF or LF diagnostics, as they do not capitalize on the idea that
phases are independent at the interfaces. Both have to do with uninterpretable
features. If phase heads drive syntactic computation, and uninterpretable
features are the syntactic engines, uninterpretable features have to be associated with phases. Thus the questions in (19ab) can also be used to diagnose
phases.
(19) a. Is XP a domain for feature valuation?
b. Is X the source of uninterpretable features?
If phase heads are the only locus of uninterpretable features, and as such, they
literally drive syntactic computation, the only way a non-phase head such as T or
V can acquire uninterpretable feature is through Feature Inheritance. We have
seen above (in Section 2.3) that in many accounts, being the locus of uninterpretable features is in fact the denitional property of phases. Recall the Phase
Condition of Gallego (2010), repeated below.
(20) Phase Condition
Uninterpretable features signal phase boundaries. (Gallego 2010: 151)
{DESKTOP}NEW FOLDER//9781107040847C03.3D
69
Syntactic diagnostics
69
best without an account of islands that is complete with its basic tenets. And
islands have indeed received a fair amount of attention in phase-theoretical
literature (see, for example, Mller 2011 and Gallego 2010 for specic proposals
on how certain islands can be derived). However, since our focus here is on
introducing Phase Theory and justifying phases themselves, an adequate discussion of islands from a phase-theoretical perspective, which merits a book-length
discussion of its own, would take us too far astray.13 Let me nevertheless comment briey on the status of islands in Phase Theory.
Locality, which is captured by means of the Phase Impenetrability Condition, is
an important aspect of Phase Theory but, strictly speaking, there is no direct
correlation between islandhood and and phasehood. This makes islands not very
useful for the purposes of diagnosing phases. First, it is not simply the category of a
given phrase (CP versus TP, or DP versus NP) that determines whether it is an island
or not. For example, only certain types of CPs and DPs are islands, namely interrogative CPs and denite DPs. Appealing to a lled edge position (lled [Spec,CP]
position and a lled [Spec,DP] position in these two cases) is not going to be helpful,
since adding an extra specier to a phase head is always an option, and a lled
specier does not always lead to island violation (as also pointed out by Boeckx &
Grohmann 2007). For example, a lled [Spec,vP] does not make that vP an island.
All Phase Theory says is that movement out of phases has to proceed through phase
edges; it does not say that movement out of phases (or non-phases, for that matter) is
impossible. Furthermore, as we know, the answer to the question of whether a CP or
DP is an island depends on many independent factors. DP is an island only for
certain movements (in certain languages, as we will see in Chapter 4): movement of
a specier of D, for example, leads to an island violation but movement of (or from)
the complement of D only does so sometimes. Furthermore, whether DP or CP is an
island depends on what positions it occupies: if it is a complement, movement out of
it is possible, but if it is a specier or an adjunct, it is not. This is Huangs (1982)
Condition on Extraction Domains, which distinguished objects on the one hand,
from subjects and adjuncts on the other hand, by appealing to proper government, an
obsolete theoretical concept. In phase-theoretical terms, CED effects have been
derived from the Phase Impenetrability Condition, combined with certain assumptions about features and activity (see Mller 2011 and Gallego 2010 in particular).
To illustrate briey, Mller proposes that the (optional) EPP feature (edge feature in
his terms) can be added to a phase head only if the phase head is still active, where
being active means not having had all its features checked. This includes features
checked by both Internal and External Merge. Extraction from subjects is banned in
his system because the subject is the last element merged in the vP phase. This means
that the EPP feature cannot be added after the subject is merged. At that point, the v
head is no longer active because it has had all its features checked. Without that EPP
feature, movement is going to violate the Phase Impenetrability Condition. Adding
13
I refer the interested reader to Mller (2011) and Boeckx (2012) for book-length discussions of
islands specically.
{DESKTOP}NEW FOLDER//9781107040847C03.3D
70
70
Phasehood diagnostics
the EPP feature before the subject is merged is possible because the v head is still
active; this is what allows movement of the object (or out of the object). But this is
not going to help with extraction out of the subject since the subject has not been
merged yet.
Setting islands aside, let me conclude the current chapter by listing the
three types of diagnostics, as they are the ones that will be used as the basis of
the discussion in the next two chapters, where we look at specic categories and
determine whether they pass as phases according to these diagnostics. In the next
chapter (Chapter 4), we will look at three classic phases CPs, vPs and DPs
whose phasehood status is fairly well established in the literature on phases.
(21) PF phasehood diagnostics
a. Does X trigger Spell-Out?
b. Does XP determine phonological domains?
c. Can the complement of X be elided?
(22) LF phasehood diagnostics
a. Can an element moving out of XP be interpreted in the edge position?
b. Is XP a target for Quantier Raising?
(23) Syntactic phasehood diagnostics
a. Can an element moving out of XP be pronounced (partially or completely) at the
edge?
b. Can an element moving out of XP strand anything at the edge?
c. Is XP a domain for feature valuation?
d. Is X the source of uninterpretable features?
This is different from proposals that claim that every phrase is a phase (as suggested by Mller
2004, for example).
{DESKTOP}NEW FOLDER//9781107040847C04.3D
71
4
Classic phases
In this section we will also look at the properties of the C head, particularly its
featural makeup (whether it has any uninterpretable features and whether it
1
Many of the phenomena I discuss in this chapter are also discussed by these authors, and are
generally well-known from the literature on successive cyclic movement. Lahne (2008), for
example, also discusses reconstruction, agreement, quantier oat, partial movement and whcopying. Her conclusions, however, are different from the conclusions reached here, as she takes
all phrases to be phases. Felser (2004) focuses on partial movement and wh-copying. Legate (2003)
discusses reconstruction, parasitic gaps, QR and nuclear stress assignment (focusing on vPs rather
than CPs).
However, there is more to establishing whether a CP (or any other category) is a phase than
establishing whether it has the edge property, and we have seen many phasehood diagnostics
that do not make reference to edges. We will discuss non-edge diagnostics in this section
as well.
The title of Lahnes (2008) dissertation Where There Is Fire There Is Smoke captures the spirit of
the diagnostics considered here; we are looking for the smoke.
71
{DESKTOP}NEW FOLDER//9781107040847C04.3D
72
72
Classic phases
triggers Spell-Out). Both of these properties are associated with phase heads only.
We will defer the discussion of CPs as potential phonological domains till
Chapter 7.
Perhaps the most straightforward illustration of the edge property comes from
the domain of long-distance wh-movement of the kind represented schematically
in (2). In such congurations, the wh-phrase moving out of the lowest CP has to
pass through (at the very least) each [Spec,CP]. For now, we will not worry about
whether it has to pass through any other positions; this is the issue we turn to in
the next section.
(2) [CP WH [C C [TP
v
v
V
TP Spell-Out
CP
C
TP
... WH...
{DESKTOP}NEW FOLDER//9781107040847C04.3D
73
CPs as phases
73
There are a number of ways to tell that wh-movement does indeed target the
edge of CP. Perhaps the simplest one is to look at what happens if the edge is
lled. And, indeed, unless a language has some other strategy (such as the
availability of multiple speciers), the result is ungrammatical and constitutes a
familiar example of a wh-island violation:
(4) a. *What did you think [CP when Jan devoured what when]?
b. *This is the book which Jan wondered [CP who wrote which].
c. *How did you wonder [CP who xed the sink how]?
Interestingly, languages can vary in whether they obey the Wh-Island Condition,
as noted by Rizzi (1982) for Italian and by Rudin (1988) for Bulgarian, for
example. The examples in (5ab) show that in these languages movement from
wh-islands does not result in ungrammaticality.
mi
domando che
storie abbiano
(5) a. tuo fratello, a cui
your brother, to whom myself ask.1SG which stories have.3PL
raccontato. [Italian]
told
your brother, to whom I wonder which stories they told (Rizzi 1982: 50)
oveka, kojto
b. C
the man who.REL
ne znae
NEG
kakvo kazvat
know.2S what
ce
kupil
[Bulgarian]
The man who you dont know what they say that he bought.
The common explanation for Bulgarian is that its CPs allow multiple speciers,
as evidenced by the well-known the fact that Bulgarian (along with other Slavic
languages) is a multiple wh-fronting language.4 From a phase-theoretical perspective, what such languages teach us is that a phase edge can contain multiple
speciers (at least in those languages in which all wh-phrases move to [Spec,CP],
such as Bulgarian and Romanian).5 This explanation, however, is not going to
work for Italian, which is not a multiple wh-fronting language, and, to make
things even more intriguing, is not even a language that allows English-style
multiple wh-questions with one of the wh-phrases in situ, noted by Calabrese
(1984) and discussed more recently by Stoyanova (2008).6 The classic account of
the grammaticality of wh-island violations in Italian is that of Rizzi (1982); it
4
The literature on multiple wh-fronting is vast, starting with Wachowiczs (1974) observation that
there are indeed languages that can front all their wh-phrases overtly, and Rudins (1988) seminal
work on the distinction between two types of multiple wh-fronting languages, which, in most
general terms, differ with respect to whether they treat all the fronted wh-pronouns alike or
privilege the highest one. See also Boeckx & Grohmann (2003), Bokovi (2002b) and Richards
(2001), among others, for more recent proposals and renements of the early empirical
generalizations.
We have already seen that v, another candidate for a phase head, also allows multiple speciers. There is
a stage in the derivation in which the subject is in the inner specier and the object in the outer one.
Stoyanova also includes Somali, Irish and Berber in her discussion of languages with no multiple
wh-questions, and attributes the lack of multiple wh-questions to the existence of a unique focus
position in such languages.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
74
74
Classic phases
On a methodological note, it is worth noting that what looks like the same phenomenon can have
very different explanations from language to language.
The phenomenon of reconstruction is quite complex; there are many factors that play a role, such as
the differences in reconstruction possibilities between A and A-bar movement, arguments and
predicates, and questions and relative clauses, to name just a few. The very assumption that
reconstruction is a syntactic mechanism is far from being uncontroversial. There is also a possibility
that different binding principles might exhibit reconstruction behavior. For example, Lebeaux
(2009) argues that Binding Principles A, B and C apply at different points in the derivation (contra
the standard minimalist view, on which they all apply at LF only), and that the negative principles
(Principles B and C) have to apply throughout the derivation while the positive principle (Principle
A) applies at LF. As he himself admits, examples of the following sort are a problem, as before
movement they violate Principle C. He resorts to a kind of late insertion mechanism to avoid such
problems.
(i) John seems to himself to like cheese.
Since my goal here is not to argue for a specic analysis of binding, I will abstract away from these
controversies. All that matters for our diagnostic purposes here is that reconstruction be a syntactic
mechanism, indicating the presence of a lower copy.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
75
CPs as phases
75
(6) Jan4 asked [CP2 [which picture of himself2,3,4]1 Piotr3 thought [CP1 which picture of
himself1 that Adam2 liked [which picture of himself2,3,4]1.
Fox (1999, 2000) looks at slightly more complex cases to establish that
in successive cyclic movement contexts, wh-pronouns can be interpreted in
intermediate [Spec,CP] positions. Building on the work of Lebeaux (1994)
and Heycock (1995), he looks at the interaction of scope reconstruction with
Binding Principle C, and concludes that reconstruction feeds Principle C. In
cases where reconstruction would lead to a Principle C violation, reconstruction
is blocked. Fox examines successive cyclic movement cases in which the whphrase undergoing movement contains both an R-expression and a bound
pronoun.9 The pronoun, in order to receive a bound variable interpretation,
has to be c-commanded by the quantied expression that binds it. The
R-expression, on the other hand, cannot be c-commanded by the pronoun
coindexed with it. In the conguration given in (7a) there exists a position
(indicated by ) in which the wh-phrase can be c-commanded by the quantied
DP but not by the pronoun. This is not an option in (7b), as reconstruction to
either of the two available positions will result in a Principle C violation.10
(7) a. [which. . . pronoun1. . . r-expression2] . . ..QP1 pronoun2 . . . * . . .
b. [which . . . pronoun1 . . . r-expression2] . . . pronoun2 . . .*. . . QP1 . . .*. . .
(Fox 1999: 173)
The actual examples are given in (8ab). In (8a), the wh-phrase can be interpreted
in the intermediate [Spec,CP] position (indicated by t), in which the pronoun he1
will be bound by every student, but the R-expression Ms. Brown will not be
bound by she2. In (8b), on the other hand, this is not an option; reconstruction
even to the intermediate position (t) will yield a Principle C effect. This is only
possible if movement takes place through the edge of CP, which in turn is only
possible on current assumptions if CP is a phase.
(8) a. [Which (of the) paper(s) that he1 gave to Ms. Brown2] did every student1 hope t
that she2 will read t?
b. *[Which (of the) paper(s) that he1 gave to Ms. Brown2] did she2 hope t that
every student1 will revise t? (Fox 1999: 173, citing Lebeaux 1990)
This suggests the derivation in (9) (rst pass) for long-distance movement.
The questions that we will come back to are whether there are any other
intermediate landing sites, and what exactly drives the movement to the intermediate [Spec,CP] positions. For now, let us assume the answer to the latter
question is some uninterpretable feature with the EPP property.
10
Foxs goal, however, is to establish that Principle C applies at LF, rather than to argue for the
phasehood of CPs.
The star here indicates a position to which reconstruction will result in a Principle C violation.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
76
76
Classic phases
b.
CP
whoiwh[wh]
CiForce[Q],uwh[wh],EPP TP
DPuC[Nom] T
shei[3sg]T u[3sg],EPP vP
she
v
v
VP
CP
who
CiForce[Decl],EPP
TP
DPuC[Nom]
D
Johni[3sg]Tu[3sg],EPP vP
John
v
v
VP
whoiwh[wh]
Another well-known range of arguments in favor of successive cyclic movement taking place through the specier of CP comes from kinds of questions
in which the wh-phrase with a matrix scope can either be pronounced in the
embedded [Spec,CP] or leave something behind in this position. There are a
number of such cases to consider, illustrated below in pseudo-English. (10a)
is allowed only in some varieties of English, (10b) in childs English, and
(10c) and (10d) not at all. In all of them, the scope of the wh-phrase who
or which movie is in the matrix [Spec,CP]; however, the embedded [Spec, CP]
contains some portion of it (in the form of the stranded quantier or the nominal),
the wh-phrase itself (in case of the partial wh-movement construction) or its copy.
(10) a.
b.
c.
d.
wh-quantier stranding
[CP who did John think [CP who all Bill saw who all]
[CP who did John think [CP who Bill saw who]
wh-copying
[CP what did John think [CP who Bill saw who]
partial wh-movement
[CP which movie did John think [CP which movie Bill saw which movie]
split wh-movement
Now, let us turn to some actual data, starting with the phenomenon dubbed whquantier oat by McCloskey (2000) in the varieties of English he refers to
as West Ulster English. What distinguishes West Ulster English dialects from
standard English is that they allow the quantier all to be associated with a whpronoun (as shown in (11a)), and more interestingly for our purposes allow
the quantier to be separated from the wh-pronoun, as shown in (11b). In this
respect, these varieties differ from standard English, which only allows quantier
oat under A-movement, thus providing a nice window into the inner workings
of successive cyclic wh-movement.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
77
CPs as phases
77
Crucially, in long-distance questions, the quantier can appear in the intermediate [Spec,CP] positions (as in (12a)), in addition to being stranded in
its base position (as in (12b)), or pied-piped all the way to the matrix [Spec,
CP] (as in (12c)).
(12) a. What did he say all (that) he wanted?
b. What did he say (that) he wanted all?
c. What all did he say (that) he wanted t? (McCloskey 2000: 61)
11
12
The stranding analysis is by no means the only analysis of quantier oat. Alternatives include
treating the oated quantier as an adjoined adverbial of sorts. I refer the interested reader to
Bobaljik (2003) and the references therein for further discussion. The claim that quantiers can be
stranded in their base positions is also not uncontroversial. Bokovi (2004a), for example, argues
that stranding in thematic positions is impossible.
Abels version of antilocality is different from Grohmanns. Grohmann (2003, 2011) divides clause
structure into what he dubs prolic domains (theta domain, case/agreement domain, operator
domain) and argues that movement within the same domain (as opposed to movement across
domains) is illegitimate.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
78
78
Classic phases
(13) a.
DP
D
whati
D
all
b.
ti
DP
D
whati
D
all
XP
X
ti
We might ask here whether quantiers are the only elements that can be stranded
in intermediate landing sites. Two other phenomena come to mind: preposition
stranding and what we might call nominal stranding. Felser (2004), following
Du Plessis (1977), shows that Afrikaans, for example, allows P-stranding in
intermediate positions. The availability of such intermediate P-stranding would
have to be parameterized, as not all languages that in principle allow P-stranding
allow it in intermediate positions. English, for example, does not, as noted by
Postal (1972) and illustrated in (14ab).13 How to achieve such parameterization
(in a way that attributes it to some lexical property of some lexical item rather than
simply stating it as a general parameter), however, is not a trivial matter. If
stranded prepositions undergo reanalysis (as suggested, for example, by
Hornstein & Weinberg 1981), one way to explain why preposition stranding in
intermediate [Spec,CP] positions is not allowed would be to assume that the
requisite reanalysis is blocked across a CP boundary and to parameterize the
reanalysis itself.
(14) a. *Who/Whom do you believe to Mary thinks Joan talked?
b. *Who/Whom do you believe Mary thinks to Joan talked? (Postal 1972: 213)
Both quantier stranding and preposition stranding are amenable to a nonstranding analysis, in which the entire wh-phrase moves but part of it is
pronounced in [Spec,CP] and part in the pre-movement site, as illustrated
schematically in (15a) for preposition stranding and (15b) for quantier stranding. This does not affect the point made here, which is that movement targets
[Spec,CP].
13
Postal himself uses the ungrammaticality of such examples to argue against successive cyclic
movement, which is not the point I am making here.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
79
CPs as phases
79
[Polish]
(17) a. O
australijskich Maria czytaa papugach.
about Australian
Maria read
parrots
Maria read about Austrialian parrots.
b. O
papugach Maria czytaa australijskich.
about parrots
Maria read
Australian
c. O
jakich
Maria czytaa papugach?
about what.kind Maria read
parrots
What kind of parrots did Maria read about?
[German]
The idea that copies can be pronounced in a discontinuous manner is related to the idea that under
some (fairly restricted) circumstances, pronunciation of a lower copy is forced in languages with overt
movement of the relevant kind. This is the case when the pronunciation of the higher copy would lead
to a PF violation (see Franks 1998, Bokovi 2002b and the references therein).
{DESKTOP}NEW FOLDER//9781107040847C04.3D
80
80
Classic phases
[Child English]
(Thornton 1990: 204)
The Copy Theory of Movement provided a simple explanation of this construction: wh-copying could be thought of as copying with no subsequent
deletion. While there are many non-trivial questions that wh-copying raises
(especially given that movement is no longer thought to be a simple copying
operation), what matters most for our purposes is the evidence for movement
through [Spec,CP] that it provides. First, how to motivate movement through
all the intermediate positions? This is the question Felser dubs the triggering
question. This is the question that arises for all cases of successive-cyclic
movement (not just wh-copying). Second, what allows the embedded CP to
converge in spite of the fact that it seems to contain copies with unvalued
features? After all, only one copy (the highest one) determines the scope of the
wh-question and values the uninterpretable wh-feature (or Q feature) on C. This
is what Felser calls a convergence question. And third, what allows multiple
copies to be pronounced. This is the linearization question. If copies count as
identical elements with respect to linearization, the result is going to violate the
Linear Correspondence Axiom (LCA) of Kayne (1994), which prohibits structures in which the same element both c-commands and is c-commanded by
itself.15 Any account of wh-copying, phase-theoretical or not, must answer
these questions. The convergence question and the linearization question might
appear to be non-distinguishable, but they are actually quite distinct. The
convergence question focuses on why the Spell-Out of a wh-phrase with
some of its uninterpretable features unchecked/unvalued does not violate Full
Interpretation. Note that the wh-phrase does not get its uQ feature valued/
checked till it undergoes Agree with the matrix C (which is the C that bears
the corresponding iQ feature and marks the scope of the wh-question). Felser
solves the convergence problem in the following way. First, she assumes,
following a long line of research going back at least to Katz & Postal (1964)
on the morphosyntax of wh-pronouns, that they can be decomposed into the
15
Even though the specics of the LCA in its original formulation are incompatible with Bare Phrase
Structure theory, the idea behind it (the idea that linear order is not a syntactic primitive) is very
much minimalist in spirit (see, however, Chomsky 1995 and Kayne 2010 for different ways to
make the LCA compatible with BPS).
{DESKTOP}NEW FOLDER//9781107040847C04.3D
81
CPs as phases
81
quanticational part (the wh-part) and the indenite part. Second, she posits
that the two parts can be pronounced in a discontinuous manner, akin to what
saw above in discontinuous DPs (see the Polish examples in (16) and (17)).
We also alluded to it being a possible analysis for preposition stranding and whquantier stranding. In a copying construction, in a way that is similar to split
wh-questions, the indenite part is pronounced in the embedded [Spec,CP] and
the quanticational part is pronounced in the specier of the matrix [Spec,CP].
It just happens to be the case that the two parts sound the same, so to speak, in
the languages under consideration here, namely in those that allow wh-copying.
A derivation incorporating this insight, including only the relevant features for
the sake of simplicity, is given in (19).16
(19)
CP
whoiWH,iINDEF
C
C
TP
you
T
T
Spell-Out whoiWH
vP
you
v
v
VP
think v think CP
whoiWH,iINDEF C
C
TP
John
vP
Spell-Out whoiINDEF
John
v
v
VP
saw v saw
whoiWH,iINDEF
16
This also leads Felser to argue for a convergence-based view of phases. In (29), the uninterpretable wh-feature of the wh-pronoun is not going to be checked until the wh-pronoun reaches the
matrix [Spec,CP]. This suggests that the embedded CP cannot be spelled out before this feature is
valued.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
82
82
Classic phases
literature on scope marking is vast; see, among others, Bruening (2004), Dayal
(1993), Felser (2004), Horvath (1997), Lahne (2008), Lutz, Mller & von
Stechow (2000), Manetta (2010), McDaniel (1989), Stepanov (2000) and the
references therein. The examples below, in which both the real wh-phrase and the
scope marker are given in bold, provide an illustration.
(20) a. Was glaubt Hans [mit wem
Jacob jetzt spricht]? [German]
what believe Hans with whom Jacob now talks
Who do you think Maria has spoken with? (cf. McDaniel 1989: 569)
b. Jak myslisz
[co
Maria zrobi]?
[Polish]
how think.2SG what Maria do.3SG
What do you think Maria will do?
c. Kak vy
dumaete, [kogo ljubit
Ivan]?
[Russian]
how you think.2PL whom love.3SG John
Who do you think John loves? (Stepanov 2000: 1)
d. Mit
gondolsz, [hogy kit
ltott
Jnos]?
[Hungarian]
who.ACC saw.3SG John.NOM
what.ACC think.2SG that
Who do you think that John saw? (Horvath 1997: 510)
17
18
Interestingly, Felser reports that some speakers accept wh-copying with prepositional phrases.
Felser discusses other differences as well, such as the ability to conjoin two embedded clauses
and their compatibility with verbs like seem. What matters for our purposes is the fact that both
partial wh-movement and wh-copying provide evidence for movement through the embedded
[Spec,CP].
{DESKTOP}NEW FOLDER//9781107040847C04.3D
83
CPs as phases
83
latter option is favored by Mahajan (1996) for Hindi and by Stepanov (2000) for
Polish and Russian.
(21)
a.
CP
C
WHATiwh[wh]
CiForce[Q],uwh[wh],EPP
CP
who iwh
CiForce[Decl], uwh[wh]EPP
TP
who iwh
b.
CP
C
CiForce[Q],uwh[wh],EPP
DP
WHATiwh
whoiwh
CP
C
CiForce[Decl], uwh[wh]EPP
TP
whoiwh
19
In addition to relative clauses and wh-questions, which are perhaps the two best-known cases, we
see the same complementizer in equative clauses, comparative clauses and clefts (see McCloskey
2001 for examples and a detailed analysis).
{DESKTOP}NEW FOLDER//9781107040847C04.3D
84
84
Classic phases
(22) a. an ghirseach a
ghoid na siogai.
[Irish]
the girl
aL stole the fairies
the girl that the fairies stole away
b. rud
a
gheall
tui
a
dheanfa.
thing aL promised you aL do.COND-S2
something that you promised that you would do (McCloskey 2001:678)
Bik
bi--ks-l-
bbo bkulu
mwm
8what 8CA-A-give-PERF-FV 2that 2woman 1chief
What did those women give the chief in the village?
mu-mwlo?
[Kilega]
183village
(Carstens 2005: 220)
In long-distance questions, both the matrix and the embedded verb agree with the
wh-pronoun in class. This also shows that the wh-phrase has passed through the
embedded [Spec, CP].
(24)
Bik
bi--tnd-l
bna
bi--gl-l
nina-b?
8what 8CA-A-say-PERF 2child 8CA-A-buy-PERF mother-their
What did the children say their mother had bought?
(Carstens 2005: 247)
20
{DESKTOP}NEW FOLDER//9781107040847C04.3D
85
CPs as phases
85
if they inherit them from phase heads above. Such Feature Inheritance is not a
very likely option for C, especially in matrix contexts, where there is no head
(phase or non-phase) dominating it. And in embedded contexts, where C is
typically dominated by a non-phase head such as a V or an N, the presence of
uninterpretable features on C is highly indicative of C being a phase head. The
phenomenon of complementizer agreement (illustrated above with Bantu, West
Germanic and, in a slightly different form, Slavic data) also suggests that the
uninterpretable -features do not have to be obligatorily inherited by T, and that
languages can vary in whether these features stay on C (yielding complementizer
agreement) or migrate down to T (yielding subjectverb agreement). These are
two of the three options proposed by Ouali (2008): SHARE and KEEP, given in
(25ab). If C shares its features with T, the result is both complementizer agreement and subject verb agreement. If C keeps its features, the result is just
complementizer agreement.
(25) a. Cu
b. Cu
Tu SHARE
T KEEP
{DESKTOP}NEW FOLDER//9781107040847C04.3D
86
86
Classic phases
(28) a. I know that Maria talked to someone but I dont know [CP whoi [TP Maria talked
to ti].
b. Whoi did Maria talk to ti?
On the other hand, Polish (patterning together with other Slavic languages in this
respect) disallows preposition stranding in both sluicing and wh-constructions, as
shown in (29ab):21
(29) a. *Wiem,
ze
Maria
rozmawiaa z
kims,
[Polish]
know.1SG that Maria
talk.PST.3SG with
someone
ale
nie wiem
[CP kimi
[TP Maria rozmawiaa z ti] ].
but
not know.1SG whom
Maria talked
with
I know that Maria talked to someone but I dont know whom?
b. *Kimi Maria rozmawiaa z
ti?
with
Maria talk.PST.3SG with
Who did Maria talk to?
Case provides further evidence in favor of a deletion account. The case of the
fronted wh-pronoun is the case it is assigned within the elided clause. In (31a) it is
Instrumental not Nominative. This also provides an argument against an analysis
of sluicing that derives it from a cleft (illustrated in (31b)), which would incorrectly predict Nominative case instead:
(31) a. Maria zainteresowaa sie czyms,
ale nie wiem
czym/*co.
Maria got.interested REFL something.INSTR but not know.1SG what.INSTR/*NOM
Maria got interested in something but I dont know what.
b. Maria zainteresowaa sie czyms,
ale nie wiem co
Maria got.interested REFL something.INSTR but not know.1SG what.NOM
to
jest
czym
Maria
sie zainteresowaa.
it
is
what.INSTR Maria
REFL get.interested.PST.1SG.
Maria got interested in something but I dont know what (it is that Maria got
interested in).
This concludes our discussion of CPs as phases. The table in (32) summarizes the
behavior of CPs (and C heads) with respect to the phasehood diagnostics
established in Chapter 3.
21
See, however, Szczegielniak (2008) and Stjepanovic (2008) for some interesting counterexamples
to this correlation, from Polish and Serbo-Croatian, and Sato (2011) for similar evidence from
Indonesian.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
87
CPs as phases
87
Answer Evidence
Yes
Yes
Yes
VP-internal elements
inaccessible
sluicing
reconstruction
Yes
preposition stranding
(subject to crosslinguistic
variation)
quantier stranding
wh-copying
scope marking
Yes
Yes
Let me conclude this section with a brief comparison of CPs with TPs with
respect to the same diagnostics. So far, we have assumed TPs not to be phases, but let
us see if the behavior of TPs with respect to the diagnostics actually conrms this
assumption. If TPs were phases (and T heads were phase heads), we would expect a
number of things from them. First, we would expect T heads to trigger Transfer/
Spell-Out. This would put limits on Agree (since on standard assumptions, no Agree
is possible into spelled out constituents). For example, if both T and v were phase
heads in (33a), VP would contain an object whose case feature could not be valued.
