The Grammar of Focus

at the Interfaces
Aritz Irurtzun Sviaguincheva

2007
Hizkuntzalaritza eta Euskal Ikasketak Saila
Department of Linguistics and Basque Studies

Advisors:
Gorka Elordieta Alzibar
Myriam Uribe-Etxebarria Goti
ACKNOWLEDGEMENTS

Although after the defense I will write the proper acknowledgements in this
preliminary version I do not want to forget those that helped me in this trip: my
advisors, Gorka Elordieta and Myriam Uribe-Etxeberria, and all the people involved
in the Department of Linguistics and Basque Studies and the PhD. Program in
Linguistics of the University of the Basque Country, specially Alazne Landa and
Myriam Uribe-Etxebarria. Many thanks as well to Xabier Artiagoitia, Cedric Boeckx,
Luis Eguren, Ricardo Etxepare, Kleanthes Grohmann, Utpal Lahiri, Itziar Laka, Jon
Ortiz de Urbina, Beñat Oyharçabal, Vidal Valmala, Juan Uriagereka, Elena

Herburger, Lisa Selkirk, Angelika Kratzer, Žjelko Bošković, Harry van der Hulst,

Chris Potts, Serkan Şener, Nilufer Şener, Luisa Marti, Klaus Abels, Florian Wagner,
John McCarthy, Richard Larson, Ilaria Frana, Keir Moulton, Carlos Gussenhoven.
Phoevos Panagiotidis, Pilar Prieto, José Ignacio Hualde, Jon Franco, Hamida
Demirdache, Orin Percus, Nicolas Guilliot, Valérie Gautier, Milan Rezac, Mélanie
Jouitteau, Anamaria Falaus, Magda Oiry, Michael Wagner, Fernando García Murga,
Javi Ormazabal, Joseba A. Lakarra, Arantzazu Elordieta, M. Lourdes Oñederra, Pablo
Albizu, Marijo Ezeizabarrena, Luis Vicente, Brenda Laca, Angel J. Gallego, Txuss
Martin, Maia Duguine, Nerea Madariaga, Susana Huidobro, Urtzi Etxeberria, Kepa
Erdozia, Adam Zawiszewski, Euridice Arregi, Ibon Manterola, Mikel Santesteban,
Oroitz Jauregi, Julen Manterola, Javi Fernandez, Brendan Costello, Nagore Calleja,
Izaskun Villareal, Jonathan McDonald, Olga Fernández-Soriano, Raquel González,

Enric Valldiví, Lanko Marušič, Tomoko Kawamura, Bill Haddican, HIM, and the

Departments of Linguistics of UConn, UMass, Université de Nantes/Naoned,

Université de Paris 8, Universiat Autònoma de Barcelona, Universitat de Barcelona,
Instituto Universitario Ortega y Gasset, University of Deusto, and IKER-CNRS and
the Department of Basque Studies of the Université de Pau et des Pays de l’Adour.
Eskerrik asko denei.
This thesis has been developed thanks to a predectoral grant by the Basque
Government.

i
ABSTRACT

This dissertation studies the overall architecture of grammar and the type of
interactions between its different modules by providing an innovative and thorough
analysis of the nature of focus. I analyze some of the most characteristic syntactic,
morphological, phonological and semantic properties of focus and provide an
explanatory analysis for them within the general “syntactocentric” conception of
language that has characterized Generative Grammar from its origins. With that
purpose, I offer a detailed analysis of the grammar of focus in Basque, a language
where focus displays a wide variety of phonological, semantic and morphosyntactic
effects which make it a particularly well suited language for investigating the nature
of focus and the interaction of the different modules of the grammar. Further, I also
bring into the discussion a number of phenomena characteristic of focal constructions
which have been attested in other languages.
My proposal is to capture the nature of focus with a Derivational Approach to
the Focus Structure that construes the F-Structure in the narrow syntax, hence
allowing interpretive modules to operate upon the syntactically fixed F-Structure.
Thus F-Structure composition is prior to and independent from any stress placement.
As for the interfaces, in PF I propose a focus-sensitive nuclear stress rule (the FNSR)
that assigns nuclear stress to the most deeply embedded element within focus; this
accounts for all the basic patterns of nuclear stress placement in a natural way, while
it allows us to circumvent cases of mismatch between the F-Structure and the nuclear
stress placement. At logical form I propose, following Herburger (2000) that focus falls
into the scope of a quantifier over events. But I propose two types of quantifiers: a
bare existential quantifier in out-of-the-blue utterances and a δ-quantifier in
contextualized utterances. These two types of quantifiers explain naturally the
presuppositionality and exhaustivity readings of each type of construction. Besides, I
argue that the syntax-logical form inyterface of contextualized utterances is a simple
mapping whereas in contextualized utterances it involves Hornstein & Uriagereka’s
(2002) ‘reprojection’ mechanism of the binary quantifier to take nuclear scope. This
reprojection will be argued to derive from the complex CP system of contextualized

ii
utterances that trigger focus-movement in contextualized utterances. Finally, I also
analyze the answers of multiple-Wh questions as instances of split F-Structures.
The main conceptual conclusions that derive from this dissertation with
regards to the interfaces are the following: Syntax-semantics and syntax-phonology
interfaces are unidirectional: Syntax → Phonology, Syntax → Semantics. In
opposition to recent Nuclear Stress Rule-based (NSR-based) proposals on the
grammar of focus (cf. i.a. Reinhart (1995, 2006), Zubizarreta (1998), Jackendoff (2002)),
the analysis presented in this thesis provides additional evidence in favor of the T-
model of the architecture of grammar. A ‘syntactocentric’ model of the architecture of
grammar can explain all the PF and LF properties of focus in a natural way, whereas,
as I discuss at length, NSR-based proposals have strong empirical and conceptual
problems.
Then, the main conclusion that emerges from this work is that a
‘syntactocentric’ architecture of grammar constitutes the best explanation for all the
properties displayed by focus.

iii
Table of Contents

Chapter 1: Introduction: Scope and Goals of the Dissertation 5

1.1. Scope and Goals of the Dissertation 5
1.2. The Framework 9
1.2.1. Minimalist Syntax and the Architecture of Grammar 10
1.2.1.1. A Note on Phases 13
1.2.2. Event Semantics and logical form 16
1.2.2.1. Quantification over Events 16
1.2.2.2. Neo-Davidsonian Separatism 22
1.2.3. Autosegmental-Metrical Phonology 24
1.3. The Nature of Merge and the Status of Syntactic Labels 28
1.3.1. Phrase Structure Composition: Merger and Projection 29
1.3.1.1. Introduction: Structure Building in Bare Phrase Structure 30
1.3.1.2. An Attempt to Eliminate Labels: The Locus Theory 33
1.3.1.3. Empirical Evidence for Labels 35
1.3.1.3.1. Generalized Pied-piping 35
1.3.1.3.2. Islandhood 37
1.3.1.3.3. Case 38
1.3.1.3.4. Phonological Phrasing 39
1.3.1.3.5. Incorporation 40
1.3.1.3.6. Argument Structure 42
1.3.1.3.7. Quantificational Dependencies 44
1.3.1.4. Conclusion: The Need for Labels 47
1.3.2. The Proposal: Labeling as a Bare Output Condition 47
1.3.2.1. First Merge, then Project 48
1.3.2.2. The Ontology of Labels 50
1.3.2.3. Extensions 57
1.4. Summary and Organization of the Dissertation 59

1
Chapter 2: The Essence of Focus in Basque: Syntactic, Prosodic and Semantic Properties 61
2.1. Syntactic Properties of Focus in Basque 62
2.1.1. The Basic Pattern 62
2.1.2. Focus in Embedded Contexts 68
2.2. Phonological Properties of Focus in Basque 71
2.2.1. Basque Dialectal Variability 72
2.2.1.1. The Stress-Accent System: Central Basque 73
2.2.1.1.1. Focal Accent and its Distribution 76
2.2.1.2. The Pitch-Accent System: Northern Bizkaian Basque 81
2.2.1.2.1. Focal Accent and its Distribution 83
2.2.2. Conclusions 87
2.3. Semantic Properties of Focus in Basque 87
2.3.1. Focus and Presupposition 88
2.3.1.1. From Presuppositions to P-sets 89
2.3.1.2. Focal Presuppositions Revisited 93
2.3.2. Focal Contrast: Information Focus vs. Identification Focus 97
2.4. Summary and a Question Concerning the Different Types of Foci 103

Chapter 3: A Derivational Approach to Focus Structure 105

3.1. The [+F] feature as an Optional Formal Feature 105
3.2. Deriving the F-Structure 107
3.3. F-Structure Licensing 112
3.4. Focus and Nuclear Stress 113
3.4.1. Nuclear Stress Placement in Basque 114
3.4.2. Focus and the Nuclear Stress Rule 116
3.4.3. A Revision of the Nuclear Stress Rule: The Focal NSR 124
3.4.4. ‘Schmerling Examples’ and Other Difficulties 132
3.5. Conclusions 135

2
Chapter 4: Focus at the Syntax-Semantics Interface 136
4.1. The Syntax of Focus in Basque: The Data and Discussion of Previous Literature 137
4.1.1. Introduction: The Cartographic and NSR-based Approaches to Focus 138
4.1.1.1. The Cartographic Approach to Focus in Basque 138
4.1.1.2. The NSR-based Approach to Focus in Basque 140
4.1.2. Evidence in Favor of the Cartographic Approach 145
4.1.2.1. Movement Patterns 146
4.1.2.2. Long-Distance Focus Movements 150
4.1.2.3. Pied-piping Phenomena 155
4.1.2.4. Islandhood 158
4.1.2.5. Dialectal Variation 160
4.1.2.6. Focus in Infinitival Constructions 162
4.1.2.7. Rightward Foci 166
4.1.2.8. Fragmental Answers 168
4.1.3. Summary and Conclusions 169
4.2. The logical form of Focus 170
4.2.1. A Neo-Davidsonian Semantics for Focus 170
4.2.2. Syntactic Complexity and logical form Complexity 175
4.2.2.1. Two Types of Expressions 175
4.2.2.2. Two Types of Quantifiers for Events 179
4.2.2.2.1. ∃ in out-of-the-blue utterances 180
4.2.2.2.2. δ in contextualized utterances 182
4.3. The road to logical form 185
4.3.1. The Introduction of the Event Quantifier 185
4.3.2. The Syntax-Semantics Interface 188
4.3.2.1. Out-of-the-blue Sentences 188
4.3.2.2. Contextualized Sentences 189
4.3.2.2.1. Dynamic Label Reprojections 190
4.3.2.2.2. A New Proposal for the Syntax-Semantics Interface
of Contextualized Utterances 192
4.4. Summary 196

3
Chapter 5: Interface Properties and theArchitecture of Grammar 197
5.1. Theories of F-Structure based on the Nuclear Stress Rule 197
5.2. NSR-based Theories and the Architecture of Grammar 202
5.2.1. Empirical Evidence Against NSR-based Approaches 202
5.2.1.1 Phonological Evidence 203
5.2.1.1.1 Categorially Different Nuclear Pitch Accents 203
5.2.1.1.2. Focus-induced Phonological Phrasing 205
5.2.1.1.3. Lack of Focal Accent 207
5.2.1.2. Syntactic Evidence Against NSR-based Theories of F-Structure 210
5.2.1.2.1. Failure to Meet the Most Embeddedness Condition 211
5.2.1.2.2 Lack of ‘Focus Projection’ 215
5.2.1.3. Conclusion: Unambiguous F-Structures in the Computation 217
5.2.2. Conceptual Arguments Against NSR-based Approaches 218
5.2.2.1. Architectural Issues 218
5.2.2.2. The Role and Nature of the Focal Feature in Building the 226
F-Structure
5.2.2.2.1. Theories that Do Not Postulate Any [+F] Feature 227
5.2.2.2.2. Theories that Assume One [+F] Feature 229
5.2.2.2.3. The Problematic Nature of the Focus set 232
5.3. Summary 233

Chapter 6: The Derivational Approach to Split Focus Constructions 235

6.1. Split F-Structures 235
6.2. An Earlier Analysis: Split F-Structures as Contrastive Topic + Focus 240
6.3. The Properties of Split Foci Revisited 244
6.3.1 Intonational Properties of Split Foci 244
6.3.2. Semantic Properties of Split Foci 248
6.3.3. Morphosyntactic Properties of Split Foci 252
6.3.3.1. ‘Emphatic’ Morphemes and Information Structure 255
6.4. Summary and Conclusions 260

Chapter 7: Summary and General Conclusions 261

References 268

4
CHAPTER 1. INTRODUCTION: SCOPE AND GOALS OF THE
DISSERTATION

1.1. Scope and Goals of the Dissertation

This dissertation studies the overall architecture of grammar and the type of
interactions between its different modules by providing an innovative and thorough
analysis of the nature of focus.
The grammar of focus has always been a controversial issue. In fact, focus has
usually been thought of as something which escapes core grammar, partially
pertaining to discourse or stylistic areas, even if it displays a well known series of
grammatical effects.
From a historical perspective, even if focalization rules are proposed from the
outset of generative grammar (cf. the TAc transformation of Chomsky (1955: 446-448)),
focus as such has not had a clear status within mainstream proposals on the design
and organization of grammar1. For instance, in Chomsky (1995b), when the general
minimalist version of the inverted-T model of the architecture of grammar is
described, focus is presented as having no specific place within it:

“Notice that I am sweeping under the rug questions of considerable

significance, notably, questions about what in the earlier Extended
Standard Theory (EST) framework were called “surface effects” on
the interpretation. These are manifold, involving topic-focus and
theme-rheme structures, figure ground properties, effects of
adjacency and linearity and many others”
[from Chomsky (1995b: 220)]

1
Remarkably, focus has been vindicated as having a definite place within the architecture of grammar
in the early seventies, when the issue was to argue for post-Deep Structure semantics and focus
seemed to provide good arguments for that, shifting focus-induced interpretation to the level of
Surface Structure (cf. i.a. Chomsky (1970)).

5
Furthermore, focus is also typically associated with a wide range of
phonological, semantic and morphosyntactic properties crosslinguistically. This
makes it difficult to explain the interface properties of focus. Works dealing with the
grammar of focus are often forced to modify substantially the overall architecture of
grammar in order to account for the phonological, morphosyntactic and semantic
effects of focus. This is the case, for instance, of the Nuclear Stress Rule (NSR)-based
trend of analysis stemming from Cinque’s (1993) seminal work on phrasal accent. In
order to explain the multiple grammatical effects of focus, this approach maintains
that focus is intimately related to nuclear stress and that it is the need to satisfy such
a necessary correlation between focus and nuclear stress that triggers a series of
syntactic and phonological operations. These operations are taken to have an effect
on the semantic interpretation of the sentence. This leads to a model of grammar with
global computations whereby any module can affect any other.
The complex and intricate nature of focus has had a direct influence in our
conception of how grammar is organized leading to some proposals that advocate
radical changes in the standard T-model assumed in generative studies. In this
dissertation I will reanalyze some of the most characteristic syntactic, morphological,
phonological and semantic properties of focus and provide an explanatory analysis
for them within the general “syntactocentric” conception of language that has
characterized Generative Grammar from its origins. In particular, the main goal of
this thesis is to provide a way to capture the interface effects of focus in a coherent
and principled way; that is, to find the locus of focus within the architecture of
grammar by analyzing the syntactic, semantic and phonological properties displayed
by focus and by discovering the principles underlying the relations between the
linguistic interfaces. The conclusion that emerges from this work is that a
‘syntactocentric’ architecture of grammar constitutes the best explanation for all the
properties displayed by focus.
With that purpose, I offer a detailed analysis of the grammar of focus in
Basque. Focus, or galdegaia (as it is called in Basque), is a much debated issue in
Basque linguistics since Basque is a language that marks focus with a specific
syntactic configuration besides nuclear stress and morphological marking in specific

6
environments (strong pronouns, do support, etc.) (cf. i.a., Azkue (1923), Altube (1929),
Mitxelena (1981), Ortiz de Urbina (1989) and Etxepare & Ortiz de Urbina (2003)).
Given that focus displays a wide variety of phonological, semantic and
morphosyntactic effects in Basque, this language is a particularly well suited
language for investigating the nature of focus and the interaction of the different
modules of the grammar.
Although the discussion is mainly based on the analysis of focus in Basque, I
also bring into the discussion a number of phenomena characteristic of focal
constructions which have been attested in other languages. These include
phenomena like (a) nuclear stress placement in languages such as English, Bengali,
Italian, Greek and European Portuguese, (b) phonological phrasing associated with
focus in languages like Japanese, Swedish, Korean, Greek and Chicheŵa, (c) the use
of syntactic dislocation in languages like Dutch, Hungarian, Russian and Zapotec or
(d) the specific morphological particles used in Korean, Tuki, Manding or Japanese.
This will help us have a more complete picture of the way in which the different
modules of the grammar interact, or do not interact, when generating the patterns
that are attested in focus constructions crosslinguistically.
The theoretical assumptions I adopt in this work (which I explain in section
1.2) are the following:
(i) A minimalist approach to syntax which adopts syntactic labels and
allows ‘reprojection’ mechanisms.
(ii) A Neo-Davidsonian semantics of focus whereby focus is mapped
into the nuclear scope of a quantifier over events.
(iii) An Autosegmental-Metrical phonology of intonation with
categorically different pitch accents in the intonational lexicon.

The specific theoretical conclusions that follow from this dissertation are the
following ones:

7
Syntax:
(i) The NSR-based approach to focus is unable to explain the word
order facts related to focus and raises important conceptual problems
regarding the interface interaction between different modules.
(ii) The cartographic approach can give a natural explanation to the
changes in word order associated with focus.
(iii) F-Structure is fixed in the narrow syntax via the Derivational Approach
to F-Structure, that coonstrues F-Structure through the elements
having a [+F] feature in the Numeration.
(iv) There is no need to appeal to focus sets nor focus projection (grammar
does not have focus sets nor focus projection mechanisms).
(v) Several isolated F-Structures might arise deriving in split-foci.

Semantics:
(i) The semantic effects of focus can be captured with a Neo-
Davidsonian approach.
(ii) The differences in presuppositionality and exhaustivity between
contextualized utterances and out-of-the-blue utterances are associated
to a different quantifier that binds the event variable; a δ quantifier
that creates definite descriptions of events in contextualized utterances

and a ∃ quantifier in out-of-the-blue utterances.

(iii) The semantic differences exhibited by both types of constructions go

hand in hand with the syntactic and phonological differences
observed in both types of constructions.

Phonology:
(i) Nuclear stress is just a cue of focus, and not its source. It is a way of
interpreting syntactically built F-Structures.
(ii) Not all languages need to mark focus with nuclear stress.

8
(iii) The association of focus with nuclear stress can be accounted for
with the Focal NSR, which assigns nuclear stress to the most deeply
embedded element within a F-Structure.

Finally, the main conceptual conclusions that derive from this dissertation
with regards to the interfaces are the following:
(i) Syntax-semantics and syntax-phonology interfaces are
unidirectional: Syntax → Phonology, Syntax → Semantics.
(ii) In opposition to recent proposals (cf. i.a. Reinhart (1995, 2006),
Zubizarreta (1998), Jackendoff (2002)), the analysis presented in this
thesis provides additional evidence in favor of the T-model.
(iii) A ‘syntactocentric’ model of the architecture of grammar can explain
all the PF and LF properties of focus.

But before we start analyzing the properties of focus, in section 1.2 I will
present the syntactic, semantic and phonological frameworks that constitute the basis
for this dissertation. Then in section 1.3 I address two questions that play an
important role in the discussion of focus in chapters 3 and 4: the nature of the
structure-building operation Merge, and the status of syntactic labels. Finally, in
section 1.4 I provide a summary of the general discussions as well as the main ideas
put forth in each chapter.

1.2. The Framework

In the following sections I will succinctly present the syntactic, semantic and
phonological framework where I will develop my dissertation: (a) Minimalist syntax
(section 1.2.1), (b) Neo-Davidsonian semantics (section 1.2.2) and (c) Autosegmental-
Metrical phonology (section 1.2.3).

9
1.2.1. Minimalist Syntax and the Architecture of Grammar

The syntactic analysis and discussions in this dissertation are framed within
the so-called Minimalist Program (cf. Chomsky (1993 et seq.). In this section I will
sketch the basic tenets of this program of research2.
Generative grammar takes language to be a property of the mind/brain of an
individual, an attained state of certain knowledge. As a mental object its design is
constrained in various ways. The minimalist program tries to provide an answer to
one crucial question, which I borrow from Chomsky (2005: 9-10): To what extent
does language approximate an optimal solution to conditions that it must satisfy to
be usable at all, given extralinguistic structural architecture?
The goal of the Minimalist Program is to go beyond explanatory adequacy and
discover the core nature of language by pursuing two strategies (i) minimizing the
assumptions and technical machinery employed in grammatical description (i.e.
searching for the most economical and elegant explanation), but also (ii) analyzing
how optimal the design of the faculty of language itself is, given the external
restrictions that it has to face (the so-called bare output conditions or legibility
conditions).
Regarding the language design question, Chomsky (2005: 6) argues that the
faculty of language is determined by three factors:

(1) Three factors in language design:

(i) Genetic endowment: apparently nearly uniform for the species, it interprets
part of the environment as linguistic experience, and it determines the
general course of the development of the language faculty.
(ii) Experience, which leads to variation, within a fairly narrow range, as in the
case of other subsystems of the human capacity and the organism generally.
(iii) Principles not specific to the faculty of language. Among them, we can
distinguish two types:

2 Good overviews of this framework can be found in Marantz (1995), Lasnik, Uriagereka & Boeckx
(2005), Hornstein, Nunes & Grohmann (2005), Hinzen (2006a), Boeckx (2006) as well as Uriagereka
(1998).