(33) a. Piotrowi podobaja sie te
Peter.DAT please
REFL
Peter likes these owers.
b.
kwiaty.
owers.NOM
[Polish]
TP
T
Tu [ ], EEP
vP
DP
v
VP
v
V
DPuC[
{DESKTOP}NEW FOLDER//9781107040847C04.3D
88
88
Classic phases
Second, we would expect ellipsis to be able to target vP. While Merchant does
analyze pseudogapping as vP ellipsis, this could be due to the fact that T might have
inherited whatever features make it a licensor for vP ellipsis from C. Thus we are
only going to look at cases in which T is not dominated by (a nite) C, and it could
not have inherited any features from it. Raising innitives (both subject to subject
raising and exceptional case marking) are a case in point. Unlike control innitives,
they are assumed to be TPs, not CPs. Evidence to support this distinction comes
from tense considerations. Stowell (1982) argues that raising innitives are nonnite and non-tensed (i.e. involve a T[-nite, -tensed]), whereas control innitives are
non-nite but tensed (i.e. involve a T[-nite, +tensed] ). The tense of an embedded
control innitive clause is interpreted as unrealized or future with respect to the
tense of the matrix clause. By contrast, the tense of an embedded raising innitive
clause is interpreted as simultaneous with respect to the tense of the matrix clause.
This is shown by the contrast in (34). In (34a), the time of parrots ying in the
rainforest is typically understood to be future with respect to the time of their desire
to do so. In (34b), on the other hand, the two events (my expectation and the event
of parrots ying) are understood to be simultaneous.
(34) a. Parrots want to y in the rainforest.
b. I expect parrots to y in the rainforest.
See Lobeck (1995) for a classic account of the factors which determine which heads do (and which
ones do not) license ellipsis.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
89
CPs as phases
89
The reason (38a) is grammatical is that which pictures of himself can be interpreted in the intermediate position (bolded in (39)), in which John is the closest
binder for the anaphor himself contained within the wh-trace/copy:
(39) [CP Which pictures of himselfi [C0 did [TP it seem to Johni [CP ti that [TP Mary liked ti
]]]]]?
23
b.
c.
PIC does not discriminate between A-bar and A-movement. The fact that passive movement does
not target the [Spec,vP] as an intermediate landing site does show that passive vPs are not phases, or
at least the same kinds of phases as active vP. This is the point I discuss in the next section.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
90
90
Classic phases
Epstein & Seely (2006) and Grohmann, Drury & Castillo (2000) argue explicitly
against A-movement proceeding in a successive-cyclic fashion. Many of their
arguments are quite theoretical in nature, involving the general status of EPP in
the grammar, or the nature of chains formed by this movement. Empirical
considerations of the familiar kind, such as binding of quantier stranding,
however, do seem to suggest that A-movement takes place through the intermediate speciers of TP.24 Consider the following example:
(42) a. Johni seems to Mary to appear to himselfi to be ill.
b. *Mary seems to Johni to appear to himselfi to be ill.
(Grohmann, Drury & Castillo 2000: 160)
If John and Mary moved directly from their thematic positions, as shown in
(43ab) below, there would be no way to account for this contrast. In both
cases, they would be moving from the positions in which they do not bind the
anaphor.
(43) a. Johnj seems to Maryi [TP to appear to himselfi [TP to be tj ill] ]
b. *Maryj seems to Johni [TP to appear to himselfi [TP to be tj ill] ]
24
Grohmann, Drurg & Castillo (2000) attribute this observation to David Pesetsky, who in turn
credits Danny Fox. They themselves do not take it to indicate successive cyclic movement.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
91
vPs as phases
91
b. [Hisi mothersj bread]k seems to every mani [TP tk to be known by herj [TP tk to be
tk the best there is.]] (cf. Bokovic 2002: 180)
The table in (47) summarizes the behavior of TPs with respect to phasehood
diagnostics.
(47) TPs as phases
Diagnostic Question
Answer Evidence
No
No
No
Yes
No
No
{DESKTOP}NEW FOLDER//9781107040847C04.3D
92
92
Classic phases
and (ii) do not select an external argument. This is Burzios Generalization, which
becomes easy to account for if there are no vPs in passives and unacccusatives.
However, irrespective of whether passives and unaccusatives involve the structures in (49b) and (50b) (with a vP) or the structures in (49c) and (49c) (with no
vP), the question of whether they constitute phases is still valid. It just becomes a
question about passive and unaccusative VPs rather than vPs.
(49) a. Tom arrived.
b.
TP
T
Tu [
],iT[pst], EPP
vP
VP
c.
V
arrived
TP
DPi [3sg,masc],uC[
Tom
T
Tu [
],iT[pst], EPPVP
V
arrived
DPi [3sg,masc],uC[
Tom
b.
TP
T
Tu [
was
], iT[pst], EPP
vP
VP
v
V
chosen
c.
TP
T
Tu [
was
], iT[pst], EPP
V
chosen
VP
DPi [3sg,masc], uC[
Tom
{DESKTOP}NEW FOLDER//9781107040847C04.3D
93
vPs as phases
93
Let us rst look at the properties of v heads in order to determine whether v is the
locus of uninterpretable features and whether v determines Spell-Out.
The answer to the rst question seems fairly straightforward, at least for
transitive (and unergative) vPs. The idea that in transitive clauses v is responsible
for structural Accusative case and that the Agree relationship between v and the
object in its domain, shown in (51ab), values Accusative case is standard, and I
will not dwell on it here.
(51)
a.
vP
vu [
VP
V
read
DPi [3sg],uC[
a book
b.
vP
VP
vu [3sg]
V
read
b.
CP
C
TP
Tom
Tu [3sg], EP vP
did
VP
SPELL-OUT DOMAIN
talk to Bill
Since v in this case is empty, we might wonder why ellipsis couldnt target vP
instead. Merchant (2008) shows that VPE is indeed deletion of a VP (and not a vP),
{DESKTOP}NEW FOLDER//9781107040847C04.3D
94
94
Classic phases
based on the fact that it allows voice mismatches in which the antecedent clause is
active and the elided clause is passive or vice versa.25
(53) a. This problem [VP1 was to have been looked into], but obviously nobody did
[VP2 look into this problem].
PASSIVE VP1 /ACTIVE VP2
b. The janitor must [VP1 remove the trash] whenever it is apparent that it
should [VP2 be removed].
ACTIVE VP1 /PASSIVE VP2
(cf. Merchant 2008: 169)
Merchant accounts for the availability of such mismatched examples by assuming that v (which includes voice information) is not included in the elided portion,
which is what allows the v head in the antecedent clause to be passive and the one
in the elliptical clause to be active, as shown in (54).
(54)
&P
&
CP
but
CP
C
vP
vP
vivoi[pass] VP
look_into
nobody
DP
this problem
vivoi[act]
VP
If pseudogapping, as Merchant (2008) argues (based on the fact that unlike VPE it does not allow
voice mismatches), is vP (rather than VP) ellipsis, the generalization that only complements of
phase heads allow ellipsis breaks down. Bokovi argues that in pseudogapping cases, it is not a v
but a higher (verbal) projection that is a phase head, which makes vP ellipsis the ellipsis of the entire
phase head. Bokovis relativized approach to phases is something we will discuss in more detail
in Chapter 6.
95
{DESKTOP}NEW FOLDER//9781107040847C04.3D
vPs as phases
95
A related question is whether [Spec,vP] is ever the nal (as opposed to the
intermediate) landing site for wh-movement, as shown in (56a) for a monoclausal wh-question and in (56b) for wh-movement that spans across clausal
boundaries.
(56) a. [CP [C C [TP[vP WH [v v . . . [VP . . .WH . . . ]]]]] ]
b. [CP [C C [TP[vP WH [v 0 v . . . [CP WH [C0 C [TP [vP WH [v 0 v . . .WH . . . ]]]]]]]]]]
This is one way to think about so-called short wh-movement, overt whmovement to a clause-medial position. The examples in (57ae) provide an
illustration: this is how Jayaseelan (2001) analyzes Malayalam, Aldridge
(2010) archaic Chinese, Manetta (2010) HindiUrdu, and Citko (2010) a subset
of wh-questions in Polish.26, 27 It is also a plausible way to think of Hungarian
wh-questions.
(57) a. nin-ne aar aTiccu?
[Malayalam]
you-ACC who beat.PST
Who beat you? (Jayaseelan)
b. Wu shei qi?
Qi
tian
hu?
[Archaic Chinese]
I
who deceive deceive Heaven Q
Who do I deceive? Do I deceive Heaven?
(cf. Aldridge 2010: 2, from Analects 9)
c. Hamid-ne kya: par.ha:?
[HindiUrdu]
Hamid-ERG what read
What did Hamid read?(Manetta 2010: 4)
d. Co
Ewa komu
daa?
what.ACC Ewa who.DAT gave
What did Ewa give to whom?(Citko 2010: 46)
e. Jnos kit
mutatott
be Marinak?
[Hungarian]
26
The idea that Polish wh-movement can target a clause medial position goes back to Dornisch
(1998), who also points out that this medial position has to be an A-bar position, based on the fact
that it licenses parasitic gaps:
(i) Za
co
by
Piotr
kogoi
wyrzuci
nie
Piotr
who
throw.out
not
wysuchawszy pgi
przedtem?
listen.PART
before
For what would Peter throw out whom without having listened to? (Dornisch 1998: 160)
for
27
what
COND
However, many researchers working on Hungarian take this position to be a more specic focusdesignated projection (see, for example, Kiss 1987). Cable (2008), however, argues against this
position.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
96
96
Classic phases
28
Example (59) could also involve the derivation in (i), in which there are no deleted copies in the
speciers of CP, just Agree between C heads and the wh-phrases in [Spec,vP]:
(i) [CP [C C [TP[vP WH [v v . . . [CP [C C [TP [vP WH [v v . . .WH . . . ]]]]]]
29
30
Manetta also gives independent evidence to argue against the claim that HindiUrdu is a wh-in-situ
language, coming from adverb placement, for example. The preverbal position that the wh-word
occupies is also a generalized focus position, something which is also the case in other languages
with this type of short wh-movement.
Manetta describes the scope marker kya: as a minimal wh-word.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
97
vPs as phases
97
31
32
{DESKTOP}NEW FOLDER//9781107040847C04.3D
98
98
Classic phases
Conversely, arguments that do not agree with the verb cannot be extracted. This
is the well-known extraction restriction of Austronesian syntax. Aldridge
attributes it to the Phase Impenetrability Condition: if movement out of a
phase proceeds through the phase edge, only elements at the edge can undergo
further movements.
Let us look at a couple of other examples. Chung (1994) analyzes
wh-agreement in Chamorro as agreement with a trace. The wh-agreement
morphology is sensitive to the case of the moved element, which, among other
things, leads her to conclude that wh-agreement is not a reex of C agreeing with
a wh-phrase in its specier. The contrast between (62a) and (62b) provides an
illustration. When the wh-trace is Nominative, the wh-agreement is realized as
the inx -um-; when it is objective (or oblique), it is realized as -in-.
(62) a. Humllum si Maria [na ha-pnak si Juan i
ptgun].
AGR.assume Maria
COMP AGR-spank Juan
the child
Maria assumes that Juan spanked the child.
[Chamorro]
Yet another way to establish the edge property of vPs is to look at stranding
possibilities. The question here is whether quantiers, prepositions or nominals
can ever be stranded in [Spec, vP] positions, the same way they can in [Spec,CP]
positions.
(64) a. [CP WH [C C [TP . . . [vP WH all [v v . . . [VP . . .WH . . . ] ] ] ] ] ]
b. [CP WH [C C [TP. . . [vP P WH [v v . . . [VP . . .WH . . . ] ] ] ] ] ]
c. [CP which [C C [TP. . . [vP which NP [v v . . . [VP . . .WH . . . ] ] ] ] ] ]
McCloskey shows that West Ulster English, a language that allows quantier
stranding in intermediate [Spec,CP] positions (as we saw in the previous section),
disallows analogous stranding in [Spec,vP] positions.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
99
vPs as phases
99
Preposition stranding yields similar results. It should not come as a surprise that
prepositions in English cannot be stranded in [Spec,vP] positions, given that
English does not allow P-stranding in [Spec,CP] positions either:
(66) Whoj did John [vP [PP to tj]i talk ti ] ]?
We saw, however, in Section 4.1 above that Afrikaans does allow P-stranding in
the intermediate [Spec,CP]. However, as noted by Rackowski & Richards (2005),
it does not allow stranding at the edge of vP:
(67) *Wat dink julle [vP oor dink die bure
stry
ons]? [Afrikaans]
what think you about think the neighbors argue we
What do you think the neighbors think we argue about?
(Rackowski & Richards 2005: 593)
kupi
bought
swojej
his.DAT
zonie t]]]]?
wife.DAT
So far we have seen mixed results regarding the pronunciation of the whpronoun (or its remnant) at the edge of vP. On the one hand, we have seen
that it can be the site of the wh-copy in HindiUrdu, that it can be the landing
site of wh-movement in languages like Hungarian, and that the nominal can be
stranded in this position in languages with Left Branch Extraction like Polish.
On the other hand, we have seen that the wh-related quantier all cannot be
stranded in this position and that prepositions cannot be stranded in this
position. I do not take these to be a fatal blow; both phenomena are known to
be sensitive to prosodic factors, which could inuence the grammaticality of
33
{DESKTOP}NEW FOLDER//9781107040847C04.3D
100
100
Classic phases
(65) or (66), as claimed explicitly by McCloskey (1998) with respect to whquantier stranding in West Ulster English. Furthermore, if P-stranding
requires reanalysis, it is also conceivable that such reanalysis is only possible
if the verb is in a certain relationship with the stranded preposition, and the
specier of vP is simply not such a position.
Let us now turn to the evidence for vPs being phases on the interpretative side.
Here the question is whether the wh-phrase can be interpreted at the edge of vP
even if it is not its ultimate resting place. The evidence that bears on this issue
comes from various types of reconstruction effects (see Agero-Bautista 2001,
Fox 1999, 2000, Lahne 2008, Legate 2003 among others).
The reconstruction argument is very similar in spirit to the argument made
in Section 4.1 to establish reconstruction into the intermediate [Spec,CP]
position. The relevant conguration is similar to the congurations we considered in the previous section (where we established that wh-movement targets
[Spec,CP] in a successive-cyclic manner). This time, however, the reconstruction site we are interested in is the edge of vP, not CP. The examples in (69)
provide evidence that wh-movement does indeed proceed through this edge. In
all of them, the fronted wh-phrase contains a bound pronoun and an Rexpression. They are all grammatical because there exists a reconstruction
site, marked with a checkmark, below the quantier and above the pronoun.
This is the edge of the vP.34
(69) a. [Which of the books that he1 asked Ms. Brown2 for] did every student1 __ get
from her2 * ?
b. [Which (of the) paper(s) that he1 wrote for Ms. Brown2] did every student1 __
get her2 * to grade? (Fox 1999: 175)
If we reverse the position of the quantier and the pronoun, as in (70) below, the
results become ungrammatical because there is no reconstruction site such that
Principle C is obeyed and the variable is bound; interpreting the wh-phrase in the
specier of the matrix vP results in a Principle C violation and a lack of variable
binding.
(70) a. *[Which of the books that he1 asked Ms. Brown2 for] did she2 * give every
student1 * ?
b. *[Which (of the) paper(s) that he1 wrote for Ms. Brown2] did she2 get * get
every student1 * to revise? (Fox 1999: 174)
Den Dikken (2006a) raises the possibility that the reconstruction site could also be the edge of TP
(thus providing an argument in favor of TPs being phases). It is not clear whether the moved whphrase could be tucking in below the subject. He also points out that the argument relies on the
assumption that reconstruction is the only way to get the bound variable reading. This does seem to
be the simplest assumption.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
101
vPs as phases
101
crucial assumption that this argument relies upon is that in questions containing a
universal quantier, the wh-phrase has to be interpreted in a position in which it is
c-commanded by the quantier. Thus, the distribution of PL readings in such
questions will help us determine possible reconstruction sites. The baseline
contrast is given in (71ab). In (71a) the PL interpretation is possible because
the wh-pronoun can reconstruct to a position below the quantied subject. In
(84b) there is no such position; hence the PL reading is not available. The only
available reading is Single Answer (SA) reading.
(71) a. Who do you think everyone saw __ at the rally? SA, PL
b. Who __ thinks everyone saw you at the rally? SA, *PL
(Agero-Bautista 2001: 142)
Yet another well-known argument comes from the domain of parasitic gaps. It
relies on a particular account of parasitic gaps, namely one that involves complex
predicate formation, defended in most detail by Nissenbaum (2000), who in turn
builds on Larsons (1988) insights. Such an analysis, Nissenbaum shows, is only
possible if the wh-movement licensing the parasitic gap moves through the edge
of vP. (73) below illustrates the conguration necessary for parasitic gap formation, with irrelevant details omitted.35
35
I refer the interested reader to Nissenbaums work (and the references therein) on how such an
account captures the properties typically associated with parasitic gaps, listed in (i) (and to the
contributions in Culicover & Postal (2000) for a discussion of some counterexamples to these
generalizations). Nissenbaum himself notes, for example, that covert movement can license
parasitic gaps in multiple wh-questions if the overtly moved wh-phrase also licenses a parasitic gap. Null operators might seem suspect and there are quite a few of alternative accounts
that do not involve null operators (or complex predicates, for that matter). See Hornstein &
Nunes (2002) for a sideward movement account, and Kasai (2007) for a multidominant
account.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
102
102
Classic phases
(73)
CP
which paper C
did
TP
T
Johni
T
vP
which paper
vP1
vP2
John file which paper
PP
OPi without PRO reading ti
The null operator movement inside the adjunct PP forms a predicate, and movement of the wh-phrase to the edge of vP2 forms a predicate out of this vP. Thus,
only if vP2 contains a trace of the moved wh-phrase, the vP and PP can combine
yielding a complex (conjoined) predicate via predicate modication. This predicate can subsequently apply to the wh-pronoun. If vP didnt contain a trace of the
wh-phrase, it could not combine with the adjunct containing the parasitic gap,
because it would not be of the right semantic type.
Antecedent-Contained Deletion (ACD) provides yet another argument in
favor of movement targeting the edge of vP. This time, the movement in question
is Quantier Raising (see also Bruening 2001 for arguments that vP can be the
target for QR). The issue is how to avoid the so-called innite regress in cases of
VP ellipsis in which the antecedent VP contains the missing VP (hence the term
Antecedent-Contained Deletion). Consider a simple example in (75a). It contains
a missing VP (indicated by ). The only VP that could help us recover the content
{DESKTOP}NEW FOLDER//9781107040847C04.3D
103
vPs as phases
103
of the missing one is the one containing the missing VP, namely saw every movie
Tom did. This leads to innite regress.36
(75) a. Bill saw every movie Tom did [VP ]
b. Bill saw every movie Tom did [see every movie Tom did [VP ]]
c. Bill saw every movie Tom did [see every movie Tom did [see every movie Tom
did [VP ]]
d. Bill saw every movie Tom did [see every movie Tom did [see every movie Tom
did [see every movie Tom did [VP ]]]]
QR solves the innite regress in the following way.37 First the quantied object
undergoes QR, as in (76b). In this particular example, it does not matter whether
QR targets vP or TP. Bruening (2001) argues in favor of the former option based
on economy considerations: adjunction to vP involves shorter movement than
adjunction to TP, and is thus more economical. Second, the missing VP gets
copied, as in (76c).
(76) a. Bill saw every movie Tom did [VP ]
b. [TP Bill [vP [DP every movie Tom did [VP ]]i [vP v [VP saw ti] ] ] ]
c. [TP Bill [vP [DP every movie Tom did [VP saw ti]]i [vP v [VP saw ti] ] ] ]
QR
COPY VP
Legate (2003) ingeniously solves the question of why QR could not always
target the matrix TP (as is perhaps more standardly assumed) by making the
quantied element a Negative Polarity Item (NPI), or by making the subject
quantied and testing if the subject can have wide scope with respect to the
36
Copying just see every movie, as in (i) below, is not an option, as this string is not a constituent;
irrespective of the specics of the structure (complementation versus adjunction) and derivation
(raising versus deletion) for relative clauses, the string saw every movie Tom did is a constituent, but
saw every movie is not.
(i) Bill saw every movie Tom did [VP see every movie]
(ii)
VP
V
saw
DP
D
every
NP
NP
37
CP
{DESKTOP}NEW FOLDER//9781107040847C04.3D
104
104
Classic phases
QR-ed object. The rst case is illustrated by (77), modeled upon Legates
examples. In (77a), the only way to resolve innite regress without moving
the NPI out of the scope of negation is to adjoin any movies that Tom did to vP,
as shown in (77b), and then copy the missing VP, as shown in (77c).
(77) a. Bill didnt [VP1 see [DP any movies that Tom did [VP2 ]]]
b. Bill didnt [vP [DP any movies that Tom did [VP2 ]]i [vP [VP1 see ti]]]
c. Bill didnt [vP [DP any movies Tom did [VP1 see ti]]i [vP [VP1 see ti]]]
QR
COPY
VP
Likewise, in (78a), the way to ensure that some linguist maintains wide scope
over every paper Bill did is to adjoin the latter to vP, as shown in (78b).
(78) a. Some linguist [VP1 read [DP every paper Bill did [VP2 ]]]
b. Some linguist [vP [DP every paper Bill did [VP2 ]]i [vP [VP1 read ti]]]
QR
c. Some linguist [vP [DP every paper Bill did [VP1 read ti]] [vP [VP1 read ti]]]
COPY
VP
All the examples and arguments discussed in this section involved vPs
containing transitive and unergative verbs. For Chomsky, such vPs are complete in terms of their argument structure, whereas their passive and unaccusative counterparts are defective, which leads him to conclude that they are not
phases. This conclusion, however, has been challenged by Legate (2003), who
applies by now familiar diagnostics to argue that vPs headed by passive and
unaccusative verbs are phases. And, interestingly, the conclusion she reaches is
that with respect to these diagnostics, unaccusative and passive verb phrases
behave the same way as transitive and unergative ones do. This, in principle,
could mean one of two things: the diagnostics are ill-conceived (and are
perhaps tests for something else); or that passive and unaccusative verb phrases
are phases as well, contrary to Chomskys view. As I alluded to at the very
beginning of this section, the issue is independent of the issue of whether
passive and unaccusative verb phrases have a vP or not. The presence of v in
these structures seems pretty innocuous, but perhaps this is the reason to be
suspicious of its existence; since it does not value Case and does not select an
external argument, it might be dispensable in the same way that Agreement
projections became dispensable in the mid nineties (see Chomsky 1995, for
example). However, there might be independent reasons in favor of having v
projections in every verbal structure. This is the view taken by Marantz (2007),
who argues explicitly that phase heads are responsible for providing categorial
status to categoryless roots. On this view, associated with the Distributed
Morphology view of the lexicon, such category-neutral roots acquire their
verbal (or nominal) status only by virtue of merging with an appropriate light
functional head (a verbal v or a nominal n, for example). This leaves us with the
{DESKTOP}NEW FOLDER//9781107040847C04.3D
105
vPs as phases
105
structures in (79b) and (80b) as the only possible structures. What is at issue
here is not so much the VP versus vP status of passive and unaccusative verbal
projections, but the phase versus non-phase status of such verbal projections,
be they VPs or vPs. The most signicant (and syntactically most visible)
consequence of treating VP/vP as a phase is that passive (or unaccusative)
movement will proceed through the phase edge, as shown below. This
could also be taken as an argument in favor of the presence of the v head,
as movement of the complement of V to the specier of V would violate
antilocality, which, as we saw above, prohibits movements that are too short.
However, movement from a complement of v to the specier of vP would not
constitute a violation of antilocality.
b.
TP
T
TiT[pst] u[3sg],EPP vP
vP
VP
v
V
arrived
DPi[3sg],uC[Nom]
Tom
c.
TP
T
TiT[pst] u [3sg],EPP VP
V
V
arrived
DPi [3sg],uC[Nom]
Tom
{DESKTOP}NEW FOLDER//9781107040847C04.3D
106
106
Classic phases
TP
T
TiT[pst] u[3sg], EPP vP
was
vP
VP
V
chosen
DPi[3sg], uC[Nom]
Tom
c.
TP
T
TiT[pst] u [3sg], EPPVP
V
V
chosen
Most of Legates arguments come from phenomena that diagnose the presence of
this extra movement step. The logic of the argument is familiar by now; if the whpronoun moved to the edge position, we should nd some evidence of this movement either on the phonological side (in the form of a stranded quantier, a copy or
an expletive wh-phrase), or on the semantic side (in the form of scope reconstruction, for example). In (70) above, we saw evidence for movement through [Spec,vP]
coming from the interaction of Principle C reconstruction and variable binding.
Those examples all involved transitive or ditransitive verbs like get, read or ask. We
also saw that in all those cases, there was evidence for reconstruction to the vP edge
position, which was a position that allowed for a bound variable interpretation
without inducing a Principle C effect. Parallel passive examples are given in
(81ab); they also show that reconstruction to the edge of vP has to be possible.38
(81) a. [CP [At which of the parties that hei invited Maryj to]k was [TP every mani [vP ti
introduced to herj * ] ] ]?
b. [CP [At which of the parties that hei invited Maryj to] was [TP shej [vP * introduced
to every mani * ] ] ]? (Legate 2003: 507)
38
I refer the reader to Legates paper for parallel data involving unaccusative verbs. And to Den
Dikkens (2006a) response for a discussion of some issues stemming from Legates proposal.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
107
vPs as phases
107
We also saw above that the edge of vP can be a landing site for Quantier Raising.
The examples involved Antecedent-Contained Deletion (ACD) with an NPI to
ensure that QR does not target the matrix TP. If it did, the NPI would no longer be
licensed. The grammaticality of such examples showed that there has to be a QR
site below negation, which makes the edge of vP a plausible candidate. Legate
shows that passive vPs behave the same way. Example (82a), modeled on
Legates example, provides evidence. It shows that QR has to target vP; if it
targeted a TP instead, the negative polarity item would be outside the scope of
negation.
(82) a. The movie wasnt [VP1 seen by [DP anyone that wanted to [VP2 ]]]
b. The movie wasnt [ [ anyone that wanted to [
vP DP
VP2 ]]i [vP [VP1 see ti]]]
QR
c. The movie wasnt [vP [DP anyone that wanted to [VP2 see ti]]i [vP [VP1
seen by ti]]]
COPY VP
Legate also brings forth nuclear nuclear stress as an argument in favor of passive
and unaccusative vPs being phases. I defer the discussion of nuclear stress (and
the question of how phases may play a role in the distribution of nuclear stress) till
Chapter 7.
Before moving on to DPs, let me summarize the behavior of various types of
verb phrases with respect to various types of phasehood diagnostics.
(83) vPs as phases
Diagnostic
Is vP a domain for feature
valuation?
Is v the locus of uninterpretable features?
Does movement out of vP
proceed through the edge?
Is the complement of v a
Spell-Out domain?
Yes
No
No
Yes
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Yes
No
No
The evidence that transitive and unergative vPs are phases is quite strong;
they behave as phases with respect to the diagnostics established in Chapter 3.
The evidence in favor of passive and unaccusative vPs being phases is somewhat less unequivocal, however. They do not trigger Spell-Out but they do
require movement to proceed through the phase edge. This is the reason why
they are sometimes referred to as weak phases, and thus distinguished from
strong (transitive and unergative) vP phases.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
108
108
Classic phases
4.3
DPs as phases
In this section, we turn to the phase status of DPs. There are two in principle
independent, questions to ask. One is whether DPs themselves are phases, and
the other one is whether DPs contain other phases. The latter question arises
especially in the context of the many well-documented similarities between noun
phrases (be they DPs or NPs) and clauses (be they TPs or CPs). If CPs contain other
phases (vPs, whose phasehood we established in the previous section), it may be the
case that DPs contain phases as well.
The well-documented similarities between DPs and CPs involve argument
structure, agreement and case morphology, and extraction possibilities. See
Bernstein (2001a) for a succinct summary (and Abney 1987, Chomsky 1970,
Hiraiwa 2005, Szabolcsi 1983, among many, many others, for the original
observations). In many languages, possessors and subjects bear the same case.
A classic illustration comes from Szabolcsis (1983) work on Hungarian; she
shows that in Hungarian DP-internal possessors and clausal subjects bear the
same (Nominative) case:
(84) a. (a) Mari-
the Mary-NOM
Marys guest
vendg-e-
guest-POSS-3SG
b. Mari-
alud-t-
Mary-NOM
sleep-PAST-3SG
Mary slept. (Szabolcsi 1983: 8990)
The distinction between A and A-bar positions, familiar from a clausal domain, is
operative inside DPs as well. Valois (1991), Szabolcsi (1983) and Tellier (1988),
among others, show that the specier of DP is an A-bar position, providing an
escape hatch for movement, and Bernstein (2001b) shows that there is a DP-internal
focus position.39 We will look at some of their evidence shortly. For now, sufce it
to note that given the many similarities between noun phrases and clauses, it is not
unreasonable to expect DPs to be phases, paralleling CPs in this respect.40
The answer to the question of whether DPs contain phases, and if so, what
these DP-internal phases might be, depends largely on what the internal structure
of DPs is. So before delving into DP phasehood, let us digress into DP-internal
syntax. There are a number of DP-internal projections that have been proposed
over the years, mainly to accommodate all the DP-internal elements, and to
capture the relative ordering of these elements as well as the crosslinguistic
39
40
On a related note, there is also interesting evidence that CP and DP-internal movements can be
equally impaired in agrammatism (see, for example, Rausch, Butchert & De Bleser 2005 for a case
study involving a German agrammatic aphasic).