10
(a) Principles of data analysis that might be used in language
acquisition and other domains.
(b) Principles of structural architecture and developmental
constraints that enter into canalization, organic form, and
action over a wide range, including principles of efficient
computation, which would be expected to be of particular
significance for computational systems such as language.
[adapted from Chomsky (2005: 6)]

The idea behind Chomsky’s minimalist approach is that if a given trait or

principle of language is independently explainable, then we need not (and, hence,
should not) postulate it as part of the faculty of language (in the narrow sense; cf.
Hauser, Chomsky & Fitch (2002)). This is a fundamental explanatory move, for, if
some features participating in the shaping of the design of natural language are also
observed to be part of other ‘cognitive modules’ then we do not have to postulate
them as being intrinsic to language. This is of utmost importance when
characterizing the interfaces.
The Aristotelian conception of language portrayed it as a vehicle to relate
sound and meaning. This is a truism; the faculty of language interfaces with at least
these two systems that human beings have independently of language. These two
systems are called the Articulatory-Perceptual system (A-P) (sometimes also called
Articulatory-Motor), and the Conceptual-Intentional system (C-I) (the thought). So
our conception of the architecture of grammar should depict language as interfacing
with those two systems, at least. What else is needed? Chomsky (1995 et seq.) argues
that not much; just access to a lexicon, a repository where linguistic units (lexical
items) will be stored. Those three interfaces (A-P, C-I, and lexicon) are virtual
conceptual necessities without which no theory of language can expect any
explanatory success.
Within those limits, the language faculty could be thought of as a
computational system that takes lexical items and via an operation called ‘Merge’
generates expressions that are interpretable at both interface levels (cf. section 1.3

11
below for an in-depth analysis of Merge). Conditions imposed by the language-
external systems will dictate which types of expressions are ‘readable’ by them. In
this dissertation I will call ‘PF’ (for Phonological Form) the linguistic representation
that interfaces with the A-P system and logical form the representation that interfaces
with the C-I system. Since there is no interaction between the C-I and A-P systems,
the architecture of grammar has the shape of an inverted-T, where the syntactic
module generates pairs of expressions that will be interpreted in the independent A-
P and C-I systems3.

(2) Architecture of grammar (T-model):

Lexicon

Language

[Phonological Form] [logical form]

A-P System C-I System

The lack of a direct link relating the PF representation and the logical form
representation captures the fact that a semantic operation does not affect a
phonological operation (and vice-versa). This is a central idea at the heart of this
dissertation, and I will discuss it at length (specially in chapter 5).
A brief note is in order here. Recently, the existence of the traditional level of
Logical Form has been questioned with the argument that maybe there is not even a
level of representation that interfaces with the C-I system as such (cf. Chomsky (2005)
and specially Hinzen (2006a) for discussion). I am not convinced of the feasibility of
that approach4. At any rate, the notion of logical form that I am going to use here

3 The first formulation of such a model is due to Chomsky & Lasnik (1977). Here PF is the equivalent

to the ‘Universal Phonetics’ of Chomsky & Lasnik (1977).

4
Even if syntax were ‘crash-proof’ (for C-I), that would not entail that any linguistic expression is
interpretable by the C-I System. In section 1.3 I will suggest that precisely this is the case, and that

12
should not be identified with the level of ‘Logical Form’ put forth by May (1985) and
others within the Principles & Parameters theory. Logical Form was conceived as a
syntactic level of representation, obtained with syntactic operations and subject to
syntactic restrictions like the ECP and so on. The term ‘logical form’ I will use here is a
representational term, closer to the ‘logical syntax’ or ‘logical form’ used by some
philosophers of language (cf. Chomsky (1976) or Hornstein (1995, 2002) for
discussion and section 1.2.2 for my specific assumptions).
Remarkably, the minimalist line of research that has been pursued in the last
years has led to a much deeper understanding of the nature of human language (cf.
the references in footnote 2). The conceptual division that Martin & Uriagereka (2000)
make between a ‘methodological minimalism’ and an ‘ontological minimalism’ has
proven to be very fruitful, since research looking for elegant and optimal theories
(methodological minimalism) has shown that the design of language itself is in many
aspects optimal for the constraints that it has to face (ontological minimalism).
I would like to frame my dissertation within this view of grammar and the
language faculty in general, with the hope that it will shed more light on the
grammar of focus, and beyond it, on our understanding of the overall architecture of
grammar.

1.2.1.1. A Note on Phases

In the last decade, the model of the architecture of grammar has been
substantially modified passing from a global to a phase-based computation like the
one depicted in (3) (cf. Uriagereka (1999), Chomsky (2000 et seq.), Epstein & Seely
(2002) for three different conceptions of phases)5.

some expressions formed by syntax (at a given derivational stage) are not interpretable by the C-I
system.
5 I will use the cover term ‘phases’ for all those approaches that postulate that the computational

system spells out chunks of structure at different times (and not at a single point of Spell Out). In
order to refer to Chomsky’s (2000 et seq.) specific model I will use the term ‘phases’, with italics.

13
(3) Architecture of grammar (phase model)6:

Lexicon

Language

[Phonological Form] [logical form]

A-P System C-I System

Arguably, a system that works in phases is computationally more economical

than a global one given that it reduces memory load: all operations apply within
reduced cycles. In fact, the research undergone in this framework has been very
fruitful and illuminating in many aspects (notably within the domain of syntax cf. i.a.,
Gallego (2007), Fortuny (2007)). However, my concern is the following: what
happens with the spelled out chunks? That is, which chunks constitute the input to
the interfaces and how is a PF or LF parsetree obtained from them? Let me illustrate
my point. Assume we have the syntactic tree in (4), with some phases defined:

(4) α
3
β γ Phase 2
3
δ ε
3 Phase 1
ζ η

Assuming for simplicity that full phases are spelled out, how are the outputs
of Phase 1 and Phase 2 assembled in PF? Are there postsyntactic “Merge”
operations? What are the predictions of such a model for the interfaces?

6I represent the most widely held model of simultaneous phases; for an alternative proposal cf.
Marušič (2005).

14
I think there are two possible ways of approaching this question:
(i) Different conceptions of the phase and spell out make different
predictions about the nature of the representations at the
interfaces. If that were the case we should know what those
predictions are, and then look for the output of syntactic phases
at the interfaces.
(ii) Different conceptions of the phase and spell out do not make
different predictions about the nature of the representations at
the interfaces. If that were the case, then why bother?

Syntactic phase theorists have not paid much attention to these issues7. If
syntactic phases have determined interface correlates as Chomsky (2000 et seq.) has
pointed out, then we can use these properties for testing phasehood. In particular, it
has been pointed out that phonetic isolability is one of the correlates of phases. The
question is that whenever this line of research has been explored potential phases
appear everywhere (DPs, PPs, VPs, vPs, TPs, CPs…) and in fact, a quick look at the
crosslinguistic literature on phonological phrasing reveals all sorts of potential
phases. On the semantics side, Chomsky has suggested that phases are somehow
complete, for instance, they are assumed to have a full-fledged argument structure (for
a v*P) or a propositional interpretation (for a CP). This is hard to evaluate for it is
definitional of v*P and CP that they have a full argument structure and propositional
interpretation (depending on our conception of the completeness of argument
structure and propositionality, of course). And furthermore, there are many phrases
that are equally complete semantically (like a DP, a V, a NP, a VP…). Another test
that has been suggested is that phase edges provide the loci of reconstruction. But
again, whenever this line has been pursued it has led to the postulation of all sorts of

phases (cf. i.a., Legate (2003), Marušič (2005)).

7 There are, however, a number of works on the interfaces that assume in some way or another a
phase-based model; cf. i.a. Seidl (2001), Marvin (2001), Dobashi (2003), Ishihara (2003), Legate (2003),
Revithiadou & Spyropoulos (2003), Kahnemuyipour (2005), Fox & Pesetsky (2005), Marušič (2005),
Kratzer & Selkirk (2007). Unfortunately, each of these works has a different conception of what phases
are, so we cannot get a clear picture of what the predictions of spelling something out might be.

15
In the end, the picture of phase theory with respect to the interfaces is,
unfortunately, one of a lack of definition. So, my position here will be a cautious one
and I won’t make any appeal to phases in my analysis (cf. Boeckx & Grohmann
(2006) for a critical review). But I hope that the theory that I will present here will still
be compatible with phase-based systems, whenever we know more about what
exactly a phase is.

1.2.2. Event Semantics and logical form

Throughout the dissertation I will assume and develop a Neo-Davidsonian

event semantics8. This approach is characterized by the analysis of propositions as
providing descriptions of events and the analysis of semantic composition as
predicate conjunction with total separation of predicates and their arguments. In this
section I will briefly introduce the original Davidsonian approach (section 1.2.2.1)
and a proposal for a modification of that approach (section 1.2.2.2).

1.2.2.1 Quantification over Events

In this section I will briefly review the original Davidsonian proposal for the
semantics of “action sentences”. An “action sentence” like (5), where the subject is
related to the ‘agency’ of what is going on can be paraphrased as in (6):

(5) Brutus stabbed Caesar.

(6) There was a stabbing of Brutus by Caesar.

Davidson (1967a) starts from the following observation: intuitively, (6) says
that something happened and that it was a stabbing of Brutus by Caesar. Thus, Davidson

8 Good introductions to Davidson’s program in semantics for natural language can be found in

Hernández (1990) and Lepore & Ludwig (2005, 2007). For introductions to his semantics of events see
Larson & Segal (1995), Pianesi & Varzi (2000), Higginbotham (2000), Kearns (2000) and Pietroski
(2003b).

16
proposes that we should specify the meaning of (6) as involving a quantificational
construction along the lines in (7)9:

(7) There was an event such that it was a stabbing of Brutus by Caesar.

In order to capture that, Davidson proposes that we should change the

ordinary representations of predicates like stab from (8a), which is satisfied by an
ordered pair <x, y> iff x stabs y to the three-place predicate of (8b) that includes an
event variable, which, by stipulation will be satisfied by a triple <e, x, y> iff e is a
stabbing of y by x10:

(8a) Stab2(x, y)
(8b) Stab3(e, x, y)

Accordingly, the logical form representation of the sentence in (5) would be

along the lines in (9)11:

(9) ∃e[Stab3(e, Brutus, Caesar)]

This analysis, besides capturing the possibility of paraphrasing (5) as (6), has
interesting semantic consequences. To begin with, the explicit introduction of an
event allows us to refer to it anaphorically as in (10), where the pronoun ‘it’ refers to
the event of Jones buttering the toast:

(10) Jones buttered a piece of toast. He did it slowly, deliberately, in the bathroom,
with a knife, at midnight.

9 By the term ‘event’ I will refer here to the ordinary notion of (spatio-)temporally restricted entities
that happen and have causes and consequences. For discussion on the metaphysics of events see the
papers in Casati & Varzi (1996), Lombard (1998) and Pianesi & Varzi (2000).
10 Following customary notation, here I represent the adicity of a predicate with a subscript so that

‘Predicate1(x)’ is a monadic predicate, ‘Predicate2(x, y)’ is a dyadic predicate and so on.

11 Let me use simple representations to illustrate the basic idea. Truly, in a transitive sentence like (5)

we would have a complex event description with subevents in a mereological relation (cf. Parsons
(1990), Pietroski (2003a)).

17
Furthermore, this analysis provides a natural way to understand the import of
modifiers. Observe the sentences in (11)-(13):

(11) Brutus stabbed Caesar in the back

(12) Brutus stabbed Caesar with a knife
(13) Brutus stabbed Caesar on the Ides of March.

Intuitively, sentence (11) will be true iff there was a stabbing of Caesar by
Brutus and it was done in the back. Likewise, sentence (12) will be true iff there was a
stabbing of Caesar by Brutus and it was done with a knife, and sentence (13) will be
true iff there was a stabbing of Caesar by Brutus and it was done on the Ides of
March. These intuitions can be captured if we postulate that those modifiers are
predicates of the event that are introduced by conjunctions. In particular, sentence
(11) would have the logical form of (11’), where ‘In’ is a binary predicate satisfied by
an ordered pair <e, x> iff e occurred within the spatial location of x.

(11’) ∃e [Stab3(e, Brutus, Caesar) & In2(e, the back)]

Likewise, sentence (12) would have the logical form representation in (12’),
where ‘With’ is a binary predicate satisfied by an ordered pair <e, x> iff e was done
using x as an instrument:

(12’) ∃e [Stab3(e, Brutus, Caesar) & With2(e, a knife)]

Finally, the logical form of (13) would be along the lines of (13’), where ‘On’ is a
binary predicate satisfied by an ordered pair <e, x> iff e occurred within the temporal
interval x:

(13’) ∃e [Stab3(e, Brutus, Caesar) & On2(e, the Ides of March)]

18
The event analysis allows us to account for the modification that those
adverbial phrases bring about the event at the same time that it captures the analytic
relationship existing among any of the sentences in (11)-(13) and (5). That is, the fact
that English speakers know that if the content expressed by (11) is true, then (5) is
true as well, if (12), then (5) and if (13), then (5). The relationship between those
sentences is not mysterious since the adverbial phrases are separate predicates of the
event. Then, the analytic relation is but a case of entailment, for the logical form

representations of all of (11)-(13) entail the logical form of (5); i.e., ∃e [Stab(e, Brutus,

Caesar)].
Notice that we could, in principle, modify the lexical entry for ‘stab’ to make it
a four-place predicate with a modifier gap as in (14), which will be satisfied iff e is a
stabbing of y by x with a given instrument i (like, for instance, the knife in (12)):

(14) Stab4(e, x, y, i) iff e is a stabbing of y by x with i.

This type of analysis has evident problems of variable polyadicity, for the
number of modifiers that a predicate can take is potentially unbounded and we
should add a potentially unbounded number of meaning postulates to relate the
independent predicates Stab4 to Stab3 etc. (cf. Parsons (1990), Pietroski (2003a, 2005),
Borer (2005) and Irurtzun (2007) for discussion).
But moreover, with the analysis of quantification over events and conjoined
modifiers we can also account for cases of non-implications. Taylor (1985), extending
an idea suggested to him by Gareth Evans, proposes the following picture: imagine
that Shem hit Shaun twice at the same time, i.e., it is an ambidextrous battery; one
hitting is done violently with a shillelagh, and the other one is done mildly with a
cudgel. And these were the only hittings of Shaun by Shem. Then, at that single time
both (15a) and (15b) become true, whereas (16a) and (16b) are false:

(15a) Shem hit Shaun with a shillelagh violently.

(15b) Shem hit Shaun with a cudgel mildly.
(16a) Shem hit Shaun with a cudgel violently.

19
(16b) Shem hit Shaun with a shillelagh mildly.

We can say, following Davidson, that ‘Shem hit Shaun’ is true in case
∃e[Hit3(e, Shem, Shaun,)] and that the manner adverbs ‘violently’ and ‘mildly’ are
monadic predicates of the event. Then the logical forms of the sentences in (15) are the
following:

(15a’) ∃e[Hit3(e, Shem, Shaun) & With2(e, a shillelagh) & Violently1(e)]

(15b’) ∃e[Hit3(e, Shem, Shaun) & With2(e, a cudgel) & Mildly1(e)]

Having these existential quantifications over events, we can account for the
fact that in the situation above ‘Shem hit Shaun’ is true twice over, for each of the
sentences in (15) introduces its own description. Moreover, we can also explain why
both sentences in (16) are false since, even if ‘Shem hit Shaun’ is true, as entailed by
the logical forms of (15a’-b’), none of these representations entails the logical forms of
the sentences in (16):

(16a’) ∃e[Hit3(e, Shem, Shaun) & With2(e, a cudgel) & Violently1(e)]

(16b’) ∃e[Hit3(e, Shem, Shaun) & With2(e, a shillelagh) & Mildly1(e)]

As can be observed, appeal to covert quantification over events and a

conjunctivist composition for adjuncts provides a natural explanation for these
(non)entailments as conjunction reduction.
The event analysis also provides an attractive analysis of perceptual reports
like (17):

(17) Nora heard Fido bark.

In (17) the truth value is preserved if we change ‘Fido’ for a coreferential

expression (which is not evident for the sentential report “Nora heard that Fido

20
barked”). Within the event tradition, we can analyze sentence (17) as having the
logical form in (18), with ‘e’ and ‘f’ being different events12:

(18) ∃e∃f[Hear3(e, Nora, f) & Bark2(f, Fido)]

That is, that there is a hearing of Nora of a barking by Fido. As Vlach (1983)
argues, introducing the reported event makes natural an explanation of why
sentence (17) can be continued as in (19), with an anaphor coreferent with the event
of Fido barking (cf. as well Higginbotham (1983)).

(19) Nora heard Fido bark. It was loud.

Likewise, as Pietroski (1999) argues, the event analysis explains in a very

intuitive way the surface ambiguity of sentences like (20), for the adjunct might be a
predicate of any of the events, as represented in (21):

(20) Nora heard Fido bark in her apartment.

(21) ∃e∃f[Hear3(e, Nora, f) & Bark2(f, Fido) &In2(e/f, her apartment)]

And obviously, introducing event variables also permits analyzing temporal

relations between events. For instance, sentence (22) would have the logical form in
(23), where ‘Before(e, f)’ is a binary predicate satisfied by ordered pairs of events iff e
takes place before f (cf. Davidson (1985)):

(22) Eleanor of Aquitaine married Louis VII before she married Henry II.
(23) ∃e[Marry3(e, E-of-A, Louis VII) & ∃f[Before2(e, f) & Marry3(f, E-of-A, Henry II)]]

Talk of events allows us to explain all of these issues and a cornucopia of other
phenomena in a very natural way (cf. i.a., Parsons (1990), Larson & Segal (1995),

12For the time being I will abstract away from tense.

21
Pietroski (2005)). Still, in the next section I will introduce a modification to the
original approach of Davidson that postulates a subatomic semantics.

1.2.2.2. Neo-Davidsonian Separatism

A soon as D. Davidson presented his proposal for action sentences, H.N.
Castañeda pointed out that the analysis could be more fine-graded. In particular,
Castañeda (1967) observes that the arguments concerning the nature of entailments
applied as well to other constructions13. Consider the examples in (24)-(28):

(24) I flew my spaceship to the Morning Star.

(25) I flew to the Morning Star.
(26) My spaceship was flown to the Morning Star.
(27) I flew.
(28) My spaceship was flown.

Castañeda (1967) argues that example (24) entails all the examples in (25)-(28).
The entailments in (25) and (27) involve the dropping of the direct object “my
spaceship” whereas the entailments in (26) and (28) involve the dropping of the
subject “I”. Thus, he suggested, the subject and the object could be equally removed
from the argument structure for “fly” and be expressed as separate conjuncts.
Compare Davidson’s analysis for the sentence (24) in (29) with Castañeda’s analysis
in (30)14:

(29) ∃e[Flying3(e, I, my spaceship) & To2(e, the Morning Star)]

13 Although Davidson rejects this modification in the beginning (cf. Davidson’s reply to Castañeda in
Davidson (1967b)), he adopts it in later work (cf. Davidson (1985)).
14 I use the term ‘Theme’ instead of Castañeda’s ‘Patient’ given that ‘Theme’ has a broader tradition.

Sometimes the thematic predicates of Agent(e, x) and Theme(e, x) are substituted for syntactic
predicates providing representations like (30’) or (30’’):
(30’) ∃e[Flying(e), Subject(e, I) & Object(e, my spaceship) & To(e, the Morning Star)]
(30’’) ∃e[Flying(e), External(e, I) & Internal(e, my spaceship) & To(e, the Morning Star)]
For ease of exposition, I will use thematic representations in this dissertation, but see section 1.3,
footnote 31 for details concerning these questions.

22
(30) ∃e[Flying1(e) & Agent2(e, I) & Theme2(e, my spaceship)

& To2(e, the Morning Star)]

So, according to Castañeda the verb won’t be a ternary predicate but a unary
predicate of the event (Flying(e)). Like adjuncts, arguments will be introduced via
predicate conjunction. However, there will be an essential difference between
arguments and adjuncts. As we saw, some adjuncts like ‘violently’ and ‘mildly’ can
be thought of as direct monadic predicates of the event (violently1(e), mildly1(e)).
Furthermore this analysis can be extended to all types of adjuncts, after all, the type
of modification that PP adjuncts like ‘to the Morning Star’ bring about is the same as
the modification by an adverb like ‘there’. So, the representation in (30) can be
reformulated as in (31), where the adjunct ‘to the Morning Star’ is a monadic
predicate of the event:

(31) ∃e[Flying1(e) & Agent2(e, I) & Theme2(e, my spaceship)

& To-the-Morning-Star1(e)]

The representation in (31) roughly says that there was an event of flying,
whose agent was I, its theme was my spaceship, and it (the flying) was to the
Morning Star. As can be easily verified, argument DPs cannot be direct predicates of
the event as are the adjuncts, for we need some sort of linkage between the argument
DPs and the event. Thus, the formula in (32) is incorrect:

(32) ∃e[Flying1(e) & I1(e) & My-spaceship1(e) & To-the-Morning-Star1(e)]

In consequence, adjuncts are direct predicates of the event whereas arguments

are introduced with “thematic” binary predicates that relate the arguments with the
event.

23
This conjunction analysis of arguments is what Schein (1993) calls essential
separation, given that in this approach all arguments are “severed out” from the
verb’s semantics.
Thus, the Neo-Davidsonian approach has appeared to be very fruitful in
allowing us to account for a variety of phenomena that range from entailments and
non-entailments to mereological relations and event complexities (cf. i.a., Parsons
(1990), Schein (1993), Pietroski (2002, 2003a, 2005)).
I will leave Neo-Davidsonian semantics here and resume in section 1.3 to an
analysis of predicate-argument relations. Then, in chapters 4 and 6, and following the
theory put forth in Herburger (2000), I will develop a Neo-Davidsonian semantics for
focus.
In the next section I will present the Autosegmental-Metrical theory of
intonational phonology.