Chomsky initially did not include DPs in the list of phases, but suggests this as a possibility in later
writings.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
109
DPs as phases
109
variation with respect to this ordering (see Alexiadou, Haegeman & Starrow 2007,
Bernstein 1993, 2001a, 2008, Cinque 2002, Ritter 1992, 1993, Valois 1991,
Zamparelli 1996, among many, many others). Svenonius (2007b), based on a
crosslinguistic study of the ordering of articles, demonstratives, plural markers,
classiers, numerals and adjectives, arrives at the following hierarchy:
(85) Dem > Art > Num > unit > Pl/sort > Adj > n > N
Below are some of the more common nominal functional projections posited to
accommodate these elements; others include Kase Phrase (Bittner & Hale 1996),
Quantier Phrase, Person Phrase, and Demonstrative Phrase.41
(86) a. Number Phrase (NumP) (Ritter 1992)
b. Gender/Classier Phrase (GenP/ClassP)42 (Picallo 1991)
c. Possessive Phrases(PossP) (Valois 1991)
d. nP
(87)
D
D
PossP
Poss
Poss
NumP
Num
Num
nP
n
n
41
42
43
NP
The list gets bigger if we assume that nouns of different semantic types project different structures.
Zamparelli (1996) adds Kind Phrases, Predicative Phrases and Strong DPs to the inventory of
nominal projections.
See, however, Ritter (1993) for an argument that there is no syntactic Gender projection, and
gender can be either part of the Number head or the Noun head itself.
This hierarchy abstracts away from the relative position of different types of adjectives with respect
to each other and with respect to other DP-internal elements. If adjectives occupy speciers of
functional heads, the structures will get exponentially more complex (see Scott 2002, Cinque 2010,
Laenzlinger 2005, Bernstein 1994 for some representative accounts). An example hierarchy, based
on Scott (2002), is given in (i). For him, this adjectival hierarchy reects the hierarchy of heads
licensing them, making the internal structure of DP quite elaborate.
(i) Ordinal > Cardinal > Subject Comment > Evidential > Size > Length >Height >
Speed > Depth > Width > Temperature > Wetness > Age > Shape > Color > Nationality/Origin >
Material
{DESKTOP}NEW FOLDER//9781107040847C04.3D
110
110
Classic phases
Some of the diagnostics are subsumed by others; the way we can tell whether
movement out of DP proceeds through the edge of DP is by looking at reconstruction (is there a copy in this position?), ordering (can a copy be pronounced in
this position?), stranding (can something be stranded at this position?). Similarly,
the way we can tell whether D triggers Spell-Out is by looking at DP-internal
ellipsis and stress assignment.
The discussion of DPs as phases in this section is to a large extent informed by
Matushansky (2005), who discusses the behavior of DPs with respect to many of
the diagnostics listed in (89), and concludes that since DPs behave as phases with
respect to some diagnostics (PF diagnostics) but not with respect to others (LF
diagnostics), the behavior of DPs does not bode well for phases (Matushansky
2005: 179).
Two of the diagnostics in (88) involve uninterpretable features: only phase
heads can be inherent hosts of uninterpretable features (although non-phase heads
can inherit them), and only phases can be domains for feature valuation. I take it
to be uncontroversial that DPs have case features, and at least for structural case,
that these case features are uninterpretable.45 However, the uninterpretable case
features on DPs are different from the uninterpretable -features on C and v
heads. An uninterpretable Case feature is not a feature that makes DP (or its head)
a Probe, which is what we saw with uninterpretable -features on v or C heads.46
Instead, an uninterpretable case feature of a DP makes it an active Goal.
This brings us to the next question, the question of whether DPs are (or
contain) domains for feature valuation. There is evidence that DPs can contain a
44
45
46
The list of questions is ordered slightly differently to reect the order of presentation in this
section.
This is independent of whether a case feature is analyzed as a sui generis feature (uC[ ] feature, as
we have assumed throughout) or an uninterpretable counterpart of some other interpretable feature
(uT[ ] for Nominative case and uAsp[ ] for Accusative case, the possibility we alluded to in
Section 1.3).
Strictly speaking, a phrase (as opposed to a head) cannot be a Probe.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
111
DPs as phases
111
separate case domain, which suggests that either a D (or some DP-internal head)
has to be the source of this case. Genitive case is a prime example. We can see it
even in a very morphologically impoverished language like English, if we take
the preposition of to be a marker of genitive case (see Chomsky 1986, for
example). A more straightforward illustration comes from Slavic languages, in
which descriptively speaking nouns assign Genitive case (so-called adnominal
genitive), as illustrated below with Polish data. In (89a) the adnominal genitive is a
possessor, but this genitive is not limited to possessors, as shown in (89bd).
(89) a. ksiazka
studenta
book.NOM student.GEN
a students book
b. artyku
profesora
article.NOM
professor.GEN
an article (written) by a professor
c. kon
wysokiej klasy
horse.NOM high.GEN class.GEN
a high-class horse
d. objecie
wadzy
assumption.NOM power.GEN
assumption of power
e. kawaek
chleba
piece.NOM bread.GEN
a piece of bread
[Polish]
POSSESSOR
AGENT
ATTRIBUTE
THEME
PARTITIVE
Another illustration comes from the domain of numerals. Slavic numerals higher
than ve are also case assigners; the lower ones (one, two, three, four) agree in
case with their complements (as shown in (90a)), and only get Genitive case if the
entire DP gets Genitive case, as shown in (90b) with respect to the so-called
Genitive of negation. The higher numerals, on the other hand, assign Genitive
case (the so-called Genitive of quantication) to their complements, as shown in
(91a). Not surprisingly, this Genitive case stays Genitive under negation.47
47
This pattern is only found in structural case contexts. The situation changes in lexical case contexts:
(i) Jan
zainteresowa
sie
dwoma
jezykami
sowianskimi.
Jan
got.interested
REFL
two.INSTR
languages.INSTR
Slavic.INSTR
zainteresowa
sie
szescioma
jezykami
sowianskimi.
Jan
got.interested
REFL
six.INSTR
languages.INSTR
Slavic.INSTR
{DESKTOP}NEW FOLDER//9781107040847C04.3D
112
112
Classic phases
ksiazek.
books.GEN
ksiazek.
books.GEN
The distribution of adnominal genitive case thus suggests that there has to be a
distinct case domain within DP, and is compatible with some DP-internal head
(like the Num head) being the source of this genitive case.
A different argument, also pointing towards DPs being self-contained
domains, comes from binding.48 Well-known examples of the kind given in
(92) show that only certain types of DPs (namely the ones that contain possessors) count as domains for binding. This raises the possibility that maybe only
certain types of DPs are phases (or strong phases), the possibility we considered
in the last section regarding different types of vPs.
(92) a. *Johni likes [Marys descriptions of himselfi].
b. Johnj likes [Marysi descriptions of herselfi].
c. Johni likes these descriptions of himselfi.
The contrast in (93ab), due to Chomsky (1986), is also compatible with this
view if DPs can contain a PRO possessor/subject, which is coindexed with the
matrix subject in (93a) but not in (93b).
(93) a. The childreni heard [PROi stories about each otheri]
b. The childreni heard [PROj stories about themi]
48
This argument builds on Quinolis (2008) proposal that binding is determined cyclically at a phase
level. His focus, however, is on CPs and vP as binding domains, rather than DPs. While a full
review of his arguments goes beyond the scope of this chapter, he does show, for example, that
Huangs (1993) predicate fronting cases, such as the one given in (i), receive a natural explanation
if binding is determined derivationally and vPs, being phases, constitute binding domains.
(i) [CP [vP t1 criticize himself1/*2] John2 never thought [tvp that [Bill1 would tvP] ] ]
Quinoli also uses the ambiguity of examples like (ii) to argue against TPs being binding domains
(and phases). If TP were a binding domain, anaphoric interpretation would be xed at the
embedded TP level.
(ii) Johnj wonders which pictures of himselfi/j Billi saw. (Quinoli 2008: 313)
{DESKTOP}NEW FOLDER//9781107040847C04.3D
113
DPs as phases
113
The next set of diagnostics refers to what we called above the edge property.
Here, the question is whether movement from a DP has to proceed through the
edge of DP. We know that movement out of a phase has to proceed through the
edge of a phase. In cases of other phases, we looked at reconstruction, Quantier
Raising, various types of stranding, copying and agreement phenomena to
determine movement paths. For independent reasons, however, not all of them
will be applicable to DPs. For example, since DP-internal elements often agree in
-features and case (due to whatever mechanism underlies case concord), it is not
a surprise that they will continue to agree if a subpart of DP gets extracted. This is
shown in (94ac) for Polish. (94a) shows that DP-internal elements agree, and
(94bc) show that movement preserves this agreement. Thus, agreement on
extracted elements is not going to tell us whether movement took place through
the specier of DP or not.
(94) a. Jan przeczyta ciekawa
ksiazke
.
Jan read
interesting.3SG.FEM.ACC book.3SG.FEM.ACC
Jan read an interesting book.
b. Ciekawa
i
Jan przeczyta [DP ti ksia
zke].
interesting.3SG.FEM.ACC Jan read
book.3SG.FEM.ACC
c. Ktrai
Jan przeczyta [DP ti ksia
zke]?
which.3SG.FEM.ACC Jan read
book3SG.FEM.ACC
Which book did Jan read?
[Polish]
The fact that movement from a DP is blocked if the specier of DP is lled might
be taken as evidence that movement does indeed have to proceed through the
edge of DP.
(95) *Whoi did Jan like Marias picture of ti?
49
50
{DESKTOP}NEW FOLDER//9781107040847C04.3D
114
114
Classic phases
Mari-
Mary-NOM
b. Mari-nak
Mary-DAT
a
the
vendg-e-
guest-POSS-3SG
vendg-e-
guest-POSS-3SG (Szabolcsi 1983: 8991)
Furthermore, Szabolcsi shows that the specier of DP is a landing site for DPinternal operators and that this DP-internal movement turns the entire DP into an
operator:
(98) a. *(a) ki-
vendg-e-
the who-NOM guest-POSS-3SG
whose guest
b. ki-nek
a
vendg-e-
who-DAT the guest-POSS-3SG
whose guest (Szabolcsi 1983: 91)
{DESKTOP}NEW FOLDER//9781107040847C04.3D
115
DPs as phases
115
Thus, only wh-phrases that are dative-marked possessors can undergo further
movement, as shown by the following contrast with respect to topicalization.
(99) a. Pter-neki, csak
Peter-DAT only
Mari
Mari
ltta [DP ti
saw
a
the
kalap-jt]
hat-POSS.3SG.ACC
51
52
Gavruseva also discusses Aissens (1996) analysis of Tzotzil, in which regular possessives appear
postnominally, but wh-possessors can appear pre-nominally. Aissen analyzes it as movement of a
wh-possessor to [Spec,DP].
The question of whether movement out of DPs proceeds through the specier of DP is different
from the question of whether the edge of DP can be the nal landing site. Hungarian dative
possessors suggest that it can indeed be a landing site. English DP inversion structures of the kind
given in (i) have also been analyzed in this way (see Valois 1991, for example).
(i) How big a nut can a parrot eat?
{DESKTOP}NEW FOLDER//9781107040847C04.3D
116
116
Classic phases
terminological choice. If d is a phase head the lower head (D head in (100)) could
inherit its uninterpretable -features.
(100)
dP
d
duOp[
], u [ ]
DP
D
D
NP
However, since the behavior of the empty operator in these parasitic gap examples is similar in this respect to the behavior of wh-phrases undergoing overt
movement (as evidenced by the contrast in (102ab)), it might also be the case
that the issue is not a lled versus empty [Spec,DP] position but other factors
(specicity, deniteness etc.):55
(102) a. What did John enjoy showing off at the party?
b. What did John enjoy Marys showing off at the party? (Emonds 2001: 104)
53
54
55
Valois (1991), following Tellier (1988), also analyzes nominal parasitic gaps as involving empty
operator movement to the specier of DP.
{DESKTOP}NEW FOLDER//9781107040847C04.3D
117
DPs as phases
117
result is a semantic type mismatch when the adjunct combines with the vP). An
analog in the DP domain would involve wh-movement from a DP internal position
to the edge of DP licensing a DP internal parasitic gap, as shown in (103b).
(103)
a.
CP
WHj
vP
WHj
vP
Adjunct
vP
tj
b.
OPiti
CP
WHj
DP
DP
WHj
DP
Adjunct
tj
OPiti
Common examples in which the parasitic gap is DP-internal, such as (104a), are
not going to be relevant, as they are licensed by another DP-external gap.
However, (104b) is the right conguration; it involves real extraction from
DP, licensing a parasitic gap inside the adjunct relative clause.
(104) a. Marys the kind of woman OPi that [DP people who meet pg] usually end up
inviting ti into their homes.
b. *Whoi did Mary [DP tell a story about ti [CP that really impressed pg] ]?
(Matushansky 2005: 168)
While I agree with Matushanskys judgment regarding (104b), the reasons for its
degraded status might be non-syntactic in nature. (105) below, involving an
analogous conguration, improves markedly:
(105) Whoi did Mary take [DP pictures of ti [CP that werent that attering to pg ] ]?
The next edge diagnostic concerns Quantier Raising. Here the question is
whether the edge of DP can be a landing site for QR. To determine that, let us look
at cases of complex DPs containing two quantiers to see if the lower quantier
can take scope over the higher one without scoping outside the DP. Discussion of
such cases abounds in the literature (see Charlow 2009, Larson 1985, May 1977,
{DESKTOP}NEW FOLDER//9781107040847C04.3D
118
118
Classic phases
Sauerland 2005, among many others). Consider the example in (106). The most
felicitous reading for it is the inverse scope reading, in which every team scopes
over one member.
(106) [DP One member of every team] won an MVP (Most Valuable Player) award.
A more complex example is given in (106). Its most natural reading is one in
which the relative scope of the three quantiers is the mirror image of their linear
ordering. This reading can naturally be paraphrased as There is a large California
city in which every freeway has some exits that are badly constructed.
(107) [Some exits from every freeway in a large California city] are badly constructed.
(Larson 1985: 1)
The issue is whether inverse scope in such cases is the result of the lower
quantier adjoining to DP or moving out of DP and adjoining to TP, as shown
in (108ab) respectively.
(108) a.
TP
DP
DPi
DP
vP
b.
TP
DPi
TP
every state DP
two delegates from ti T
T
vP
The availability of the bound variable reading in (109) suggests the latter
possibility; however, given Kaynes denition of c-command, both adjuncts
and speciers (which are indistinguishable for him) can c-command out of DPs
containing them.
(109) Someone from every city despises it. (Larson 1985: 1, citing May 1977)
However, if inverse scope were the result of movement of the lower quantier to a
TP-adjoined position (as opposed to a DP-adjoined position), we would expect
the scope of the lower quantier to be independent of the scope of the higher
quantier in cases involving yet another (DP-external this time) quantier.
Larson (1985) also discusses such cases, concluding that the scope of the lower
quantier is dependent on the higher one. To illustrate with a concrete example,
in (110), every city can have scope over two politicians, but only if someone has
{DESKTOP}NEW FOLDER//9781107040847C04.3D
119
DPs as phases
119
scope over two policitians. This is unexpected if every city could move out of DP
and adjoin to TP directly.
(110) Two politicians spy on someone from every city. every city> someone > 2
politicians (Larson 1985: 5)
The literature on inverse scope centers around the issue of whether DP is a scope
island or not, with examples of the kind given in (110) suggesting that it is.56
However, this issue is independent of the issue at stake here, which is the issue of
whether the edge of DP is a landing site for QR. And the fact that every city can
scope over someone does suggest that QR can target the edge of DP. This brings
us to the question of whether QR can ever target a DP-internal position. If so, it
might suggest the presence of another DP-internal phase. The following example,
from Matushansky, shows that there has to be such a site; the quantier a/any
foreign country has to undergo QR in order to be interpreted. But at the same
time, being a Negative Polarity Item, it has to remain within the scope of the
negative quantier no.
(111) No student from a/any foreign country was admitted.
(Matushansky 2005: 168)
Thus the edge diagnostics do show that DPs are phases and possibly contain other
phases. We also saw evidence that DPs are case and binding domains, which is
also compatible with DPs being phases. What is left is the semantic status of DPs
(are they semantically complete in the same way CPs and vPs are?) and their
phonological status (do D heads determine Spell-Out?). Let us discuss these
questions in turn. For CPs and vPs, the criterion of semantic completeness had
to do with their propositional status and full argument structure. The obvious
parallelism between CPs like Romans destroyed the city and DPs like Roman
destruction of the city suggests that DPs are complete in the requisite sense.
Matushansky, however, interprets semantic completeness slightly differently,
focusing of semantic types instead. The question of whether DPs are semantically
complete, as she points out, is complicated in view of the fact that DPs come in
different semantic guises, with referential DPs being of type e, quantied DPs of
type <e, <e,t>>, and predicative DPs of type <e,t>. If we take completeness to be
saturation, only referential DPs are complete, and on the Generalized Quantier
Theory (which treats all DPs as generalized quantiers), no DPs would ever be
semantically complete.
56
Sauerland (2005) argues that DP is not a scope island (see also Marui 2009, Charlow 2009 for
relevant discussion), based on the fact that examples of the kind given in (i) allow the interpretation
in which the most deeply embedded quantier these two countries can scope over want and
someone.
(i) Mary wants to marry someone from these two countries. (Sauerland 2005: 306)
{DESKTOP}NEW FOLDER//9781107040847C04.3D
120
120
Classic phases
If we take the higher element to occupy either the specier of DP or its head, and
the lower one to occupy either the specier of NumP or its head (as in (114),
adapted from Lobeck 1995: 87), these case will involve ellipsis of the complement of the Num head, thus pointing towards Num being a phase head as well.
(114)
DP
Possessor
D
all
both
these
s
D
NumP
Num
Num
six/few
NP
{DESKTOP}NEW FOLDER//9781107040847C04.3D
121
DPs as phases
121
DP
Fred
D
the
both
those
which
D
NumP
Num
Num
many
few
several
NP
dwarfs
A potential problem for the idea that D heads trigger Spell-Out involves
case. It is well known that in a language like Polish, which marks case
morphologically, DP internal elements usually (but not always, see the discussion of adnominal genitive or genitive of quantication above) agree in case.
This is the familiar phenomenon of concord, which involves not only case, but
gender, number and person features as well (cf. Danon 2011, Carstens 2000
and the references therein for a detailed discussion of concord from a more
crosslinguistic perspective).
ta
ciekawa
ksiazke.
[Polish]
read.1SG this.3SG.FEM.ACC interesting.3SG.FEM.ACC book.3SG.FEM.ACC
I am reading this interesting book.
b. Nie przeczytam
tej
nudnej
ksiazki.
not read.PERF.1SG this.3SG.FEM.GEN boring.3SG.FEM.GEN book.3SG.FEM.GEN
I will not read this boring book.
(116) a. Czytam
The simple question, raised by Matushansky (2005), is how case spreads onto all
the DP-internal elements if D is a phase head. In more technical terms, the
question is what the uninterpretable case feature (uC feature) is a feature of.
The logical possibilities to consider are: the uC feature could be a feature of D
(percolating to DP), N, or both, as illustrated in (117ac) for an uC feature of a
direct object.57
57
The latter two options are compatible with the idea that only phase heads can have uninterpretable
features, as N can inherit its features from D. See the discussion of Feature Inheritance in Chapter 2
(Section 2.6).
{DESKTOP}NEW FOLDER//9781107040847C04.3D
122
122
Classic phases
(117)
a.
vP
v
vu [
VP
DP
DuC [
b.
NP
vP
v
vu [
VP
DP
V
D
c.
NP
NuC [
vP
v
vu [
VP
DP
DuC [
NP
NuC [
Each of these options raises questions. If we assume the structure in (117a), the
question is how case spreads downwards to N. If NP (being a complement of the
phase head D) is spelled out, the question about (117b) is how case spreads
upwards to D. And, nally, if we assume the structure in (117c), the question is
how v can value uninterpretable case feature on two elements simultaneously. If it
values uC feature of D rst, then this D head with a valued case feature might act
as an intervener, and block Agree between v and N. We could appeal to Hiraiwas
(2001) Multiple Agree mechanism, which allows a single Probe to undergo
Agree with multiple Goals. However, in typical cases of Multiple Agree, the
two Goals are in the same Spell-Out domain. If D is a phase head, its complement
NP is spelled out as soon as the next phase head (v in the case at hand) gets
merged. This would also result in the uC feature on N being unvalued.
Matushansky (implicitly) assumes the option in (117a), pointing out that it is
not clear how case gets onto the N head, if NP gets spelled out when the v head
is merged. But it does seem reasonable to assume that case spreads DP
{DESKTOP}NEW FOLDER//9781107040847C04.3D
123
DPs as phases
123
Evidence
Yes
Yes
Yes
Yes
N ellipsis
picture noun phrases
genitive case
inverse scope
The next chapter examines the idea of extending the idea of phasehood to
three other categories, all of which have been argued to be phases. The three are:
Predicatation Phrases (PrPs), Prepositional Phrases (PPs) and Applicative
Phrases (ApplPs). The discussion in this chapter will be somewhat more speculative for two reasons. First, the very existence of two of the three categories is
not uncontroversial. A PrP structure is not the only possible structure for small
clauses (see Citko 2011c for an overview of possibilities), and an Applicativetype structure is certainly not the only structure that has been proposed for double
object constructions. While the very existence of PPs is fairly settled, there are
quite a few of unsettled matters regarding the types, structure, and lexical vs.
functional status of prepositions.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
124
5
Other ph(r)ases
Den Dikken differs from Bowers in that he assumes that predication relationships can be mediated
by a number of functional categories (which he collectively refers to as relators), rather than a
predication-specic functional category.
See also Carnie & Barss (2006) for a slightly different view. For them, phases have to contain some
predicative element, a single argument, and a temporal operator that situates the predicate and
argument in time and space.
124
{DESKTOP}NEW FOLDER//9781107040847C05.3D
125
PrPs as phases
125
b.
TP
T
T
is
PrP
DP
Pr
Icarus
Pr
DP
a parrot
b.
TP
T
T
vP
v
VP
consider
PrP
DP
parrots Pr
Pr
AdjP
smart
The Predication Phrases in both cases are certainly semantically complete since
they contain a predicate with all its arguments. This is the behavior we expect of
phases. Tanaka & Yokogoshi (2010) also bring forth general architectural considerations as an argument in favor of Predicatation Phrases being phases. For
them, treating PrP as a phase ts in well with the conguration of phases
envisaged by Chomsky (2001): F-XP, where a functional category F selects a
substantive category XP, e.g. C-TP and v-VP (Tanaka & Yokogoshi 2010: 26).
However, it is not clear what exactly they mean by substantive category. A TP
does not seem more substantive than a CP, both being functional. We have also
seen some issues for the characterization of phases as semantically complete;
what counts as complete may vary from category to category. Therefore, let us
look at more concrete diagnostics, listed in (3), to determine the phasehood of
PrPs. These are the same diagnostics we used in previous chapters to determine
the phasehood of other categories.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
126
126
Other ph(r)ases
And to tell whether movement out of PrP proceeds through the edge of PrP, we
will ask the more specic questions in (4).
(4) a. Can an element moving out of PrP be interpreted in the edge position?
b. Can QR target PrP?
c. Is stranding at the edge of PrP possible?
Many languages have what can be referred to as a predicative case, i.e. case
assigned to predicate nominals that cannot be attributed to a default case mechanism or a case agreement mechanism (see Matushansky 2008 and the references
therein for a typological overview of case patterns on predicates). To illustrate with
one concrete example, predicate nominals are marked with Instrumental case in
Slavic languages like Polish or Russian, as shown in (5a) and (5b), respectively.3
jest studentka/*studentka.
Maria.NOM is
student.INSTR/*NOM
Maria is a student.
ego
durakom/*duraka.
b. Ja scitaju
I
consider him.ACC fool.INSTR/*ACC
I consider him a fool. (Bailyn & Citko 1999: 19)
(5) a. Maria
[Polish]
[Russian]
Bailyn & Rubin (1991) propose that the Pr head heading the small clause is
responsible for Instrumental case assignment. In phase-theoretical terms, their
claim can be restated as a claim about the Pr head valuing the uninterpretable case
feature on the predicate as Instrumental. This is only possible if Pr is a phase head.
It can value case, which means that it has to have uninterpretable features which
make it an active Probe, and since it is not a complement of a phase head (in (1b)
PrP is a complement of T and in (2b) of V), it could not have inherited its
uninterpretable features from a phase head above it. There is also no (other)
Instrumental nominal in either sentence with which the predicate could agree, and
there are no reasons to consider Instrumental case to be default, even if we grant
the existence of default case in general.
Interestingly, in Russian (and to a much lesser degree in Polish), predicate
nominals can also be marked with Nominative case, as shown in (6ab). This
alternation (Instrumental versus Nominative) is a very well-documented case
alternation in Slavic linguistics (see Bailyn & Citko 1999, Citko 2008b, Harves
2002, Matushansky 2000, Pereltsvaig 2007, Richardson 2007, among many
This is not the only option. I turn to the other option, predicates marked with Nominative case,
shortly.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
127
PrPs as phases
127
(6) a. Brat
Interestingly, if the head of the PrP is lled by an overt predicator (such as the
particle za as in the examples below), the predicate nominal is no longer
Instrumental. Instead, it agrees with the DP it is predicated of: in (7ab) the
adjective is predicated of the direct object and both are Accusative.
(7) a. Uwazam
Marie
za
I.consider Maria.ACC as
b. Ja prinimaju ego
I
take
him.ACC
I consider him as a fool.
madra
studentke/*madra studentka.
clever.ACC student.ACC/*INSTR
za duraka/*durakom.
as fool.ACC/*INSTR
(Bailyn & Citko 1999: 22)
[Polish]
[Russian]
studentka.
student.NOM
TP
T
Tu:[3sg],EPP
PrP
DPuC:[Nom],i:[3sg] Pr
Maria
Pru:[3sg]
DPuC:[Instr],i:[3sg]
a student
{DESKTOP}NEW FOLDER//9781107040847C05.3D
128
128
Other ph(r)ases
b.
TP
T
Tu:[1sg],EPP
vP
DPuC:[Nom],i:[1sg] v
I
VP
vu:[3sg]
consider
PrP
DPuC:[Acc],i:[3sg]
Maria
Pr
Pru:[3sg]
DPuC:[Instr],i:[3sg]
a student
The fact that the Pr head can lose its ability to value Instrumental case suggests
that it can lose its uninterpretable -features, and, consequently, stop being
a phase head.4 The fact that neither the pronominal copula to nor the
prepositional predicator za as agrees with the subject or the predicate is
consistent with Pr lacking -features. Instead, predicate nominals agree in
case with the subject of predication, and end up either with Nominative
case (by Multiple Agree with T) or with Accusative case (by Multiple Agree
with v):
(10) a.
TP
T
T u:[3sg],EPP
PrP
DP uC:[Nom],i:[3sg] Pr
John
Pr
DPuC:[Nom],i:[3sg]
a student
The insight that a lled Pr head loses the ability to assign Instrumental case is due to Bailyn & Citko
(1999), who capture it under a slightly different (non-phase-theoretical) set of assumptions. The
insight is reminiscent of the classic account of passive morphology as absorbing Accusative case
(see Baker, Johnson & Roberts 1989).
{DESKTOP}NEW FOLDER//9781107040847C05.3D
129
PrPs as phases
b.
129
vP
vu:[3sg],EPP VP
consider
PrP
DPuC:[Acc],i:[3sg] Pr
John
Pr
DPuC:[Acc],i:[3sg]
a student
The ability of Pr to value Instrumental case (or the lack thereof) has been shown
to correlate with the possibility of extracting the complement of a Pr head. The
contrast in (11ab), discussed in Citko (2006, 2008b), shows that copular clauses
with Nominative and Instrumental predicates behave differently with respect to
movement: Instrumental predicates can undergo wh-movement, but Nominative
ones cannot. Russian behaves similarly in this respect, as discussed by
Matushansky (2000) and Harves (2002), among many others.5
(11) a. Kimi
chcesz,
zeby Maria bya ti?
who.INSTR want.2SG COMP Maria was
Who do you want Maria to be?
b. *Ktoi
chcesz,
zeby Maria (to) bya ti?
who.NOM want.2SG COMP Maria DEM was
Who do you want Maria to be?