1.2.3. Autosegmental-Metrical Phonology

In this section I will briefly review the framework of Autosegmental-Metrical
Phonology that stems from Liberman (1975), and Bruce’s (1977) work and gets a full
development in Pierrehumbert (1980)15. This is the framework on which I will base
my discussion of the intonation of focus. Although the main objective of this thesis is
not to develop an analysis of the intonation of focus, throughout the dissertation I
will discuss the most relevant aspects of the phonology of focus intonation in various
languages. My discussions will be based on the Autosegmental-Metrical descriptions
of various languages.
As the name of the framework shows, Autosegmental-Metrical Phonology is
based on the combination of two theories of prosodic phonology; Autosegmental
Phonology and Metrical Phonology. Thus, the framework of Autosegmental-Metrical
Phonology is an eminently phonological theory of intonation (as opposed to a
phonetic one). From Metrical Phonology it takes the idea that phonological units
(syllables, words, phrases…) are grouped in units of relative prominence, and from

15The name of “Autosegmental-Metrical Phonology” was introduced by Ladd (1996). Comprehensive

introductions can be found in Ladd (1996), Beckman (1996), Shattuck-Hufnagel & Turk (1996), Hualde
(2003), Prieto (2003b) and Gussenhoven (2004), cf. as well Pierrehumbert, Beckman & Ladd (1997).

24
Autosegmental Phonology it takes the idea that there are separate tiers of
representation for segmental units and tonal units. So, for instance, the prosodic
representation of a word like the pentasyllabic word arrigarriya ‘surprising’ in
Errenteria Basque will be as depicted in (33) (cf. section 2.2.1.1 for further details).
This representation has separate tiers for the pitch accent (abstractly marked as T*)
and the syllables:

(33) ω Word

σ σ σ σ σ Syllable
.........................................................................................
T* Tone-level

..........................................................................................
a rrí ga rri ya Phonetic level

The main characteristics of Autosegmental–Metrical Phonology are (i) that it is

a strongly sequential model, generating intonational contours just as a result of the
concatenation of phonological units (the tonal events) and (ii) it is a radical level-
based theory in that it only observes two basic levels: a high tone (H) and a low tone
(L)16.
Further, there are a number of hierarchically organized prosodic constituents.
Pierrehumbert & Beckman (1988), for instance, propose the following schema17:

(34)
υ Utterance
ι Intermediate Phrase
α Accentual Phrase
ω Word
σ Syllable
μ Mora

16 Previous formal analyses of intonation assumed more levels (like for instance, mid tones (M)). Bruce
(1977) and Pierrehumbert (1980) showed the inadequacy of that analysis both on empirical and
conceptual grounds.
17 Similar proposals can be found in Selkirk (1980, 2000), Booij (1983), Nespor and Vogel (1986) or

Hayes (1989), among others.

25
However contrary to more global models of intonational phonology (cf. i.a.,
Fujisaki (1983)), in the Autosegmental-Metrical phonology the intonational contour
will not be generated from a hierarchically structured representation of tunes.
Rather, F0 will be derived as the implementation of the phonetic realization of tonal
targets and the interpolation between the targets.
Works like Pierrehumbert (1980) and Ladd (1983) propose two types of tonal
interpolations. (i) A direct “monotonic” interpolation between tones of different
levels or two L tones, and (ii) a “sagging” interpolation between two H tones. This is
represented in (35)18:

(35)
Monotonic Interpolations: Sagging Interpolations:
H H H H

L L L L

The theory further describes three types of tonal events: pitch accents,
boundary tones (aligned to the left or right edge of an intonational unit) and phrase
accents that might be associated to boundary tones. Pitch accents are
autosegmentally aligned with metrically accented syllables and might be simple (T*)
or bitonal (T*+T or T+T*). In case of bitonal pitch accents, one of the tones is taken to
be strictly aligned with the accented syllable whereas the other tone is either a
leading tone preceding it or a trailing tone following it. These differences are
represented notationally with the star association, which identifies the tone that is
strictly aligned with the accented syllable. Thus, for instance a L+H* pitch accent is a
complex accent composed by a leading L tone and a H tone strictly aligned with the
accented syllable, whereas a H*+L pitch accent has a H tone strictly aligned with the
accented syllable followed by a trailing L tone19. Similarly, a L*+H accent is an accent

18 See however Gussenhoven (1984) for a monotonic treatment of the interpolation of two H tones (cf.
as well Ladd & Schepman (2003) for a revision of raising tones in English).
19 Recently, some researchers have questioned the star association, i.e. the idea that just one of the

tones composing a complex tone is strictly aligned with the accented syllable (cf. Atterer & Ladd

26
composed by a L tone strictly aligned with the accented syllable and a trailing H
tone, and a H+L* is a pitch accent composed by a leading H tone and a L tone
aligned with the accented syllable. Compare the abstract representations in (36), for a
postinitial pitch accent:

(36)
(a) L+H* (b) H* (c) H*+L

σ σ σ σ σ σ σ σ σ

(d) L*+H (e) H+L*

σ σ σ σ σ σ

Among the pitch accents in a sentence there is one, called ‘nuclear accent’ (or
‘nuclear stress’) that is perceptually more prominent that the rest of the accents and
structurally it might be different from them in some languages. For instance,
prenuclear accents in a language might be L*+H whereas the nuclear accent is L+H*.
A focalized constituent is usually assigned a nuclear accent and accompanied by a
compressed pitch range in the following stretch that normally does not contain any
other pitch accent.
These are the components of a grammar in the Autosegmental-Metrical
theory. A grammar of a given language, then, will have to specify the types of tonal
events that it contains (i.e., the intonational lexicon), and the rules of association with
prosodic units. For instance, Pierrehumbert (1980: 29) formulates the following finite
state grammar for English intonation:

(2004)). See as well Arvaniti et al. (2000), G. Elordieta & Calleja (2005), Prieto et al. (2005) and Face &
Prieto (2007) for discussion on the star notation.

27
(37) Finite-State Grammar of English Intonation20:

Boundary Tone Pitch Accent Phrase Accent Boundary Tone

H*

H% L* H- H%

L*+H
O O O O O

L+H*

L% H*+L L- L%

H+L*

H*+H

Thus the Autosegmental-Metrical theory proposes that a F0 contour will be

the phonetic implementation of the realization of the tonal targets autosegmentally
aligned with the segmental tier.
In the next section I address the properties of the operation Merge and its
consequences.

1.3. The Nature of Merge and the Status of Syntactic Labels

In this section I present the analysis of phrase structure composition that I will
adopt in this dissertation. The theory of F-Structure that I will develop in chapter 3
will be a ‘syntactocentric’ one, whereby F-Structure is composed within the narrow

I have updated the notation of leading and trailing tones to current customs deleting the hyphen, so
20

what Pierrehumbert (1980) encoded as T*+T- and T-+T* appear here as T*+T and T+T*, respectively.

28
syntax and can be interpreted by the interfaces. In order to develop that theory, it
will be essential to explore the mechanism of syntactic composition found in natural
languages.
In section 1.3.1 I develop an analysis of the mechanism of syntactic
composition based in Chomsky’s (1995a) Bare Phrase Structure. I provide a range of
empirical evidence that shows that syntactic phrases have labels. Then, in section
1.3.2, I propose the hypothesis that the presence of labels is required by interface
legibility conditions, and I argue for a set-theoretic implementation of the labeling
mechanism. As I will argue, labels provide the step from a primitive syntax of
adjunction to a more complex syntax of predicate-argument relations at the same
time that they bring a fundamental economy property of human language: recursion.
This Bare Phrase Structure analysis will play an important role for the syntax-
semantics interface of focus I will defend in chapter 4.

1.3.1. Phrase Structure Composition: Merger and Projection

In this section I analyze the syntactic structure-building operation Merge as a
two step process. Based on the Bare Phrase Structure theory (cf. Chomsky (1995a)), I
propose that Merge, as such, is a simple operation that creates unlabeled sets, but I
also defend that in certain domains (notably, predicate-argument relations), a label is
required by the interfaces for legibility. Under this view, labeling should be regarded
as a two step operation: the first one creates a set and the second one provides a label
to it. Thus, I argue, it is only in this way that we jump from a simple, conjunctive,
and label-less adjunct-like syntax, to a labeled higher order predicate-argument
syntax (cf. Hinzen (2006a-b), Hornstein, Nunes & Pietroski (2006) and section 1.3.2).
But before I introduce my analysis of phrase structure building, I will first
present the theoretical background where my proposal is framed. Section 1.3.1.1 is a
brief introduction to the structure building operation Merge as presented in
Chomsky (1995a). In section 1.3.1.2 I review a recent approach to Merge that
proposes getting rid of labels from syntactic theory (cf. Collins (2002), Seely (2006)).
Next, in section 1.3.1.3, I will discuss a variety of phonological, syntactic and
semantic phenomena that show the need for postulating labels. Finally, in section

29
1.3.2 I will propose to reanalyze the label-interpretation mechanism of Bare Phrase
Structure as deriving from a legibility condition that requires sets with coherent
intensions. I also discuss some of the implications of such a move (notably, bringing
about recursion and the derivation of predicate-argument relations as an emergent
property due to a type shifting operation).

1.3.1.1. Introduction: Structure Building in Bare Phrase Structure

Chomsky (1995a) proposes a strongly principled phrase structure theory.
Regarding human language as a generative system that constructs pairs of symbols
(π, λ) that are interpreted at the Articulatory-Perceptual (A-P) and Conceptual-
Intentional (C-I) interfaces, respectively, his proposal is that a simple operation called
Merge is sufficient to generate interpretable phrase markers. More specifically, his
proposal is the following:

(38)
“Applied to two objects α and β, Merge forms the new object γ. What is
γ? γ must be constituted somehow from the two items α and β; the only
alternatives are that γ is fixed for all α, β or that it is randomly selected,
neither worth considering. The simplest object constructed from α and β
is the set {α, β}, so we take γ to be at least this set, where α and β are the
constituents of γ. Does this suffice? Output conditions dictate otherwise;
thus verbal and nominal elements are interpreted differently at LF and
behave differently in the phonological component […]. γ must therefore
at least (and we assume at most) be of the form {δ, {α, β}}, where δ
identifies the relevant properties of γ, call δ the label of γ.”
[Chomsky (1995a: 396)]

Thus, the merger of two elements creates a labeled set that contains both
elements. However, now a question arises: what constitutes the label of the set?
Chomsky (1995a) assumes that, given minimalist principles, the label should be

30
defined on the basis of the elements already present in the derivation; this would be a
parsimonious way of observing the so-called Inclusiveness Condition:

(39)
“A ‘perfect language’ should meet the condition of inclusiveness: any
structure formed by the computation (in particular, π and λ) is
constituted of elements already present in the lexical items selected for
N [the Numeration]; no new objects are added in the course of
computation apart from rearrangements of lexical properties…”.
[Chomsky (1995b: 228)]

Strictly adhering to these set-theoretic realms, Chomsky (1995a) considers

three potential candidates for the creation of a label when merging two objects α and
β. If the label has to be derived from the elements being merged (say, α and β), it
should be one of the three possibilities presented in (40a-c):

(40a) The intersection of α and β.

(40b) The union of α and β.
(40c) One or the other of α, β.

According to Chomsky (1995a), the options in (40a) and (40b) are to be

excluded, since they would create anomalous objects. On the one hand, an algorithm
along the lines in (40a) creates uninterpretable null outputs in cases in which the
intersection of α and β is empty. On the other hand, (40b) generates self-
contradictory outputs when α and β have different values in some feature (the
normal case in human language syntax). Thus, Chomsky argues, the label δ of the
newly created set {δ, {α, β}} should be as in (40c); either one or the other of α, β. This
set-theoretic implementation of structure-building captures in an elegant way the X’-
theoretic idea of the projection of the head: the output of merging β (a head) and
α (its complement), is a βPhrase; a projection of β (cf. (41a-b)). A category that doesn’t
project anymore is then called a Maximal Projection (notated informally as βP in

31
(41b)), and a terminal category (a category that doesn’t dominate any other category)
is called a Minimal Projection (αP in (41b)).

(41a) Bare Phrase Structure (41b) X’ Theory

{β, {α, β}} βP

3 3
α β αP β

Likewise, if the head is still active, the output of merging the complex {β, {α,
β}}/βP with another Maximal Projection (i.e., a specifier like δ/δP in (42a-b)) will
also be a projection of β; that is, a βPhrase. Now, the former Maximal Projection of β
in (41) loses that status, and the new element (42) becomes the Maximal Projection of
β (cf. Epstein et al. (1998) for extensive discussion):

(42a) Bare Phrase Structure (42b) X’ Theory

{β, {δ, {β, {α, β}}}} βP

3 3
δ {β, {α, β}} δP β’
3 3
α β αP β

Thus, the goal of the proposal in Chomsky (1995a) is to capture in a minimalist

and derivational way the structure of complex phrases using only the items being
merged and a simple structure building operation that creates sets (i.e., Merge).
As mentioned above, the justification for the postulation of labeled structures
relies on phenomena observed at the interfaces (see Chomsky’s (1995a) quotation in
(38)). And, although Chomsky (1995a) isn’t very specific about these phenomena, the
evidence at both interfaces is, in my view, pervasive enough to justify the need for
postulating labels in the computational system. But before I enter into that
discussion, I would like to review a recent alternative proposal in which it is argued
that we should get rid of labels. First I will succinctly present the label-free proposal
(section 1.3.1.2). Then in section 1.3.1.3 I will provide ample empirical evidence that
shows that the label-less analysis is misguided.

32
1.3.1.2. An Attempt to Eliminate Labels: The Locus Theory

In recent years some scholars have proposed a reanalysis of the operation

Merge, trying to show that labeled representations are not necessary and arguing for
a simpler structure-building mechanism (cf. i.a., Collins (2002) and Seely (2006)).
Basically, their proposal is that the output of merging two elements α, β is a bare
two-membered set {α, β}, instead of the more complex labeled structure {δ, {α,
β}} presented in section 1.3.1.1. In this section I will briefly review the arguments of
this trend of analysis and then, in section 1.3.1.3, I will resume to a discussion
arguing that, pace Collins (2002) and Seely (2006), labels are indeed necessary in order
to capture a number of interface phenomena.
The idea of a label-less syntax has been originally put forth by Collins (2002).
This author proposes a reanalysis of the operation Merge in Chomsky (1995a). His
proposal is that we can get rid of labeled projections and take Merge(α, β) as creating
an unlabeled phrase/set like {α, β}. Thus, according to him the simplest (hence,
correct) representation of the structure in (42) above will not be (42a), but (43):

(43)
{δ, {α, β}}
3
δ {α, β}
3
α β

Collins (2002) proposes that once we have (43), all we need in order to
continue with the derivation is what he calls the Locus Principle (cf. as well Chomsky
(2006)):

33
(44) Locus Principle: Let X be a lexical item that has one or more probe/selectors.
Suppose X is chosen from the lexical array and introduced into the derivation.
Then the probe/selectors of X must be satisfied before any new unsaturated
lexical items are chosen from the lexical array. Let us call X the locus of the
derivation.
[from Collins (2002: 46)]

Given the locus principle, we can consider a lexical item saturated whenever it
does not have any probe/selector. And, inversely, a lexical item that has a
probe/selector will be unsaturated.
Therefore, provided the locus principle, Collins (2002) argues that there is no
need for labels. According to this label-free system, there will be no need to project
labels given that the unsaturation of a lexical item will be enough to account for the
subcategorization and selectional properties of syntactic phrases, that is, the locus
principle suffices to keep the derivation active.
As an illustration, Collins (2002) discusses the impossibility of having a
convergent derivation like (45) from the Numeration in (46), even if the principle of
‘Merge over Move’ would require the insertion of Compº (cf. Chomsky 2000: 138 for
discussion):

(45) *That will John arrive.

(46) NUM: {that, will, John, arrive}

Within the label-free syntax, the deviance of (45) is explained as follows:

(47) (a) Choose ‘John’ from the lexical array.

(b) Choose ‘arrive’ from the lexical array.
(c) Merge (John, arrive) = {John, arrive} satisfying the theta-role of ‘arrive’.
(d) Choose ‘will’ from the lexical array.
(e) Merge (will, {John, arrive}) = {will, {John, arrive}} satisfying the
subcategorization feature of Infl ‘will’.
(f) Choose ‘that’ from the lexical array.

34
(g) Merge (that, {will, {John, arrive}}) = {that, {will, {John, arrive}}} satisfying
the subcategorization feature of Comp ‘that’.

Here, Collins (2002) argues, the problem arises in step (f), because at step (e)
‘will’ is still unsaturated (it has an active EPP feature). Hence, with the introduction
of the complementizer ‘that’ in step (f), the derivation incurs in a violation of the
locus principle.
Recently, Seely (2006) has implemented Collins’ (2002) proposal of a label-less
syntax concentrating on the operation Merge and reinforcing the argument that if
bare merge (just set-creation as in (43)) is sufficient to provide all the properties that
the complex operation of the ‘Labeling Merge’ of Chomsky (1995a) provided
(notably, derivational c-command), then we should assume the simplest option.
I agree with that minimalist motto, but as I will discuss in the next section,
there is ample evidence that shows that labels are present at the interfaces (cf. as well
Boeckx (2007)).

1.3.1.3. Empirical Evidence for Labels

In this section I will provide empirical evidence for the presence of labels at
the interfaces. It will be shown that there is a rich variety of phenomena that seem to
make reference to labeled projections. Hence, this section should be regarded as
providing support against label-free theories (I will discuss this issue in section
1.3.1.4). These phenomena will also provide the ground for the labeling proposal I
will make in section 1.3.2. As I said before, this analysis of labels will be important
for the syntax-semantics interface of focus that I will develop in chapter 4 of this
dissertation, for it is based on label “reprojections”.

1.3.1.3.1. Generalized Pied-piping

One of the core advancements of generative syntax has been to envisage and
provide a logic to the property of displacement, a phenomenon characterized by the
fact that a syntactic phrase has the properties of two different syntactic positions: the
position it occupies in the surface and another one. Putting it in derivational terms,
basically, displacement constructions are those where an expression E that at some

35
derivational stage D was in a configuration C appears in a different configuration C’
at a different derivational stage D’. For instance, in the tree diagrams of (48)-(49) the
XPi that at the derivational stage D was in sisterhood with Wº, at the derivational
stage D’ acquires as well the properties of being associated in a sisterhood relation
with S’ —whichever properties these are (scopal, linear or whatever):

(48) Stage D (49) Stage D’

SP
qp
SP XP S’
3 3 3
Sº WP YP X’ Sº WP
3 3 3
UP W’ Xº ZP UP W’
i
3 3
Wº XP Wº XPi
3
YP X’
3
Xº ZP
i

There are many technical ways to implement displacement, among others:

specific transformations, movements motu propio, internal merge, the outcome of
attraction or agree relations with OCC, EPP features, etc. However, no matter how
we implement the displacement operation, the generalized pied-piping property
observed in displacements strongly suggests that there are labeled structures in
syntax. What moves constitutes an XP, a maximal projection of a certain head21.
Even if the nature of generalized pied-piping is still poorly understood,
having labeled structures gives some logic to it. Assuming that displacement takes
place to check a feature, in displacement configurations more material than the
(seemingly) necessary is moved. For instance, in Wh-questions it is the whole XP
containing the Wh-phrase instead of just the lexical item containing the [+Wh] feature
that moves. Thus, in (50) the [Wh] feature sitting in the Wh-word ‘which’ triggers the
pied-piping of the whole XP (‘DP’) where it is contained:

21 I abstract away from head movement here.

36
(50) [Which book on linguistics]Obj. did you burn t?

Leaving aside what forces the whole XP to move, the fact that XPs/DPs move
suggests that labels (and, therefore, maximal projections) are visible during the
derivation and at the interfaces.
Under a label-free approach to syntax it is difficult to account for the fact that
it is a whole XP that moves.

1.3.1.3.2. Islandhood
This point is related to the previous one. Certain syntactic configurations are
islands for movement. One of the best known island constructions involves
coordination structures like those in (52) and (53) (where the unpronounced copies
are shaded):

(51) They are going to eat [calamari and cod].

(52) *What are they going to eat [w
whhaatt and cod]?
(53) *What are they going to eat [calamari and w
whhaatt]?

Although it is not clear why Coordination Phrases are islands for extraction
this shows that some syntactic domains which, apparently, can only be defined in
terms of labels (like ‘CoordP’) are closed domains for extraction. Without labels it is
hard to imagine even a description of the data.22
Finally, notice that remarkably, a way of solving the need to move an element
from within an island is to pied-pipe the whole island (cf. as well section 4.1.2.3). This
brings us back to the problem mentioned in the previous section: if we deny labeling
and projection, then there is no way to specify which phrase moves since there is no
‘phrase’ to move.

22Authors like Uriagereka (1999) and Hornstein, Lasnik & Uriagereka (2007) have proposed to derive
some islands from the dynamics of the derivation. I am completely sympathetic to these approaches.
However, the Coordinate Structure constraint represented above doesn’t seem to be explainable in
those terms.

37
1.3.1.3.3. Case

One of the most widely held assumptions in generative grammar is that whole
DPs (and not Dºs, nor D’s) are assigned case. In some languages case surfaces
morphologically at the edge of the whole DP. This is illustrated in Basque, a
language with SOV word order. In Basque, the surface word order within the DP is
numeral-noun-adjective-article and case marks the whole DP and appears attached
to the rightmost element, as illustrated by the ergative in (54):

(54) [Bi txapel beltz-e]-k

two hat black-ART-ERG
“The two black hats”

Hungarian, on the other hand is a SVO language, and the word order within
the DP is determiner-numeral-adjective-noun. However, as in the case of Basque,
case is assigned to the whole DP and surfaces morphologically at the right-edge of
the DP. Observe the example in (55), from Szabolcsi (1994), with accusative case:

(55) [A ket fekete kalap]-ot

the two black hat –ACC
“The two black hats”

What follows from this quick overview of Basque and Hungarian is that
syntactically DPs get case, and that the case-marking is overly realized at the edge of
the DP, no matter whether the case marker happens to be left-adjacent to Dº (like
Basque) or whether Dº and the case marker appear on opposite edges (like in
Hungarian). Unless we have labels like ‘DP’s we cannot explain which constituents
get case and where it appears morphologically. We need to appeal to labelled
constituents to define the edges.

38
1.3.1.3.4. Phonological Phrasing
The common assumption since Chomsky, Halle & Lukoff’s (1956) seminal
work on the relationship between syntax and phonology is that prosodic units above
the word level are constructed upon syntactic constituents, and that some purely
phonological processes might also take place to reshape the phrases phonologically.
In Selkirk’s (1986) terms we would have, then, a ‘phonosyntactic’ subcomponent, and
a ‘phonological’ one derived from it (cf. as well section 5.2.2.1 for discussion). In this
vein, the most widely held analysis of prosodic phrasing is the Optimality Theoretic
one put forth in works like Selkirk (1995b) and Truckenbrodt (1995, 1999) (cf. i.a.
Gussenhoven (2004), G. Elordieta, Frota & Vigário (2005), G. Elordieta (2007b)). The
crucial point here is that this process that creates prosodic constituents (just like
previous proposals based on a serial rule application) necessarily makes reference to
labeled structures. As an illustration, observe in (56) the structure of accentual
phrases in Lekeitio Basque. As I will explain in section 2.2.1.2, in this variety, and
focus matters aside, syntactic DPs composed exclusively by lexically unaccented
words are mapped into accentual phrases characterized by an initial %L tone aligned
with the first syllable, and a H- phrasal tone aligned with the second syllable that
spreads rightwards until a H*+L pitch accent is found aligned with the final syllable.
Thus, the simple accentual phrase (α1) ‘the flutist’ and the more complex one
containing a locative genitive (α2) have the very same phonological structure:

(56) Accentual Phrase in Lekeitio Basque (cf. G. Elordieta (1998))

α1 α2

%L H- H*+L %L H- H*+L
(spreading) (spreading)

σ1 σ2 σ3 σ4 σ5 σ 1 σ 2 σ3 σ 4 σ5 σ6 σ7 σ8 σ9
txis. tu. la. ri. dxá. mar. ki. ña. ko. txis. tu. la. ri. dxá.
(the flutist) (the flutist from Markina)

Thus, no matter what its internal constituency is, it is the whole DP that is
taken into account for phonological phrasing. That is why DPs with different internal

39
structure display the same phonological pattern by being mapped into a
phonological phrase23. Specifically, Selkirk (1995b) proposes to implement the end-
based approach to phonological phrasing developed in Selkirk (1984, 1986) within
the Generalized Alignment system of McCarthy & Prince (1993). She proposes two
alignment constraints to account for prosodic phrasing facts that will align syntactic
XPs (not X’s nor Xºs) with the edges of phonological phrases:

(57a) ALIGN-XP,R: ALIGN(XP, R; P, R)

For each XP there is a P such that the right edge of XP coincides with the right edge
of P.

(57b) ALIGN -XP,L: ALIGN (XP, L; P, L)

For each XP there is a P such that the left edge of XP coincides with the left edge of P.

Based on this account, Truckenbrodt (1995) has proposed a further markedness

constraint requiring prosodically ‘wrapped’ XPs (i.e., maximal projections):

(58) WRAP-XP
Each XP is contained in a phonological phrase.

The conclusion to be derived from this is clear: syntactic phrases have to be

labeled; otherwise the phonological component cannot use this information for
phonological phrasing. In order to capture these facts, having a bare {α, β} set won’t
suffice, for we would lose the categorial information of the phrase and the PF
component wouldn’t know where the XP begins and where it ends.

1.3.1.3.5. Incorporation

The gist of this section is that in order to explain syntactic incorporation

processes like the ones observed in languages like Onondaga (an Iroquoian

23As I said before, I am simplifying here since there are also specific complexity constraints on the
maximum number of syllables, prosodic words or syntactic branches that a prosodic phrase might
contain (cf. G. Elordieta, Frota & Vigário (2005)).

40
language) we also need some sort of labels. Consider (59a-b), both grammatical
sentences in Onondaga (cf. Baker (1988) for an extensive analysis of these
constructions):

(59a) Pet wa?-ha-htu?-t-a? ne o-hwist-a ?.

Pet PAST-3MS/3N-lost-CAUS-ASP the PRE-money-SUF.
“Pet lost money”

(59b) Pet wa?-ha-hwist-ahtu-?ta-?.

Pet PAST-3MS-money-lost-CAUS-ASP
“Pet lost money”

According to Baker (1988: 77), both sentences are exact thematic paraphrases
of one another (the same verb with the same selectional requirements relates the
same arguments with the same thematic-roles). The difference between (59a) and
(59b) is that in the former each of the event participants constitutes its own
phonological word (the subject ‘Pet’ and the object ‘ne ohwista’), whereas in the latter
the subject still constitutes a word on its own but the direct object appears
sandwiched between verbal affixes, and constituing a morphological word with the
verb (that is, ‘wa?hahwistahtu?ta?’). Baker (1988) proposes that these facts can be
explained if we assume that the verbal complement head has been incorporated into
the verb, thus subsuming itself within a verbal projection. Thus, he argues that the
structure for the sentence in (59b) would be along the lines in (60):

(60)
S
qp
NP VP
g 3
Pet V NP
3
N V
g g
moneyi -lose ti

41
The only point I want to make here is that after incorporation the newly
created complex element is unambiguously interpreted as verbal, and not as nominal
(hence, for instance, the verbal prefixes of past tense and agreement as well as the
suffixes of causation and aspect appear left-attached to it). Remarkably, these are the
very same ones as those appearing in the non incorporated case in (59a). This fact can
easily be explained if we assume that ‘hwist’ (money) is within a verbal constituent
(the complex verb in (60)) and that the affixes attach to that verbal phrase (i.e., to
‘money-lose’). However, in order to have this interpretation, we need some sort of
projection of the verb; clearly, a bare {N, V} set without a label won’t suffice given
that in that case the verbal prefixes would be attaching to a noun24.

1.3.1.3.6. Argument Structure

One more piece of evidence in favor of a label-based phrase structure comes
from configurations of argument structure and theta-role assignment.
From the mid eighties, with the development of what has been called the VP
Internal Subject Hypothesis (VISH) it is a common assumption that even subjects are
base-generated within the verbal projection. There are many reasons for that change
of point of view; among others, some properties of ‘wanna-contraction’, floating
quantifiers, and clausal idioms (cf. i.a. Contreras (1987), Zagona (1987), Sportiche
(1988), Kuroda (1988), Koopman & Sportiche (1991)). But the most important fact for
my concerns here is that introducing the subject within the VP/vP provides a local
configuration for all the predicate-argument relations when they first merge in D-
Structure25. Nowadays, it is a widely held assumption that the structure of a
transitive predicate is as illustrated in (61):

24 Moreover, if we take the process of incorporation to be a narrow-syntactic one (as Baker (1988)
does). With this assumption, only a projection of the verb would guarantee the correct semantic
interpretation of the outcome of the incorporation.
25 Note here that I am using the term ‘D-Structure’ just for expository purposes, and no ontological

commitment should be inferred from the use of that term. See below for a reanalysis of the status of
the D-Structure component (cf. as well Uriagereka’s (2006) proposal for a different view).

42
(61)
vP
qp
DPθ−AGENT v’
3
v VP
3
V DPθ−THEME
5
John broke a glass

Within this configuration, the DP object ‘a glass’ in sisterhood with the V gets
interpreted as the Theme, and the DP subject ‘John’ in the specifier of v, i.e., the sister
of v’, gets interpreted as the Agent. With this in mind, the only way of generalizing
over the theta-positions occupied by arguments is to claim that arguments are
assigned theta-roles in a very local configuration with the verb: a verbal projection.
But for that, labels are needed; no labeling, no projection26.
Likewise, ditransitive predicates point out in the same direction. Assuming a
larsonian shell structure like (62) (cf. Larson (1988, 1990), Chomsky (1995b)), theta-
role assignment takes place in the following configurations: (i) direct objects like ‘a
book’ get interpreted as Themes in sisterhood with the V’s first Merge, (ii) indirect
(oblique) objects like ‘Mary’ get interpreted as Goals in Spec-VP (i.e., in sisterhood
with V’), and (iii) subjects like ‘John’ get interpreted as Agents in Spec vP (the sister
of v’).

26 Notice as well that it is the whole DP that is assigned the theta-role, not only the Dº. See section
1.3.2.2.

43
(62)
vP
3
DPθ v'
g 3
John v VP
3
DPθ V'
g 3
Mary V DPθ
g 5
give a book

Again, the only generalization that can be derived for these facts is that
arguments take their theta-roles in a very local relation with a predicate; i.e., within
its projection. As far as I know, there is no explanation for this locality constraint.
Usually, the locality observed in theta-relations is lifted to a principle that requires
arguments to be in a local relation for theta-role assignment (cf. i.a. Williams (1994),
Frank (2002)), but there is no explanation for why this should be so. In section 1.3.2 I
attempt an explanation of this issue of locality and argument structure. By now, the
generalization above (that theta-roles are assigned within the maximal projection of a
verbal head) suffices to vindicate the need of labeled structures: VPs (and vPs) are
XPs, and labels are necessary to account for their phrasal properties.

1.3.1.3.7. Quantificational Dependencies.

This point is closely related to the previous one. A natural way of viewing
quantificational dependencies is to take quantifiers (Q) as being predicative in
nature. Under this view, the restriction and scope of a binary quantifier are
considered to be the quantifier’s arguments. Larson (1991) and Larson & Segal (1995)
offer a proposal along these lines. According to these authors, quantifiers are
basically predicates of sets and binary quantifiers are binary predicates in that they
take their “arguments” (the Q’s restriction and scope) in a syntactically ordered way.
For instance, we can say that when interpreting the sentence in (63), we create a
subset/superset relation between two sets; the set of ‘whalehood’ and the set of
‘mammalhood’ (64):

44
(63) All whales are mammals.
(64) {y : y is a whale} ⊆ {x : x is a mammal}

As argued by these authors, a natural way of capturing this asymmetry is to

take the quantifier ‘all’ to be a predicate that takes the sets of ‘whalehood’ and
‘mammalhood’ as its internal and external arguments respectively:

(65) ALL({y : y is a whale}) ({x : x is a mammal})

Thus, the truth conditional difference between a sentence like (63) and a
sentence like (66) would derive from the ordering of the arguments. Crucially this is
a syntactic proprerty of binary Qs (that is, it is the syntactic sister of the Q that
constitutes its restriction) 27:

(66) All mammals are whales.

But here we face a puzzle: the external argument of the Q is not in any obvious
way in its local domain (in the case of (63), the phrase ‘t are mammals’ is not in the
projection of the Q ‘all’). Thus, in order to guarantee the locality of the relation
between the Q and the arguments it takes, Larson (1991) proposes that a silent pro is
merged in the specifier of Q as in (67). This provides the locus for the nuclear scope
of the Q:

27Recall that a unary quantifier like ‘some’ would not induce any truth conditional difference between
analogous examples as in (i) and (ii):
(i) Some whales are mammals.
(ii) Some mammals are whales.
Thus, these quantifiers have to be treated as monadic predicates (cf. Hornstein & Uriagereka (2002:
124-125).

45
(67)

QP
3
pro Q’
3
all whales

He further proposes that this pro will get coindexed with the scopal phrase that gives
it a value at LF, after quantifier raising (QR) takes place. This is depicted in (68):

(68)
TP
qp
QPk T’
3 6
proi Q’ vP
3 6
all whales [tk are mammals]i
COINDEXATION

Thus, the projection requirements of the predicative Q are satisfied, since it

takes its restriction (‘whales’) by first Merge, and its nuclear scope when the Q gets
the pro in its specifier and it is coindexed with the scopal phrase that gives it a value
at LF. As can be observed, this is a very natural way of capturing the conservativity of
natural language quantifiers (or, in Barwise & Cooper (1981)’s terms, what the Q lives
on)28.
Again, without labels and projections, the syntactic nature of the structure of
binary quantification would be hard to explain for we would lose the generalization
on predicate-argument relations.

28 In chapter 4 I will develop this idea adopting Hornstein & Uriagereka’s (2002) “reprojection”
mechanism and applying it to the analysis of the syntax-semantics of focus. See Pietroski (2005) for an
in-depth analysis of conservativity in a conjunctivist, Neo-Davidsonian semantics.

46
1.3.1.4. Conclusion: The Need for Labels
As we saw, Chomsky’s (1995a) idea that syntactic phrases have labels is well
grounded empirically. I have discussed a number of interface phenomena that show
that labels are necessary and that they are present during the derivation and at the
interfaces. The conclusion that derives from this section then is that, while Collins
(2002) raised a fair question regarding the need of labels, labels seem to be necessary
in order to explain a wide range of empirical phenomena. Although I have not
analyzed the intricacies of the specific proposals in Collins (2002) and Seely (2006)
(see for a critical review Boeckx (2007)), I have discussed the main implications of the
label-free theory and shown that it cannot explain many basic features of grammar
like displacements, phonological phrasing or the thematic/selectional properties of
Vs and Qs.
Labels are necessary in order to capture a range of syntactic, semantic and
phonological properties of human language during the computation and at the
interfaces. This, obviously raises interesting questions for, as Collins (2002) and Seely
(2006) reasonably argue, labels seem to be more an imperfection than a feature of
optimal design (after all the bare set Merge operation that they propose is simpler
than Chomsky’s (1995a) labeling Merge). I analyze these questions in the next
section.

1.3.2. The Proposal: Labeling as a Bare Output Condition

In this section I present a formal reanalysis of the labeling mechanism of Bare
Phrase Structure, and as a corollary, I propose that argument structure is an
'emergent' property. My proposal will rely on the formulation of the hypothesis in
(69):

(69) Interface Legibility Hypothesis: Interfaces require sets with coherent type
intensions.

Given such a restriction, labeling operations will be explained as repairing strategies:

the label provides a coherent intension to the set it labels.

47
First, in section 1.3.2.1., I advocate for an analysis of the mechanism of labeling
as resulting from a complex operation: a dual application of Merge. Then, in section
1.3.2.2, I face the question of the ontology of labels (in particular, why they are there
and how they label the phrase they label). I propose that labels arise as repairing
strategies for legibility. Finally, in section 1.3.2.3., I explore some extensions of this
idea and its relation with recursion. This derivational analysis of labeling and
argument structure is what I will use in chapter 4 in my explanation of the syntax-
semantics interface of focus.

1.3.2.1. First Merge, then Project

Once we assume Chomsky’s (1995a) system in which a label identifies the set
created after the merger of two elements, the following questions arise:
(i) How is the identity of the label decided?
(ii) How does the label identify the phrase/set?
Question (i) is presented as unproblematic in Chomsky (1995a), since, as said before,
given two elements α and β, the simplest assumption is that the label δ of the set
constituted by α and β will be either α or β ((70a) or (70b) respectively):

(70a) (70b)
{α, {α, β}} {β, {α, β}}
3 3
α β α β

However, this assumption is clearly problematic in complex cases involving

specifiers like (71):

(71a) (71b)
{α, {δ, {α, {α, β}}}} XP
3 3
δ {α, {α, β}} YP X’
3 3
α β X ZP

In the cases of (71), we have the same structure we have in (70a) plus another
element δ merged as its specifier. Assuming that the correct representation of the

48
phrase structure of such a construction is as in (71), it will be incorrect to say that
either of the mergees projects. Let me rephrase this; it happens to be factually true
that in the simple case of (70), when α and β are merged, if the label of the output of
Merge(α, β) is either α or β, then the label is one of the mergees. However, if we
maintained that the label of a set created via Merge will be either of the mergees, the
conclusion would be that when merging δ with {α, {α, β}} in (71a), we would have
the following two options for labeling the output structure:

(72)
(i) δ
(ii) {α, {α, β}}

As can be observed, the labeling algorithm proposed in Chomsky (1995a), (cf.

section 1.3.1.1) will not be able to derive a structure like (71a). If, when merging {α,
{α, β}} with δ, it is δ that projects (option (72i)), the structure obtained is as in (73a),
which corresponds to the X’ representation in (73b), i.e., a structure where δ is the
head, and {α, {α, β}} is its complement:

(73a) (73b)
{δ, {δ, {α, {α, β}}}} YP
3 3
δ {α, {α, β}} Y XP
3 3
α β X ZP

Likewise, option (72ii) won’t provide the structure in (71) either. In particular,
notice that option (72ii) would produce something like (74a), whose corresponding
X’ theoretic representation would be one where the label of the structure is an XP,
something like the representation in (74b):

49
(74a) (74b)
{{α, {α, β}}, {δ, {α, {α, β}}}} (XP)P
3 3
δ {α, {α, β}} Y XP
3 3
α β X ZP

Besides it does not provide the intended result of (71), option (72ii) is also
problematic given the assumption that the label of a phrase is just a simple head, not
complex phrases (cf. the ‘uniqueness principle’ of Brody (1998) and López (2001). Thus,
if δ is to become a specifier with the Bare Phrase Structure representation in (71)), then
it is the label of the already constructed complex set ({α, {α, β}}) that will project, not
the whole set (as acknowledged in Chomsky (2000, 2001)). That amounts to saying
that the bare labeling algorithm of Chomsky (1995a) that we saw in (40c) doesn’t
suffice: something else is needed in order to derive the intended representation.
Furthermore, and addressing question (ii), notice that the notion of a label
does not derive directly from set-theoretic assumptions; there is nothing like a ‘label’
in axiomatic set theories (cf. i.a., Quine (1963)). Thus, given the phrase {α, {δ, {α, {α,
β}}}} of (71a), the somehow privileged status of the α label (which set-theoretically is
just one of the two members of a (71a)) will have to derive from something else. This
issue is the topic of the next section.

1.3.2.2. The Ontology of Labels

Chomsky (2005) proposes explicitly that the operation called Merge is nothing
more than an operation of set-formation (cf. Chomsky (2005) and Hinzen (2006b) for
discussion)). According to their analysis, using that simple operation we can derive
the successor function of positive integers à la von Neumann, starting from the
empty set Ø = { }. An application of Merge to that single object yields the singleton
{Ø}, different from the empty set. An application of Merge to that newly created
object provides the set that contains the singleton containing the empty set, i.e. {{Ø}},
a two-membered set since it contains both {Ø} and Ø. A further application of Merge
yields {{{Ø}}}, a three-membered set, and so on:

50
(75)
Ø=1
Merge1 Ö {Ø} = 2
Merge2 Ö {{Ø}} = 3
Merge3 Ö {{{Ø}}} = 4
Merge4 Ö {{{{Ø}}}} = 5

Thus, the potential unboundedness of the Merge operation amounts to the

successor function, providing the algebra to generate the set of positive integers.
However, it should be noted that this conception of binary Merge qua set-
formation entails the conclusion that in and of itself, the merger of α and β cannot
give a labeled structure, but a simpler {α, β} set. So, the only way to get labeled
structures using Merge and the lexicon is to conclude that in these instances we have
in fact two applications of Merge, where the first one creates a set and the second one
provides it a label. These two steps are represented in (76):

(76)
Step1: α↔β = {α, β}

{V, DP}
3
V DP

Step2: α↔{α, β} = {α, {α, β}}

{V, {V, DP}}

3
V DP

I have reviewed in section 1.3.1.3 some of the empirical evidence

demonstrating the presence of labels (i. e. the Step2 of (76)), but again, questions
about the emergence and ontology of these labels arise:

51
(77) I - Why do syntactic phrases have labels?
II - How do labels appear derivationally?
III - How do labels identify the set they label?

As I advanced, in order to answer these questions, my proposal will rely on

the following hypothesis:

(78) Interface Legibility Hypothesis: Interfaces require sets with coherent type
intensions.

As I will discuss presently, the adoption of (78) will answer automatically the
question in (77-I). Furthermore, assuming such a restriction, labels will identify their
sets providing them with an intension (question (77-III)). In this respect, we capture
the core property ascribed to labels, stating what is known as the is-a relation (cf.
Chomsky (1955, 1957, 1965), Chomsky & Miller (1963), Lasnik & Kuppin (1977)). Let
me explain this.
Let’s take back the case of (76), represented here as (79) for convenience. At
step1, where the simple {V, DP} set is created we have the following situation:

(79)
Step1:
{V, DP}
3 = {Kiss(e) & Mary(y)}
V DP

The question is that, at this point, the set {V, DP} obtained is heterogeneous: a
set of two independent predicates (one over events ‘Kiss(e)’ and the other one over
individuals ‘Mary(y)’). Therefore, by (78), it is illegible. And notice here that, by
assumption, this malformation is purely combinatorial (i.e., the outcome of the
merger of two elements). The problem is not in the denotata of those elements, but in
their combinatorial properties: all we have in (79) is a monadic event predicate
‘Kiss(e)’ conjoined with a predicate over individuals ‘Mary(y)’, which, as such, is

52
independent from the verb (i.e., the phrase {Kiss(e) & Mary(y)}). That is, the DP is not
a participant of the event at this stage. Obviously, this formula cannot have a
propositional interpretation, and should create a semantic crash if sent to the C-I
interface (cf. section 1.2.2 of this dissertation). That is, providing the (simplified)
lexical entries in (80) where ‘kiss’, ‘Mary’ and ‘John’ are lexically monadic predicates,
a correct logical form representation for the sentence in (81a) would be (81c), where
predicate participants are introduced by theta-roles. On the other hand, (81b) is
obviously ill formed (cf. section 1.2.2)29:

(80) kiss: Val(e, kiss) iff kissing(e) (predicate over events)

John: Val(x, John) iff John(x) (predicate over individuals)
Mary: Val(x, Mary) iff Mary(x) (predicate over individuals)

(81a) John kissed Mary.