[Polish]
This contrast has also been accounted for by taking the Pr head selecting an
Instrumental predicate (but not one selecting a Nominative predicate) to be a
phase head.6 Matushansky (2000) and Harves (2002) analyze Russian in a
similar way; however, for them the relevant difference between the two types
of Pr heads is the difference between defective and non-defective phase heads,
rather than phase versus non-phase heads. What both types of accounts
share is the insight that the extraction asymmetry between Nominative and
Instrumental predicates is due to the presence of an EPP feature on the phasal
The Polish complementizer eby is a subjunctive complementizer, selecting past tense complements (see Tomaszewicz 2012 for a recent discussion of its many interesting properties). The
reason I use subjunctive clauses here is that for some speakers, extraction out of indicative clauses
is degraded.
The idea that there are two variants of the same head, one phasal and one not, is not new. This is a
common way to think of v heads, for example. In Section 5.3 of this chapter, we will look at two
types of applicative heads, which also vary in terms of phasehood.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
130
130
Other ph(r)ases
Pr head, which allows the Instrumental wh-predicate to move to the outer edge
of PrP, as shown in (12).7
(12) Structure of (11a)
CP
C
C
TP
T
Tu:[3sg],EPP
PrP
PrP
DPuC:[Nom],i:[3sg] Pr
Maria
Pru:[3sg],EPP
DPuC:[Instr],i:[3sg]
who.INSTR
If Pr lacked the EPP feature, the wh-predicate would not be able to move past
the subject, as is the case with Nominative predicates. This movement would
be blocked either by the Phase Impenetrability Condition if Pr remains a phase
head, or by the presence of the intervening subject either in [Spec,PrP] or
[Spec,TP] if Pr is no longer a phase head. The two options are given in (13a)
and (13b), respectively. In (13a), the subject intervenes and C cannot undergo
Agree with the wh-phrase.8 In (13b), the Merge of C (the next phase head up)
triggers the Spell-Out of the complement of Pr (the nominal predicate). Since it
is impossible to prevent a phase head from having an EPP feature in a nonspeculative way, I will not pursue this option any further.
Note that the movement of the Instrumental DP from the complement of Pr head to its (outer)
specier is a violation of antilocality. This is not a problem if Instrumental DPs involve a richer
structure, in which what moves is not a complement of the phase head but a complement of a
complement of phase head).
This conguration is very similar to the conguration involving object wh-movement out of
simple transitive clauses. It is important to note that the reason movement is not allowed in (13a)
is not PrP not being a phase, as movement out of non-phases is certainly possible. The reason is
the intervention of the subject. One might wonder why the subject, not being a wh-phrase itself,
blocks wh-movement. This I take be similar to the cases we saw in Section 2.6, in which the whobject in the outer [Spec,vP] blocked Agree between T and the subject.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
131
PrPs as phases
(13)
131
CP
C
C
TP
T
Tu:[3sg],EPP
PrP
PrP
DPuC:[Nom],i:[3sg] Pr
Maria
Pr
DPuC:[Nom],i:[3sg]
who.NOM
b.
CP
C
C
TP
T
Tu :[3sg],EPP
PrP
DPuC:[Nom],i:[3sg] Pr
Maria
Pru :[3sg]
DPuC:[Nom],i :[3sg]
who
{DESKTOP}NEW FOLDER//9781107040847C05.3D
132
132
Other ph(r)ases
VP
V
PrP
Pr
Maria
Pr
AdjP
smart
b.
vP
v
VP
V
V
PrP
Maria
Pr
Pr
AdjP
smart
{DESKTOP}NEW FOLDER//9781107040847C05.3D
133
PrPs as phases
133
TP
T
T
PrP
DP
Pr
Maria Pr
DP
my best friend
b.
CP
C
TP
T
T
PrP
DP
Maria Pr
Pr
DP
my best friend
{DESKTOP}NEW FOLDER//9781107040847C05.3D
134
134
Other ph(r)ases
We have seen that in Polish a language that allows Left Branch Extraction the
position of the nominal can mark the movement path. The grammaticality of
(19a), then, is also consistent with the nominal being stranded in its underlying
position (shown in (19b)) or in the matrix object position (shown in (19c)), so it
does not tell us much about movement through the edge of PrP.10
(19) a. Ktrego
which.ACC
uwazasz
studenta
za
najlepszego
skadniowca?
consider.2SG
student.ACC
as
best.ACC
syntactician.ACC
[Polish]
Extraction from the predicate position will thus be more telling. If movement
proceeds as schematized in (20ab), we should expect that the nominal can be
strandable in the bolded outer [Spec,PrP] position.
(20) a. [CP WH [TP [vP WH v [VP [PrP WH [PrP SUBJECT [Pr Pr [ WH]]]]]]]]
b. [CP WH [TP [PrP WH [PrP SUBJECT [Pr Pr [ WH ]]]]]]
Since the subject moves from the specier of PrP in both cases (to the matrix
[Spec,VP] position in (20a) and to the matrix [Spec,TP] in (20b)), the surface
congurations we end up with are as shown in (21ab), which again makes it
difcult to distinguish some of these cases from cases in which the moving
wh-predicate strands something in its base position.
(21) a. [CP WH [TP [vP WH v [VP SUBJECTi [PrP WH [PrP ti [Pr Pr [ WH ]]]]]]]]
b. [CP WH [TP SUBJECTi [PrP WH [PrP ti [Pr Pr [ WH ]]]]]]
10
I do not have information on the status of the following examples in Irish English. However, if they
are acceptable, a similar caveat applies to them; the nominal could be stranded in situ or in the
matrix object position.
(i) Who do you consider all smart?
{DESKTOP}NEW FOLDER//9781107040847C05.3D
135
PrPs as phases
135
head of PrP.11 (22d) and (22e) show that the nominal can be stranded. (22d) is
compatible with the nominal being stranded in its underlying position. However,
in (22e) the matrix subject remains inside PrP (as evidenced by the position of the
adverbial), which makes the outer specier of PrP the most likely culprit for the
position of the stranded nominal.
(22) a. Uwazam
Marie
za najlepsza przyjacike.
consider.1SG Maria.ACC as best.ACC
friend.ACC
I consider Jan to be (my) best friend.
b. Za jakai
przyjacike uwazasz
Marie
ti?
as what.ACC friend.ACC
consider.2SG Maria.ACC
c. *Jakai
przyjacike uwazasz
Marie
za ti?
what.ACC friend.ACC
consider.2SG Maria.ACC as
What kind of friend do you consider Maria to be?
d. Za jakai
uwazasz
Mariej [PrP tj przyjacike ]?
as what.ACC consider.2SG Maria.ACC
friend.ACC
What kind of friend do you consider Maria to be?
uwazasz
po
tym [PrP [ti przyjacike]k [PrP Marie tk ] ]?
e. Za jakai
as what.ACC consider.2SG after this
friend.ACC
What kind of friend do you consider after this Maria to be?
Maria.ACC
11
This does not seem to have anything to do with za as being prepositional (and Polish disallowing
P-stranding), as the English as resists stranding as well, as pointed out by Emonds (1985), for
example:
(i) ??Who do you consider John as?
{DESKTOP}NEW FOLDER//9781107040847C05.3D
136
136
Other ph(r)ases
In cases of other phases (vPs and DPs in particular), we also looked at Quantier
Raising to determine whether movement proceeds through the phase edge. Let us
now apply this diagnostic to PrPs. Ambiguity of examples like (24) below is not
directly relevant; the surface scope reading is in principle compatible with the
quantied predicate nominal remaining in its surface position, and the inverse scope
reading is compatible with the lower of the two quantiers adjoining to TP.
(24) Everyone is someone.
EVERYONE
>
SOMEONE, SOMEONE
>
EVERYONE
However, the rst option disappears if quantied elements need to QR for independent semantic reasons even if this movement does not establish a new scopal
relationship. Then someone will have to move in both cases, and it will have to
move to the closest position of type t, as is commonly assumed in the literature on
QR (see Bruening 2001, for example). In copular constructions such as (24), the
closest category of the right semantic type to which someone could adjoin is PrP.12
Negative Polarity Licensing also suggests that PrP is a potential QR site. In
(25), for example, anyone has to undergo QR for semantic reasons, and it has to
remain within the scope of negation to be licensed as a Negative Polarity Item.
Thus the only possible QR site is the PrP adjoined position.
(25) a. John isnt anyone.
b. [TP Johnj [T isnt [PrP anyonei [PrP tj [Pr Pr ti ]]]]]
The last diagnostic we will consider involves the ability of Pr heads to determine
Spell-Out. One way to diagnose this is by looking at whether the complement of
Pr can be deleted. Here, however, the evidence is somewhat inconclusive. The
grammaticality of (27a) does not necessarily involve deletion of the complement
of the Pr head, and the ungrammaticality of (27b) seems to be related to a more
general ban on as stranding illustrated in (27c):
(27) a. John is my best friend but Bill isnt .
b. *Mary considers Bill as her best friend and Ann considers Mike as .
c. *Whoi does Mary consider Bill as ti?
This concludes our discussion of PrPs as potential phases. The table in (28)
provides a brief summary of the main ndings in this section. We saw that the
evidence for the phasehood of PrPs is less unequivocal than the evidence for the
12
These cases are different from the cases of frozen scope in examples of the kind given in (i),
discussed by Hornstein (1995), Den Dikken (2007), among others, which show that the smallclause-internal elements cannot scope out of the small clause.
(i) Someone considers every parrot endangered.
SOMEONE
SOMEONE
>
EVERY PARROT,
*EVERY
PARROT
>
{DESKTOP}NEW FOLDER//9781107040847C05.3D
137
PPs as phases
137
phasehood of other categories, and that different types of PrPs behave differently
with respect to some of the diagnostics. Nevertheless, I hope to have established
in this section that at least some languages have PrPs that do behave like phases
with respect to most phasehood diagnostics.
(28) PrPs as phases
Diagnostic Question
Answer Evidence
Yes
No
5.2
Yes
PPs as phases
The category we focus on in this section, with an eye towards determining its
phase status, is a Prepositional Phrase. What makes determining the phasehood of
PPs somewhat more elusive than determining the phasehood of CPs or vPs is the
fact that there are quite a few background issues that have to be settled before we
can even gure out whether PPs are phases. We rst need to establish: (i) are
prepositions functional or lexical?, (ii) what is the internal structure of PPs?,
(iii) do different types of PPs have different structure?13, 14 Thus a fair amount of
the discussion in this section will revolve around these background questions.
This is not to say that these questions do not matter for other phasal categories; the
answers to them, however, seem relatively more settled and less controversial.
13
14
Perhaps an even more fundamental question is what prepositions are to begin with, as the line
distinguishing them from case particles or verbal particles is not always clear. It is not even clear if
all languages have prepositions, which, if we establish that prepositions are phases and that some
languages lack prepositions, would bear directly on one of the more general issues concerning
phases under consideration here, namely the issue of whether there is any crosslinguistic variation
with respect to phasehood. Typically, however, the issue is whether a given category such as
ApplP or DP counts as a phase in a given language, which is different from the issue of whether a
given language might lack a potentially phasal category.
See Cinques Mapping Spatial PPs: An Introduction in Cinque & Rizzi (2010) for a very
illuminating review of these (as well as a host of other) issues.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
138
138
Other ph(r)ases
The logical possibilities concerning the phasehood of PPs, all of which have
been defended in the literature, are given in (29).
(29) a. Prepositional phrases are phases
(Bokovic 2004b, Drummond, Hornstein & Lasnik 2010, Kayne 1999, 2004)
b. Prepositional phrases are not phases (Den Dikken 2010)
c. Prepositional phrases in some languages are phases (Abels 2003)
First, however, let us digress a bit into the background issues alluded to above: the
functional versus lexical status of prepositions, as well as the structural differences between different types of prepositions. These are directly relevant for the
following reason: if only functional categories are capable of being phases,
whether or not PPs are phases might depend on whether prepositions are functional or lexical.15 And if different types of PPs involve different structures, we
might expect only certain types of PPs to be phases.
The assumption that only functional categories are capable of being phases
has been implicit throughout the discussion in this book. Being a functional
category, however, would not make a very useful phasehood diagnostic, as the
implication goes only one way: while all the phase heads we have encountered so
far are functional, not all functional heads are phase heads. Examples of uncontroversially functional heads that are not considered to be phasal include (but are
not limited to): Tense heads, Aspect heads or various functional heads within the
(extended) nominal domain.
Prepositions have always straddled the lexical/functional divide, which
reects the fact that different types of prepositions might behave differently
with respect to some of the criteria distinguishing lexical from functional categories. For some researchers, prepositions are essentially lexical (but may be
dominated by a functional layer) (e.g. Den Dikken 2010, Jackendoff 1977); for
15
The view that some or all prepositions are lexical (thus non-phasal) is compatible with the idea that
they are dominated by some P-related phasal functional head (little p perhaps, by analogy to a little
v dominating VP), as is sometimes suggested.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
139
PPs as phases
139
others, they are functional (e.g. Baker 2003, Emonds 1985, Grimshaw 1991,
Kayne 2004); and yet for others some prepositions are lexical and others functional (e.g. Abraham 2010, Van Riemsdijk 1998, 1990, Yadroff 1999)16
Abney (1987) characterizes functional elements as belonging to closed
lexical classes, lacking descriptive content and being phonologically and morphologically dependent. Even though there are more prepositions than articles
or conjunctions, the number of prepositions in English, to choose a familiar
example, is much smaller than the number of nouns or verbs (less than a
hundred, according to most estimates).17 Prepositions also pattern with other
functional categories with respect to the remaining criteria: they tend to be
phonologically and morphologically dependent, they do not undergo derivational processes, they often cliticize, and they tend to lack descriptive content.18 Baker (2003) also notes that they cannot be incorporated, which he
attributes to the Proper Head Movement generalization, a generalization that
makes it impossible to move from a functional head to a lexical one (see Baker
2003: 306).
A slightly different host of arguments in favor of treating prepositions as functional and/or phasal comes from the parallelism between prepositions and other
(uncontroversially) functional heads, most notably C heads. This is what Emonds
(1985), Kayne (1999) and Abels (2003), among others, do. Complementizers often
develop from prepositions, with the English preposition for or the Romance de/di
serving as good examples.19
16
17
18
19
Strictly speaking, Van Riemsdijk treats prepositions as belonging to the so-called semi-lexical
category.
Baker lists English as having around fty prepositions, and points out that in some languages, the
class of prepositions is really small: Mohawk has four simple adpositions, Edo three and Chichewa
three or ve (depending on the criteria used). However, small does not mean closed.
Psycholinguistic evidence also seems to points towards treating prepositions as functional. Froud
(2001), as reported by Cinque (2010), shows that they pattern with functional categories in aphasia
studies. Frederici (1982), as reported by Muysken (2008), shows that with respect to some types of
speech errors, prepositions pattern with lexical classes, but with respect to others they pattern with
other (functional) categories.
Emonds (1985) actually treats complementizers as prepositions. Kaynes (1999) main arguments
in favor of treating prepositions and complementizers alike comes from the fact that both can act as
Probes and attract phrases to their speciers. More specically, he argues that prepositional
complementizers (such as di in (i)) do not form a constituent with their clausal complements.
Instead, the main verb merges with the innitival clause and the prepositional complementizer di is
merged later. It attracts the innitive clause to its specier. Next, it moves to a higher head, which
subsequently attracts the VP to its specier:
(i) Gianni
John
ha
tentato
di
cantare.
has
tried
di
sing.INF
[Italian]
{DESKTOP}NEW FOLDER//9781107040847C05.3D
140
140
Other ph(r)ases
20
21
without
behind
with, from, out of
on
from
Lexical prepositions
miedzy
zamiast
powyzej
poza
podczas
between/among
instead of
above
except
during
To be accurate, Yadroff argues that there is no such thing as a preposition in Universal Grammar
(Yadroff 1999: 26). For him, functional prepositions occupy a functional slot within the extended
projection of N/D (Yadroff 1999: 162). Yadroff & Franks (2001) take a similar view, arguing that
prepositions are an epiphenomenon: functional prepositions arise through ssion of a functional
projection dominating NP, and lexical prepositions are bleached lexical categories.
Abraham (2010) argues that the two types are merged in different positions: lexical prepositions are
merged high and serve as Probes, and grammatical ones are merged low.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
141
PPs as phases
141
Lexical prepositions can be stressed and are often morphologically complex (for
example, podczas during morphologically consists of pod under and czas
time).
The upshot of the discussion so far is that at least some prepositions are
functional. On the assumption that only functional heads are phase heads, we
might expect only these prepositions to be phase heads. However, the existence
of two types of prepositions is also consistent with all PPs being phases, if lexical
prepositions are not phase heads but are dominated by functional heads that are.
Before turning to this possibility, let me introduce another distinction: the
distinction between locative and directional prepositions, as this distinction is
often correlated with a difference in structure. Many prepositions are ambiguous
and can be either locative or directional, depending on the context. This is shown
in (33ab).
(33) a. The statue sat on/under/behind the table.
b. John put the statue on/under/behind the table.
LOCATIVE
DIRECTIONAL
P
P
[Polish]
One way to think about it is too assume that directional prepositions are
structurally more complex than locative ones, and that perhaps directional
prepositions are literally built from locative ones. Structural implementations
{DESKTOP}NEW FOLDER//9781107040847C05.3D
142
142
Other ph(r)ases
of this general idea range from very simple ones as in (37a) to cartographically
rich ones as in (37b).22
(37) a. [PP Pdir[PP Ploc DP]]
b. [CP(Place) (er) [C(Place) [DegP(Place) (two meters) [Deg(Place) [PlaceP (er) [Place [PP (P)
[AgrP pronouns Agr [PP (P) DP]]]]]]]]] (cf. Koopman 2000: 61)
22
Maria.
Maria.LOC
Many proposals build on this basic insight, proposing a ner-grained structure in which there are a
number of projections dominating PP, the speciers of which provide landing sites for PP-internal
movement of pronouns and adjunction sites for different types of measure phrases. Koopmans
structure in (37b) is one example. Other representative examples, which also distinguish structurally locative from directional prepositions, are given in (iii).
(i) [PPStat (at) [DPplace [AxPartP under [PP P [NPplace the table [PLACE]]]]]] (Cinque 2010: 5)
(ii) [DirP (DegP) Dir [LocP (DegP) Loc [DP N/CP]]] (cf. Biskup 2009b: 22)
Den Dikken (2010) takes Koopmans structure as a starting point but modies it substantially. For
him, both locative and directional prepositions are lexical categories that project an extended
functional layer, similarly to what other lexical categories do. Thus, verbs, nouns and prepositions
all have a C-like projection, a Deixis-like projection and an Aspect-like projection. The precise
nature of this projection varies depending on the nature of the head.
(iii) a. [CP C[FORCE] [DxP Dx[TENSE] [AspP Asp[EVENT] [VP V ]]]]
b. [CP C[DEF] [DxP Dx[PERSON] [AspP Asp[NUM] [NP N ]]]]
c. [CP C[SPACE] [DxP Dx[SPACE] [AspP Asp[SPACE] [PP P ]]]] (Den Dikken 2010: 100)
Radkevich (2010) also posits a more exploded structure for PPs, in which the core P-meaning
(Path, Location) can be dominated by three levels of functional projections, subject to
crosslinguistic variation. While there is no variation with respect to the availability of the core
structure and all languages have the locative/directional complex, Radkevich shows that languages
can differ with respect to how much functional structure they allow in PP. The diagnostics
Radkevich uses to diagnose the level of functional structure involve the compatibility of PPs
with measure phrases, the availability of quantier oat, and the behavior of PPs in a given
language with respect to binding.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
143
PPs as phases
b. Wszyscy poza
Maria
all
except Maria.LOC
Everyone except Maria left.
143
wyszli.
left
So according to the ability to value cases, all prepositions pass as phase heads.
Prepositions can also show agreement. The following examples from Scottish
Gaelic, due to Adger & Ramchand (2003), show that prepositions agree in
deniteness with their complements:
(39) a. ri
tidsear
with.INDEF teacher
with a teacher
b. ris
an tidsear
with.DEF the teacher
with the teacher (Adger & Ramchand 2003: 348)
[Scottish Gaelic]
The next host of diagnostics concerns extraction from PPs. There are two
questions to consider here. One is whether movement out of PPs is possible at all
(especially movement of the complement of P, which gives rise to preposition
stranding), and the other one is whether movement proceeds through the edge of
PP (or some other functional head within the extended P projection). If the answer
to both is yes, the P head is a phase head.
Preposition stranding is considered to be typologically a rare phenomenon. Van Riemsdijk (1978) is quite explicit about not allowing any
crosslinguistic variation with respect to extraction out of PPs, taking all PPs
to be islands, and allowing extraction out of PPs only under exceptional
circumstances:
(40) It is impossible to relate X, Y in the structure
Xi [P [P Y ] ] Xj
unless (a) Y = r-pronoun
(b) Y = modifying clause
(c) Y = motional postposition (Van Riemsdijk 1978: 159)
Van Riemsdijk also points out that English has R-forms, such as therewith, whereby, hereafter.
These, however, are more restricted and typically frozen.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
144
144
Other ph(r)ases
[Dutch]
b.
PP
P
eri
P
ti
The same R-rule is operative with wh-pronouns, and accounts for the contrast in
grammaticality between (42a) and (42b). In (42a), the R-rule applies rst and
P-stranding is allowed, whereas in (42b), the R-rule does not apply and
P-stranding is disallowed.
(42) a. Waari heb
je
[PP ti op ti] gerekend?
where have you
on
counted
What did you count on?
b. *Wiei heb
je [PP op ti ] gerekend?
who
have you
on
counted
Who did you count on? (Van Riemsdijk 1978: 1357)
[Dutch]
Preposition stranding, even in a language like English, which allows it in principle, is not without
exceptions. As also pointed out by Van Riemsdijk, prepositions like during, despite or notwithstanding resist it. Others require it:
(i) a. *As what is this house famous?
b. What is this house famous as? (Emonds 1985: 276)
25
Strictly speaking, Van Riemsdijk allows two distinct mechanisms to permit preposition stranding:
reanalysis for pseudopassives and movement through the escape hatch for wh-movement.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
145
PPs as phases
145
The ability to strand prepositions (or the lack thereof) does not seem to distinguish between lexical and functional prepositions or locative and directional
ones. In English all four types in principle allow it:
(45) a. Whati were you thinking of ti?
b. Whati were you hiding under ti?
c. Whati was the statue resting against ti?
d. Whoi were you running towards ti?
FUNCTIONAL
LEXICAL
LOCATIVE
DIRECTIONAL
P
P
P
P
And in a language like Polish, which disallows P-stranding, all four types
disallow it, as shown in (46ad). The pied-piped counterparts of (46ad) are all
perfectly well formed.
(46) a. *Kimi
rozmawiaas z ti?
who.LOC talk.PST.2SG
with
Who did you talk to?
b. *Kogoi posaas
Jana
zamiast ti?
who.GEN send.PST.2SG Jan.ACC instead.of
Who did you send Jan instead of?
c. *Jakich wakacjachi
Jan
by
na ti?
what.LOC vacation.LOC Jan.NOM was on
What kind of vacation was Jan on?
d. *Jakie
wakacjei
Jan
pojecha na ti?
what.ACC vacation.ACC Jan.NOM went
on
What vacation did Jan go on?
FUNCTIONAL
LEXICAL
LOCATIVE
DIRECTIONAL
26
The fact that swiping is only allowed in sluicing constructions led Craenenbroeck (2010) to a
different analysis of (44b), in which the entire PP moves to [Spec,CP] rst, and, next, the whpronoun moves to the specier of a higher CP in a CP shell-like structure:
(i) [CP whati [C C [CP [PP about tj]i [C C [TP ti ] ] ] ]
{DESKTOP}NEW FOLDER//9781107040847C05.3D
146
146
Other ph(r)ases
head) rst, which Abels argues is too short: it does not establish any new
relationship between H and XP, since the two are already in a very local
relationship.
(47)
HP
H
XP
H EPP
For Abels, movement that does not stop at the edge of HP is also not an option, as
it is a violation of the Phase Impenetrability Condition. Whether the specier of
HP is lled or not should not matter, as an extra optional specier is in principle
allowed.27
(48)
ZP
Z
Z
HP
(YP)
H
H
XP
The ban on movement in (48) does not take into account the distinction
between the two versions of PIC, which, interestingly, make different predictions in this case. In particular, PIC1 disallows any movement of XP (or out of
XP), as XP is spelled out as soon as HP is complete (namely when the next
head Y is merged):
(49)
ZP
Z
Z
HP
(YP)
H
H
XP
Spell-Out domain
27
This depends of whether the lled specier is an A or A-bar position. A specier of vP occupied by
a subject does not block wh-movement out of vP, but a specier of CP occupied by a wh-phrase
does block wh-movement of another wh-phrase out of this CP.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
147
PPs as phases
147
But PIC2 does not spell out XP until the next phase head is merged. Thus, it
allows XP in (50a) to move to a higher specier position without stopping at the
edge of HP, as long as this higher specier position is not a specier of a phase
head. Spell-Out does not occur until later, as shown in (50b).
(50) a.
YP
Y
Y
H
(YP)
H
H
b.
XP
ZP
Z
YP
Y
XPi
Y
HP
(YP)
H
H
ti
Spell-Out domain
This predicts the existence of languages that might allow preposition stranding
with A movement but not with A-bar movement. These are languages in which
the following contrast would hold:28
(51) a. J O H N W A S
b. *WHOi D I D
T A L K E D A B O U T ti.
W E T A L K A B O U T ti?
28
I use small caps here to indicate that these are hypothetical/pseudo-English examples.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
148
148
Other ph(r)ases
CP
C
TP
Johni
T
vP
VP
v
V
PP
P
ti
Spell-Out domain
Another prediction of the Stranding Generalization (and the account of preposition stranding that relies on it), examined in detail by Abels, concerns
extraction from as opposed to of the complement of the P head. This is
allowed: (54), for example, involves movement of the PP ber welches
Thema about which topic from the complement of the PP headed by malnach
after.
(54) ber welches Themai hast du mich noch [PP malnach ti ] einem Buch gefragt?
about which
topic
have you me again after
a
book asked
Which topic did you ask me about a book on again? (Abels 2003: 211)
Furthermore, if some prepositions are more complex than others (for example, if directional prepositions are more complex than locative ones), we might
expect to nd cases in which locative prepositions can be stranded but directional
ones cannot. This does not seem to be the case either in P-stranding or non-Pstranding languages.
The next diagnostic concerns binding: can PPs be domains for binding?
Interestingly, as often noted, different types of prepositions do behave differently in this respect (see Bring 2005, Chomsky 1986, Radkevich 2010,
Reinhart & Reuland 1993, among many others). For example, PPs headed by
prepositions like near or apart from can constitute a distinct binding domain,
but PPs headed by prepositions like about or with do not.30, 31
29
30
31
The passive v-V complex, irrespective of whether it is a phase or not, does not determine Spell-Out.
The examples are from Reinhart & Reuland (1993); I have added the indices and bolding.
Radkevich (2010) notes a fair amount of crosslinguistic variation in this respect and shows
that languages that allow coindexed pronouns inside PPs have more functional structure than
languages that do not. PPs in Polish, Czech and Russian are structurally impoverished, and
{DESKTOP}NEW FOLDER//9781107040847C05.3D
149
PPs as phases
149
lack the functional layer that would allow them to count as a separate binding domain. SerboCroatian and Slovak, on the other hand, behave like English; they allow the coreferential
pronoun inside the PP, which suggests more structure, and potentially a clausal boundary.
(i) Mariai
Maria
poozya
bron
koo
siebiei/*nieji.
put
gun
near
herself/her
[Polish]
poloila
zbran
blizko
sebai/?neji.
put
gun
near
herself/her
[Slovak]
{DESKTOP}NEW FOLDER//9781107040847C05.3D
150
150
Other ph(r)ases
[Icelandic]
alla ti.
I
talked yesterday with the-students all
Lit. I talked yesterday with all the students. (cf. Bokovic 2004b: 102)
The simplest derivation that would yield (61b) is the one in (62), in which the NP
simply moves to [Spec,PP], and the quantier is stranded in D.32
(62) a. [PP with [DP all students]]
b. [PP studentsi [P with [DP all ti ]]]
c. [pP with [PP studentsi [P t [DP all ti]]]]
The last two diagnostics we will consider involve Quantier Raising and
ellipsis. If QR can target PP, we have extra evidence that movement can
proceed through the edge of PP. However, PPs are not of the right semantic
type to serve as the target for QR. If the complement of P can undergo ellipsis,
we have evidence that P can determine Spell-Out. Here, the evidence also
suggests that P heads cannot license ellipsis, but it is reducible to independent
factors. Thus, the ungrammaticality of (64a) in Polish is due to the lack of
preposition stranding. And the ungrammaticality of (64b) in English, a language with preposition stranding, could be due to the lack of argument ellipsis
in English.