(81b) *∃e[kissing(e) & John(x) & Mary(y)]

(81c) ∃e[kissing(e) & Agent(e, John) & Theme(e, Mary)]

Hornstein, Nunes & Pietroski (2006) and Hinzen (2006a-b) suggest that theta
roles are correlated with syntactic labels and the order of the merger, given that
argument DPs need theta-roles but adjuncts do not, for they are pure predicates of
the event, as we saw in section 1.2. As an illustration, observe the logical form of the
sentence (82a), represented in (82b):

(82a) John kissed Mary passionately.

(82b) ∃e[kissing(e) & Agent(e, John) & Theme(e, Mary) & passionately(e)]

29 The representations in (80) just says that ‘a value of e will be kiss iff e is a kissing’ and so on. This

assumes that a predicate can apply to various things that might not form a set, which is not important
for my discussion here (cf. Higginbotham (1986), Larson & Segal (1995), Pietroski (2005)).

53
I would like to provide a principled explanation for the idea that predicate-
argument relations involve labels but adjunctions do not. My proposal is based in the
Interface Legibility Condition proposed in (78). As I said earlier, having the {V, DP} set
in (79), the structure is computationally illegible. Thus, my proposal would be that in
order to gain legibility, the strategy that human language deploys is labeling the
structure: i.e., promoting the locus head to a privileged status to give the phrase/set a
coherent intension. This could be accomplished remerging the V head (the active
locus) to the previous structure as in (83), the step2 of (76). This provides an answer to
the question in (77-II)30:

(83)
Step2:
{V, {V, DP}}
3 = {Kiss(e), {Kiss(e) & Mary(y)}}
V DP

Now, the moment the V is remerged with the {V, DP} set, an asymmetry
emerges in the new set created with this labeling and projection. Crucially both
members of the newly created {V, {V, DP}} set (V and {V, DP}) will have a verbal
character, for both of its members contain a verbal element. Thus, as such, the set {V,
{V, DP}}v can be labeled with a verbal type to be readable at the interfaces as a phrase
that predicates over events31.
We are left with a last problem though: {V, DP} (a member of the newly
created {V, {V, DP}}v) is still an incoherent object at the interfaces, since the symmetry
broken by the head’s projection arises only in the highest set (i.e., V’s projection; the
set obtained by Step1 still violates (78)). Notice that after the projection of V, we have

30 The representation in (83) is a shorthand of the structure in (83’) below where the locus is explicitly
copied and remerged with the previous structure. Another representational homonymous to (83) and
(83’) is the multi-dominance structure in (83’’):
(83’) (83’’)
{V, {V, DP}} {V, {V, DP}}
3 gi
V {V, DP} g {V, DP}
3 g3
V DP V DP
31 I represent set intensions informally with subscripts: {V, {V, DP}}
v

54
a DP that is interpretable as such (i.e, it is interpretable as an individual (Val(x, Mary)
iff Mary(x))) embedded within a verbal environment; a VP (or {V, {V, DP}}v).
However, given the assumption of the Interface Legibility Condition in (78), the
embedded {V, DP} set within this convergent VP is illegible, since it lacks a coherent
intension. Obviously, then, recursion on the strategy of remerging the locus wouldn’t
solve the problem for it only applies to the root of the tree. Hence, I would like to
suggest a repairing strategy at this stage. I would like to propose that the fact that
now it is contained within a verbal-typed set ({V, {V, DP}}v) forces the DP inside {V,
DP}, to lift its type (à la Pietroski (2005)) to accommodate its type to that of the
intension of the highest set that contains it, (84). This repairing strategy is what will
allow the jump from a pure conjunctive, adjunct-like syntax to a higher-order syntax
of predicate-argument relations. Given the assumption that what interfaces require is
a verbal set (or any other coherently typed expression), the DP complement will lift
its type from an individual-denoting type to an event-participant one. This is the third
step32:

(84) Type shifting:

Step3: Val(x, Mary) iff Mary(x) Ö Val(e, int-Mary) iff Internal(e, Mary))

This type-lifting turns an individual-denoting expression like ‘Mary’ into an

argument of a verb by relating the value of the internal Mary to that of the verb. The
label, thus, serves as a bridge for establishing predicate-argument relations.
As a result, given this adjustment, both syntax and semantics get the correct
representation: in syntax, a coherent set-theoretic object is created ({Vv, {Vv,
[DPθ]v}v}v)33 and in semantics, we get the correct compositionality of the event
predicate and its internal argument (Kiss(e) & Internal(e, Mary)). With this model,
then, adjunction syntax is the basic pattern, and predicate-argument syntax is the

32 Here I am using Pietroski’s (2005) terminology: the term ‘Int(ernal)’ stands for a generic
participation relation “Internal of”. Then each verb will specify the type of internal participants that it
takes (a rheme, resultee, undergoer etc.). As Pietroski (2005: 53) puts it “such verbs will “infuse”
grammatical relations (which by themselves have purely formal significance) with specific thematic
significance”. See as well Carlson (1984), Dowty (1989) and Schein (2002) for discussion.
33 I use the notation [DP ] for the DPs that after lifting their type become participants of an event (i.e.,
θ v
become arguments of the verb that get a theta-role).

55
marked (or derived) one. Both arguments and adjuncts are introduced via the same
operation: Merge.
The move I just proposed argues for a change in the point of view of the goal
of inquiry; rather than looking for explanations for the lack of connectivity effects in
adjuncts, we should be looking for the reasons for the connectivity of arguments.
Presumably, as the proposal just made suggests, they are due to their dependent
nature.
For concreteness, the external argument will be introduced in quite in the
same way as the internal argument: when a V-typed set (the VP) merges with v, the
new phrase created will be convergent, for both elements undergoing Merge ({V, {V,
[DPθ]v}v}v and v) are verbal (predicates of the event):

(85)
{v, {V, {V, [DPθ]v}v}v}v
3
v {V, {V, [DPθ]v}v}v
3
V DP

The introduction of the subject DP will require the same labeling + type
shifting strategy that we saw for the internal DP in order for the structure to be
readable (remerging the locus v to the whole structure and adjusting the type of the
external argument):

(86)
{v, {[DPθ], {v, {V, {V, [DPθ]v}v}v}v}v}v
3
DP {v, {V, {V, [DPθ]v}v}v}v
3
v {V, {V, [DPθ]v}v}v
3
V DP

Finally, notice that the type lifting provides the v with its external argument at
the same time that it explains the mysterious necessity for a local configuration for
predicate-argument relations that I alluded to in section 1.3.1.3.6. As I argued there,

56
predicates take their arguments in very local configurations. Thus, for instance,
Williams (1994) argues that in the sentence in (87), both Bill and mackerel are
arguments of the predicate like, but fish is not a predicate of any verb, but rather a
predicate or a theme of the whole proposition34.

(87) As for fish, [Bill likes mackerel].

[Williams (1994: 27)]

With the phrase structure building mechanism presented here, it won’t be the
case that verbs have to take arguments in local configurations of the ‘D-Structure
Component’ (cf. Uriagereka (2006)), but quite the opposite: argument structure
emerges as the outcome of a repair strategy when heterogeneous {V, DP}-like sets are
readjusted for convergence at the interfaces.
The corollary of this idea is that argument structure is an emerging property,
derived from an interface legibility condition on the sets created via a two stepped
Merge (or two applications of Merge) and a third step of type-adjustment. Thus, the
proposal I just made defends the natural marriage of minimalist syntax and
conjunctive Neo-Davidsonian semantics (cf. as well Hornstein (2002), Pietroski
(2003a).

1.3.2.3. Extensions
The statement I made in (78), repeated here as (88) is a strong one:

(88) Interface Legibility Hypothesis: Interfaces require sets with coherent type
intensions.

34 For clarification, even if in the example in (87) we have a hyperonymy/hyponymy relation between
fish and mackerel this is not the same case as those conflation cases discussed by Hale & Keyser (2002)
like (i):
(i) Bill danced a Scottish dance.
In the case of (87) the hyperonymy/hyponymy relation is not necessary, and we could have, for
instance, something like (ii) (implying that the wine was chosen according to Bill’s menu or so):
(ii) As for the wine, Bill likes mackerel.

57
The prediction is that we should observe the same predicate-argument
relations in other cases of mismatch. I believe the approach is promising. As I
advanced in section 1.3.1.3.7 and will extend in chapter 4, QPs (and CPs) also should
be understood as predicational in nature, given that they take their arguments in a
syntactically ordered way. Likewise, recent analyses also lead towards a
predicational analysis of other categories like Ps (cf. Svenonious (2004)) or Ts and
Asp-s (cf. Demirdache & Uribe-Etxebarria (2000, 2007)).
Notice, finally, that taking adjunction syntax to be more basic than predicate-
argument syntax might also have an evolutionary basis, providing a way to analyze
the origin of the complex language capacity from a more primitive syntax. Authors
like Hurford (2001, 2003) have argued that the concepts of protothought were just
one place predicates of the form PREDICATE(x). Unifying this idea with the
assumption that label-less adjunction syntax is more primitive than predicate-
argument syntax provides a seemingly plausible hypothesis for the structure of
protolanguage as having monadic predicates and adjunction syntax (i.e. bare Merge),
but not any predicate-argument relation. Then, a crucial step towards the complex
properties of natural language syntax would have been provided by the appearance
of the Interface Legibility Condition in (88), and the ability to type shift predicates.
Notice as well that what I termed “Interface Legibility Condition” could also be
understood as one of Chomsky’s (2005) “third factor” effect; one of the “principles of
data analysis” that is built-in the faculty of language. Then, the phrasal adjustments
triggered by (85) wouldn’t incur in any look ahead. In fact, regarding it as a syntactic
property, one of the predictions of the idea that I presented is that as a result of the
acquisition of the labeling option the computational system of the human language
gained a fundamental economy treat: recursion35. Labels permit to keep the derivation
active while forgetting about the inner constituents of the phrase. Then, the recursive
trait of natural language syntax, which apparently, is uniquely human (cf. i.a.,
Hauser, Chomsky & Fitch (2002), Tomalin (2007)) would constitute a cognitive tool

35Authors like G. Lorenzo suggest that recursion might not be an innovation but rather it could have
been adapted from previously available tools for the “Theory of Mind”: asymmetric embedding of
propositional attitudes like “beliefs”, “fears” and so on (cf. Lorenzo (2006a), (2006b)).

58
that boosts computation36. As Chomsky (2000: 135) puts it “Computation is driven by
a probe/selector of a label, which projects. Hence, no operation can be contingent on
application of earlier ones. […] Both label determination and operations are “first-
order Markovian”, requiring no information about earlier stages of derivation”37.
Obviously, these speculations raise important conceptual concerns with
respect to the ontology of the C-I interface, for as Hinzen (2006a) has defended, there
might not be anything like a Logical Form level of representation38. As I said in
section 1.2.1, I am agnostic. The issue is extremely interesting but, unfortunately, I do
not have much of interest to add to it.
I will leave these issues aside and resume to the grammar of focus in the next
chapter. But first, I would want to offer a summary of the dissertation.

1.4. Summary and Organization of the Dissertation

In chapter 2, I present the basic patterns of focus in Basque. I will describe
three effects of focus: (a) a specific syntactic configuration (left-adjacent to the verb),
(b) a specific prosodic realization through pitch accent assignment and phonological
boundaries and (c) the dual semantic effects of focus regarding presuppositionality
and exhaustivity.
In chapter 3 I will develop the Derivational Approach to F(ocus)-Structure
whereby F-Structure is constructed within the narrow syntax and can be interpreted
by the interfaces (thus triggering the phonologic and semantic effects I describe in
chapter 2). I will show that the Focus Nuclear Stress Rule I that I will present there
derives the facts of nuclear stress placement straightforwardly given that it applies
upon the syntactically construed F-Structures.
Chapter 4 is devoted to the syntax-semantics interface of focus. I propose that
the overt syntactic displacements that we observe in languages like Basque provide a
direct mapping from syntax to the logical form. This is obtained with the combination

36 Thanks to Itziar Laka for stimulating discussion on these issues.

37 A. Clark’s thesis that labels enable humans to form mental representations that would otherwise be
much harder to form by serving as “conceptual anchors” points in the same direction (cf. Clark (1997)
and Lupyan (2006) as well as Longa (2006)).
38 Thanks to Cedric Boeckx for pointing out this issue to me.

59
of a Neo-Davidsonian event semantics à la Herburger (2000) and the labeled syntax I
developed in chapter 3.
In chapter 5 I compare my analysis with the NSR-based approaches to F-
Structure. I show that these approaches to F-Structure have many empirical and
conceptual drawbacks and that the theory I have proposed in chapters 3 and 4
circumvents them in a very natural way.
In chapter 6 I analyze the patterns of answers to multiple Wh-questions, which
can be accounted for naturally within the theory presented in chapters 3 and 4 as
split foci. I show that previous approaches to these constructions are paradoxical,
whereas the theory I develop accounts for them in a straightforward way.
All in all, the gist of this dissertation should be regarded as an attempt to
capture the essence of one of the core functional properties of natural languages (i.e.,
focus) within a highly restricted formal framework. The corollary will be that such an
epistemological move proves not only plausible but highly clarifying both for the
study of focus and for the reinforcement of the explanatory adequacy of
“syntactocentric” architectures of grammar and the theory of Bare Phrase Structure,
that will be essential throughout this dissertation.

60
CHAPTER 2. THE ESSENCE OF FOCUS IN BASQUE: SYNTACTIC,
PROSODIC AND SEMANTIC PROPERTIES

In this chapter I present the basic data that will provide the empirical grounds
for this dissertation. I start with an introduction to the syntax of focus in Basque in
section 2.1. The exposition here will be very descriptive, its goal being to present the
basic properties of the syntax of focus in Basque. For expository purposes, I will use
the mainstream analysis of these facts which is Ortiz de Urbina’s (1989) left
peripheric analysis. Later on, these facts will be reanalyzed in a detailed way in
chapter 4, where I compare the Nuclear Stress Rule (NSR)-based approach and the
“left peripheric” or “cartographic” approach to focus while developing a new
analysis of the syntax-semantics interface of focus.
In section 2.2 I present the phonological realization of focus in two
typologically different dialects of Basque: Central Basque (variety of Errenteria) and
Northern Bizkaian Basque (variety of Lekeitio). The phonological properties
displayed by focus in these two dialects will be used later on to analyze the relation
between syntax and phonology and the architecture of the grammar (cf. chapters 3
and 5 of this dissertation).
Section 2.3 is dedicated to the semantic nature of focus in Basque. In
particular, I study the contrastivity and presuppositionality triggered by focus. I
show that focus triggers both contrastive interpretations and presuppositional
readings, the only exceptions being out-of-the-blue utterances.
Finally, in section 2.4 I summarize the main conclusions of the chapter and
discuss a more theoretical point regarding what it would mean to have different
types of foci in the grammar. The issue is important since it is not uncommon to find
analyses that postulate different types of foci with different phonological,
morphological and/or syntactic realizations as well as different semantic
interpretations. I will argue that before proposing several types of foci, we should
define what it means to be a type. In fact, in chapter 4, I will defend that the different

61
focus constructions found in Basque follow from different CP systems, and I will
suggest that the different types of focus might be the epiphenomenal result of that
difference.

2.1. Syntactic Properties of Focus in Basque

In this section I present the main features of the syntax of focus in Basque. As I
said in the introduction, for ease of exposition I will follow Ortiz de Urbina’s (1989)
analysis but I will use it very descriptively. The reason for this is the existence of
completely different analyses of focus in Basque that postulate different kinds of
movements and completely different conceptions of the architecture of grammar;
among the most relevant ones we must mention the “Cartographic” approach and
the “NSR-based approach”. The “cartographic approach” I will use here proposes a
leftward movement of the focal XP to a functional specifier, whereas the “NSR-based
approach” proposes movements of nonfocal material instead, so that the focalized
constituent ends up being the most embedded element, thus getting nuclear stress.
For the time being, I just want to present the basic syntactic data with the aim of
setting the empirical phenomena that I will analyze in the rest of the dissertation. I
will resume to a more theoretically grounded discussion of the data in chapters 4 and
5, where I compare the “cartographic” and “NSR-based approaches” and show the
superiority of the cartographic approach over the NSR-based approach.

2.1.1. The Basic Pattern

One of the well-known characteristics of Basque, along with its ergative
morphosyntax and the pitch-accent typology of some of its varieties (cf. section 2.2),
is that it is a ‘Discourse-Configurational’ language. In this type of languages, the
(discourse-)semantic function ‘focus’ is realized through a particular structural
relation (cf. i.a. Altube (1923), de Rijk (1978), Mitxelena (1981) and Etxepare & Ortiz
de Urbina (2003) on Basque data and Kiss (1995) on discourse-configurationality).
Thus, the word order of the clause is affected by discursive functions. For instance,
an out-of-the-blue utterance, where the whole sentence is interpreted as focus (cf.
section 2.3), necessarily has the word order S-O-V, where V represents the verbal

62
complex [V + Aux], (1)1. On the other hand, a sentence with focus on the subject -a
potential answer to a question like (2a)- cannot be S-O-V, as shown in (2b). Rather, it
has to be either S-V-O (2c), or O-S-V (2d), in which case the direct object is
interpreted as being topicalized:

(1) [Bidasoak maila galdu du]F [out-of-the-blue utterance]

Bidasoa category loose AUX Broad Focus
“[Bidasoa lost the category]F” S-O-V

(2a) Nork galdu du maila?

who loose AUX category
“Who lost the category?”
(2b) *[Bidasoak]F maila galdu du
Bidasoa category loose AUX
“[Bidasoa]F lost the category”
(2c) [Bidasoak]F galdu du maila [contextualized utterances]
Bidasoa loose AUX category Narrow Focus on the Subject
“[Bidasoa]F lost the category” (OTop)-S-V-(O)
(2d) [Maila]T [Bidasoak]F galdu du
category Bidasoa loose AUX
“The category, [Bidasoa]F lost it”

If, on the other hand, we want to focalize the object instead of the subject, the
word order can be just S-O-V as illustrated in (3b) (just like in the out-of-the-blue
sentence of (1)). But in this case, as in the case of (2d), all the material preceding the
focus (in this case, the subject) is interpreted both semantically and intonationally as
topicalized. This is illustrated in (3b), as an answer to the question in (3a):

1 These sentences are also called “broad foci” whereas sentences with focus on just one constituent are

called sentences with “narrow foci” or contextualized utterances (cf. section 2.3).

63
(3a) Zer galdu du Bidasoak?
what loose AUX Bidasoa
“What did Bidasoa loose?”
(3b) [Bidasoak]T [maila]F galdu du
Bidasoa category loose AUX
“Bidasoa, it lost [the category]F”

Another constituent order for marking focus on the object is O-V-S, in this
case, with no topic effect on the subject, (3c):

(3c) [Maila]F galdu du Bidasoak

category loose AUX Bidasoa
“Bidasoa lost [the category]F”

The word order of constituents in (3c) with the verb to the left of the subject
strongly suggests that both the focal object and the verb undergo leftward movement
over the subject.
Oblique objects (4) and adjuncts (5) do not alter this pattern. The neutral word
order of a sentence with a ditransitive predicate is S-IO-DO-V (cf. i.a. A. Elordieta
(2001), E. Arregi & Ormazabal (2003)). Since the focalized element must be
immediately adjacent to the verb, the neutral word order cannot be used when focus
is on the indirect object:

(4a) *Jonek [Mireni]F musu bat eman dio.

Jon Miren.to kiss one give AUX
“Jon gave [Miren]F a kiss”

In order to express focus on the indirect object the order of constituents must
be altered. The focus is moved to the clause initial position, followed by the verbal
complex (main verb and auxiliary) as in (4b), or, while keeping this focus-verb
adjacency, some material appears preceding them. In this case, as in the preceding

64
examples of subject and object focus, the material preceding the focus will be
interpreted as topicalized (4c)2:

(4b) [Mireni]F eman dio Jonek musu bat.

Miren.to give AUX Jon kiss one
“Jon gave [Miren]F a kiss”
(4c) Jonek musu bat [Mireni]F eman dio .
Jon kiss one Miren.to give AUX
“Jon gave [Miren]F a kiss”

As I said, the pattern is exactly the same for adjuncts. Consider, as an

illustration of this, the case of clausal adverbs like “atzo” (yesterday). These adverbs
have quite a free distribution within the clause, and in out-of-the-blue contexts they
can appear in different positions: sentence initially (5a), or sentence internally (5b) (cf.
Osa (1990), A. Elordieta (2001)).

(5a) [Atzo Jonek Miren ikusi zuen]F.

Yesterday Jon Miren see AUX
“[Jon saw Miren yesterday]F”
(5b) [Jonek atzo Miren ikusi zuen]F.
Jon yesterday Miren see AUX
“[Jon saw Miren yesterday]F”

However, if we want to focalize the adverb, it must appear in either of the two
aforementioned configurations; either left-adjacent to the verb and sentence-initially
(5c), or, while keeping the same adjacency with the verb, in a non-initial position
with the material preceding it interpreted as topic (5d).

2 For ease of exposition, in the translations and glosses of the examples here and hereafter I will not

specify the topical interpretation of the material preceding the focal XP.

65
(5c) [Atzo]F ikusi du Jonek Miren.
Yesterday see AUX Jon Miren
“Jon saw Miren [yesterday]F”
(5d) Jonek Miren [atzo]F ikusi du.
Jon Miren yesterday see AUX
“Jon saw Miren [yesterday]F”

Cases in which the focussed adverb and the verbal complex are not adjacent are
ungrammatical, as illustrated in (5e) and (5f):

(5e) *[Atzo]F Jonek Miren ikusi du.

Yesterday Jon Miren see AUX
“Jon saw Miren [yesterday]F”
(5f) *Jonek [atzo]F Miren ikusi du.
Jon yesterday Miren see AUX
“Jon saw Miren [yesterday]F”

The generalization that can be drawn from the data so far is that the focal XP
necessarily appears left adjacent to the verbal complex, in a clause initial position.
The focus can be preceded by topicalized material.
The traditional way of analyzing these facts stems from the Principles &
Parameters framework. The idea is that the focal phrase undergoes movement to a
position situated at the left periphery of the clause followed by movement of the
verb. The main implementation of this analysis of focus in Basque is Ortiz de Urbina
(1989). Based on the syntactic and semantic parallelism between Wh-movement and
focus in Basque, Ortiz de Urbina proposes that in this language, both Wh-elements
and foci move to Spec-CP, which he takes to be a left headed projection in Basque.
The occupation of Spec-CP triggers an I-to-C movement that creates the V2 pattern
observed in these constructions. According to this analysis, sentence (2c), repeated
here as (6a), is derived as represented in (6b):

66
(6a) [Bidasoak]F galdu du maila.
Bidasoa loose AUX category
“[Bidasoa]F lost the category”

(6b)
CP
qp
[Bidasoak]F C’
qp
galdu du InflP
qp
t Infl’
qp
VP Infl
qp t
maila t

Likewise, a sentence like (3c), repeated here as (7a), would be derived as

represented in (7b):

(7a) [Maila]F galdu du Bidasoak

category loose AUX Bidasoa
“Bidasoa lost [the category]F”
(7b)
CP
qp
[maila]F C’
qp
galdu du InflP
qp
Bidasoak Infl’
qp
VP Infl
qp t
t t

In the next section I discuss the syntactic properties of focus in embedded contexts.

67
2.1.2. Focus in Embedded Contexts
Focus can also appear in embedded contexts. A complex out-of-the-blue
sentence is represented in (8a), where the order of constituents is S-[S-O-V]CP-V:

(8a) [Jonek [Mirenek ura edan duela]CP esan du]F

Jon Miren water drink AUX.that say AUX
“[Jon said that Miren drank water]F”

In Basque, there are two basic possibilities to focalize an element of an

embedded clause. (8b) and (8c) with focalization of the embedded subject of (8a)
illustrate these two possibilities. (i) The first strategy fronts the focalized element —in
this case, the subject ‘Mirenek’— to clause-initial position in the matrix CP, as in (8b).
This strategy has been analyzed by Ortiz de Urbina (1989) as a case of long distance
movement of the focal element. (ii) The second strategy, illustrated in (8c), consists
on pied-piping the whole CP that contains the focalized element. The embedded CP
in (8c) thus moves to the position that is occupied by the focus in the cases of long
distance movement illustrated in (8b), and must be followed by the verb of the
matrix clause3.

(8b) [Mirenek]F esan du Jonek edan duela ura.

Miren say AUX Jon drink AUX.that water
“Jon said that [Miren]F drank water”

3 Actually, sentences like (8c) are a bit marked with the elements of the embedded clause in postverbal
position. They ameliorate if we topicalize those elements over the focus, as in (8c’):
(8c’) Ura [Mirenek]F edan duela esan du Jonek.
water Miren drink AUX.that say AUX Jon
“Jon said that [Miren]F drank water”
The reasons for the preferability of (8c’) over (8c) are obscure to me; it might be related to a processing
effect of having an element left-adjacent to the main verb, which is the canonical position for focus.
This is an issue that deserves a close exploration and I will leave it for further research. Here, for the
ease of exposition, I will use sentences like (8c), where focus is the first element in the pied-piped
clause for the discussion.

68
(8c) [Mirenek]F edan duela ura esan du Jonek.
Miren drink AUX.that water say AUX Jon
“Jon said that [Miren]F drank water”

According to Ortiz de Urbina (1989), the long distance strategy will extract the
focus phrase from its base-generated position and move it to Spec-CP successive
cyclically, hence, triggering I-to-C movement in all clauses where Spec has been
occupied, as represented in (8d):

(8d) CP1
ei
[Mirenek]F C’
3
esan du IP
3
Jonek I’
3
VP t
3
CP2 t
ei
t C’
3
edan duela IP
3
t I’
3
VP t
3
ura t

In (8d) the focal subject of the embedded clause undergoes movement to the
Spec-CP of the matrix clause but the movement takes place in a cyclic way; from
Spec-CP to Spec-CP all the way up. The occupation of the Specifier of CP2 triggers I-
to-C movement in the embedded clause (hence, the Object-Verb inversion). The focus
moves from Spec-CP2 to Spec-CP1 triggering the same V2 pattern in the matrix
clause.

69
Notice that these operations take place systematically no matter how
embedded the focalized element is. This is illustrated in (8b’), with long distance
movement of a doubly embedded CP:

(8b’) [[Mirenek]F esan du Jonek [uste duela Peruk [edan duela ardoa]]]
Miren say AUX Jon think AUX-that Peru drink AUX-that wine
“Jon said that Peru thinks that [Miren]F drank the wine”

Alternatively, the pied-piping strategy in (8c) will extract not only the focus
phrase but the whole embedded clause containing it. Notice that the first step in the
strategy illustrated in (8c) is exactly the same one as that illustrated in the long
distance movement case in (8b). First, the focus phrase is extracted to the Specifier of
CP2; this triggers the V2 effect within CP2. This is represented in (9a). Then the whole
CP2 is pied-piped to Spec-CP1 triggering the I-to-C movement in CP1, as represented
in (9b)4:

(9a) CP2
3
[Mirenek]F C’
3
edan duela IP
3
t I’
3
VP t
3
ura t

4
See footnote 3 of this chapter.

70
(9b) CP1
qp
[[Mirenek]F edan duela ura]CP2 C’
3
esan du IP
3
Jonek I’
3
VP t
3
t t

These are the basic patterns of the syntax of focus in Basque. At this stage, I
will leave the discussion on the syntax of focus of Basque at this descriptive level and
in chapter 4 I will resume to a more theoretically-grounded evaluation. There I will
compare and discuss the two main approaches to these data (the ‘Cartographic’
approach that I have just sketched and the alternative NSR-based approach).
Let us now turn to another aspect of focus in Basque; its phonological
properties.

2.2. Phonological Properties of Focus in Basque

In this section I will review the phonological properties of focus in Basque. I
will provide a description of the dialectal variation observed in two typologically
different word-prosodic systems of Basque and review the basic facts concerning the
intonation of focus. The issue of the dialectal variability is of a major importance
given that despite the fact that we can find completely different word-prosodic
systems across dialects, they all display a very similar syntactic pattern regarding
word order. These facts thus offer us a very nice area to test the relations and
interactions between syntax and prosody and will be crucial in understanding the
interaction between the syntactic and the phonological components within the
architecture of grammar.

71
2.2.1. Basque Dialectal Variability
It is a well known fact that the Basque language shows a wide dialectal
variation within a rather small diatopic distribution. With respect to accentuation, we
can find two typologically different word-prosodic systems: Northern Bizkaian
Basque (NBB) is typologically a pitch-accent variety (a ‘non-stress’ language in
Beckman’s (1986) terms) whereas the rest of the varieties are stress-accent systems5.
In this thesis I will use evidence from these two varieties in order to analyze the
syntax-phonology interface. I will take the variety of Basque spoken in Lekeitio as an
example of the pitch-accent variety of NBB (cf. i.a. G. Elordieta (1997, 1998, 2003),
Hualde, G. Elordieta, Gaminde & Smiljanic (2002), Gussenhoven (2004)) and the
variety of Basque spoken in Errenteria as an example of the stress-accent variety of
Central Basque (cf. i.a. Hualde (1997), Irurtzun (2003, 2006)). As an illustration,
observe the location of Errenteria and Lekeitio with respect to the cities of San
Sebastian and Bilbao. I provide the locations of Errenteria and Lekeitio (with Bilbao
and Donostia for refeence) over the dialectal map designed by the prince L.L.
Bonaparte in 1863; the Carte des sept provinces basques montrant la délimitation actuelle de
l’euscara et sa division en dialectes, sous-dialectes et variétés.

5 Although Hualde, G. Elordieta, Gaminde & Smiljanic (2002) and Hualde (2007) report somehow

‘hybrid’ patterns typologically in between the pitch-accent and the stress-accent varieties.

72
Bilbao Errenteria

Donostia

Lekeitio

In the next sections I present the basic data of the accentual pattern of Central
Basque (2.2.1.1) and NBB (2.2.1.2).

2.2.1.1. The Stress-Accent System: Central Basque

What I will call ‘Central Basque’ corresponds to the variety spoken in most of
the province of Gipuzkoa. This variety has been analyzed in works like Mitxelena
(1972, 1976, 1985), Holmer (1991), Hualde (1997, 1999) and Irurtzun (2003, 2006).
Here, I will review the pattern as described in Irurtzun (2003, 2006). Typologically,
Central Basque is characterized by the following facts:
(i) There are two prosodically distinct sets of words: (a) an “unmarked” set of
words composed by the majority of morphemes and (b) a limited set of “marked”
words, mostly loanwords and words composed by specific suffixes. The unmarked
words follow a postinitial accentual pattern, which is very uncommon

73
crosslinguistically: in the sample of 444 languages studied by Hyman (1977) only 12
languages (2’7% of the sample) are reported to have a postinitial accent, and only 15
out of the 500 languages (3%) mentioned in the World Atlas of Language Structures
(Haspelmath et al. (2005))6. On the other hand, marked words have initial accent, a
quite common pattern crosslinguistically. As illustrated in (10), there are some
minimal pairs of unmarked-marked words in the variety spoken in Errenteria:

(10) Unmarked/Marked Words in Errenteria Basque

Unmarked: Marked:
basúa ‘the forest’ básua ‘the glass’
iltzía ‘the nail’ íltzia ‘to die’
aríya ‘the thread’ áriya ‘the ram’

(ii) Whichever the type of the root (marked (11a) or unmarked (11b)), the
accent is characterized by a F0 peak aligned with the accented syllable, followed by a
fall. This is illustrated in (11a, b), taken from Irurtzun (2003)7:

6 Hyman (1977) doesn’t include Basque in the list of languages with a postinitial accent pattern; the
languages he reports are Tolai, Tsaxur, Araucanian, Baure, Ignaciano, Dakota, Northern and Southern
Paiute, Central Pomo and Seneca. Standard Mongolian and Vogul (Tavda dialect) are also classified as
languages with postinitial accent, despite the pattern in those languages is not that clear. Basque is
classified as not having any dominant stress pattern (maybe due to the rich dialectal variability found
in Basque (cf. Hualde (1999)). On the other hand, Haspelmath et al. (2005) mention 16 varieties as
having a postinitial accentual pattern, but two of them are varieties of Basque that I analyze as
corresponding to the variety of Central Basque: Bidasoa Valley Basque and Oñati Basque (cf. Hualde
(1997) for discussion). The other 14 languages mentioned by Haspelmath et al. (2005) as having a
postinitial accentual pattern are Alyawarra, Araona, Arrernte (Western), Dakota, Lamu-Lamu,
Mapudungun, Mbabaram, Paiute (Southern), Pomo (Eastern), Siroi, Stoney, Thaypan, Tolai and
Uradhi.
7 All the pitch tracks in this section are taken from Irurtzun (2003).

74
(11a) Unmarked: a.rrí.ga.rri.ya ‘surprising’

[Irurtzun (2003: 23)]

(11b) Marked: í.ru.tan ’three times’

[Irurtzun (2003: 30)]

As I suggested in Irurtzun (2003), the tonal structure of the pitch accent could
be analyzed as H* for, if there are enough postonal syllables, F0 descends in a smooth
slope8:

8 G. Elordieta (p.c.) suggests to me an alternative analysis in terms of L+H*, with possible truncation of

the leading L tone in utterance-initial marked words. That is an interesting option which is worth
exploring. In any event that would not alter the conclusions to be drawn for the grammar of focus.

75
(12)
H* H*

´
σ σ´ σ σ σ σ

However, what is important for us is that in Central Basque the choice of pitch
accent changes with the focal status of the item, as I show in the next section.

2.2.1.1.1. Focal Accent and its Distribution

After reviewing the basic accentual patterns observed in Central Basque, in

this section I present the basic data concerning the prosodic realization of focus.
Within full-fledged clauses, there is more variability with regards to the pitch
accent inventory than the one described in the previous section. In particular, in
broad focus sentences (cf. section 2.3 below for details), there tends to be a F0 peak
aligned with the accented syllable of the immediately preverbal word, followed by a
substantial pitch reduction in the postnuclear stretch:

(13a) Broad Focus: H*

[Mirenek liBUrua eraman du]F
Miren book take AUX
“[Miren took away the book]F”

76
[Irurtzun (2003: 59)]

In Irurtzun (2003) I propose the following (simplified) tune-structure for a

broad focus (out-of-the-blue) utterance (where υ = utterance, ι = intermediate phrase,
α = accentual phrase and ω = prosodic word (cf. section 1.2.3)):

(13b) Broad Focus:

υ
qp
ι ι
g 3
α α α
g g 2
ω ω ω ω
Mirenek liBUrua eraman du
H- H* L-

If we analyze narrow focus constructions (where it is not the whole clause but
just one subconstituent that is focussed) we observe that they have a quite similar
tune. The difference is that the topical elements preceding the focal XP have a more
marked final high tone (maybe a H%) and, crucially, the pitch accent of the focal XP
is not a simple H*, for the F0 undergoes a sharp fall in the postonic syllable. Narrow
focus is also characterized by a compressed pitch range in the postfocal region, even

77
more so than in out-of-the-blue cases. Thus, as I argued in Irurtzun (2003), the pitch
accent of these focus constructions must be a compound H*+L tone:

(14a) Narrow Focus: H*+L

Amagoiak [orDEnagailua]F eraman zuen9
Amagoia computer take AUX
“Amagoia took away [the computer]F”

[Irurtzun (2003: 70)]

Again, the corresponding tonal structure is the one depicted in (14b):

9 I use the Standard Basque orthography in the written representation of the examples. The actual

pronunciation is given in the word tier of the pitch tracks.

78
(14b) Narrow Focus:

υ
qp
ι ι
g ei
α α α
g g 2
ω ω ω ω
Amagoyak orDEnagaillua eraman tzun
H% H*+L L-

Looking at the distribution of these pitch accents, the generalization that

emerges is as follows: in out-of-the-blue statements, the nuclear H* accent appears
aligned with the accented syllable of the immediately preverbal constituent:

(15a) [S Ó V]F
[Jonek liBUrua eraman du]F
Jon book take AUX
“[Jon took the book away (with him)]F”

(15b) [S IO DÓ V]F
[Jonek Mireni liBUrua eraman dio]F
Jon Miren.to book take AUX
“[Jon gave the book to Miren]F”

When the preverbal constituent is a complex XP as in the examples in (16),

then the nuclear stress appears on the rightmost element of that complex XP; that is,
the complement in cases of head + complement construction like (16a), or the head in
phrases composed by a specifier and a head, like that in (16b). In a complex phrase

79
composed by a specifier, a head, and its complement, the accent falls on the
complement, as shown in (16c)10:

(16a) [Jonek [liburu loDIa] irakurri du]F

Jon book thick read AUX
[Jon read the thick book]F
(16b) [Jonek [bere liBUrua] irakurri du]F
Jon his book read AUX
[Jon read his book]F
(16c) [Jonek [bere liburu loDIa] irakurri du]F
Jon his book thick read AUX
[Jon read his thick book]F

The distribution of the nuclear stress in contextualized utterances (narrow focus)

is very similar to that of out-of-the-blue utterances: it falls on the rightmost constituent
of the focal XP, as illustrated in the examples of (17), variants of the out-of-the-blue
sentences of (16) with focus on the object:

(17a) [Liburu loDIa]F irakurri du Jonek

book thick read AUX Jon
“Jon read [the thick book]F”
(17b) [Bere liBUrua]F irakurri du Jonek
his book read AUX Jon
“Jon read [his book]F”

10 For the purposes of this discussion, I will assume, adapting Goenaga’s (1978) analysis, that this is

the structure of a DP in Basque (cf. Artiagoitia (2006) for a recent and different analysis of the structure
of DPs in Basque):
DP
3
Possessor D’
3
NP Det
3
Noun Adjective

80
(17c) [Bere liburu loDIa]F irakurri du Jonek
his book thick read AUX Jon
Jon read [his thick book]F

As can be observed, the nuclear stress appears on the rightmost element of the
focal XP (the immediately preverbal one).
Having reviewed the main features of the prosody of the stress-accent system
of Central Basque, let us now review the pitch-accent variety of Northern Bizkaian
Basque.

2.2.1.2. The Pitch-Accent System: Northern Bizkaian Basque

Northern Bizkaian Basque (NBB) is a variety spoken on the North and
Northeastern coast of the province of Bizkaia and in a band of around 20 kilometers
inland from the coast. This is the most deeply studied variety of Basque (cf. i.a.,
Hualde (1989, 1999), G. Elordieta (1997, 1998, 2003, 2007a-b) Jun & G. Elordieta (1997)
and Gussenhoven (2004, §9)). Typologically NBB can be characterized as follows:
(i) It has a non-stress accent word-prosodic system (cf. Beckman (1986) for
discussion on the typological background). In NBB, the only correlate of accent is
pitch or F0, other potential correlates like duration and intensity are not consistent;
not even in corrective focalization nor downstep environments (cf. G. Elordieta &
Hualde (2001), G. Elordieta & Hualde (2003)).
(ii) Similar to what happens in Central Basque, there are two sets of words, an
“unmarked” set composed by the majority of roots (ama ‘mother’, euli ‘fly’, iturri
‘fountain’…), and a “marked” set composed mainly by most loanwords (presidénte
‘president’, arbóla ‘tree’, boligráfo ‘pen’…) and some suffixes (the comparative /-
’ago/, the superlative /-’en/, the diminutive /-‘txu/...). However, unlike in the
Central variety where all words can surface with a pitch accent, in NBB unmarked
words lack any accent of their own whereas marked words have a lexically specified
accent. The position of the accent of the marked words varies depending on the
subvariety of NBB, but it is always a prefinal syllable (on the final syllable in most
varieties and on the penultimate syllable in a few local varieties like Markina and

81
Ondarroa). In the subvariety of Lekeitio, which is the one I will concentrate on here,
the accent of “marked” words falls on the penultimate syllable. Unmarked words
only receive an accent when they occur immediately preceding the verb or when
uttered in isolation.
In Lekeitio Basque words combine in bigger prosodic units called “accentual
phrases”, which are characterized by a %L initial tone, as well as by a H- phrase
accent aligned with the postinitial syllable that spreads rightwards creating a plateau
which extends until it reaches a pitch accent (cf. Jun & G. Elordieta (1997), G.
Elordieta (1998)). The tonal structure of the lexical pitch accent of marked words is
characterized by a sharp fall in F0, which G. Elordieta (1997, 1998) analyzes as
deriving from a H*+L pitch accent. We can find, then, minimal pairs like the one in
(18a-b), where the only distinguishing cue is the accent exhibited by (18b) which is
lexically attached to the plural morpheme /’-en/:

(18a) Unmarked (unaccented): lagunen diruá ‘the money of the friend’

[G. Elordieta (1998: 555)]

82
(18b) Marked (accented): lagúnen dirua ‘the money of the friends’

[G. Elordieta (1998: 555)]

This system, thus, is very similar to the one found in Tokyo Japanese (cf.
Pierrehumbert & Beckman (1988)).
Interestingly, unmarked words like the one in (18a) might as well bear a pitch
accent in some specific constructions. I address this issue in the next section.

2.2.1.2.1. Focal Accent and its Distribution

We have just seen that in NBB there are two sets of words, a set of lexically
unaccented words and a small set of lexically accented words. When a phrase
composed by accented words is focal, the distribution of nuclear accent in the
sentence is very similar to the one observed in Central Basque: nuclear stress appears
on the immediately preverbal word. Likewise in out-of-the-blue statements. However
there is another source for accentuation in NBB: if a phrase composed solely by
unaccented words is in narrow focus (i.e., it is the focus of a contextualized utterance),
then a nuclear pitch accent is assigned to the last syllable of the last word of that
phrase (this is what Jun & G. Elordieta (1997) call a derived accent). The same happens
if the phrase is uttered in isolation. For instance, consider the example of the word
laguna in (19), composed by the unaccented noun lagun ‘friend’ and the unaccented
singular article ‘-a’ (the word ‘umiágas’ has the lexically marked suffix /’-gas/

83
(‘with’) that assigns an accent to its preceding syllable regardless of the position of
the word). So, in the sentences in (19) the word laguna will get a derived accent only
when it appears left-adjacent to the verb (cf. (19b), (19c) vs. (19f), (19d)). Not assigning
an accent to it when it occupies the preverbal position, as in (19a), results in
ungrammaticality. Likewise, assigning laguna a derived accent when it does not
occur in the immediately preverbal position results in ungrammaticality as well (cf.
(19e), (19g) (cf. G. Elordieta (2007a-b) for discussion):

(19a) *Umiágas [laguna]F etorri da.

child.with friend come AUX
“[The friend]F came with the child”
(19b) Umiágas [laguná]F etorri da.
child.with friend come AUX
“[The friend]F came with the child”
(19c) [Laguná]F etorri da umiágas.
friend come AUX child.with
“[The friend]F came with the child”
(19d) Laguna [umiágas]F etorri da.
friend child.with come AUX
“The friend came [with the child]F”
(19e) *Laguná [umiágas]F etorri da.
friend child.with come AUX
“The friend came [with the child]F”
(19f) [Umiágas]F etorri da laguna.
child.with come AUX friend
“The friend came [with the child]F”
(19g) *[Umiágas]F etorri da laguná.
child.with come AUX friend
“The friend came [with the child]F”

So, the pattern that emerges is clear:

84
(i) Accented words always bear an accent.
(ii) Unaccented words get a derived accent only when they they appear
in the immediately preverbal position, or when uttered in isolation.
As I said before, in other cases unaccented words do not bear any accent and
create F0 plateaus with the spreading of the phrasal H- tone of the second syllable
until the next pitch accent. Let us look at the schemas in (20).