(64) a. *Jan zadzwoni do Marii a
Piotr poszed do .
Jan
phoned
to Maria and Piotr went
to
Jan called Maria and Piotr went to see her.
b. *John talked about a holiday in Paris and Bill dreamt about .
[Polish]
As we can see, neither QR nor ellipsis provides an argument in favor of PPs being
phases.
32
33
The derivation in which students moves to [Spec,DP] stranding all in D is excluded as a violation
of antilocality.
This is more in line with Bokovis actual derivation, given in (i). For Bokovi, however, there
are extra steps involved, as he disallows stranding in thematic positions:
(i) [PP withj [OP studentsi tj [AgrpP [all ti] tj [PP tj ti(cf.
]]]] Bokovic 2004b: 108)
{DESKTOP}NEW FOLDER//9781107040847C05.3D
151
ApplPs as phases
151
This concludes our discussion of PPs as phases. The behavior of PPs with
respect to phasehood diagnostics is mixed, as summarized in the following table.
(65) PPs as phases
Diagnostic Question
Answer Evidence
Yes
34
There are many alternative structures for double object constructions, and many issues surrounding
double object constructions, which I will abstract away from here. For example, the relationship
between the two objects in a double object construction is in principle independent from the issue
of what the relationship is between the so-called double object and dative constructions (expressed
by means of the preposition to in English). More specically, the issue is whether the two sentences
in (i) and (ii) are related transformationally or not.
(i) Jan gave Piotr a cookie.
(ii) Jan gave a cookie to Piotr.
I refer the interested reader to Green (1974) and Oehrle (1976) for some early diagnostics and
observations about the differences between the two. To illustrate briey, double objects are
infelicitous with locations (as shown in (iiab)) and imply a successful transfer of possession
{DESKTOP}NEW FOLDER//9781107040847C05.3D
152
152
Other ph(r)ases
(66)
ApplP
IO
Appl
Appl
DO
This straightforwardly captures the fact that the indirect object c-commands the direct
object, as can be shown by applying the familiar Barss & Lasnik (1986) c-command
diagnostics, which I will refrain from doing here. The structure in (66) also captures
the fact that in some languages, the applicative head is realized overtly, as shown by
the following examples from Bukusu (a northeastern Bantu language).35
(67) a. n-a-keend-el-a
omu-xasi.
1SS-tense-walk-APP-FV CL1-woman
I walked for the woman.
b. n-a-ar-ir-a
e-nyuungu luu-saala
1SS-tense-break-APP-FV CL9-pot
CL11-stick
I broke the pot with the stick.
c. wanjala e-er-er-a
en-goxo
e-yaywa
Wanjala 3sS.tense-kill-APP-FV CL9-chicken CL9-axe
Wanjala killed the chicken with the axe. (Peterson 2007: 7)
[Bukusu]
(either literal or metaphorical). Thus (iiib) implies that the students successfully mastered Polish;
no such implication is present in a DP PP frame.
(ii) a. Maria sent a book to Warsaw.
b. #Maria sent Warsaw a book.
(iii) a. Maria taught Polish to the students.
b. Maria taught the students Polish.
35
36
In languages that allow both orders of the two objects, the issue of some contention is which of the
objects is basic and which one derived (see Miyagawa & Tsujioka 2004 for a discussion of
Japanese, and Bailyn 2009 and Dyakonova 2007 for a discussion of Russian). See also the brief
discussion of dative shift in Chapter 6.
The applicative head can convey a large number of meanings: benefactive, malefactive, instrumental, locative and comitative meanings, among others. Other types of applicative markers,
discussed by Peterson (2007), include prioritive ones (which indicate that the action was completed
before a given object) or relinquitive ones, which indicate that the action was accomplished without
the (applied) object.
See also Marantz (1993) for the idea that different types of applicative heads might occupy different
levels, even in a language like English.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
153
ApplPs as phases
153
ently even within a single language. The differences involve transitivity restrictions, compatibility of an applicative structure with static verbs and the ability for
a depictive modier to modify the indirect object:
(68) High versus low applicatives
High
applicatives
Low
applicatives
*
*
*
They are not implausible semantically as applicatives; (70a) could be a benefactive and (70b) a malefactive:
(70) a. Maria ran for the benet of Jan.
b. Maria died to the detriment of Jan.
Second, English disallows applicative objects (indirect objects) with stative verbs
like hold. This is illustrated by the ungrammaticality of (71a), whose intended
interpretation is the one in (71b).
(71) a. *I held Maria the door.
b. I held the door for Maria.
In a language with high applicatives, all three applicative sentence types are
possible. Pylkknen applies these diagnostics to a variety of languages; below I
illustrate with data from Luganda.
(73) a. Mukasa ya-tambu-le-dde
Katonga.
Mukasa 3SG.PAST-walk-appl-PAST Katonga
Mukasa walked for Katonga.
b. Katonga ya-kwaant-i-dde
Mukasa ensawo.
Katonga 3SG.PAST-hold-APPL-PAST Mukasa bag
Katonga held the bag for Mukasa. (Pylkknen 2008: 20)
[Luganda]
{DESKTOP}NEW FOLDER//9781107040847C05.3D
154
154
Other ph(r)ases
c. Nd-o-shum-el-a
Katonga o
neta.
1SG-PAST-work-APPL-FV Katonga 3SG tired
I worked for Katonga while he was tired. (Pylkknen 2008: 32)
Pylknnen derives the differences between the two types of applicatives from
either structural or interpretative considerations. First, she establishes a different structure for high and low applicatives. In a high applicative structure,
given in (74) below, the direct object is the complement of the verb, and the
indirect object is the specier of a high applicative head (which itself takes a
VP as a complement). A high applicative head establishes a relationship
between the indirect object and the event described by the VP containing the
direct object:
(74)
ApplP
IO
Appl
Appl
VP
V
DO
(75)
V
ApplP
IO
Appl
Appl
DO
Pylknnen argues that the differences illustrated above follow from this difference in structure. Given that the low applicative head takes a direct object as its
complement and the indirect object as its specier, the ungrammaticality of
applicative structures in which the direct object is missing becomes a matter of
lexical selection; in the English examples in (69ab) above the selectional
requirements of the applicative head are not satised. Since in a high applicative structure, the applicative head selects the VP, it does not matter if this VP is
transitive or not.
The next two differences between high and low applicatives, namely the fact
that only high applicatives are compatible with static verbs and allow the indirect
object to be modied by a depictive phrase, follow from the semantics of the two
types of applicative heads.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
155
ApplPs as phases
155
So far, nothing we have said about the two types of applicatives has any
bearing on their phasal status, as the diagnostics distinguishing the two types
are independent of the diagnostics that distinguish phase heads from non-phase
heads. However, McGinnis (2001) argues that high applicatives differ from
low applicatives in that the high applicative head is a phase head but a low one
is not. As such, it can have the EPP feature, which allows movement of the
direct object over the indirect one. What this captures nicely is the distinction
between languages in which only the higher of the two objects (the indirect
object) can passivize, and languages in which either the direct or the
indirect objects can passivize. The examples in (76ab) show that Bukusu is
of the latter type: in benefative applicatives, either object can undergo passive
movement:37
(76) a. omu-xasi
a-kul-il-w-a
sii-tabu
nee-wanjala.
[Bukusu]
CL1-woman
3SS-buy-APP-PASS-FV CL7-book by-Wanjala
The woman was bought the book by Wanjala.
b. sii-tabu
sy-a-kul-il-w-a
omu-xasi
(?nee-wanjala).
CL17-book
CL7S-tense-BUY-APP-PASS-FV
CL1-woman
by-Wanjala
The book was bought for the woman (?by Wanjala). (Peterson 2007: 8)
37
38
The language McGinnis focuses on is Kinyarwanda, another Bantu language in which either object
can in principle passivize. In this language (as in many Bantu languages), different types of
applicatives behave differently with respect to passivization. For example, in locative applicatives,
only the locative argument can passivize.
McGinnis speculates that perhaps more generally the difference in phasal status of the two types
of applicatives is due to the fact that a sister of a VP can head a phase only if it has an argument in
its specier. This correctly distinguishes high applicatives from the low ones, but it does not
seem to be a general fact about phase heads; for example, it is not the case that C heads only
constitute phases if the specier of CP is lled. Perhaps this is due to the fact that the specier of
CP is not a thematic position, which would mean that McGinniss generalization could be
restated as a statement that the sister of VP counts as a phase head only if it selects an external
argument. If, however, passive and unaccusative vPs are phase heads, as argued by Legate
(2003) (whose arguments we reviewed in Section 4.2), this might not be the right characterization, either.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
156
156
Other ph(r)ases
], EPP
ApplP
ApplP
DOi
IO
Appl
VP
ApplEPP
V
ti
b. Low applicative
TP
DOi
T
VP
V
ApplP
Appl
IO
Appl
ti
Movement of the direct object to the outer specier of the ApplP (over the indirect
object), which is analogous to what we have seen above in a high applicative
structure, can be ruled out in two in principle independent ways. We could
appeal to the idea that the low applicative head, being a non-phase head, does not
have the EPP feature, which is the feature driving the movement of the direct
object to the outer specier.39 Alternatively, we could appeal to antilocality,
which is what Jeong (2006) does.
We have seen above what allows passivization of direct objects in high
applicatives. We have yet to see what allows passivization of indirect objects; if
39
Feature Inheritance is not an option either, since a low applicative phrase is dominated by a nonphase head (the V head).
{DESKTOP}NEW FOLDER//9781107040847C05.3D
157
ApplPs as phases
157
the direct object were always to move across the indirect object, passivization of
the direct object should be the only possibility. We could either assume that the
direct object does not move to [Spec, ApplP], which implies that the EPP feature
on the high applicative head is optional (not an unreasonable assumption), or that
it does move, but that it does so in a tucking-in fashion, below the indirect object,
the latter option being the one favored by McGinnis.40 These two possibilities are
schematized in (78ab).
(78) a.
TP
IO
T
T
ApplP
Appl
ti
Appl
VP
V
b.
DO
TP
T
IOi
T
ApplP
ti
ApplP
DOj
Appl
ApplEPP
V
VP
tj
So far, the discussion in this section has established that high and low applicatives differ with respect to extraction, and that passivization of the lower
object in a high applicative proceeds through the edge of the ApplP. The
extraction diagnostic thus distinguishes between low and high applicatives,
40
Equidistance is a logical possibility, too. However, since equidistance became obsolete with
Feature Inheritance in place, I will not elaborate on it here.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
158
158
Other ph(r)ases
but it is not clear that other diagnostics do. For example, the Appl head is
implicated in case valuation in both high and low applicatives.41
To complete the discussion of high Applicative Phrases as phases, let us apply
the remaining phasehood diagnostics to high applicatives. Polish is a language
which, according to Pylkkanens diagnostics, has high applicatives (see Citko
2010, 2011b for a more detailed discussion). Dyakonova (2007) makes a similar
point about Russian, and the Polish examples below are based on her Russian ones.
(79) a. Bede wam
spiewac i
tanczyc.
[Polish]
be.FUT you.DAT sing.INF and dance.INF
I will sing and dance for you.
b. Siedzisz mi
na sukni.
sit.2SG
me.DAT on dress
Youre sitting on my dress.
c. Maria odebraa
nam
klucze juz
troche pijanym.
Maria took.away us.DAT keys
already bit
tipsy.INSTR
Maria took away the keys from us (while we were) already a bit tipsy.
Direct objects can be passivized, which implies movement through the outer
specier of ApplP. The ungrammaticality of (80b) is unexpected in a high
applicative structure, but receives an independent explanation based on case:
indirect object is dative and dative objects do not passivize, something I discuss in
detail in Citko (2011b).42
(80) a. Ksiazkai zostaa wysana Marii ti
(przez
book.NOM became sent
Maria.DAT by
A book was sent to Maria
b. *Mariai
zostaa wysana ti ksiazke.
Maria.NOM became sent
book.ACC
Maria was sent a book.
Jana).
Jan.ACC
41
42
This is something we have not discussed in a lot of detail here. But it is quite standard to assume that
the Applicative head values the case feature of the indirect object (often as dative).
Even though the dative indirect object cannot move, it could still act as an intervener (of a defective
kind), so similar considerations apply.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
159
ApplPs as phases
159
When both move, the ordering is free, which is not a surprising fact for a language
with no superiority effects:
(82) a. Coi
komuj
Jan
wysa tj ti?
what.ACC who.DAT Jan.NOM sent
What did Jan send to whom?
b. Komuj
coi
Jan
wysa tj ti?
who.DAT what.ACC Jan.NOM sent
What did Jan send to whom?
I assume the underlying order is one in which the indirect object precedes the
direct one, even though this is an issue of some controversy in the literature on
Slavic double object constructions (see, for example, Bailyn 2009 and
Dyakonova 2007 for opposing views on this matter). This implies that the whmovement of the direct object is always going to proceed through the edge of
ApplP, and that the structure of (83a) is that of (83b).
(83) a. Coi
Jan
wysa Marii ti?
what.ACC Jan.NOM sent
Maria.DAT
What did Jan send to Maria?
b. [CP whati [TP Jan [vP sent [ApplP ti [ApplP Maria [VP ti ]]]]]]
The edge of ApplP may also be the shifted position for scrambled orders in
which the direct object precedes the indirect one:
(84) a. Jan wysa ksiazkei Marii ti.
Jan sent
book.ACC Maria.DAT
Jan sent a book to Maria.
b. [TP Jan [vP sent [ApplP a booki [ApplP Maria [VP ti]]]]]
What is relevant for our purposes is the fact that a complex wh-phrase moving
through the edge of ApplP can strand a nominal in that position:
(85) a. Ilei
Jan
wysa ksiazek
Marii ti?
how.many.ACC Jan.NOM sent
books.GEN Maria.DAT
How many books has Jan sent Maria?
b. [CP how.many [TP Jan [vP sent [ApplP how.many books [ApplP
Maria [VP how.many books]]]]]]
The fact that the English example in (i) does not readily allow the interpretation in (ii) is thus
consistent with English Applicative Phrases not being phases.
(i) John sent Maria letters and Peter Eve.
(ii) John sent Maria letters and Peter sent Eve letters.
{DESKTOP}NEW FOLDER//9781107040847C05.3D
160
160
Other ph(r)ases
(86) a. Jan
wysa Marii
listy,
a
Piotr
Ewie.
Jan.NOM sent
Maria.DAT letters.ACC and Piotr.NOM Ewa.DAT
Jan sent Maria letters and Peter (sent letters to) Eve.
b. [TP Piotr.NOM [vP [ApplP Ewa.DAT [Appl Appl [VP sent letters ]]]]]
The behavior of (high) Applicative Phrases with respect to phasehood diagnostics is summarized in (87).
(87) High ApplP as phases
Diagnostic Question
Answer Evidence
Yes
Yes
Yes
{DESKTOP}NEW FOLDER//9781107040847C06.3D
161
6
Variation in phasehood
6.1
Dynamic phases
This chapter focuses on the idea that the phase status of a given category
can change throughout the derivation. This is very different from the view
that has been implicit throughout this book, which is that phasehood, once
xed, remains constant. The ideas we will examine in this chapter allow
categories to become phases or stop being phases. Such a dynamic
approach to phasehood has been articulated most explicitly by Den Dikken
(2007) and Gallego (2010), and in what follows, I focus on their
proposals. What they share is the basic idea that a non-phase head can become
a phase head under certain well-dened circumstances. Where they differ is in
the details of implementation, as well as in their fundamental assumptions
about the nature of phases. For both Den Dikken and Gallego, the process
responsible for the variability of the phase status of a given head is head
movement. Both thus rely on head movement being a syntactic movement, as
opposed to a post-syntactic PF movement. Den Dikken, whose formulation is
given in (1), refers to this process as Phase Extension, and Gallego refers to it as
Phase Sliding.
(1) Phase Extension
Syntactic movement of the head H of a phase to the head X of the node
dominating extends the phase up from to ; loses its phasehood in the
process, and any constituent on the edge of ends up in the domain of the derived
phase as a result of Phase Extension. (Den Dikken 2007: 1)
161
{DESKTOP}/9781107040847C06.3D
162
Variation in phasehood
(2)
a.
XP
PHASE
X
HP
ZP
H
YP
SPELL-OUT DOMAIN
EDGE
PHASE
b.
XP
WP
X
X
HP
EDGE
SPELL-OUT DOMAIN
ZP
H
H
YP
Logically speaking, it is also possible for the lower head not to lose its phase
status. On this scenario, we end up with two phase heads (X and H), two edges
(circled below), and two Spell-Out domains:
(3)
XP
X
WP
HP
SPELL-OUT DOMAIN 2
ZP
H
H
YP
SPELL-OUT DOMAIN 1
There are two arguments against adopting the scenario in (3), in which a phase
head moves up to another phase head and keeps its phase status. First, it
is incompatible with the assumption that every phase head has to dominate a
non-phase head, as argued by Richards (2008), whose reasoning we reviewed in
the section on Feature Inheritance in Chapter 2 (Section 2.6). Second, it will not
allow Agree between X and YP, which as we will see shortly was Den
Dikkens main empirical motivation for Phase Extension.
Phase Extension is not the only innovation in Den Dikkens system. As
we saw earlier, his denition of a phase also departs crucially from the
standard denition. For Den Dikken, it is neither propositionality, nor complete
argument structure, nor being the locus of uninterpretable features that denes a
phase. For him, what denes a phase is a subjectpredicate conguration. Such
{DESKTOP}NEW FOLDER//9781107040847C06.3D
163
Dynamic phases
163
If CP is not a phase, wh-movement through (or to) [Spec,CP] can only take place if there is v-T-C
movement; something that happens universally at LF, Den Dikken argues.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
164
164
Variation in phasehood
I will focus here on copular constructions, since we have devoted a fair amount of
attention to them in the previous chapter. The PrP structure presented there
is different from Den Dikkens structure, but it is consistent with his assumptions that the relationship between the subject and the predicate is mediated by
a functional projection. On Den Dikkens assumptions, all the b examples in (4)
(6) involve inversion of the predicate around the subject, as illustrated in (7ab):
(7) a. Spixs Macawi is [PrP ti [Pr Pr the worlds rarest parrot] ].
b. The worlds rarest parroti is [PrP Spixs Macaw [Pr Pr ti] ].
A similar inversion mechanism applies to locative inversion and dative shift cases,
the only difference being that the functional head mediating the relationship between
the subject and the locative PP in (5ab) or the two objects in (6ab) is different.3
Den Dikken argues that in all of them, the relationship between the subject and the
predicate is mediated by a functional category he refers to as a relator, which projects
its own phrase, the Relator Phrase (RP). In the cases at hand, the RP is dominated
by an FP (where FP stands for any phrase headed by a functional head F). In copular
clauses, RP and FP stand for PrP and TP, respectively:
(8)
FP(TP)
F/T
RP(PrP)
SUBJECT
R(Pr)
R(Pr)
PREDICATE
Den Dikkens main theoretical argument in favor of Phase Extension involves the
following issue: if RP (TP) in (9) is a phase, how is it possible for the functional
head F(T) to establish an Agree relationship with the predicate, and, consequently, for the predicate to move to [Spec,TP]?
(9)
FP/TP
F(T)
F/T
RP(PrP)
SUBJECT
R(Pr)
R(Pr)
PREDICATE
We might wonder whether dative shift (which in Den Dikkens terms is inversion around the
preposition to) is compatible with the approach to double object verbs advocated in Section 5.3
above, which relied on the presence of an applicative head. While this is not the approach Den
Dikken takes, nothing in principle would prevent the inversion from taking place around the
applicative head.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
165
Dynamic phases
165
If, however, R raises to F, as shown below, then RP stops being a phase and F
(with R adjoined to it) becomes one. Now, on Den Dikkens assumptions, F can
undergo Agree with the predicate; the complement of the derived phase head (i.e.
RP) is not spelled out until FP is merged with a higher head.
(10)
FP(TP)
F/ T
RP(PrP)
R(Pr)
SUBJECT
R(Pr)
PREDICATE
(11)
PREDICATEi
F/ T
F/ T
RP(PrP)
SUBJECT
R(Pr)
R(Pr)
ti
Now the (original) subject of PrP becomes part of the Spell-Out domain when a
higher head is merged:5
Den Dikken relies on equidistance in order to explain why the subject in the specier of RP does not
intervene.
In order to rule out the derivation in which the subject moves to the edge of FP rst, Den Dikken
appeals to a ban against adjunction to empty categories.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
166
166
Variation in phasehood
(12)
HP(CP)
H(C)
FP/ TP
PREDICATEi
F/ T
F/ T
RP(PrP)
R(Pr)
SUBJECT
R(Pr)
ti
Phase Extension can not only explain what allows the predicate to move
to [Spec,FP], but also why the postverbal subject (the original subject in [Spec,
RP]) cannot move. The immobility of postverbal subjects in inverted copular
constructions, illustrated in (13ab), is well documented in the literature (see Den
Dikken 2006, Moro 1997, 2000, among others). On the Phase Extension account,
the ungrammaticality of the b examples is a consequence of the structure and
derivation in (11), where the original subject is buried inside the Spell-Out
domain of the derived phase head T.6
(13) a. Which parroti do you think is ti the worlds rarest parrot?
b. *Which parroti do you think the worlds rarest parrot is ti?
(14) a. Which parrotsi do you think ti ew up the river?
b. *Which parrotsi do you think up the river ew ti?
Let me conclude this section with a brief comparison of Den Dikkens Phase
Extension with Gallegos Phase Sliding. Phase Sliding is similar in spirit, but
it relies on a different (more standard) denition of a phase, and it applies to
a different empirical domain. Gallego shares Den Dikkens misgivings
about interface-based criteria for distinguishing phases from non-phases, suggesting a need for a denition of phases that does not rely on the interface diagnostics.
Den Dikken extends this reasoning to DPs, and takes the ban on extraction out of specic noun
phrases of the kind illustrated in (i) to be a consequence of a DP becoming a phase by extension due
to the raising of the relator head (the head mediating the relationship between picture of who and
maria) to D, as shown schematically in (iiac). After Phase Extension, the wh-phrase who is
buried inside the complement of the derived phase headed by the complex head D with an R
head adjoined to it. Since DP is an argument, adjunction to it is blocked by assumption.
(i) Whoi did Jan see Marias picture of ti?
(ii) a. [RP picture of who R Maria]
b. [DP R+D [RP picture of R Maria] ]
c. [DP Mariais R+D [RP picture of ti] ]
{DESKTOP}NEW FOLDER//9781107040847C06.3D
167
Dynamic phases
167
Gallego thus does not redene phases; he still takes core phases to be CPs and vPs,
and he shares with Den Dikken the idea that head movement can extend phases.
His empirical focus is on v to T movement, and the contrast between Romance
languages (Spanish and Catalan primarily) and English in this respect. He frames
his discussion in more general terms, suggesting that Phase Sliding could be what
distinguishes Null Subject Languages (NSLs) from non-Null Subject Languages.
I illustrate his proposal below with data from Polish, also a NSL. As is well known,
NSLs differ from non-NSLs in that they do not show that-trace effects:
(16) a. Ktoi Jan chce, zeby ti przeczyta ten
who Jan wants COMP
read
this
Who does Jan want to read this poem?
b. *Whoi do you think that ti read this poem?
wiersz?
poem
[Polish]
wiersz.
poem
[Polish]
For Pesetsky & Torrego (as well as for Gallego), these two facts are related. C has
an uninterpretable T feature, which can be checked either by the complementizer that or by the movement of the subject, which also has the uT feature
(namely, its Nominative case feature). With Phase Sliding, the subject (being
inside vP) is spelled out as soon as C is merged, as shown in (18).9 The subject
thus cannot move to [Spec,CP] to check the uT feature of C, which is what
happens in English, according to Pesetsky & Torrego. What forces the presence
7
8
9
This suggests that T in NSLs either has no EPP feature, or that its EPP feature is checked by verb
movement to T, as proposed, for example, by Alexiadou & Anagnostopoulou (2001).
A postverbal position is not the neutral position for subjects in Polish.
Gallego thus assumes the weak PIC (PIC2), in contrast to Den Dikken.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
168
168
Variation in phasehood
(18)
CuT [
TP
TiT [val]
vP
SUBJECTUT[NoM]
v
DP
{DESKTOP}NEW FOLDER//9781107040847C06.3D
169
Crosslinguistic variation
169
The rst possibility is for a given language to simply lack a category that is
considered to be a phase universally. The second one might be a misnomer,
since, technically, there is no variation on this view. Independent factors can
affect phases in ways that ultimately lead to different categories behaving like
phases in different languages.11 The third option is to simply assume that
variation in phasehood does occur, and posit a parameter to capture this variation
accordingly.
A straightforward illustration of the rst option (and not unexpected,
given the discussion in Section 5.3) comes from the domain of double
object constructions. We saw that only certain types of applicative heads
are phase heads (i.e. so-called high applicative heads), which can explain why
passivization of either the higher (i.e. the indirect) object or the lower (i.e. the
10
11
We also saw in Chapter 3 that phase heads determine Spell-Out. Crosslinguistic variation with
respect to what heads determine Spell-Out (and thus are phase heads) would thus have consequences for determining case valuation domains and ellipsis types. And both are places in the
grammar where we do nd some crosslinguistic variation. Logically speaking, however, crosslinguistic variation with respect to ellipsis could mean one of two things: either that the inventory of
phase heads (which are the heads that license ellipsis of their complements) can vary from language
to language, or that ellipsis is not a reliable phasehood diagnostic either.
Processes like Phase Extension or Phase Sliding can endow new heads with phasehood characteristics,
without changing the denitional characteristics of phases. As a result of Phase Extension, TP can be a
phase in one language but not in another. For Den Dikken, similar considerations apply to CPs: a CP
can only be a phase if a phase head moves to it. On this view, there is no (or very little) crosslinguistic
variation with respect to what categories count as inherent phases. Gallego (2010), in fact, is quite
explicit about not making the claim that there is crosslinguistic variation with respect to phasehood.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
170
170
Variation in phasehood
direct) object is possible in high applicative constructions.12 We also saw that the
existence of high applicatives in a language can be independently established by
the behavior of its applicatives with respect to three diagnostics, due to Pylkknen
(2008): compatibility with intransitive verbs, compatibility with stative predicates, and the ability for the indirect object to be modied by a depictive phrase.
What allowed passivization of the lower object was the EPP feature on the high
applicative head, as proposed by McGinnis (2001), something that we took to be
possible only if the high Applicative head is a phase head.13
A common place to look for variation with respect to phasehood is in the
domain of noun phrases (or determiner phrases), given the long line of research
on the crosslinguistic validity of the DP Hypothesis. The logic of the argument
is very simple: if DPs are phases (as we saw in Section 5.2), and if some
languages lack DPs (as is often argued for languages with no overt articles),
12
13
We also saw in Section 5.1 above that Slavic languages like Polish or Russian have two types of
Predication Phrases, one of which is a phase head and one is not. This makes it quite plausible that some
languages might lack the Pr head that is a phase head and only allow the one that is not. English might
seem like a plausible candidate for such a language: predicate nominals do not receive a designated
predicative case. However, unlike Slavic PrPs with Nominative predicates, they do allow extraction.
In principle, there could be another way for a high Applicative head to acquire its EPP feature,
without necessarily being a phase head. Given that high applicatives are dominated by a vP, the
high Applicative head could also acquire its uninterpretable features via Feature Inheritance, by
analogy to T inheriting its uninterpretable features from C. However, strictly speaking, this would
not be a case of Feature Inheritance, but rather of feature spreading (since v has to retain its
uninterpretable features), and it would also have to be possible with passive vPs. The two options
are also hard to distinguish empirically, since in both cases we end up with uninterpretable features
on both v and Appl heads:
(i)
vP
vu [
], (EPP)
ApplP
Appl
IO
Applu [
], (EPP)
V
(ii)
VP
DO
V AND
APPL
AS PHASE HEADS
vP
vu [
], (EPP)
ApplP
IO
Applu [
Appl
], (EPP)
VP
DO
APPL
HEAD
{DESKTOP}NEW FOLDER//9781107040847C06.3D
171
Crosslinguistic variation
171
such languages are predicted to lack one phase when compared to languages with
overt articles. This is the case that I will focus on in what follows.
Bokovi (2005, 2009, to appear) revives the idea that languages can vary with
respect to how much nominal structure they project: languages with overt articles
project their nominals Traditional Noun Phrases (TNPs), in Bokovis terms
up to DP, whereas languages with no articles projects their TNPs only up to NP.