(20) Accentual Phrase in Lekeitio Basque (cf. G. Elordieta (1998))

(a) Derived Accents (α1, α2):

α1 α2

%L H- H*+L %L H- H*+L
(spreading) (spreading)

σ1 σ2 σ3 σ4 σ5 σ1 σ2 σ3 σ4 σ5 σ6 σ7 σ8 σ9
txis. tu. la. ri. dxá. mar. ki. ña. ko. txis. tu. la. ri. dxá.
(the flutist) (the flutist from Markina)

(b) Lexical Accent (α3) & Derived Accent (α4)

α3 α4

%L H- H*+L %L H- H*+L

σ1 σ2 σ3 σ4 σ5 σ6 σ7 σ8 σ9
on. da. rrú. ko. txis. tu. la. ri. dxá.
(the flutist from Ondarroa)

In (20a) we have two accentual phrases (α1 and α2). The first one is composed
of just one unaccented word (txistularidxa ‘the flutist’), and it bears a pitch accent
given that it is uttered in isolation (it is, somehow, ‘focused’). The second one, on the
other hand, is composed of the same word and an unaccented locative genitive
(markiñako ‘from Markina’) preceding it. As shown in the schemas in (20), the tonal
events are the same in both accentual phrases: an initial %L, and a phrasal H- aligned

85
with the second syllable that spreads rightwards up until we find a H*+L pitch
accent (which in the case of the examples in (20a) are derived accents, for txistularidxa
is a lexically unaccented word). In (20b), on the other hand, we have two accentual
phrases (α3 and α4): α3 is composed of an accented word ondarrúko ‘from Ondarroa’
and the α4 is composed of a lexically unaccented word that gets a derived accent by its
phrase-final position. Notably, the phrase ‘the flutist from Markina’ in (20a) and the
phrase ‘the flutist from Ondarroa’ in (20b) are syntactically analogous, but they have
a different prosody due to the fact that ondarrúko is a lexically accented word,
whereas markiñako is unaccented.
As can be gathered from the examples in (19) and (20), the only way an
unaccented word can get an accent is by appearing in the immediately preverbal
position (or by being uttered in isolation).
However, there is an important exception to the pattern of derived accent
assignment described above: when a narrowly focused unaccented word appears on
a position other than the immediately preverbal position, then it will not be able to
get a derived accent (cf. Hualde, G. Elordieta & A. Elordieta (1994), and G. Elordieta
(2007a-b) for an analysis of these facts). Imagine a case where we want to know the
origin of a given flutist that came. Then, we could ask the question in (21a), but,
crucially, its answer cannot be the one in (21b) with a derived accent on the focussed
genitive locative (which is a lexically unaccented word). Rather, a proper answer to
the question in (21a) must be as in (21c), where the focal markiñako bears no accent:

(21a) Núngo txistularidxa etorri da?

where.from flutist come AUX
“The flutist from where has arrived?”

(21b) *[Markiñakó]F txistularidxá etorri da.

Markina.from flutist come AUX
“The flutist [from Markina]F came”

86
(21c) [Markiñako]F txistularidxá etorri da.
Markina.from flutist come AUX
“The flutist [from Markina]F came”

The descriptive generalization of these facts is that in order for an unaccented

word to get a derived accent it has to be the last word of the XP where it is contained;
i.e., the one left-adjacent to the verb, as we saw in section 2.1. Thus, even if
[markiñako]F is the narrow focus in (21c), it cannot get a derived accent because it is not
the last word of the XP where it is contained [[Markiñako]F txistularidxa]DP. We cannot
split the DP ‘Markiñako txistularidxa’ syntactically and, hence, the focal genitive is not
left adjacent to the verb. As a result of this, it cannot get the derived accent. Rather, it is
txistularidxa that gets the derived accent by virtue of being the immediately preverbal
word11.

2.2.2. Conclusions
As we observed, there is quite a regular correlation between focus and pitch
accent in different dialects of Basque: the focal XP is realized through the assignment
of a specific pitch accent. There are, however, some exceptions: in NBB, the pitch-
accent variety, a lexically unaccented word in narrow focus that appears in a position
other than the immediately left-adjacent to the verb cannot get a derived accent.
In chapter 3, section 3.2, once I have presented my proposal of a derivational
approach to the focus structure, I will offer an analysis of the nuclear stress
placement that derives the patterns that we saw so far and overcomes some of the
major problems that previous proposals must face. But before doing that, in the next
section I will analyze the semantic contribution of focus in Basque.

11 There are as well other more complex cases where accentuation of a lexically unaccented and

focused word cannot get a derived accent, notably, within embedded clauses that are islands for
extraction (cf. A. Elordieta (2002, 2004, 2006) and Arregi (2007) for data and discussion.

87
2.3. Semantic Properties of Focus in Basque

In this section I analyze the nature of focus regarding its semantic contribution
to the interpretation of the sentence. There is a general debate in the literature
concerning the nature and semantic import of focus. This debate is articulated along
two main axes of (dis)agreement (cf. i.a., Krifka (2004), Molnár & Winkler (2006)):

(i) The presuppositionality / non-presuppositionality of focus.

(ii) The informative / contrastive nature of focus.

In this section I will discuss both axes and argue that there is a clear interpretive
distinction between out-of-the-blue statements and contextualized statements12: out-of-
the-blue statements do not trigger any focal presupposition nor contrast. On the other
hand, contextualized utterances have always an exhaustive interpretation and trigger
a focus induced presupposition. I will start in section 2.3.1 with an analysis of the
issue of whether focus induces presuppositions and leave the question of the
contrastive nature of focus for section 2.3.2.

2.3.1. Focus and Presupposition

In this section I analyze the issue of focus-induced presuppositions. I start

with a brief review of the literature concerning focal presuppositions (2.3.1.1) and
then I argue in favor of a presuppositional account of the meanings introduced by
focus (2.3.1.2). Then, in section 2.3.2, I analyze the question of the contrastivity of
focus. I will clarify the point made in 2.3.1.2. and argue that out-of-the-blue statements
and contextualized statements behave in different ways regarding both their
constrastivity and the focal presuppositions that they introduce: out-of-the-blue
statements won’t trigger any focally induced presupposition and their interpretation
will be just informative, whereas contextualized statements do trigger focal
presuppositions and contrastive readings.

12As mentioned earlier, traditionally out-of-the-blue sentences are also known as ‘broad foci’ whereas
contextualized utterances correspond to ‘narrow foci’.

88
2.3.1.1. From Presuppositions to P-sets
The trend of analysis in generative grammar relating focus to presupposition
dates back at least to Chomsky (1970, 1976). When analyzing A’-dependencies,
Chomsky (1970, 1976) proposes that focus creates a binary partition of a clause into
the focal phrase and a focally induced presupposition. According to his proposal, the
sentence in (22a) with focus on the subject would have the underlying partition in
(22b). Likewise, the sentence in (23a) with focus on the object would have the
partition in (23b):

(22a) [John]F writes poetry.

(22b) John [is the one who writes poetry].

(23a) John writes [poetry]F.

(23b) Poetry [is what John writes]

As can be observed, the intuition about focally induced presuppositions is

rather strong; (22b) and (23b) constitute appropriate paraphrases of (22a) and (23a),
respectively. Furthermore, assuming such a partition, we could imagine that it is
obtained via a covert transformation that extracts the focus XP from its base position
and replaces it with a variable, as in (24), which would be the transformation behind
(22a)-(22b):

(24) John(x) [ x writes poetry]

Moreover, Chomsky remarks that this extraction is overtly realized in many

discourse-configurational languages. One such language would be Basque, where the
focal XP appears dislocated from its first-merge position and surfaces at the left edge
of the clause (cf. sections 2.1. and chapter 4 of this dissertation for discussion). In fact,
a good prediction of such an extraction analysis was that it captured

89
straightforwardly the striking crossover effects observed in sentences with focus, like
the one in (25):

(25) The woman hej/*i loved betrayed [Johni]F.

In this sentence, the focal XP ‘John’ cannot be understood as correferential

with the pronoun ‘he’. If, as Chomsky (1976) defends, we take focus to be an operator
that undergoes leftward movement (in the case of English, a covert one), in (25) we
would have the crossover configuration in (26a), which is absolutely parallel to the
one that obtains with Wh operators that undergo overt movement in English (cf.
(26b)):

(26a) *Opi pronouni ti

(26b) *Whoi did the woman hei loved betray ti?

Thus, the presuppositional analysis of focus which began from a mere

semantic observation about its interpretation gains plausibility from different
syntactic facts: (i) the typological evidence of overt displacements of the focus phrase
observed in discourse-configurational languages, and (ii) the behavior that we expect
from those displacements with respect to weak crossover effects, among others.
In this line, Jackendoff (1972) makes one of the most influential analyses of the
semantics of focus. He starts out from the intuition in Chomsky (1970, 1976), taking
focus to introduce an existential presupposition. Jackendoff (1972) formalizes this
notion proposing that the presuppositions introduced by foci are bona fide existential
presuppositions. Thus, according to this analysis, the sentence in (27) would
introduce the presupposition in (28):

(27) John bought [potatoes]F

(28) ∃x [John bought x]

90
However, even if the intuition of focus presuppositions is strong, and its
analysis elegant, until recently the mainstream assumption in the semantic literature
has been that focus does not trigger presuppositions (cf. i.a, Rooth (1985), Krifka
(1991)). This is due to the fact that if we take focus to trigger presuppositions, we
immediately run into problems when we turn to negative contexts. In fact,
Jackendoff (1972) notes that as soon as we build an example with focus on a negative
polarity item like the one in (29), the analysis based on focal presuppositions
becomes problematic:

(29) John bought [nothing]F

It is evident that the assertion of (29) would be contradictory with any type of
representation along the lines in (28): if the proposition expressed by (29) is true, then
there is no thing such that John bought it, and hence, the existential presupposition of
(28) is not expressed in (29). Thus, Jackendoff (1972) concludes that the claim that a
sentence involving focalization can be partitioned into a focus and a presupposition
is too strong. Rather, he proposes to analyze focus-induced meanings as weaker
objects with no existential import, like the λ-abstract in (30):

(30) λx [John bought x]

Jackendoff (1972) postulates that this λ-abstract provides what he calls a ‘p-set’,
where the ‘p-set’ is characterized as follows:

(31) is a coherent set

is well defined in the present discourse
λx Presupps (x)
is amenable to discussion
is under discussion

According to Jackendoff’s (1972) analysis, this characterization has all the

virtues of the presuppositional analysis in that the meaning induced by focus has a

91
familiarity reading (it is something well defined in the present discourse, amenable to
discussion, and under discussion), but without the ontological compromise of an
existential quantification.
Extending the jackendoffian non-presuppositional analysis of the p-set, Rooth
(1985) proposes a very successful and productive analysis of the semantics of focus.
He proposes that a sentence with focus has two denotations: the ‘Ordinary Semantic
Value’ [[Φ]]O, the proposition obtained by compositional semantics and without any
influence of focus, and the ‘Focus Semantic Value’ [[Φ]]F, a set of propositions obtained
by the substitution of the focal XP with discursively available alternatives that match
it in semantic type (basically, the denotation of the question that the sentence could
answer in a Hamblin-Karttunen semantics of questions (cf. Hamblin (1973),
Karttunen (1977)). For instance, the example in (32a) with focus on the subject would
have the two denotations in (32b):

(32a) [Mary]F loves Paula.

(32b) [[Φ]]O: {love(m, p)} = [[Mary loves Paula]]
[[Φ]]F: {love(x, p)| x ∈ E} = {[[Mary loves Paula]], [[John loves Paula]], [[Peter
loves Paula]], [[Sarah loves Paula]], [[George loves Paula]]...}

The semantic role of focus, thus, is to raise the set of possible alternatives;
computing the two semantic values of a sentence (its ordinary semantic value and its
focus semantic value) we get the contrastive nature of focus.
Finally, Krifka (1999) proposes to implement the p-set with an assertion
operator that cancels the availability of the alternative values. This operator is
sketched in (33):

(33) ASSERT(M, A, c) (a sentence with meaning M and alternatives A in a context c is

asserted):
- the speaker claims M (in c).
- for every alternative M’ ∈ A, M’ ≠ M, the speaker explicitly does not
claim M’ (in c).

92
This assertive operator is what will cancel the rest of alternatives, providing
focus with its core contrastive interpretation.
Summarizing the discussion so far, the current general trend in the semantic
literature on focus, stemming from Jackendoff’s (1972) work, is that focus does not
trigger any presupposition. This point of view, however, has been questioned
recently.

2.3.1.2. Focal Presuppositions Revisited

Recently, Geurts & van der Sandt (2004) have challenged this view on focally
induced presuppositions claiming that, in fact, foci do trigger presuppositional
meanings. They propose the following Background-Presupposition Rule:

(34) Background-Presupposition Rule: Whenever focusing gives rise to a

background λx.ϕ(x), there is a presupposition to the effect that λx.ϕ(x)
holds of some individual.

One of their arguments in favor of this presuppositional analysis is based on

the fact that the meanings introduced by foci behave like other well known
presuppositions regarding their variability under filters, holes and plugs. For instance,
conditional constructions have been described as being presupposition ‘filters’ (cf. i.a.
Karttunen (1973, 1974)), where the presuppositions of a proposition of the form “If A,
then B” (with respect to a context C) will be the ones of (35):

(35) (i) All the presuppositions of A (with respect to C)

(ii) All the presuppositions of B (with respect to C) except those entailed by
the set C∪A and not entailed by B alone.

Thus, in both sentences of (36a-b), for instance, we have the presuppositions

triggered by the definite description ‘Fred’s wife’ and the focal presupposition ‘that

93
someone stole the tarts’ (cf. von Fintel (1994), Rooth (1999), Kadmon (2001) and
Geurts & van der Sandt (2004)):

(36a) If [Fred’s wife]F stole the tarts, then Fred is innocent.

√ Presupp: Fred has a wife.
√ Presupp: Someone stole the tarts.

(36b) If Fred is innocent, then [his wife]F stole the tarts.

√ Presupp: Fred has a wife.
√ Presupp: Someone stole the tarts.

However, since conditionals are presupposition filters, in the cases of (37a-b), the
relevant presuppositions are “absorbed” by the information in the antecedent:

(37a) If Fred has a wife, then Fred’s wife stole the tarts.
# Presupp: Fred has a wife.

(37b) If someone stole the tarts, then [Fred’s wife]F stole the tarts.
# Presupp: Someone stole the tarts.

This is a remarkable fact that shows that an analysis of focus as triggering

presuppositions is on the right track. In fact, we can even extend the argument of
presupposition filters to counterfactual conditionals, constructions that imply that
the content of their antecedent does not hold. For instance, in a counterfactual
sentence like “If I were you, I would do corpus studies’ it is somehow expressed that
‘that I am you is false’. Now, consider the scenario in (38), previously discussed in
some form or another in Dretske (1972), von Fintel (1994) and Rooth (1999):

(38) SCENARIO: Pat had two daughters, one named Bertha; the other was named
Aretha and was indispensable to him in his business. He had made a
commitment to marry one of the daughters to one of the sons of a man who

94
once saved his life. There were two such sons, the elder son Clyde and the
younger son Derek. According to a custom of the society and period, an elder
son had to marry before his younger brothers; this was known as seniority.
Given the contract, seniority, and the desirability of leaving Aretha free to run
his business for him, he figured out that the best thing to do was to marry
Bertha to Clyde, and that is what he did.

Now, given the scenario in (38), counterfactual conditionals vary in truth

conditions depending on the placement of focus. Consider the cases of (38) with their
corresponding truth values:

(38a) If he hadn’t married [Bertha]F to Clyde, Aretha couldn’t have continued to run
the business.
TRUE
(38b) If he hadn’t married Bertha [to Clyde]F, Aretha couldn’t have continued to run
the business.
FALSE
(38c) If he hadn’t married [Bertha]F to Clyde, seniority would have been violated.
FALSE
(38d) If he hadn’t married Bertha [to Clyde]F, seniority would have been violated.
TRUE

As can be observed, depending on the placement of focus the sentences state a

different reason for the content expressed by the conclusion. Crucially,
notwithstanding the antipresupposition triggered by the counterfactual conditional,
we still get the presupposition triggered by the antecedent in (38a), i.e. ‘that he
married someone to Clyde’. In fact, the same content can be expressed with
statements of reason:

95
(38a’) The reason he married [Bertha]F to Clyde was that Aretha was indispensable in
the business.
TRUE
(38b’) The reason he married Bertha [to Clyde]F was that Aretha was indispensable in
the business.
FALSE
(38c’) The reason he married [Bertha]F to Clyde was that he wanted to obey seniority.
FALSE
(38d’) The reason he married Bertha [to Clyde]F was that he wanted to obey seniority.
TRUE

And, furthermore, we can combine both counterfactual conditionals and

statements of reason to make an explicit assertion like, for instance, (38a’’):

(38a’’) The reason he married [Bertha]F to Clyde was that if he hadn’t married
[Bertha]F to Clyde, Aretha couldn’t have continued to run the business.
TRUE

The observation here is that the presupposition triggered in the antecedent of

the conditional (38a) (‘that he married someone to Clyde’) is coherent and persistent
in both assertive paraphrases ((38a’) and (38a’’)).
The conclusion that follows from these facts is that foci do trigger
presuppositions. Obviously, now the cases of ‘presupposition failure’ of negative
contexts ask for an explanation (i.e., sentences like (29) above). Geurts & van der
Sandt (2004) propose to overcome this problem by analyzing these contexts as cases
of focus on the polarity of the proposition or ‘verum foci’. On the basis of Herburger
(2000), I shall give an alternative analysis of these puzzles in chapter 4.

96
2.3.2. Focal Contrast: Information Focus vs. Identification Focus:
Contrastivity and its grammaticalization are among the most controversial
issues concerning information-packaging. Szabolcsi (1981) proposed that in
Hungarian, a SVO language, there are two types of foci in different syntactic
positions: (i) a ‘new information focus’ realized in situ, and (ii) an ‘identification
focus’ realized in a left dislocated position (cf. as well Kiss (1998)). To distinguish
semantically among both types, she proposes the two tests that are summarized
below:
(i) First Exhaustivity Test: Given a pair of sentences in which the first sentence
contains the coordination of two XPs as the focus, and the second sentence differs
from the first one in that it only has one of the two XPs as the focus, if the second
sentence is not a logical deduction from the first one, then the focus of the second
sentence has an exhaustive interpretation.
For instance, consider the pair of sentences in (39a), where the object has
dislocated from its postverbal position. The sentence pair is awkward because the
second one is not a proper continuation of the first one:

(39a) Mari [egy kalapot és egy kabátot]F nézett ki magának.

Mari a hat and a coat picked out herself-to
“It was a hat and a coat that Mary picked out for herself”

#Mari [egy kalapot]F nézett ki magának.

Mari a hat picked out herself-to
“It was a hat that Mary picked out for herself”

However, according to Szabolcsi (1981), leaving the direct object in situ makes
the sentence pair completely natural:

(39b) Mari ki nézett magának [egy kalapot és egy kabátot]F.

Mari out picked herself a hat and a coat
“Mary picked out [a hat and a coat]F for herself”

97
Mari ki nézett magának [egy kalapot]F.
Mari out picked herself a hat
“Mary picked out [a hat]F for herself”

The second exhaustivity test is the mirror image of the first one:
(ii) Second Exhaustivity Test: It consists on building a pair of sentences in which
the first sentence has under negation an XP as the focus, and the second sentence
differs from the first one in that it has the coordination of two XPs as the focus (one
of them the same XP of the first sentence) and there is no negation. Given this setting,
if the second sentence is a natural continuation of the first one then the focus has an
exhaustive interpretation.
Consider the cases in (40a)-(40b):

(40a) Janos nem [Marit]F hívta meg. [Marit és Évát]F hívta meg.
Janos neg Mari invite part Mari and Eva invite part
“It was not [Mary]F that John invited. It was [Mary and Eve]F that John Invited”

(40b) #Janos nem hívta meg [Marit]F. Meghívta [Marit és Évát]F.

Janos neg invite part Mari. invite Mari and Eva
“John did not invite [Mary]F. He invited [Mary and Eve]F”

(40a) with left-peripheric focus is a coherent sentence pair but (40b) is not; the
second sentence of (40a) is a natural correction of the first one but in (40b) such a
correction has no sense. Thus, Szabolcsi’s conclusion is that, as we saw with the first
test, in Hungarian left peripheric focus is interpreted contrastively whereas in situ
focus is interpreted as new information.
Applying the first test to Basque data we see that the content of the second
sentence with a single-XP focus is not entailed by the first one; rather, an assertion of

98
the second sentence after the first one seems to contradict it or to be in disagreement
with it13:

(41) [Txapel bat eta soineko bat]F erosi zituen Mirenek.

hat a and coat a buy aux Miren
“Miren bought [a hat and a coat]F”

#[Txapel bat]F erosi zuen Mirenek.

hat a buy aux Miren
“Mary bought [a hat]F”

In fact, the contrastivity of focus can be tested with the agreement or

disagreement nature of both utterances: in (41) it would be natural to introduce the
second sentence with a disagreement denial (41’) but it would be unnatural to
introduce it with an agreement assertion (41’’):

(41’) [Txapel bat eta soineko bat]F erosi zituen Mirenek.

hat a and coat a buy aux Miren
“Miren bought [a hat and a coat]F”

Ez, [Txapel bat]F erosi zuen Mirenek.