However, the claim that the lack of overt articles implies the lack of a DP projection
is far from uncontroversial; see Citko (2010), Pereltsvaig (2007), Progovac (1998),
Rappaport (2001), Rutkowski (2002) and Rutkowski & Progovac (2005), among
many others, for a defense of the universality of the DP Hypothesis.14
14
A complete discussion of the arguments in favor of (and against) the Universal DP Hypothesis
would take us too far astray here and is somewhat tangential to the issue of crosslinguistic variation
in phasehood. Let me nevertheless mention a couple of potential issues for the Parametrized DP
Hypothesis. While languages at the core of the debate (Slavic languages like Polish, SerboCroatian or Russian) lack overt articles, they do have other elements that are typically analyzed
as D heads: demonstratives, possessive pronouns and quantiers, as shown in (iac).
(i) a. z
with
ta
papuga
this.3SG.FEM.INSTR
parrot.3SG.FEM.INSTR
[Polish]
moja
papuga
with
my.3SG.FEM.INSTR
parrot.3SG.FEM.INSTR
with my parrot
c. z
kazda
papuga
with
every
parrot.3SG.FEM.INSTR
Bokovi analyzes such elements essentially as adjectives, based on the fact that they show the
same agreement with the nouns they modify as adjectives do. He also points out that they can cooccur with each other and that some of them can appear in predicative positions:
(ii) ta
this
(iii) Ta
this
moja
papuga
my
parrot
[Polish]
papuga
jest
moja.
parrot
is
my
This, however, is not a convincing argument unless we have independent evidence that DP-internal
concord is limited only to adjectives.
A different argument against treating TNPs as Noun Phrases (as opposed to Determiner Phrases)
comes from contrasts of the following kind, which can be attributed to the fact that pronouns
occupy D positions (see Progovac 1998, Rutkowski 2002, for example). The contrast between (iii)
and (iv) shows that pronouns have to follow nominal modiers like sam alone. This follows
nicely from an analysis in which pronouns either occupy a D position to begin with, or raise to it.
(iii) Ona
she
sama
przysza
na
zebraine.
alone
came
to
meeting
[Polish]
{DESKTOP}NEW FOLDER//9781107040847C06.3D
172
172
Variation in phasehood
orzechy?
nuts
ona
przysza
na
zebranie.
alone
she
came
to
meeting
16
The fact that there are languages that allow violations of the Left Branch Condition was noted by
Ross (1967).
Bokovi also points out that such an explanation will not explain why LBE violations sometimes
emerge, for example, when extraction takes place from the left branch of the complement of the
noun.
(i) *Cijei
whose
je
on
vidio [NP [N
prijatelja [NP ti
majke]]]?
is
he
seen
friend
mother
[Serbo-Croatian]
While the equivalent of (i) in Polish is also degraded, it improves markedly if the Genitive precedes
the head noun:
(ii) ?*Czyjeji
Jan
zobaczy
przyjaciela ti
matki?
whose
Jan
saw
friend
mother
Jan
zobaczy ti
matki?
przyjaciela
whose
Jan
saw
mother
friend
[Polish]
{DESKTOP}/9781107040847C06.3D
Crosslinguistic variation
173
DP
AdjP
D
Adj
NP
b.
NP
AdjP
NP
To try and settle the question of whether DPs are phases in languages with no
articles, let us apply to them the same diagnostics we applied to English DPs in
Section 4.3. As usual, I will use Polish to illustrate. The examples below show
that with respect to these diagnostics, Polish Traditional Noun Phrases (be they
DPs or NPs) behave like phases. First, they contain heads that are capable of
valuing so-called adnominal Genitive case:
(22) a. kawa
Jana
coffee.NOM Jan.GEN
Jans coffee
17
[Polish]
LBE, being a fairly straightforward case of A-bar movement, could not be easily reanalyzed as
head movement, which could in principle be possible out of the structure in (21a).
{DESKTOP}/9781107040847C06.3D
174
Variation in phasehood
b. kawa
dobrego
coffee.NOM good.GEN
good quality coffee
gatunku
quality.GEN
Second, Polish TNPs are domains for the valuation of uninterpretable features.
DP-internal concord provides evidence:
(23) a. Jan zobaczy piekna
zielona
papuge.
[Polish]
Jan saw
beautiful.3SG.FEM.ACC green.3SG.FEM.ACC parrot.3SG.FEM.ACC
Jan saw a beautiful green parrot.
b. Piekna
zielona
papuga
green.3SG.FEM.NOM parrot.3SG.FEM.NOM
beautiful.3SG.FEM.NOM
przeleciaa
Janowi
nad
over
ew
Jan.DAT
A beautiful green parrot ew over Jans head.
gowa.
head
[Polish]
dwch
two.3PL.MASC.ACC
And fourth, DP-internal inverse scope is also possible, which shows that the edge
of DP can be a target for QR. The inverse scope reading is the most natural
reading for the following example.
(25) Dwch senatorw z
kazdego miasta bedzie na tym zebraniu.
two
senators
from every
city
will.be at this meeting
Two senators from every city will be at the meeting.
Logically speaking, there are two conclusions we can draw from the
behavior of Polish TNPs. One is that Polish noun phrases are DPs and behave
accordingly (thus supporting the Universal DP Hypothesis), and the other
one is that in some languages NPs can be phases as well (thus supporting the
Parameterized DP Hypothesis). Bokovi (to appear) argues for the latter,
relativized approach to phasehood, in which the highest projection is a phase
(see also Bobaljik & Wurmbrands 2005 work on relativized domains). If in
a given language TNPs are NPs, then NP is going to be a phase; if TNPs are
DPs, then DP is going to be a phase; and if TNPs are KPs, then KP is going
to count as a phase, as shown schematically in (26ac). What is impossible is
the situation illustrated in (25d), in which TNPs are KPs but a DP counts as a
phase.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
175
Crosslinguistic variation
(26) a.
VP
V
b.
NP
VP
V
c.
DP
175
VP
d. * VP
V
NP
KP
K
V
DP
KP
K
NP
DP
D
NP
phases
Bokovis proposal is interesting in that it only indirectly attributes crosslinguistic variation in phasehood to lexical variation (e.g. the presence or absence
of a D projection in a language). It does, however, show that a different inventory
of categories might yield differences in phasehood. This is what distinguishes his
proposal from others considered in this section: the number of phases remains the
same, only their identity changes.
This brings us to the third option for variation in phasehood, which simply
attributes it to variation in parameter setting. This is what Abels (2003) proposes
in order to explain crosslinguistic variation with respect to preposition stranding.
For Abels, such variation is simply due to the different settings of the following
parameter:18
(27) Parameter 1: [+/] P is a phase head. (Abels 2003: 233)
A more general question that a proposal like Abels raises is whether such a
parameter is a plausible parameter on theoretical grounds. If the phasehood of P is
subject to crosslinguistic variation, we would expect the phasehood of other
phase heads to be subject to similar parametric variation. More generally then,
the question is whether the parameter in (28), which I will call the Phasehood
Parameter, is a plausible parameter. X in (28) ranges over the categories that we
have seen to be phase heads.
(28) [+/] X is a phase head.
Capitalizing on the same intuition, Bokovi offers a slightly different take on the difference
between P-stranding and non P-stranding languages. In particular, he attributes this difference to
the possibility that PPs in languages that allow P-stranding project more structure than PPs in
languages that disallow P-stranding.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
176
176
Variation in phasehood
19
20
21
The name BorerChomsky conjecture refers to Borers seminal work on parametric variation
(Borer 1984) and Chomskys endorsement of it, and the term itself is attributed to Baker (2008a).
To be fair, Baker himself has a different take on the nature of parameters, and argues explicitly in
favor of so-called macro-parameters, which cannot easily (if at all) be stated as features of
individual functional heads.
As we saw in the earlier chapters, for Gallego uninterpretable formal features are the key formal
criterion of phasehood.
Feature Inheritance thus muddies the picture somewhat, as (29b) can only refer to heads that can
inherently count as bearers of uninterpretable features (as opposed to heads that get to inherit them).
{DESKTOP}NEW FOLDER//9781107040847C06.3D
177
Non-simultaneous phases
177
TP
DP
PF2, LF2
T
T
vP
v
ti
VP
v
V
22
PF1 LF1
DP
For the sake of clarity, I only mark the rst two cases of Transfer in the diagrams that follow.
Technically speaking, the edge of the CP phase gets transferred separately. PF1 stands for the rst
Transfer to PF, and LF1 for the rst Transfer to LF.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
178
178
Variation in phasehood
CP
C
TP
DP
T
T
PF2
LF1
VP
PF1
vP
v
ti
v
V
DP
CP
C
TP
DP
LF2
PF1
vP
ti
VP
v
V
LF1
DP
One could also imagine scenarios in which there are Multiple Transfers to
both PF and LF, but one is delayed with respect to the other. This might be the
case if the CP in (31a) or (31b) were embedded and Transfer to both interfaces
proceeded simultaneously once the embedded CP is complete.23
23
There is also another possibility, which I will not consider here: for Transfer to be totally
asynchronic.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
179
Non-simultaneous phases
179
Now that we have seen that Non-Simultaneous Multiple Transfer is conceptually plausible, we can turn to the question of how a grammar with it
compares to a grammar without it. More specically, we turn to the empirical
motivation for Non-Simultaneous Multiple Transfer.
What Non-Simultaneous Transfer is particularly well designed to handle is
a situation in which a given element is phonologically visible, so to speak,
in one position but semantically visible in another one. Marui argues that
Non-Simultaneous Transfer is what happens in cases of reconstruction (in
which a moved element is interpreted lower than it is pronounced) and covert
movement (in which a seemingly unmoved element is interpreted higher
than it is pronounced), illustrated in (32a) and (32b), respectively. In (32a),
we are interested in the non-specic reading of the indenite a Kenyan, which
can be paraphrased as It is likely that some Kenyan (or other) will win the
marathon. And in (32b), we are interested in the inverse scope reading, in
which the object every problem has wide scope over the subject some student
and we are not necessarily talking about one smart student. Both are ambiguous
and also allow interpretations in which the scope of the relevant elements
parallels their relative order.24
(32) a. A Kenyan is likely to win the marathon.
b. Some student solved every problem.
LIKELY
EVERY PROBLEM
>
>
A KENYAN
SOME STUDENT
Marui is careful about distinguishing cases like (32a), in which the entire
constituent reconstructs (total reconstruction cases), from cases in which only a
part of it does (partial reconstruction cases), and focusing on the former. The latter
is what happens with wh-movement, in which the operator part is interpreted in
its scopal position and the remainder reconstructs.
The early Transfer to LF (illustrated in (31b) above) gives rise to a situation
in which a moving element is interpreted earlier than it is pronounced.
This could be a general way to think about all cases of semantically vacuous
overt movement, of which (32a) above was one example. Another plausible
example comes from cases of scrambling which are (arguably) semantically
vacuous (see Bokovi & Takahashi 1998, Saito 2003, among many others).25
Consider the following contrast. (33a) is the unscrambled variant, (33b) shows
that the wh-pronoun can be scrambled out of the embedded CP; the scrambled
variant is nevertheless interpreted as an indirect question. Saito takes it to
mean that it has to undergo total reconstruction; if there were a trace (or copy)
24
25
Marui puts both of these cases in the more general context of attested mismatches between form
and meaning. He points towards idioms as another common example of a mismatch between
phonological and semantic units: they function as one unit semantically but multiple units
phonologically.
See, however, Bailyn (2001), Miyagawa (1997), for example, for a different take on scrambling.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
180
180
Variation in phasehood
left in the matrix clause position, the result would be a violation of the Proper
Binding Condition, the condition that requires traces to be bound.26
(33) a. John-ga
In a grammar with Multiple Non-Simultaneous Transfer, such cases of scrambling can be analyzed as involving Transfer to LF before Transfer to PF. In
(33b) the embedded vP phase containing the wh-pronoun nani-o what is
transferred to LF at a lower phase level.27 However, its phonological features
will still be accessible, since Transfer to PF has not occurred yet. This means
that the pronoun is going to be interpreted low no matter where it ends up
surfacing.
Reconstruction under A-movement of the kind illustrated in (32a) above is
different though. Sauerland & Elbourne (2002) analyze such movement as PF
movement, which, strictly speaking, would not involve reconstruction of any
type. We might be tempted to analyze such cases in a way parallel to vacuous
scrambling as also involving Transfer to LF before Transfer to PF. This is the
derivation Marui argues explicitly for, a derivation in which the indenite
subject (a Kenyan in (32a)) is transferred to LF before it is transferred to PF.
However, given that we are dealing with a raising construction, the only two
(non-DP) phases are the vP associated with the embedded verb phrase win the
marathon and the CP associated with the matrix clause, both circled in (34)
below. In (34), all three copies of the indenite subject a Kenyan are in the same
domain (the complement of the matrix C) and get transferred at the same stage.
This does not change even if verb phrases headed by verbs like seem or be likely
(or passive and unaccusative verb phrases in general) are phases; they are
thought to be weak phases; weak in that they do not trigger Transfer to the
interfaces.28
26
27
28
This is not to say that all cases of scrambling are semantically vacuous in this way. In the literature
on Japanese scrambling, it is standard to distinguish long-distance and short-distance scrambling,
and assume radical reconstruction (or some process akin to it) only for the long-distance cases.
The question is whether this necessarily means that the embedded phase has to be an LF phase but
not a PF phase, and whether there is any independent evidence for treating it as such.
This is not a problem for Marui, who treats non-nite TPs as LF phases but not PF phases. I do
not think this conclusion is generally warranted for raising TPs, and is incompatible with the
assumptions about phases defended here. Raising TPs are not complete in the requisite sense; they
are not capable of valuing Nominative case, for example.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
181
Non-simultaneous phases
(34)
181
CP
C
TP
T
DPuC[Nom]
v
likely
TP
DP
DP
a Kenyani [3sg]
VP
win the marathon
The intuition is clear; either the copy below or above the predicate be likely
gets interpreted. Depending on which one it is, we either get the reconstructed
or the non-reconstructed interpretation. However, this seems to be more a case
of Transfer targeting different features differently (PF features versus LF
features) than Transfer happening at different points in the derivation. Thus,
the interpretation of the lower copy does not seem to have much to do with
Transfer to the two interfaces happening non-simultaneously, given the
assumption that phase heads determine the points of Transfer and the complement of a lower phase head is transferred to the interfaces when a higher
phase head is merged.29
Another illustration of Non-Simultaneous Transfer, which we alluded to
above, comes from the domain of Quantier Raising. The idea here is that QR
(and perhaps covert movement more generally) involves Transfer to PF before
Transfer to LF. In (32b), repeated below as (35a), there are two Transfer domains,
with copies of the quantied object every problem present in both. Given that
every problem is pronounced low but interpreted high, it is reasonable to assume
29
The issue of reconstruction under A-movement (or the lack thereof) is quite a thorny issue. It is only
allowed with certain elements (indenites contrast in this respect with universal quantiers, for
example). That is why the mechanism that is used to capture it is sometimes different from the
mechanism used to capture A-bar reconstructions. It has been given many different accounts:
quantier lowering by May (1977, 1985); covert insertion of an expletive there by Boeckx (2001),
PF movement by Sauerland & Elbourne (2002) (see Iatridou & Sichel 2011 for a recent proposal
teasing the different complicating factors apart).
{DESKTOP}NEW FOLDER//9781107040847C06.3D
182
182
Variation in phasehood
that it the lowest copy is transferred to PF during the rst PF cycle. In the diagram
that follows I indicate quantication type with subscripts on the relevant DPs.30
(35) a. Some student solved every problem.
b.
CP
C
TP
DP
TP
DP
PF2
T
T
vP
LF1
DP
vP
DP
v
PF1
VP
v
V
DP
Bachrach & Katzir assume a multidominant structure for RNR, in which the RNR
element is literally shared between the two conjuncts, as shown in (37a).31
Furthermore, they argue that the italicized string a new documentary on
Australian parrots is not spelled out when (and where) we would expect it to
be spelled out, namely inside the embedded vP at the time when that embedded vP
is spelled out. Instead, it is spelled out during the next Spell-Out cycle, above the
coordination level.
30
31
Marui treats these cases slightly differently, deriving the effects of QR from DPs being PF phases
but not LF phases.
See Citko (2011a,b and the references therein) for a defense of such structure for RNR, and for a
more general discussion of multidominance in the grammar.
{DESKTOP}NEW FOLDER//9781107040847C06.3D
183
Non-simultaneous phases
(37)
a.
183
CP
C
&P
&
and
TP
Terryi
TP
Lesliei
vP
T
T
vP
ti
ti
VP
VP
Transfer 1
recommended
watched DP
a new documentary on Australian parrots
b.
CP
&P
&
and
TP
T
Terryi
T
Transfer 2
TP
T
Lesliei
T
vP
ti
v
v
vP
ti
VP
recommended
v
v
VP
watched
DP
For Bachrach & Katzir, Transfer can only affect completely dominated
constituents, which is what allows the shared element to escape early (or standard, perhaps more accurately) Transfer; in (37ab) it is not completely dominated by either VP.32 Such a multidominant structure for RNR (coupled with
Delayed Transfer) can explain why the shared element (often referred to as the
pivot in the literature on RNR) has to be right-peripheral; otherwise the result
would be a linearization contradiction, with the pivot both preceding the verb
watched and following it, in violation of the antisymmetry requirement on linear
order.33 It also captures (and reconciles) two seemingly irreconcilable properties
32
33
The rst node that completely dominates the shared DP is the &P node.
Since neither linearization of multiply dominated structures or constraints on RNR are our primary
focus here, I will refrain from digressing into the details and refer the reader to the literature on the
{DESKTOP}NEW FOLDER//9781107040847C06.3D
184
184
Variation in phasehood
of RNR: the fact that it is impervious to islands (as noted by Wexler & Culicover
1980 and discussed in much later literature on RNR) and the fact that the pivot
can scope over material inside both conjuncts.34 These two properties come
together, somewhat paradoxically, in (38). The lack of an island violation implies
a non-movement analysis but the wide scope interpretation for the pivot implies a
movement analysis.35
(38) John knows [someone who speaks __ ], and Bill knows [someone who wants to
learn __], every Germanic language.
( > , > )
(Sabbagh 2007: 367)
On Bachrach & Katzirs proposal, the pivot is transferred only when it becomes
fully integrated into the structure (where full integration amounts to complete
dominance), which is when the two conjuncts are combined to form a coordinate
phrase. Only then will the pivot be fully dominated (by the &P node). It means
that it will only incur island violations from this point on (hence the grammaticality of (37)) and that it will be interpreted in this high position (hence wide
scope for the pivot).
Even though the concept of Delayed Transfer illustrated here is different from
the concept of Non-Simultaneous Transfer, the two are alike in that they constitute departures from the standard Multiple Transfer architecture. I take this
exibility to be a welcome consequence of Phase Theory, rather than a potential
issue for it, given the facts it can capture. In the next chapter, we turn to the
relationship between syntax and phonology on the one hand, and syntax and
semantics on the other hand, and we will look at this relationship from the
perspective of Phase Theory. We will examine the evidence in favor of phases
being (or determining) both phonological and semantic domains, focusing on the
following questions: (i) how do phases affect linearization, (ii) how do phases
affect prosodic structure?, and (iii) how do phases affect interpretation?
34
subject (see, for example, Citko (to appear) for a handbook-style overview of these (and related)
issues surrounding RNR and the references therein for the details of the specic accounts of RNR).
The contrast below between the a examples, involving run-of-the-mill wh-movement, and the b
examples, involving RNR, provides an illustration:
(i) a. *Whati did Leslie wonder [WH-ISLAND when Terry saw ti]?
b. Leslie asked [WH-ISLAND when John directed ti] and Terry asked [WH-ISLAND when Bill saw ti ] a new
documentary on Australian parrotsi.
(ii) a. *What does Leslie know [COMPLEX DH-ISLAND someone who directed ti]?
b. Leslie knows [COMPLEX DP-ISLAND someone who directed ti] and Terry knows [COMPLEX DP-ISLAND someone who
saw ti] a new documentary on Australian parrots.
35
Covert movement is not an option here since it would have to be covert Across-the-Board (ATB)
movement of the pivot, and covert ATB movement is generally disallowed (see Bokovi & Franks
2000 and Citko 2000, 2005, among others, for data and analysis).
{DESKTOP}NEW FOLDER//9781107040847C07.3D
185
7
Phases and the interfaces
185
{DESKTOP}NEW FOLDER//9781107040847C07.3D
186
186
illustrate how Order Preservation works. In (1) below, there is a single Spell-Out
domain (D1) consisting of three elements, X, Yand Z. The ordering statements for
this domain D1 are given in (1b).
(1) a.
X
Y
Z
D1
X
Y
Z
Given the Order Preservation Principle, a number of things could happen next;
crucially, however, the relative order of elements established within D1 cannot
change.1 This means that X, being at the edge of D1, can freely move out of D1.
After movement, it will still precede all the elements within D1.
(3) a. Move X
D2
Xi
D1
ti
Y
See, however, Richards (2007) for a proposal that phase-internal movements are order preserving
but transphasal movements are not.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
187
Phases and PF
187
The ordering created at the next Spell-Out domain (D2) does not change the
ordering previously established, it merely adds new orderings to the ones established at D1. This is how Fox & Pesetsky derive the fact that movement from a
given domain has to proceed through the edge of that domain. If, on the other
hand, movement from D1 takes place from a non-left-edge position, as in (4a), the
new linearization statements are going to contradict the previously established
ones. Here both precedes Y (by virtue of preceding D1, which includes Y) and
follows Y (due to Y having moved out of D1).
(4) a. Move X
D2
Y1
D1
X
ti
However, if it is essentially PF that rules it out, as Fox & Pesetsky point out, we
expect to nd ways to x it. And this is indeed what happens. If both X and Y
move but their relative order is preserved, the new linearization statements are not
going to contradict the previously established ones.2
(5) a. Move X and Y
D2
Xj
Yi
D1
tj
ti
This is how Fox & Pesetsky derive Holmbergs Generalization, illustrated by the
contrast in (6). This contrast shows that the object can undergo object shift only if
the verb has shifted even further and the order of the two is preserved:
Ellipsis is another way to rescue a structure in which a non-nal element moves out of a domain D.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
188
188
(6) a. [CP
In the grammatical (6a), this is what happens. The ordering established at the CP
level does not contradict the ordering established at the VP level:
(7) a. Ordering within VP: KISSED > HER
b. Ordering within CP: I > KISSED, KISSED > HER, HER > NOT, N O T >
VP
However, in the ungrammatical (7b), the ordering established at the CP level does
contradict the ordering established at the VP level: henne precedes kysste at the
CP level but follows it at the VP level.
(8) a. Ordering within VP: K I S S E D > H E R
b. Ordering within CP: THAT > I, I > H E R ,
HER
> NOT,
NOT
> VP H E R >
KISSED
Fox & Pesetskys Cyclic Linearization proposal establishes a direct link between
Spell-Out domains and linearization domains. It is not clear, however, what in their
system prevents the object from moving to the edge of VP rst, in which case it
could continue to move to the next Spell-Out domain with the verb remaining within
VP without violating the Order Preservation Principle. The relationship between
Fox & Pesetskys Spell-Out domains and phases (or complements of phase heads) is
also not quite transparent. Given that vP is a phase, the Spell-Out domain should be a
VP, which is consistent with their proposal. However, on standard assumptions
about phases and successive cyclic movement, movement takes place through the
edge of a phase (vP) not through the edge of the Spell-Out domain (VP).
In the rest of this chapter, I will be concerned with the relationship between
phases (or Spell-Out domains) and phonological domains, and the issue of
whether phases correspond to (or help determine) phonological domains. If so,
what are these phonologically relevant domains, and what phonological processes apply to them?
The commonly assumed prosodic units are listed in (9) (see Nespor & Vogel
1995: 11). The question is which of them correspond to Spell-Out domains. Since
phases correspond to larger chunks of structure, the most likely culprits are
higher-level units such as intonational phrases.
(9) a. phonological utterance
b. intonational phrase
c. clitic group
d. phonological word
d. foot
e. syllable
3
This follows from transitivity. Given that henne precedes inte and inte precedes VP (and kysste is
included within VP), henne will end up having to precede kysste.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
189
Phases and PF
189
We will focus below on sentential stress (also referred to as nuclear stress), which
is a good example of a phonological process that applies to larger units and which
has been shown to be sensitive to syntactic domains. The more specic questions
about nuclear stress we will ask are:
(10) a. Do phases play a role in determining domains for nuclear stress assignment?
b. If so, does movement across phasal domains affect nuclear stress?
c. How can we account for the crosslinguistic variation (or the lack thereof) with
respect to nuclear stress?
* )
(
* )
(* ) ( * )
like NUTS
Cinque (1993) points out both conceptual and empirical problems with
Halle & Vergnauds formulation and suggests the revision given in (12). The
crucial innovation in his proposal lies in the fact that it makes no reference
whatsoever to the directionality of stress assignment (right versus left).
(12) a. Interpret boundaries of syntactic constituents as metrical boundaries.
b. Locate the heads of line N constituents on line N + 1.
c. Each rule applies to a maximal string containing no internal boundaries.
d. An asterisk on line N must correspond to an asterisk on line N 1.
(Cinque 1993: 244)
{DESKTOP}NEW FOLDER//9781107040847C07.3D
190
190
element that bears nuclear stress irrespective of the word order type involved
(SVO, SOV or VSO):4
(13) a. John read a BOOK.
b. Ali
KETAAB
ye
Ali
a
book
Ali bought a book.
SVO
[Persian]
xarid.
bought
c. Chuala
Seonag
CALUM.
heard
Seonag
Calum
Seonag heard Calum (Kahnemuyipour 2004: 12)
SOV
[Scottish Gaelic]
Other factors, however, do play a role. For example, as often noted, unaccusative
and unergative verbs behave differently: with unergative verbs, stress can fall
either on the verb or on the subject (see Kahnemuyipour 2004, Selkirk 1995,
Zubizaretta & Vergnaud 2005 and the references therein). With unaccusative
verbs, on the other hand, only the subject can bear nuclear stress:
(14) a. The PARROT is screeching.
b. The parrot is SCREECHING.
(15) a. The PARROT disappeared.
b. *The parrot DISAPPEARED.
NSR is a cyclic rule: stress is assigned cyclically (domain by domain) and the
highest stress within a given domain projects to the next level. Furthermore, the
contrast between the a and be examples in (1719), due to Bresnan (1971), who
attributes the rst pair to Newman (1946), suggests that the stress assigned within
one domain can sometimes survive as the highest stress if the stressed element
moves out of that domain. All the a examples involve noun complement clauses
(thus no movement from the embedded CP), whereas all b examples involve
relative clauses (which on the Head Promotion Account means that the head
moves from within the relative CP). Interestingly, the relative clause head keeps
its stress.5
This is not to say that there is no crosslinguistic variation whatsoever. Zubizarreta (1998), for
example, in order to account for the differences between Romance and Germanic languages, posits
the existence of two types of nuclear stress rules, one sensitive to selection (which she dubs
selectional nuclear stress rule) and the other one sensitive to c-command. They might interact
differently in different languages (or one of them might be absent altogether in a given language),
which is what gives rise to crosslinguistic variation.
Legate reinterprets Bresnans ndings in phase-theoretical ways and takes this to mean that movement across phases does not affect nuclear stress. Kahnemuyipour takes issue with this assertation
based on the fact that the moved wh-phrase does not keep its stress.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
191
Phases and PF
191
With this as background, let us turn to the question of how Nuclear Stress Rule
might be recast in phase-theoretical terms. A natural starting point is the hypothesis, stated in various forms in (20ad), that the Spell-Out domain of the phase is
the domain for prosodic rules like the Nuclear Stress Rule.
(20) a. The spellout domain of a phase is the prosodic domain for phrase stress.
(Kratzer & Selkirk 2007: 104)
b. On the construction of each phase, the complement of the phasal head is
Spelled Out, so it is this category that undergoes phonological rules.
(Adger 2007: 247)
c. Phonological rules apply directly to the material that is spelled out at
eachphase. (Pak 2008: 10)
d. Spell-Out domains are mapped onto prosody as MaPs. (Ishihara 2007: 144)
However, even with these proposals about the mapping between Spell-Out
domains and prosodic domains, there is still the question of how exactly NSR
applies within these domains. There are two logical possibilities. One is that
the lowest (e.g. the most deeply embedded) element within the Spell-Out
domain is prosodically the most prominent one, and the other one is that it is
the highest element within a given domain that is prosodically the most
prominent one. Interestingly, both of them have been defended in the literature,
not for different languages. This is not necessarily a contradiction. However,
depending on which route we take, we will require a different set of background assumptions.
Adger (2007) provides a very straightforward implementation of the Cinquestyle Nuclear Stress Rule in a phase-based system.6 Let us look at concrete
examples, parallel to Adgers original (parrot-free) examples.
(21) The parrot ate a nut.
Adgers implementation of it, however, does avoid some of the problems Cinque faces. To illustrate,
complex noun phrases of the following sort, as pointed out by Adger, are problematic for Cinque.
(i) [the [man [from [Philadelphia]]]]s hat. (Cinque 1993: 268, crediting Richard Kayne)
Since speciers are at the edge, they escape Spell-Out. Consequently, the noun hat can be the most
prominent element in spite of the fact that it is not the most deeply embedded constituent in this noun
phrase.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
192
192
The steps in the derivation that are relevant for determining nuclear stress
(namely Spell-Out domains) are given below. (22a) shows that the complement
of the lowest phase head (D) is spelled out when the next phase head (v in this
case) is merged.
(22)
vP
VP
v
V
ate
DP
D
a
NP
nut
At this stage, the lexical stress on the noun nut projects to the next level, as
shown in (23). Similar considerations apply to the subject the parrot. If
Spell-Out domain is the domain for stress assignment, and NSR amounts to
projecting the highest stress within the Spell-Out domain. This is actually a
departure from Adgers proposal; for him NSR applies to entire phases, and
phase edges are not included in the calculation of stress due to extrametricality.
Here, NSR applies to Spell-Out domains, which eliminates the need for
extrametricality.
(23) Spell-Out NP, apply NSR to NP
*
*
nut]
[NP
*
*
[NP parrot]
The next Spell-Out occurs when C (the next phase head up) is merged. At this
point the complement of VP, being the complement of v, is spelled out.
(24) a.
CP
C
TP
DPi
the parrot T
vP
v
ti
VP
v
V
ate
DP
D
a
NP
nut
{DESKTOP}NEW FOLDER//9781107040847C07.3D
193
Phases and PF
193
[VP
*
*
*
a nut]
*
ate
In the last Spell Out-cycle, the TP spelled out, and the stress on the noun nut, still
being the highest one, is projected to the next level. This correctly derives the
placement of nuclear stress on the noun nut in The parrot ate the nut.
(25)
a.
CP
C
TP
DPi
the parrot T
vP
ti
v
VP
v
V
ate
DP
D
a
NP
nut
b.
[TP
*
*
The parrot
*
ate
*
*
*
*
nut]
This algorithm can also account for the contrast between unergatives and unaccusatives. We have seen above that unaccusatives differ from unergatives in that
they only allow stress on the subject (recall the contrast between (15) and (16)
above).7 On the assumption that unaccusative vPs are not strong phases in that
they do not trigger Spell-Out, there are two Spell-Out domains, the complement
of D and the complement of C.
Crucially, Adger assumes that the unaccusative v does not trigger Spell-Out. If it did, the verb would
be the most stressed element in this domain.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
194
194
(26)
CP
C
TP
DPi
the parrot T
vP
v
VP
V
disappeared
ti
In the rst domain the stress on parrot, being the only stress (the determiner the,
being functional, is not stressed), is the one that projects to the next level, as
shown in (27a). In the second domain, the stress on parrot is still the one that
projects to the next level, being the highest stress in this Spell-Out domain, as
shown in (27b)).8
(27) a. Spell Out NP, Project stress on parrot
*
*
[NP parrot]
*
dissappeared]
Legates view is different. She assumes there is a phase boundary here, and assumes that in those
cases it preserves the stress it was assigned at an earlier cycle. The reason parrot in (i) is stressed is
thus similar to the reason proposal is stressed in (ii).
(i) The PARROT disappeared.
(ii) I liked the PROPOSAL that George left.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
195
Phases and PF
(28) a.
195
CP
C
TP
T
DPi
D
the
NP T
parrot
vP
ti
v
v
VP
screeched
b.
*
*
*
[TP the parrot
*
*
screeched]
*
*
[TP the parrot
*
*
*
screeched]
c.
At rst glance, this seems to make wrong predictions for a language like English:
within the Spell-Out domain of a transitive v, a direct object is not the highest
element. However, it becomes one if the verb moves out of VP (or the object
moves to some higher position within this Spell-Out domain). In verb-nal
languages (of which Kahnemuyipour 2004 focuses on Persian and German),
stress falls on the direct object in simple transitive clauses, just as in English.
Kahnemuyipour accounts for it as follows; he assumes the existence of a low
aspectual projection, sandwiched between vP and VP (see MacDonald 2006,
Travis 2010 and the references therein for a justication of such a projection). He
furthermore assumes that the direct object moves to the specier of this projection, even in a language with no object shift like English. This means that the
Spell-Out domain (SPELLEE in his terms) of the phase head v is this low
aspectual projection. Thus, at the point in the derivation illustrated in (30)
{DESKTOP}NEW FOLDER//9781107040847C07.3D
196
196
(namely when C is merged), the complement of the lower phase head, in this case
an AspP is spelled out:
(30)
CP
TP
Subjectj
vP
tj
v
AspP
OBJECTi
Asp
Spell-Out Domain
VP
Asp
V
ti
Even though Kahnemuyipour (2004) and Adger (2007) agree that phases
are relevant to the determination of nuclear stress, we saw that they differ in the
details of implementation. For Adger, it is the element with the highest level of
stress that gets its stress augmented by one stress mark. For Kahnemuyipour,
it is structurally the highest element that does. They also differ in how they
handle the difference between sentences with passive/unaccusative verbs and
unergative ones. We have seen above how this difference can be derived in
Adgers system. For Kahnemuyipour (2004), this difference is tied to the
assumption that unaccusative vPs are not (strong) phases in that they do not
trigger Spell-Out. The only phase is the CP phase, and, consequently, the
subject being the highest element in the Spell-Out domain of C is the element
that bears nuclear stress.
Kahnemuyipours system gives rise to two straightforward predictions. First,
if the highest element moves out of what would otherwise become its Spell-Out
domain, it should lose nuclear stress. Specic objects in Persian are a case in
point; the contrast between (31a) and (31b) shows that specic objects differ from
their non-specic counterparts in that they do not receive nuclear stress.
(31) a. Ali
QAZAA
Ali
food
Ali ate food.
b. Ali
Ali
qazaa-ro
food-ACC
xord.
ate
[Persian]
XORD.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
197
Phases and PF
197
(32)
TP
a.
Alij
TP
b.
Alij
T
T
T
T
vP
tj
qazaa-roi
v
v
vP
QAZAA
tj
AspP
Asp
Asp
VP
xord
v
v
AspP
Asp
tj
ti
Asp
VP
XORD
ti
Second, if some (extra) element is merged higher than the non-specic direct
object (which would otherwise be the highest element inside the Spell-Out
domain, that element should bear nuclear stress. For example, if an adverb is
merged inside the Spell-Out domain, it will bear nuclear stress.
So far, we have focused in this section mostly on the Nuclear Stress Rule, and
mostly on how it works in English. However, phases and Spell-Out domains have
been implicated in a number of other phonological phenomena in a number of
other languages. Let me thus conclude this section with a couple of examples.
Paks (2008) analysis of High Tone Anticipation (HTA) in Luganda makes
crucial use of phases. HTA, illustrated in (33ab) below, is a kind a spreading in
which high tone spreads leftward across a given domain, which Pak argues is the
Spell-Out domain of a phase head. The verb and the material following it are
treated as a single domain, whereas preverbal elements belong to a separate
domain. Pak argues that preverbal elements are in the specier of C, thus by
hypothesis outside the Spell-Out domain of C.
(33) a. omulenzi a-gul-ir-a
Mukasa
kw.
1.boy
SBJ1-buy-APPL-IND
1.Mukasa 1A.coffee
The boy is buying Mukasa some coffee.
b. (mlnz) (-gl-r-
Muks
kw.
1.boy
SBJ1-buy-APPL-IND
1.Mukasa 1A.coffee
The boy is buying Mukasa some coffee. (Pak 2008: 15)
[Luganda]
{DESKTOP}NEW FOLDER//9781107040847C07.3D
198
198
two objects are in the same Spell-Out domain, whereas in (34b), which is a high
applicative, the direct object and the indirect object are in two different domains.
This follows if the Spell-Out domain of the high applicative phase head is a VP
containing the direct object but not the indirect object, which is part of the next
domain.
(34) a. [Tu-k-i-tum-ir-
omkali valin:nde.]
we-PST-T-send-APPL-FV woman
Valinande
We have just sent Valinande to the woman.
b. [Ni-mw-ndik-il-ile nu:ru:] [xti.]
SP-OP-write-APPL-FV
Nuru
letter
I wrote Nuru a letter. (McGinnis 2001: 23)
[Kinande]
[Chi-Mwhi:ni:]
Ishihara (2003, 2007) shows that in Japanese, Spell-Out domains are mapped
onto Major Phrases and that scope in Japanese cannot extend beyond Major
Phrases. The empirical support comes from the contrast, due to Miyagawa
(2003), illustrated in (35ab), showing that a universal quantier unambiguously
scopes over negation in SOV word order, but can get either wide or narrow scope
with respect to negation in an OVS word order type.
(35) a. zenin-ga sono tesuto-o
all-NOM
that test-ACC
All did not take that test.
uke-nakat-ta.
take-NEG-PST
(SOV)
*NOT>>ALL,
ALL>>NOT
In (37a), the subject and negation are not in the same Major Phrase, as schematized in (36):
(36) [TP ALLi
MaP
(SOV)
MaP
[vP
NOT TOOK] ]
ALL ti
MaP
(OSV)
{DESKTOP}NEW FOLDER//9781107040847C07.3D
199
Phases and LF
b. [ THAT TEST
CP
i
MaP
199
(OSV)
MaP
This concludes our discussion of the role phases play in determining phonological domains. The question we focused on is whether phonological cycles
correspond to syntactic cycles, and more importantly, given the focus of this
book, whether phases or complements of phase heads (being the Spell-Out
domains) either are the relevant phonological units or map onto the phonological
relevant units.We saw that phases and Spell-Out domains (as determined by
phase heads) help us understand PF phenomena such as linear order preservation
effects or nuclear stress assignment. We also saw that phases play a role in
determining phonological processes such as Vowel Lengthening or High Tone
Anticipation in Bantu, and that Major Phrases as determined by phases play a role
in constraining scope in Japanese. In the next (nal) section of this book, we ask
similar questions about the mapping between syntax and semantics, and the role
phases might play in that mapping.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
200
200
Semantic type
e (proper names)
<<e,t>,t> (quantied noun phrases)
t
<<s,t>, t> (questions)
t (declaratives)
<e,t> (relative clauses)
t
<e,t>
Semantic type
<e,t>
<e,t>
t
<e,t>
<e,t>
Thus, a more fruitful line of inquiry into the role of phases in the syntax
semantics interface might come from the question of what semantically relevant
domains phases might correspond to. For reasons that are similar to the reasons
we discussed in the last section (which dealt with phases and PF), given how
Phase Theory and Multiple Spell-Out work, it may well be the case that it is not
phases per se that are the relevant domains, but, rather, complements of phase
heads.
Interestingly, there is an important (and independently motivated) semantic
distinction that has been argued to correlate with the boundary between vP and
CP phases. It is the distinction between the nuclear scope (the domain of
existential closure) and the restrictive clause in a tripartite quanticational structure, as proposed explicitly by Biskup (2009a) (see, however, Butler 2005,
Carnie & Barss 2006 for different implementations of the same general idea).9
This is Diesings (1992) Mapping Hypothesis, stated in (39) in phase-theoretical
terms, and illustrated in (40) in arboreal terms.
Carnie and Barsss approach to phases is slightly different, though, from the approach taken here.
They argue for a relativized approach to phases, in which each phase has a predicative element, a
single argument and a temporal operator.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
201
Phases and LF
201
Restrictive Clause
TP
DPSUBJ
T
T
vP
Nuclear Scope
DPSUBJ v
VP
v
V
DPOBJ
What is important is that the two subject positions (the thematic vP-internal
position and the EPP-related [Spec,TP] position) end up in different domains.
Diesing shows how it can explain ambiguity of bare plural subjects in sentences
of the following sort:
(41) a. Parrots are loud.
b. Parrots are intelligent.
10
Diesings original formulation, given in (i), does not make reference to phases (but to VPs and IPs
instead). This, of course, is not surprising, since her formulation predates Phase Theory by a
decade.
(i) Material VP is mapped into the nuclear scope.
(ii) Material from IP is mapped into a restrictive clause. (Diesing 1992: 10)
11
To be consistent with the analysis of small clauses assumed in this book (see the discussion in
Section 5.1), I use PrPs instead of vPs here. This does not impact upon the Mapping Hypothesis in
any way: the higher (CP) phase is the restrictive clause and the lower (PrP) phase is the nuclear
scope. In fact, this ambiguity could be seen as another argument in favor of treating PrP as phases.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
202
202
TP
parrots
T
T
PrP
parrots
Nuclear Scope
Pr
Pr
AdjP
loud
b. Generic reading
CP
C
TP
Restrictive Clause
T
parrots
T
PrP
parrots Pr
Pr
Nuclear Scope
AdjP
intelligent
For Diesing, the position of the subject on the generic interpretation is due to
Quantier Lowering. Given the somewhat unclear current status of Quantier
Lowering, I take it to be due to the option of interpreting the lower versus the higher
copy. There is still the question of what disallows the lower interpretation with
individual-level predicates. For Diesing, this is due to the fact that these are in fact
control clauses and there is simply no subject in the PrP- (or vP-)internal position.
Similar considerations apply to objects. Diesing shows that indenite objects
are in principle ambiguous between specic (or denite interpretations) and
existential interpretations. Biskup (2009a), following Diesings insights, derives
this ambiguity from the fact that they can be interpreted either inside the domain
of existential closure (within the vP phase) or outside it (within the CP phase). In
languages like German (i.e. languages with overt scrambling), the two interpretations can be disambiguated by scrambling. In (43a), the object remains within
the vP (the domain of existential closure) and is interpreted existentially. In (43b),
on the other hand, the object has moved out of the domain of existential closure
(as evidenced by its position relative to the manner adverb immer always),
which makes the existential interpretation no longer available.
{DESKTOP}NEW FOLDER//9781107040847C07.3D
203
Phases and LF
203
liest
reads.
[German]
Biskup (2009a) goes a step further, and, in addition to making the explicit
proposal that Diesings Clause Partitioning corresponds to the split between the
CP and the vP phase, also argues that the boundary between these two phases
corresponds to the information-structure distinction between topic (old information) and focus (new information), or, to use his terms, background information
and informational focus. Thus, vP-internal elements are interpreted as new
information (informational focus), whereas vP-external elements (which for
subjects implies movement to [Spec,TP], and for objects scrambling above vP)
as old information (background, topic). This is illustrated by the contrast in (44a
b). In (44a), the subject Maria, being vP-internal, is interpreted as new information. This sentence would be an appropriate response to the question Who
bought a new car? However, in (44b), being vP-external, it is interpreted as
background information. This would be a felicitous response to the question
What did Maria buy?
(44) a. Nowy samochd kupia Maria.
new
car
bought Maria
Maria bought this new car.
b. Maria kupia nowy samochd.
Maria bought new car
Maria bought a new car.
[Polish]
The structures of the two are given in (45ab), with insignicant details omitted
and/or simplied.
(45) a. Structure of (44a)
CP
C
TP
a new car
bought
vP
Maria
v
v
VP
bought a new car
{DESKTOP}NEW FOLDER//9781107040847C07.3D
204
204
b. Structure of (44b)
CP
C
Maria
TP
bought
vP
v
Maria
v
Considerations of this sort show that the boundary between vP and CP phases is
relevant not only to syntax, semantics and phonology, but also to information
structure.
{DESKTOP}NEW FOLDER//9781107040847C08.3D
205
8
Summary
{DESKTOP}NEW FOLDER//9781107040847C08.3D
206
206
Summary
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
References
207
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
208
References
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
References
209
Bittner, Maria & Ken Hale. 1996. The Structural Determination of Case and Agreement.
Linguistic Inquiry 27: 167.
Bobaljik, Jonathan David. 2003. Floating Quantiers: Handle With Care. In The Second
Glot International State-of-the-Article Book, ed. L. Cheng & R. Sybesma, 10748.
Berlin and New York: Mouton de Gruyter.
Bobaljik, Jonathan David & Dianne Jonas. 1996. Subject Positions and the Roles of TP.
Linguistic Inquiry 27: 195236.
Bobaljik, Jonathan David & Susi Wurmbrand. 2005. The Domain of Agreement. Natural
Language and Linguistic Theory 23: 80965.
Boeckx, Cedric. 2001. Scope Reconstruction and A-Movement. Natural Language and
Linguistic Theory 19: 50348.
2006. Linguistic Minimalism: Origins, Concepts, Methods, and Aims. Oxford
University Press.
2008a. Understanding Minimalist Syntax: Lessons from Locality in Long-Distance
Dependencies. Malden, MA: Blackwell.
2008b. Bare Syntax. Oxford University Press.
2012. Syntactic Islands. Cambridge University Press.
Boeckx, Cedric & Kleanthes K. Grohmann 2003. Multiple Wh-Fronting. Amsterdam and
Philadelphia: John Benjamins.
2007. Remark: Putting Phases in Perspective. Syntax: 10: 20422.
Booij, Geert & Jerzy Rubach. 1987. Postcyclic versus Postlexical Rules in Lexical
Phonology. Linguistic Inquiry 18: 144.
Borer, Hagit. 1984. Parametric Syntax: Case Studies in Semitic and Romance Languages.
Dordrecht: Foris.
Bokovi, eljko. 1997. The Syntax of Nonnite Complementation: An Economy
Approach. Cambridge, MA: MIT Press.
2002a. A-Movement and the EPP. Syntax 5: 167218.
2002b. On Multiple Wh-Fronting. Linguistic Inquiry 33: 35183.
2003. Agree, Phases and Intervention Effects. Linguistic Analysis 33: 5496.
2004a. Be Careful Where You Float Your Quantiers. Natural Language and
Linguistic Theory 22: 681742.
2004b. Object Shift and the Clause/PP Parallelism Hypothesis. In WCCFL 23: Proceedings
of the 23rd West Coast Conference on Formal Linguistics, ed. B. Schmeiser, V. Chand,
A. Kelleher, & A. Rodriguez, 10114. Somerville, MA: Cascadilla Press.
2005. On the Locality of Left Branch Extraction and the Structure of NP. Studia
Linguistica 59: 145.
2007. On the Locality and Motivation of Move and Agree: an Even More Minimal
Theory. Linguistic Inquiry 38: 589644.
2009. More on the No-DP Analysis of Article-Less Languages. Studia Linguistica 63:
187203.
2011. On Unvalued Interpretable Features. In NELS 39: Proceedings of the 39th
Annual Meeting of the North East Linguistic Society, ed. S. Lima, K. Mullin &
B. Smith, 10920. Amherst, MA: GLSA.
(To appear). Now Im a Phase, Now Im Not a Phase: On the Variability of Phases with
Extraction and Ellipsis. Linguistic Inquiry.
Bokovi, eljko & Steven Franks. 2000. Across-the-Board Movement and LF. Syntax
3: 10728.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
210
References
Bokovi, eljko & Daiko Takahashi. 1998. Scrambling and Last Resort. Linguistic
Inquiry 29: 34766.
Bowers, John. 1993. The Syntax of Predication. Linguistic Inquiry 24: 591656.
2002. Transitivity. Linguistic Inquiry 33: 183224.
Bresnan, Joan W. 1971. Sentence Stress and Syntactic Transformations. Language 47:
25781.
Bruening, Benjamin. 2001. QR Obeys Superiority: Frozen Scope and ACD. Linguistic
Inquiry 32: 23373.
2004. Two Types of Wh-Scope Marking in Passamaquoddy. Natural Language and
Linguistic Theory 22: 229305.
Bring, Daniel. 2005. Binding Theory. Cambridge University Press.
Butler, Jonny. 2005. The Phase Structure of Tense. In Perspectives on Phases ed.
M. McGinnis & N. Richards, 6985. Cambridge, MA: MIT Working Papers in
Linguistics.
Cable, Seth. 2008. Wh-fronting in Hungarian Is Not Focus Fronting. Manuscript.
Available at Lingbuzz (lingbuzz/000674).
Calabrese, Andrea 1984. Multiple Questions and Focus in Italian. In Sentential
Complementation, ed. W. De Geest & T. Putseys, 6774. Dordrecht: Foris.
Carnie, Andrew & Andrew Barss. 2006. Phases and Nominal Interpretation. Research In
Language 4: 12732.
Carstens, Vicki. 2000. Concord in Minimalist Theory. Linguistic Inquiry 31: 31955.
2003. Rethinking Complementizer Agreement: Agree with a Case-Checked Goal.
Linguistic Inquiry 34: 393412.
2005. Agree and EPP In Bantu. Natural Language and Linguistic Theory 23: 21979.
Carstens, Vicki & Michael Diercks. 2011. Agreeing How? Implications for Theories of
Agreement And Locality. Manuscript. Available at Lingbuzz (lingbuzz/001389).
Castillo, Juan Carlos, John E. Drury & Kleanthes K. Grohmann. 2009. Merge over Move
and the Extended Projection Principle: MOM and the EPP Revisited. Iberia: An
International Journal of Theoretical Linguistics 1: 53114.
Cecchetto, Carlo. 2004. Explaining the Locality Conditions of QR: Consequences for the
Theory Of Phases. Natural Language Semantics 12: 34597.
Cecchetto, Carlo & Caterina Donati. 2010. On Labeling: Principle C and Head
Movement. Syntax 13: 24178.
Charlow, Simon. 2009. Can DP Be a Scope Island? Talk given at the ESSLLI09 Student
Session.
Cheng, Lisa Lai-Shen. 1991. On The Typology of Wh-Questions. Doctoral dissertation,
MIT.
Chomsky, Noam. 1970. Remarks on Nominalization. In Readings In English
Transformational Grammar, by R. A. Jacobs & P. S. Rosenbaum, 184221.
Waltham, MA: Ginn & Co.
1986. Knowledge of Language: Its Nature, Origin, and Use. New York: Praeger.
1994. Bare Phrase Structure. Working Occasional Papers in Linguistics 5. Cambridge,
MA: MIT.
1995. The Minimalist Program. Cambridge, MA: MIT Press.
2000. Minimalist Inquiries: The Framework. In Step By Step. Essays on Minimalist
Syntax in Honor of Howard Lasnik, ed. R. E. A. Martin, 89155. Cambridge, MA:
MIT Press.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
References
211
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
212
References
1992. Left Branch Extraction. In NELS 22: Proceedings of the 22nd Annual Meeting
of the North East Linguistics Society, ed. K. Broderick, 6784. Amherst, MA: GLSA.
Craenenbroeck, Jeroen van. 2010. The Syntax of Ellipsis: Evidence from Dutch Dialects.
Oxford University Press.
Craenenbroeck, Jeroen van & Liliane M. V. Haegeman. 2007. The Derivation of SubjectInitial V2. Linguistic Inquiry 38: 16778.
Culicover, Peter W. & Paul Martin Postal (eds.). 2001. Parasitic Gaps. Cambridge, MA:
MIT Press.
Danon, Gabi. 2011. Agreement and DP-Internal Feature Distribution. Syntax 14: 97317.
Davies, William D. & Stanley Dubinsky. 2003. On Extraction from NPs. Natural
Language and Linguistic Theory 21: 137.
Dayal, Veneeta. 1996. Locality in Wh-Quantication: Questions and Relative Clauses in
Hindi. Dordrecht: Kluwer.
1993. Scope Marking as Indirect Wh-Dependency. Natural Language Semantics 2:
13770.
Den Dikken, Marcel. 2006a. A Reappraisal of vP Being Phasal A Reply to Legate.
Unpublished manuscript, CUNY Graduate Center.
2006b. Relators and Linkers: The Syntax of Predication, Predicate Inversion, and
Copulas. Cambridge, MA: MIT Press.
2007. Phase Extension: Contours of a Theory of the Role of Head Movement in Phrasal
Extraction. Theoretical Linguistics 33: 141.
2010. On the Functional Structure of Locative and Directional PPs. In Mapping
Spatial PPs: The Cartography of Syntactic Structures, ed. G. Cinque & L. Rizzi,
74126. Oxford University Press.
Diercks, Michael. 2011. Against Feature Inheritance: Phase Heads Are Not Dened by
Unvalued -features. Manuscript.
Diesing, Molly. 1992. Indenites. Cambridge, MA: MIT Press.
Dobashi, Yoshihiko. 2003. Phonological Phrasing and Syntactic Derivation. Doctoral
dissertation, Cornell University.
Dogil, Grzegorz. 1987. Lexical Phonology and Floating Inection in Polish. In
Phonologica 1984, ed. W. U. Dressler, 3947. Cambridge University Press.
Donati, Caterina & Carlo Cecchetto. 2011. Relabeling Heads: A Unied Account for
Relativization Structures. Linguistic Inquiry 42: 51960.
Dornisch, Ewa. 1998. Multiple Wh-Questions in Polish: The Interactions between WhPhrases and Clitics. Doctoral dissertation, Cornell University.
Drummond, Alex, Norbert Hornstein & Howard Lasnik. 2010. A Puzzle about PStranding and a Possible Solution. Linguistic Inquiry 41: 68992.
Dyakonova, Marina. 2007. Russian Double Object Construction. ACLC Working Papers
21: 128.
Dziwirek, Katarzyna. 2000. Why Polish Doesnt Like Innitives. Journal of Slavic
Linguistics 8: 5782.
Elordieta, Gorka. 2008. An Overview of Theories of the Syntax-Phonology Interface.
International Journal of Basque Linguistics and Philology (ASJU) 42(1): 20986.
Embick, David. 1995. Mobile inections in Polish. In NELS 25: Proceedings of the 25th
Annual Meeting of the North East Linguistics Society, ed. J. N. Beckman, 12742.
Amherst, MA: GLSA.
Emonds, Joseph. 1985. A Unied Theory of Syntactic Categories. Dordrecht: Foris.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
References
213
2001. The Lower Operator Position with Parasitic Gaps. In Features and Interfaces in
Romance, ed. J. Herschensohn, E. Malln & K. Zagona, 85106. Amsterdam and
Philadelphia: John Benjamins.
En, Mrvet. 1991. The Semantics of Specicity. Linguistic Inquiry 22: 125.
Engdahl, Elisabet. 2001. Parasitic Gaps. In Parasitic Gaps, ed. Peter W. Culicover &
Paul M. Postal, 6997. Cambridge, MA: MIT Press.
Epstein, Samuel. 2007. On I(nternalist)-Functional Explanation in Minimalism.
Linguistic Analysis 33: 2053.
Epstein, Samuel D. & T. Daniel Seely. 2002. Rule Applications as Cycles in a Level-Free
Syntax. In Derivation and explanation in the Minimalist Program, ed.
S. D. Epstein & T. D. Seely, 6589. Oxford: Blackwell.
2006. Transformations and Derivations. Cambridge University Press.
Fanselow, Gisbert & Damir avar. 2002. Distributed Deletion. In Theoretical
Approaches to Universals, ed. A. Alexiadou, 65107. Amsterdam and
Philadelphia: John Benjamins.
Felser, Claudia. 2004. Wh-Copying, Phases, and Successive Cyclicity. Lingua 114: 54374.
Fox, Danny. 1999. Reconstruction, Binding Theory, and the Interpretation of Chains.
Linguistic Inquiry 30: 15796.
2000. Economy and Semantic Interpretation Cambridge, MA: MIT Press.
Fox, Danny & David Pesetsky. 2005. Cyclic Linearization of Syntactic Structure.
Theoretical Linguistics 31: 146.
Frampton, John & Sam Gutmann. 2000. Agreement is Feature Sharing. Unpublished
manuscript, Northeastern University.
2002. Crash-Proof Syntax. In Derivation and Explanation in the Minimalist Program,
ed. S. D. Epstein & T. Daniel Seely, 90105. Malden, MA: Blackwell.
Franks, Steven. 1998. Clitics in Slavic. Paper Presented at the Comparative Slavic
Morphosyntax Workshop. Spencer, IN.
Franks, Steven & Piotr Baski. 1999. Approaches to Schizophrenic Polish Person
Agreement. In Formal Approaches to Slavic Linguistics: The Seattle Meeting, ed.
K. Dziwirek, H. Coats & C. M. Vakareliyska, 12344. Ann Arbor: Michigan Slavic
Publications.
Frascarelli, Mara (ed.). 2006. Phases of Interpretation. Berlin: Mouton de Gruyter.
Frederici, Angela. 1982. Syntactic and Semantic Processes in Aphasic Decits: the
Availability of Prepositions. Brain And Language 15: 26958.
Froud, Karen. 2001. Prepositions and the Lexical/Functional Divide: Aphasic Evidence.
Lingua 111: 128.
Gallego, ngel J. 2010. Phase Theory. Amsterdam and Philadelphia: John Benjamins.
2012. Phases: Developing the Framework. Berlin and Boston: De Gruyter Mouton.
Gavruseva, E. 2000. On the Syntax of Possessor Extraction. Lingua 110: 74372.
Gengel, Kirsten. 2007. Focus and Ellipsis a Generative Analysis of Pseudogapping and
Other Elliptical Structures. Doctoral dissertation, Stuttgart University.
2009. Phases and Ellipsis. Linguistic Analysis 35: 2142.
Green, Georgia M. 1974. Semantics and Syntactic Regularity. Bloomington, IN: Indiana
University Press.
Grimshaw, Jane. 1991. Extended Projection. Manuscript, Brandeis University.
Grohmann, Kleanthes K. 2003. Prolic Domains: On the Antilocality of Movement
Dependencies. Amsterdam and Philadelphia: John Benjamins.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
214
References
(ed.) 2009a. Explorations of Phase Theory: Features And Arguments. Berlin: Walter de
Gruyter.
2009b (ed.). Explorations of Phase Theory: Interpretation at the Interfaces. Berlin:
Walter de Gruyter.
2009c (ed). Interphases: Phase-Theoretic Investigations of Linguistic Interfaces.
Oxford University Press.
2011. Anti-Locality: Too-Close Relations in Grammar. In The Oxford Handbook of
Linguistic Minimalism, ed. C. Boeckx, 26090. Oxford University Press.
Grohmann, Kleanthes K., John Drury & Juan Carlos Castillo. 2000. No More EPP. In
WCCFL 19: Proceedings of the 19th West Coast Conference on Formal Linguistics,
ed. R. Billerey & B. D. Lillehaugen, 15366. Somerville, MA: Cascadilla Press.
Haegeman, Liliane. 1992. Theory and Description in Generative Syntax: A Case Study in
West Flemish. Cambridge University Press.
Haegeman, Liliane & Marjo Van Koppen. 2011. Complementizer Agreement and the
Relation Between C And T. Manuscript, Ghent University and University of Utrecht.
Hale, Ken & Jay S. Keyser. 1993. On Argument Structure and the Lexical Expression of
Syntactic Relations. In The View From Building 20: Essays in Linguistics in Honor of
Sylvain Bromberger, ed. K. Hale & J. S. Keyser, 53109. Cambridge, MA: MIT Press.
Halle, Morris & Alex Marantz. 1993. Distributed Morphology and the Pieces of
Inection. In The View From Building 20, ed. K. Hale & J. S. Keyser, 11176.
Cambridge, MA: MIT Press.
Halle, Morris & Jean-Roger Vergnaud. 1987. An Essay on Stress. Cambridge, MA: MIT Press.
Hankamer, Jorge & Ivan Sag. 1976. Deep and Surface Anaphora. Linguistic Inquiry 7:
391428.
Harves, Stephanie. 2002. Where Have All the Phases Gone? (Non-)Defective Categories
and Case Alternations in Russian. In Formal Approaches To Slavic Linguistics 10:
The Second Ann Arbor Meeting, ed. J. Toman, 97118. Ann Arbor, MI: Michigan
Slavic Publications.
Heycock, Caroline. 1995. Asymmetries in Reconstruction. Linguistic Inquiry 26: 54770.
Hiraiwa, Ken. 2001. Multiple Agree and the Defective Intervention Constraint in
Japanese. In Proceedings of HUMIT 2000, 6780. Cambridge, MA: MIT Working
Papers in Linguistics.
Hiraiwa, Ken. 2005. Dimensions of Symmetry in Syntax: Agreement and Clausal
Architecture. Doctoral dissertation, MIT.
Holmberg, Anders. 2001. The Syntax of Yes and No in Finnish. Studia Linguistica 55:
14175.
Hornstein, Norbert. 1994. An Argument for Minimalism: the Case of AntecedentContained Deletion. Linguistic Inquiry 25: 45580.
1995. Logical Form. Oxford: Blackwell.
Hornstein, Norbert & Jairo Nunes. 2002. On Asymmetries between Parasitic Gap and
Across-the-Board Constructions. Syntax 5: 2654.
Hornstein, Norbert, Jairo Nunes & Kleanthes K. Grohmann. 2005. Understanding
Minimalism. Cambridge University Press.
Hornstein, Norbert & Amy Weinberg. 1981. Case Theory and Preposition Stranding.
Linguistic Inquiry 12: 5591.
Horvath, Julia. 1997. The Status of Wh-Expletives and the Partial Wh-Movement
Construction of Hungarian. Natural Language and Linguistic Theory 15: 50972.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
References
215
Huang, Cheng-Teh James. 1982. Logical Relations in Chinese and the Theory of
Grammar. Doctoral dissertation, MIT.
1993. Reconstruction and the Structure of VP: Some Theoretical Consequences.
Linguistic Inquiry 24: 10338.
Iatridou, Sabine & Ivy Sichel. 2011. Negative DPs, A-Movement, and Scope
Diminishment. Linguistic Inquiry 42: 595629.
Inenlikj, Petr I. & Vladimir Nedjalkov. 1973. Glagoly uvstva v ukotskom jazyke.
Lingvistieskie issledovanija 1972, 175203.
Ishihara, Shinichiro. 2003. Intonation and Interface Conditions. Doctoral dissertation, MIT.
2007. Major Phrase, Focus Intonation, Multiple Spell-Out (Map, FI, MSO). The
Linguistic Review 24: 23.
Jackendoff, Ray. 1977. X-bar Syntax: A Study of Phrase Structure. Cambridge, MA:
MIT Press.
Jaysaseelan, K. A. 2001. IP-Internal Topic and Focus Phrases. Studia Linguistica 55:
3975.
Jeong, Youngmi. 2006. The Landscape of Applicatives. Doctoral dissertation, University
of Maryland.
Johnson, Kyle. 2000. Few Dogs Eat Whiskas or Cats Alpo. In UMOP 23: Issues
in Semantics and its Interface, ed. K. Kusumoto and E. Villalta, 5982. Amherst,
MA: GSLA.
2009. Gapping Is Not (VP-)Ellipsis. Linguistic Inquiry 40: 289328.
Kahnemuyipour, Arsalan. 2003. Syntactic Categories and Persian Stress. Natural
Language and Linguistic Theory 21: 33379.
2004. The Syntax of Sentential Stress. Doctoral dissertation, University of Toronto.
2005. Towards a Phase-Based Theory of Sentential Stress. In Perspectives on Phases,
ed. M. McGinnis & N. Richards, 12546. Cambridge, MA: MIT Working Papers In
Linguistics.
Kasai, Hironobu. 2007. Multidominance in Syntax. Doctoral dissertation, Harvard
University.
Katz, Jerrold J. & Paul Martin Postal. 1964. An Integrated Theory of Linguistic
Descriptions. Cambridge, MA: MIT Press.
Kayne, Richard S. 1980. Extensions of Binding and Case-Marking. Linguistic Inquiry
11: 7596.
1981. On Certain Differences between French and English. Linguistic Inquiry
12: 34972.
1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
1999. Prepositional Complementizers as Attractors. Probus 11: 3973.
2004. Prepositions as Probes. In Structures and Beyond, ed. A. Belletti, 192-X. Oxford
University Press.
2010. Why Are There No Directionality Parameters? Manuscript, New York
University.
Kennedy, Christopher. 1997. Antecedent-Contained Deletion and the Syntax of
Quantication. Linguistic Inquiry 28: 66288.
Khomitsevich, Olga. 2007. Dependencies across Phases: From Sequence of Tense to
Restrictions on Movement. Doctoral dissertation, Utrecht: LOT.
Kinyalolo, Kasangati Kikuni Wabongambilu. 1991. Syntactic Dependencies and the
SPECHead Agreement Hypothesis in Kilega. Doctoral dissertation, UCLA.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
216
References
Kiparsky, Paul. 1998. Partitive Case and Aspect. In The Projection of Arguments, ed.
M. Butt & W. Gueder. Stanford: CSLI Publications.
Kiss, Katalin . 1987. Congurationality in Hungarian. Dordrecht: Reidel.
Koopman, Hilda 2000. Prepositions, Postpositions, Circumpositions, and Particles: The
Case of Dutch PPs. In The Syntax of Speciers and Heads: Collected Essays of Hilda
J. Koopman, ed. H. Koopman, 20460. London: Routledge.
Kratzer, Angelika & Elisabeth Selkirk. 2007. Phase Theory and Prosodic Spell-Out: The
Case of Verbs. The Linguistic Review 24: 93135.
Laenzlinger, Christopher. 2005. French Adjective Ordering: Perspectives on DP-Internal
Movement Types. Lingua 115: 64589.
Lahne, Antje. 2008. Where There Is Fire There Is Smoke: Local Modelling of SuccessiveCyclic Movement. Doctoral dissertation, University Of Leipzig.
Landau, Idan. 1999. Elements of Control. Doctoral dissertation, MIT.
2000. Elements of Control: Structure and Meaning in Innitival Constructions
Dordrecht and Boston: Kluwer.
2003. Movement out of Control. Linguistic Inquiry 34: 47198.
2008. Two Routes of Control: Evidence from Case Transmission in Russian. Natural
Language and Linguistic Theory 26: 877924.
2010. The Locative Syntax of Experiencers. Cambridge, MA: MIT Press.
Larson, Richard K. 1985. Quantifying into NP. Manuscript, Stony Brook University.
1988. Light Predicate Raising. MIT Lexicon Project Working Papers 27. Cambridge,
MA: MIT Center for Cognitive Science.
Larson, Richard & Robert May. 1990. Antecedent Containment or Vacuous Movement:
Reply to Baltin.Linguistic Inquiry 21: 10322.
Lasnik, Howard. 1992. Case and Expletives: Notes Toward a Parametric Account.
Linguistic Inquiry, 23: 381405.
1998. Exceptional Case Marking: Perspectives Old And New. In Formal Approaches
To Slavic Linguistics: The Connecticut Meeting 1997, ed. Z. Bokovi, S. Franks &
W. Snyder, 187211. Ann Arbor: Michigan Slavic Publications.
1999. Chains of Arguments. In Working Minimalism, ed. S. D. Epstein &
N. Hornstein, 189215. Cambridge, MA: MIT Press.
Lebeaux, David. 1990. Relative Clauses, Licensing, and the Nature of the Derivation. In
NELS 20: Proceedings of the 20th Annual Meeting of the North East Linguistic
Society, ed. J. Carter, R.-M. Dechaine, B. Philip & T. Sherer, 31832. Amherst, MA:
GLSA.
1991. Relative Clauses, Licensing, and the Nature of the Derivation. In Syntax and
Semantics 25: Perspectives on Phrase Structure, ed. S. Rothstein, 20939. New
York: Academic Press.
1994. Where does binding theory apply? MS, University of Maryland, College Park.
2009. Where Does Binding Theory Apply? Cambridge, MA: MIT Press.
Legate, Julie Anne. 2002. Phases in Beyond Explanatory Adequacy. Manuscript.
2003. Some Interface Properties of the Phase. Linguistic Inquiry 34: 50616.
2011. Under-Inheritance. Handout of a paper presented at NELS 42, University of
Toronto.
2012. The size of Phases. In Phases, ed. A. Gallego, 23350. Berlin: De Gruyter.
Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity: At the SyntaxLexical
Semantics Interface. Cambridge, MA: MIT Press.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
References
217
Lobeck, Anne C. 1995. Ellipsis: Functional Heads, Licensing, and Identication. New
York: Oxford University Press.
Lutz, Uli, Gereon Mller & Arnim von Stechow. 2000. Wh-Scope Marking. Amsterdam
and Philadelphia: John Benjamins.
Macdonald, Jonathan Eric. 2006. The Syntax of Inner Aspect. Doctoral dissertation,
Stony Brook University.
Mahajan, Anoop Kumar. 1989. Agreement and Agreement Phrases. In Functional Heads
and Clause Structure, ed. I. Laka & A. K. Mahajan, 21752, Cambridge, MA: MIT
Working Papers in Linguistics.
1996. Wh-Expletives and the Syntax of Partial Wh-Movement. In Papers On Scope
Marking, ed. U. Lutz & G. Mller, 16377. University of Tbingen / University of
Stuttgart.
Manetta, Emily. 2010. Wh-Expletives in Hindi-Urdu: The vP Phase. Linguistic Inquiry
41: 134.
Marantz, Alec. 1993. Implications of Asymmetries in Double Object Constructions. In
Theoretical Aspects of Bantu Grammar, ed. S. Mchombo, 11350. Stanford: CSLI
Publications.
2007. Phases and Words. In Phases in the Theory of Grammar, ed. H. Choe, 191220.
Seoul: Dong In.
Martin, Roger Andrew. 1996. A Minimalist Theory of PRO and Control. Doctoral
dissertation, University Of Connecticut.
Marui, Franc Lanko. 2005. On Non-Simultaneous Phases. Stony Brook University.
2009. Non-Simultaneous Spell-Out in the Clausal and Nominal Domain. In
Interphases: Phase-Theoretic Investigations of Linguistic Interfaces, ed.
K. K. Grohmann, 15181. Oxford University Press.
Matushansky, Ora. 2000. The Instrument of Inversion: Instrumental Case and Verb
Raising in the Russian Copula. In WCCFL 19: Proceedings of the 19th West Coast
Conference on Formal Linguistics, ed. R. Billerey & B. Lillehaugen, 10115.
Somerville, MA: Cascadilla Press.
2005. Going through a Phase. In Perspectives on Phases, ed. M. McGinnis &
N. Richards, 15781. Cambridge, MA: MIT Working Papers in Linguistics.
2008. Predication: A Case Study. In Studies in Formal Slavic Linguistics:
Contributions from Formal Description of Slavic Languages 6.5, ed. F. Marusic &
R. Zaucer, 21339. Frankfurt am Main: Peter Lang.
May, Robert 1977. The Grammar of Quantication. Doctoral dissertation, MIT.
1985. Logical Form: Its Structure and Derivation. Cambridge, MA: MIT Press.
McCloskey, James. 1998. The Prosody of Quantier Stranding under Wh-Movement in
West Ulster English. Manuscript.
2000. Quantier Float and Wh-Movement in an Irish English. Linguistic Inquiry 31:
5784.
2001. The Morphosyntax of WH-Extraction In Irish. Journal of Linguistics 37:
67100.
2002. Resumption, Successive Cyclicity, and the Locality of Operations. In
Derivation and Explanation in the Minimalist Program, ed. S. D. Epstein &
T. Daniel Seely, 184226. Malden, MA: Blackwell.
McDaniel, Dana. 1989. Partial and Multiple Wh-Movement. Natural Language and
Linguistic Theory 7: 565604.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
218
References
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
References
219
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
220
References
Richardson, Kylie. 2007. Case and Aspect in Slavic. Oxford University Press.
Ritter, Elizabeth. 1992. Cross-Linguistic Evidence for Number Phrase. Canadian
Journal of Linguistics Revue Canadienne de Linguistique 37: 197218.
1993. Wheres Gender? Linguistic Inquiry, 24: 795803.
Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris.
Roberts, Ian & Anders Holmberg. 2010. Introduction: Parameters in Minimalist Theory.
In Parametric Variation: Null Subjects in Minimalist Theory, ed. T. Biberauer,
A. Holmberg, I. Roberts & M. Sheehan, 157. Cambridge University Press.
Rooryck, Johan. 1997. On the Interaction between Raising and Focus in Sentential
Complementation. Studia Linguistica 51: 149.
Ross, John Robert. 1967. Constraints on Variables in Syntax. Doctoral dissertation, MIT.
Rouveret, Alain. 2012. VP Ellipsis, Phases and the Syntax of Morphology. Natural
Language And Linguistic Theory 30: 897963.
Rudin, Catherine. 1988. On Multiple Questions and Multiple WH Fronting. Natural
Language and Linguistic Theory 6: 445501.
Rutkowski, Pawe. 2002 The Syntax of Quantier Phrases and the Inherent versus
Structural Case Distinction. Linguistic Research 7: 4374.
Rutkowski, Pawe & Ljiljana Progovac. 2005. Classication Projection in Polish and
Serbian. In Formal Approaches to Slavic Linguistics 2004, ed. S. Franks et al.,
28999. Ann Arbor: Michigan Slavic Publications.
Sabbagh, Joseph. 2007. Ordering and Linearizing Rightward Movement. Natural
Language and Linguistic Theory 25: 349401.
Sag, Ivan A. 1976. Deletion and Logical Form. Doctoral dissertation, MIT.
Saito, Mamoru. 2003. Derivational Approach to the Interpretation of Scrambling Chains.
Lingua 113: 481518.
Saito, Mamoru, T. H. Jonah Lin & Keiko Murasugi. 2008. N-Ellipsis and the Structure of
Noun Phrases in Chinese and Japanese. Journal of East Asian Linguistics 17: 24771.
Samuels, Bridget. 2011. A Minimalist Program for Phonology. In The Oxford Handbook
of Linguistic Minimalism, ed. C. Boeckx, 57494. Oxford University Press.
Sato, Yosuke. 2011. P-Stranding under Sluicing and Repair by Ellipsis: Why Is
Indonesian (Not) Special? Journal Of East Asian Linguistics 20: 33982.
2012. Successive Cyclicity at the Syntax-Morphology Interface: Evidence from
Standard Indonesian and Kendal Javanese. Studia Linguistica 66: 3257.
Sauerland, Uli. 2005. DP Is Not a Scope Island. Linguistic Inquiry 36: 30314.
Sauerland, Uli & Paul Elbourne. 2002. Total Reconstruction, PF Movement, and
Derivational Order. Linguistic Inquiry 33: 283319.
Scheer, Tobias 2012. Chunk Denition in Phonology: Prosodic Constituency vs. Phase
Structure. In Modules and Interfaces, ed. M. Bloch-Trojnar & A. Bloch-Rozmej,
22153. Lublin: Wydawnictwo KUL.
Scott, Gary-John. 2002. Stacked Adjectival Modication and the Structure of Nominal
Phrases. In Functional Structure in DP and IP, ed. G. Cinque, 91120. Oxford
University Press.
Seely, T. Daniel. 2006. Merge, Derivation C-Command and Subcategorization in a LabelFree Syntax. In Miminalist Essays, ed. C. Boeckx, 182220. Amsterdam and
Philadelphia: John Benjamins.
Seidl, Amanda Hallie. 2000. Minimal Indirect Reference: A Theory of the Syntax
Phonology Interface. Doctoral dissertation, University of Pennsylvania.
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
References
221
{DESKTOP}NEW FOLDER//9781107040847RFA.3D
222
References
{DESKTOP}NEW FOLDER//9781107040847IND.3D
Index
Barros, Matthew, 63
Barss, Andrew, 4, 66, 74, 124, 152, 200
Belletti, Adriana, 35, 36, 208
Berber, 73
Bernstein, Judy, 108, 109
Bhatt, Rajesh, 37, 38
Biberauer, Theresa, 176
binding, 74, 75, 90, 112, 142, 148, 149
Biskup, Petr, 6, 66, 142, 200, 202, 203
Bittner, Maria, 109
Bobaljik, Jonathan, 36, 70, 77, 174
Boeckx, Cedric, 1, 2, 3, 12, 37, 67, 69, 71,
73, 181
Booij, Geert, 51
Borer-Chomsky conjecture, 74, 176
Bokovi, eljko, 19, 37, 40, 53, 54, 55, 62, 65,
70, 73, 77, 79, 90, 91, 115, 138, 149, 150,
171173, 174175, 179, 184, 199
Bowers, John, 124
Bresnan, Joan W., 190, 191
Bruening, Benjamin, 67, 82, 102, 103, 136
Bukusu, 152, 155
Bulgarian, 73
Bring, Daniel, 148
Burzios Generalization, 92
Butler, Jonny, 200
Cable, Seth, 95
Calabrese, Andrea, 73
Carnie, Andrew, 66, 124, 200
Carstens, Vicki, 4, 50, 51, 84, 121, 123
case
Accusative, 17, 91
223
{DESKTOP}NEW FOLDER//9781107040847IND.3D
224
Index
case (cont.)
case concord, 113, 121, 174
Dative, 114
Genitive, 35, 111, 173
Instrumental versus Nominative, 126128
Locative, 142
Nominative, 17, 19, 21, 108, 167
quirky, 35, 47
valuation, 21, 121123, 127
Castillo, Juan Carlos, 90
avar, Damir, 79
Cecchetto, Carlo, 12, 13, 67, 177
Chamorro, 97, 98, 115
Charlow, Simon, 117, 119
Cheng, Lisa, 19
Chi-Mwhi:ni:, 198
Chinese, 95
Chomsky, Noam, 1, 2, 3, 7, 8, 10, 11, 12, 13, 14,
16, 18, 19, 20, 21, 23, 28, 29, 30, 31, 32, 33,
43, 47, 48, 49, 50, 54, 62, 66, 67, 74, 80, 104,
108, 111, 112, 125, 148, 163, 168, 176, 189
Chukchee, 37, 38
Chung, Sandra, 64, 97, 98, 211
Cinque, Guglielmo, 109, 137, 139, 140, 142,
189, 191
Citko, Barbara, iv, 10, 12, 13, 22, 64, 95, 123,
126, 127, 128, 129, 158, 171, 182, 184, 211
Clause Partitioning, 203
Cole, Peter, 98
Collins, Chris, 13
complementizer agreement, 4, 5051, 8385
conceptual-intentional system, 8
Condition on Extraction Domains, 69
control, 88
exhaustive, 39
obligatory, 39
partial, 39
Copy Theory of Movement, 80
Corver, Norbert, 115, 172
Craenenbroeck, Jeroen van, 52, 53, 64, 145
Culicover, Peter W., 101, 102, 184, 222
cyclic linearization, 6, 185, 188
Danon, Gabi, 16, 121
Davies, William D., 114
Dayal, Veneeta, 4, 82
Den Dikken, Marcel, 3, 4, 5, 70, 100, 106, 124,
136, 138, 142, 160, 161166, 167, 169, 199
{DESKTOP}NEW FOLDER//9781107040847IND.3D
Index
features
EPP feature, 32, 47, 69, 75, 167
feature valuation, 18, 68, 110, 111112
interpretability versus valuation, 1920
interpretable, 14, 15
interpretable versus uninterpretable, 1415
uninterpretable, 3, 14, 18, 30, 31, 4647, 60,
110, 176
Felser, Claudia, 4, 6, 67, 71, 78, 79, 80, 81, 82, 177
Finnish, 17
Fox, Danny, 6, 74, 75, 90, 100, 101, 185, 187,
188, 219
Frampton, John, 9, 19
Franks, Steven, 51, 79, 140, 184, 209
Frascarelli, Mara, 1
Frederici, Angela, 139
French, 17
Froud, Karen, 139
Gallego, ngel J., 1, 3, 5, 12, 13, 31, 68, 69, 160,
161, 166, 167, 169, 176
Gavruseva, Elena, 115
Gengel, Kirsten, 64
German, 17, 79, 82, 108, 203
Goal, 20, 21, 22, 38, 40, 110, 122
Green, Georgia M., 151
Grimshaw, Jane, 139
Grohmann, Kleanthes, 1, 2, 3, 7, 32, 37, 69,
77, 90
Gutmann, Sam, 9, 19
Haegeman, Liliane, 4, 50, 51, 52, 53, 85
Hale, Ken, 66, 109
Halle, Morris, 8, 189
Hankamer, Jorge, 64
Harves, Stephanie, 124, 126, 129
Hermon, Gabriella, 98
Heycock, Caroline, 75, 214
Hindi-Urdu, 38, 83, 95, 96
Hiraiwa, Ken, 5, 21, 108, 122
Holmberg, Anders, 48, 64, 187, 208
Holmbergs Generalization, 48, 187
Hornstein, Norbert, 7, 23, 32, 78, 101,
103, 138
Horstein, Norbert, 136
Horvath, Julia, 82
Hovav, Malka Rappaport, 58
Huang, Cheng-Teh James, 69, 112
225
{DESKTOP}NEW FOLDER//9781107040847IND.3D
226
Index
{DESKTOP}NEW FOLDER//9781107040847IND.3D
Index
Polish, 16, 35, 51, 55, 79, 82, 83, 86, 87, 95, 99,
111, 113, 121, 126, 129, 134, 140, 141, 145,
149, 150, 158, 167, 172, 173, 174, 203
possessor extraction, 115
Postal, Paul, 78, 80, 101, 215
Potsdam, Eric, 37, 38
predication phrases, 124125
extraction from, 129131
Preminger, Omer, 9
preposition stranding, 78, 85, 86, 99, 143144,
145148, 168, 175
prepositions
directional, 141143, 145
functional, 140, 145
lexical, 140, 145
locative, 141143, 145
Probe, 12, 20, 21, 22, 48, 84, 110, 122, 126, 127,
139, 206
Progovac, Ljiljana, 171
Pylkknen, Liina, 151, 152, 153, 154, 170
quantier oat. See quantier stranding
quantier lowering, 202
quantier raising, 66, 67, 102, 107, 117119,
136, 150, 181
quantier stranding, 7677, 90, 98, 133134
Quinoli, Carlos, 112
Rackowski, Andrea, 97, 99
Radkevich, Nina, 5, 142, 148, 149
Ramchand, Gillian, 143
Rappaport, Gilbert C., 171
Rausch, Philip, 108
reconstruction, 4, 7475, 90, 100101, 106, 179
Reinhart, Tanya, 148, 149
Relator Phrase, 164
restrictive clause, 66
Reuland, Eric, 148, 149
Richards, Marc, 3, 18, 33, 36, 37, 41, 49, 50, 53,
60, 162
Richards, Norvin, 1, 73, 97, 99
Richardson, Kylie, 126
right node raising, 62, 182184
Ritter, Elizabeth, 109
Rizzi, Luigi, 35, 73, 137
Roberts, Ian, 128
Romani, 80
Romanian, 73
227
Rooryck, Johan, 54
Ross, John Robert, 172
Rouveret, Alain, 64
Rubach, Jerzy, 51
Rubin, Edward, 126
Rudin, Catherine, 73
Russian, 17, 82, 83, 126, 152
Rutkowski, Pawe, 171
Sabbagh, Joseph, 184
Sag, Ivan, 64
Saito, Mamoru, 179, 180
Samuels, Bridget, 185
Sato, Yosuke, 86, 97, 98
Sauerland, Uli, 117, 119, 180, 181
scattered deletion, 79
Scheer, Tobias, 185
scope, 184, 198199
scope marking, 4, 8183, 96
Scott, Gary-John, 109
Scottish Gaelic, 143, 190
Seely, T. Daniel, 13, 18, 50, 59, 70,
90, 199
Seidl, Amanda Hallie, 197
Selkirk, Elisabeth, 6, 185, 190, 191
semantic types, 66, 199
ener, Serkan, 65
sensorimotor system, 8
sentence fragments, 6162
sequence of tense, 55
Serbo-Croatian, 172
short wh-movement, 95
Shortest Link, 155
Sichel, Ivy, 181
Sigursson, Halldr rmann, 35, 47
Single Answer readings, 101
Slovak, 149
small clauses, 124, 131
Somali, 73
Spell-Out, 60, 61
Delayed Spell-Out, 182
Multiple Spell-Out, 1, 3, 4143, 177, 185
Non-Simultaneous Spell-Out, 43
Spellee, 195
Spell-Out domain, 32, 161, 185, 195
Sportiche, Dominique, 77
Starke, Michal, 134
Stepanov, Arthur, 4, 82, 83
{DESKTOP}NEW FOLDER//9781107040847IND.3D
228
Stjepanovic, Sandra, 86
Stowell, Timothy, 55, 88
Stranding Generalization, 145
Strong Minimalist Thesis, 8
subject to object raising. See exceptional case
marking
successive cyclic movement, 4, 32, 57, 67, 70,
7576, 78, 80, 83, 8891, 94
Svenonius, Peter, 5, 15, 17, 47, 109, 124
Swan, Oscar, 111
Szabolcsi, Anna, 108, 114, 115
Szczegielniak, Adam, 86
Tagalog, 97
Takahashi, Daiko, 65, 179
Tamazight Berber, 52
Tanaka, Tomoyuki, 124, 125, 131, 132, 133
Tellier, Christine, 108, 116
Thornton, Rosalind Jean, 80
Tomaszewicz, Barbara, 129
Torrego, Esther, 12, 17, 19, 20, 21, 53, 167
Transfer, 19, 30, 4142, 177, 180
Multiple Transfer, 178, 180
Non-Simultaneous Transfer, 179, 180, 181
Simultaneous Transfer, 177
transitive expletive, 36
Travis, Lisa, 195
Tsez, 37, 38
Tsujioka, Takae, 152
Turkish, 17
Index
Tzez, 38
Tzotzil, 115
Uriagereka, Juan, 1, 43, 115, 172
Valois, Daniel, 108, 109, 113, 115, 116
Value-Transfer Simultaneity, 49
van Riemsdijk, Henk, 139, 140, 143, 144
variable binding, 90, 100, 101, 106
Vergnaud, Jean-Roger, 189, 190
Vicente, Luis, 63
Vogel, Irene, 188
voice mismatches, 94
Wachowicz, Krystyna A., 73
Weinberg, Amy, 78
West Flemish, 51
West Ulster English, 76, 77, 99, 149
Wexler, Kenneth, 184
wh-agreement, 98
wh-copying, 4, 6, 76, 79, 8081
Wiland, Bartosz, 4, 99
Wurmbrand, Susi, 70, 174
Yadroff, Michael, 139, 140
Yokogoshi, Azusa, 124, 125, 131, 132, 133
Zagona, Karen, 64
Zamparelli, Roberto, 109
Zwart, C. Jan-Wouter, 4