No hat a buy aux Miren
“No, Mary bought [a hat]F”

(41’’) [Txapel bat eta soineko bat]F erosi zituen Mirenek.

hat a and coat a buy aux Miren
“Miren bought [a hat and a coat]F”

13 A. Elordieta (2001) reports different judgments from mines for Szabolcsi’s (1981) tests. G. Elordieta

as well (p.c.) tells me that he accepts the sentence pair in (41) as natural. It might be a dialectal
variation since both A. Elordieta and G. Elordieta are speakers of NBB and other Central Basque
speakers like me share my judgments. In any case, all speakers share the judgments reported for (43)-
(45).

99
#Bai, [Txapel bat]F erosi zuen Mirenek.
Yes hat a buy aux Miren
“Yes, Mary bought [a hat]F”

Likewise, the second test also shows the contrastive interpretation of focus in
Basque, since the second sentence is a coherent continuation of the first one.

(42) Ez zuen [Miren]F gonbidatu Jonek. [Miren eta Amaia]F gonbidatu zituen.
not aux Miren invite Jon Miren and Amaia invite aux
“It was not [Mary]F that John invited. It was [Mary and Eve]F that John Invited”

Clearly, in a discourse like (42) the second sentence is interpreted as a

correction of the first one or a disagreement with it.
In fact, the picture appears more clearly with plural objects. Imagine a
situation where I have five children and I am asked the question in (43a). If I answer
(43b), I am not just telling something not very informative; rather, I am lying:

(43a) Zenbat seme-alaba dituzu?

how many children have
“How many children do you have?”
(43b) [Hiru]F ditut.
three have
“I have [three]F children”

If the semantic contribution of focus were to add information non-

contrastively, (43b) would be true, but according to native speakers intuitions (43b) is
not true. The same thing happens with cumulative readings: if a stone of 325 kilos is
lifted by three friends it is blatantly false to claim something like (44b) as an answer
to (44a) (providing the answerer is one of the lifters):

100
(44a) Nork jaso du harria?
who lift AUX stone
“Who lifted the stone?”
(44b) [Nik]F jaso dut.
I lift AUX
“[I]F lifted it”

The reason why both (43b) and (44b) are false is that focus in Basque has an
exhaustive interpretation, like the left-dislocated foci in Hungarian.
In fact, this intuition gets reinforced if we test focused XPs in contexts that
require a non-exhaustive reading. For instance, in the answer sentence in (45B) the
adverb behintzat (at least) which, given its lexical force, requires a non-exhaustive
reading, turns the sentence inappropriate with focus on the subject14:

(45) A: Nork eskatu du ardoa?

who order AUX wine
“Who ordered wine?”
B: #[Nik]F behintzat eskatu dut.
I at least order AUX
“At least [I]F ordered wine”

This, arguably, is due to the inherent contrastive interpretation of focus in

Basque. The conclusion that derives from this discussion is the following:

(46) Tentative Conclusion: Focus in Basque is always exhaustive and triggers focal
presuppositions.

However, this conclusion fails to be general enough, and it will have to be

revised in order to accommodate the patterns exhibited by out-of-the-blue statements.

14 A sentence like (45B) with a different focal structure might be correct, in particular, a sentence with

prominence on the verb and a verum focus interpretation.

101
These constructions, where the whole sentence is under focus, do not seem to have
any exhaustive interpretation, and, clearly, there is no focal presupposition either.
Consider for instance the out-of-the-blue statement in (47)

(47) [Jonek patatak erosi ditu]F.

Jon potatoes buy AUX
“[Jon bought potatoes]F”

Here, there doesn’t seem to be any contrast, and there is no focal

presupposition either since the sentence was uttered in the absence of any previous
discursive context. Another difference between out-of-the-blue and contextualized
utterances is that in the former type of utterances there is no evidence to postulate
any syntactic focal movements. Thus, the pattern is the following one:

(48) Contextualized Utterances:

[Patatak]F erosi ditu Jonek. Focus Movement
Potatoes buy AUX Jon Contrastivity
“Jon bought [potatoes]F” Presuppositionality

(49) Out-of-the-blue Utterances:

[Jonek patatak erosi ditu]F. Focus in situ
Jon potatoes buy AUX New Information
“[Jon bought potatoes]F” No Presupposition

Clearly, the tentative conclusion made in (46) will have to be amended in

order to accommodate for the patterns of out-of-the-blue sentences, as in (50):

102
(50) Generalization on the Possible Interpretations of Focus: in contextualized utterances
like (48), the focus has a contrastive interpretation, there is a focal
presupposition, and the focal XP is left dislocated, whereas in out-of-the-blue
statements like (49), the focus has a non-contrastive interpretation, there is no
focal presupposition, and there is no focus movement.

2.4. Summary and a Question Concerning the Different Types of Foci

In this chapter I have presented the basic syntactic, phonological and semantic
properties of focus constructions in Basque. I showed that focus in Basque involves
different types of syntactic structure and that it has different semantic interpretations
depending on whether it is a broad focus of the out-of-the-blue type or a narrow focus
(a contextualized utterance). I showed that in contextualized utterances, the focus has a
contrastive interpretation, there is a focal presupposition, and the focal XP appears at
the left-edge of the clause immediately followed by the verb. Any material preceding
the focus in these constructions is interpreted as topical. On the other hand, in out-of-
the-blue statements the focus (i.e., the whole sentence) has a non-contrastive
interpretation, there is no focal presupposition, and there is no evidence of any
focally induced movement. As we saw in section 2.2.1, the distinction between the
two types of sentences is also phonologically realized in the variety of Central
Basque: the nuclear accent associated with out-of-the-blue utterances is H* whereas the
nuclear accent of the foci of contextualized utterances is H*+L. So we see a clear
distinction between both types of constructions in syntactic, phonological, and
semantic terms.
The main question that emerges now is whether there are different types of
foci or whether these differences are epiphenomenal. But we have an underlying and
essential question that we have to analyze first: what it means to talk about different
types of foci. In the literature there are several proposals defending that focus has a
dual nature (most of them making a distinction of contrastive vs. new information
focus). As we saw in the preceding section, in Basque as well there is a distinction
among out-of-the-blue and contextualized utterances in syntactic, semantic and
phonological terms. The question that arises here is the following one: do we need to

103
postulate different types of foci as such or can these differences be derived from
other factors?
My concern here is that postulating different types of foci begs the question of
what are the types. I believe that any work that postulates different types of foci as
different information packaging notions should address the following questions:

(51) Questions:
(i) What are the theoretical primitives?
(ii) Is there any subset/superset relation?
(iii) Can we have different types of foci in the same sentence?
(iv) Can we have different types of foci in the same phrase?
(v) Is it a matter of features?
(vi) Is it a matter of configurations?
(vii) How do we get them derivationally from the NUM to Spell Out?

I think it is important to have these questions in mind when analyzing the

variety of focus interpretations. Clearly, Ockham’s razor argument should prevent us
from postulating several theoretical primitives where one suffices, and this precisely,
is the route I will take in this dissertation when analyzing the dual grammar of focus
in chapter 4. There I will propose that the differences between out-of-the-blue and
contextualized utterances follow from different CP systems, and that the phenomenon
of different types of focus might be the epiphenomenal result of differences in the
functional structure of the left periphery of the CPs involved.

104
CHAPTER 3. A DERIVATIONAL APPROACH TO FOCUS
STRUCTURE

In this chapter I propose a Derivational Approach to F(ocus)-Structure (DAFS)

based on the dynamic Bare Phrase Structure construal presented in chapter 1. The
analysis is uphold on two main assumptions: (i) that the [+F] focal feature is an
optional formal feature that can be assigned to multiple tokens in the Numeration,
and (ii) that F-Structure is created by set composition along the theory of Bare Phrase
Structure. I introduce these two assumptions in sections 3.1 and 3.2 respectively.
Next, in section 3.3 I discuss the issue of how [+F] feature assignment should be
constrained and, finally, in section 3.4, I analyze the relation between focus and
nuclear stress in Basque and other languages.

3.1. The [+F] Feature as an Optional Formal Feature

According to Chomsky (1995b), a lexical entry is made up of three collections
of features:

(1)
(i) Phonological Features.
Lexical Item: (ii) Semantic Features.
(iii) Formal Features.

Among the formal features, there is a subset of optional features like [+plural]
that are added as the pertinent lexical item enters into the Numeration.
Within this framework, I would like to propose that the [+F] feature is one of
the optional formal features that can be interpreted both by PF and LF.
In fact, the availability of this [+F] feature might not be optional for all the
lexical items, and some lexical items might bear this [+F] feature lexically specified as
an intrinsic feature. These could be elements such as Wh-words, or the focus particles

105
of languages like Somali (cf. Lecarme (1999)) or Tuki (cf. Biloa (1995)). The other side
of this coin would be the question regarding whether there are lexical items that are
unfocussable inherently, as seems to be the case of pronominal clitics in Romance
languages1. This is completely an empirical issue. However, I think that there might
be some examples of metalinguistic focus even with clitics, as in the Spanish example
in (2) (cf. as well Artstein (2004), Potts (2007)):

(2) No se dice “me le haces”, se dice “me [lo]F haces”!

Neg cl say cl cl do cl say cl cl do
“You cannot say “me le haces” (you make it to me), you say “me [lo]F haces”!

But at any rate, if there is a restriction on the inherent unfocussability of some

lexical items, then we have to posit it for any type of theory of F-Structure, not only
for the theory to be presented here. For the Bare Phrase Structure theory of F-Structure
to be developed in this dissertation, the restriction would just be a ban on the [+F]
feature assignment for the relevant set of lexical items. Other theories like the NSR-
based theory of the F-Structure would have more problems to account for them (cf.
chapter 5 for a critical review of these theories). So I assume that, in principle, the
[+F] feature is an optional formal feature that can be assigned to any lexical item.
Now, the fact that some of them are more often marked as [+F] will depend
completely on discourse-flow and information packaging, not on their grammatical
category per se2. Thus, we will assume that all types of lexical items can bear the [+F]
feature and that the likeliness of them to be [+F] is a completely independent issue
(here again, I disagree with NSR-based theories of F-Structure in that I assume that
focusing the DP object is as economic as focusing any other XP (cf. chapter 5 for
discussion).

1Thanks to João Costa for pointing this out to me.

2 Hence, even if it is much more usual to add to our information stock expressions denoting
individuals than, say, quantifiers, this will be a completely discourse-driven fact of information
packaging: whereas it is easy (and common) to ask for the identity of an individual, it is tougher to ask
for a quantifier, and even tougher to ask for a light verb, for instance,. But there is no grammatical
restriction preventing a light verb from being focused; this being a matter of ‘informatics’ (cf. Vallduví
(1993) and section 3.3)

106
3.2. Deriving the F-Structure
Assuming that the [+F] feature is an optional formal feature, we can assume
that this feature can be assigned to multiple tokens of the lexicon selected in the
Numeration, and that the F-Structure of a sentence will be constructed out of the
tokens bearing the [+F] feature. Note here that the postulation of feature assignment
in the Numeration does not have to have any teleological flavor, and the assignment
of the [+F] feature can be completely free (like, by assumption, the assignment of the
feature [+Plural]). Thus, at a descriptive level, the lexical tokens that will ultimately
construct the F-Structure will bear the [+F] feature, whereas other tokens will not.
So the idea I would want to put forth in this section is that for an element to
bear the [+F] feature in the Numeration does not mean that that specific lexical item
will be the actual focus of the sentence, but just that it will take part in the
composition of the F-Structure: the F-Structure will be the outcome of the merger of
[+F] featured items. As an illustration, given the appropriate question/answer pair in
(3a-b), the only element in the Lexical Array of (3c) that will bear the [+F] feature in
the answer of (3b) is ‘potatoes’:

(3a) What does Ibon want?

(3b) He wants [potatoes]F.
(3c) Lexical Array (simplified) for (3b): {he, want, potatoesF}

In the same way, since no minimalist principle prohibits the insertion of

various [+F] features, and given the dialogue in (4), my proposal is that in (4b) (an
appropriate answer to the question in (4a)), we have two elements bearing the [+F]
feature: ‘potatoes’ and the verb ‘eats’, as represented in (4c):

(4a) What does Ibon do?

(4b) He [eats potatoes]F.
(4c) Lexical Array (simplified) for (4b): {he, eatF, potatoesF}

107
Given this multiplicity of [+F] featured tokens, I would like to propose that the
F-Structure is constructed as the derivation unfolds and structure is created. Hence,
instead of a PF focus projection, the F-Structure will be the outcome of phrase
structure building in Bare Phrase Structure, along the lines discussed in chapter 1.
For instance, given the configuration in (5a-b), where two [+F] marked
elements are merging together, the syntactic object that will be created will be a set
that contains only [+F] featured lexical items:

(5a) (5b)
{βF, {αF, βF}} NPF
3 3
αF βF AdjF NF

Thus, the newly created object will be read as focal by the interface
components. That is, a syntactic phrase will be focal when it is composed only by
[+F] featured lexical items. Thus, set containment will be the crucial relation for F-
Structure composition, and the label of the syntactic structure will play no role in
determining whether the phrase is focal or not (see (8) below). Likewise, as
illustrated in (6), when that object is merged with another element γ that itself bears a
[+F] feature the newly created object will be a set constituted exclusively out of [+F]
featured lexical items:

(6a) (6b)
{γF, {γF,{ βF, {αF, βF}}}} DPF
3 3
γ F {βF {αF, βF}} DF NPF
3 3
αF βF AdjF NF

On the other hand, if we have a [+F] featured element α and a [+F] featureless
element β undergoing Merge, the newly created element will not be a set containing
only [+F] featured material, and hence, the only chunk of structure to be read as focal
will be the singleton {α}. This is illustrated in (7):

108
(7a) (7b)
{γ, {γ ,{β, {αF, β}}}} DP
3 3
γ {β,{αF, β}} D NP
3 3
αF β AdjF N

As I said before, at this stage it does not matter what the label of the newly
created structure is, for the relation that is important for the F-Structure is set-
membership. Thus, even if it was β that was [+F]-marked instead of α, the F-
Structure would not project. That is, both elements undergoing merge have to be
[+F]-marked for the phrase they create to be focal:

(8a) (8b)
{γF, {γ ,{βF, {α, βF}}}} DP
3 3
γ {βF,{α, βF}} D NP
3 3
α βF Adj NF

Thus, with the adoption of this derivational theory, there will be no focus
projection as such, but just a purely syntactic F-Structure composition. This allows us
to account for the variability of F-Structure possibilities without having to postulate
optional operations of focus projection; the optionality comes for free built up in
lexical choice (a virtual conceptual necessity).
In order to show how the system works, let us say that we have the simplified
lexical array in (9)3. When the [+F]-marked object is merged with the [+F] featureless
verb the phrase created will not be focal, for it is not composed exclusively of [+F]
featured material. Thus, such a configuration would end up in a sentence like (9b)
with the direct object as its focus:

3I will simplify the presentations abstracting away from Tense and other projections that are not
directly relevant for the discussion.

109
(9a) Lexical Array: {{Mary}, {JohnF}, {kiss}, {v}}
vP
qp
Mary v’
3
kissed VP
2
tv JohnF
(9b) Mary kissed [John]F

Right in the same way, if we have the lexical array in (9a), when the object
bearing a [+F] feature is merged with the verb that itself bears the [+F] feature, the
VP created will contain only [+F] featured lexical items4. Hence, since there is no
other [+F] featured item on the lexical array of (9a), the F-Structure will be a VP, as in
illustrated in (9b):

(9a) Lexical Array: {{Mary}, {JohnF}, {kissF}, {vF}}

vP
qp
Mary v’F
3
kissedF VPF
2
tv JohnF

(9b) Mary [kissed John]F

Instead, if we have the lexical array in (10a), when the object and the verb are
merged, a new syntactic/set theoretic object is created out of elements that bear the
[+F] feature. Once this object is merged with the light verb and the new element with
the subject that itself bears the [+F] feature, vP focus is obtained:

4 I assume here for the ease of exposition that the light verb itself also bears the [+F] feature, but note

that, given the complex event structure denoted by transitive predicates this is not a necessary
assumption for the DAFS: if the little v is not [+F] marked, it just will not be part of the F-Structure,
and hence the semantic contribution of the F-Structure will be different.

110
(10a) Lexical Array: {{MaryF}, {JohnF }, {kissF}, {vF}}
vPF
qp
MaryF v’F
3
kissedF VPF
2
tv JohnF

(10b) [Mary kissed John]F

Thus, the node with the highest [+F] feature will always demarcate the F-
Structure of the sentence, i.e., a XP will be interpreted as focal when all the lexical
items that build up that XP are [+F] marked.
Then, the F-Structure will be just a product of the standard bottom-up
semantic composition. No matter which lexical items enter the derivation with the
[+F] feature, the content of the F-Structure will be the outcome of their syntactic
composition. For instance, starting with the lexical array in (11a) that will create the
sentence (11b), the derivational construal of the sentence (11c) fixes the semantic
import of the F-Structure:

(11a) Lexical Array (simplified): {Gorka, buy[+F], v[+F], cider[+F]}

(11b) Gorka [bought cider]F

(11c) vP e[Agent(e, Gorka) & Buy(e) & Theme(e, cider)]

qp
Gorka(x) v’ e[Buy(e) & Theme(e, cider)][+F]
qp
v[+F] V’ F-Structure
qp
buy(e)[+F] cider(y)[+F]

This derivational construal, as I said before, fixes the size of F-Structure in

narrow syntax. Thus, contrary to the widely accepted NSR-based theories of F-
Structure, under the proposal I defend here focus has an unambiguous PF and logical

111
form interpretation.5 In particular, according to my proposal (and contrary to the idea
of focus set of the NSR-based theories), in (12-14) we will not have a single derivation
with an ambiguous F-Structure, but rather three different sentences derived from
different Numerations and each one of them with a different F-Structure. The
question is that all of them end up with the same linearization (#Johnˆboiledˆwater#)
and nuclear stress falls on the final word:

(12) John boiled [water]F

(13) John [boiled water]F
(14) [John boiled water]F

Now, the questions that should be investigated are how we could account for
the interface properties of focus that we saw in chapter 2. I will leave issues of logical
form interpretation for the next chapter, where I analyze the syntax-semantics
interface of focus. Here I will look at evidence from PF. First, I will address the
question of F-Structure licensing (section 3.3), and then I provide an analysis of the
nuclear stress placement that predicts the position of the nuclear stress in accordance
with the nature of the syntactically fixed F-Structure (section 3.4).

3.3. F-Structure Licensing

I just showed that the Derivational Approach to F-Structure (DAFS) allows fixing
the F-Structure of a sentence in narrow syntax given that the F-Structure will be the
outcome of the syntactic composition of [+F] featured lexical items. The question that
arises now is how to constrain [+F] feature assignment. From an I-language
perspective, I think that it makes little sense to ask the question of why some token of
the Numeration bears the [+F] feature while others do not (cf. also section 3.2.1). This
would be tantamount to asking why in the sentence (15) the subject is ‘John’ instead
of ‘Michael’:

(15) John kissed Mary.

5See chapter 4 for an analysis of the derivation from narrow syntax to logical form and chapter 5 for a
critique of NSR-based theories of F-Structure.

112
This, presumably, is a matter of lexical choice, performance, and ultimately, of
free will. Likewise for [+F] feature assignment. I will just assume that this feature is
assigned to some tokens, without any teleology, and that the derivation unfolds as
such creating the F-Structure. Then, if the derivational outcome of the combination of
the lexical items in a Numeration results in a discursively inappropriate F-Structure,
this will be a matter of lack of discourse coherence. In particular, according to its F-
Structure, an out-of-the-blue sentence will be appropriate in situations where there is
no previous linguistic context. A contextualized utterance, on the other hand, will need
a more explicit discourse-setting in order to be appropriate. As I argued in chapter 2,
a contextualized utterance induces a focal presupposition, so the context where the
utterance is introduced will have to instantiate the presupposition that the utterance
brings with it. Otherwise, the presupposition will have to be accommodated (cf.
section 4.2.2.2.2). But that is a matter of pragmatics and discourse. Quite the same for
(15), if the proposition it denotes happens to be contrary to fact. If this is the case and
it is Michael that kissed Mary, the proposition expressed by (15) is just false, but no
syntactic violation occurs and no malformation of the Numeration is at stake.

3.4. Focus and Nuclear Stress

In this section I will analyze the relation between focus and nuclear stress in
Basque and other languages. As we just saw, under the proposal of the DAFS I have
put forth, the construction of the F-Structure precedes and is totally independent
from any accent placement. This, I will argue, has welcome predictions (cf. especially
chapter 5). However, since as we saw in chapter 2 focus is somehow correlated with
nuclear accent in Basque, we need to provide an analysis of the relationship between
F-Structure and nuclear stress.
In section 3.2.4.1 I will review the basic facts that we observed in section 2.2.
Then, in section 3.2.4.2 I give an overview of the main approaches to this type of
data. Finally, in section 3.2.4.3 I provide my own analysis which is based on (a) the
DAFS that sets F-Structures in the narrow syntax and (b) Cinque’s (1993) theory of
nuclear stress assignment.

113
3.4.1. Nuclear Stress Placement in Basque
We saw in chapter 2 that focus tends to be correlated with the assignment of
nuclear stress in Basque6. The pattern that we saw is the following one: in out-of-the-
blue sentences nuclear stress falls on the last word to the left of the verb. Observe the
examples in (16a-d):

(16a) [Jonek [LIBURUA]DP eraman du]F

Jon book take AUX
“[Jon took the book away]F”

(16b) [Jonek [liburu LODIA]DP irakurri du]F

Jon book thick read AUX
“[Jon read the thick book]F”

(16c) [Jonek [bere LIBURUA]DP irakurri du]F

Jon his book read AUX
“[Jon read his book]F”

(16d) [Jonek [bere liburu LODIA]DP irakurri du]F

Jon his book thick read AUX
“[Jon read his thick book]F”

Likewise, in contextualized utterances nuclear accent tends to appear on the

rightmost constituent within the focalized element:

6 Abstracting away from the examples of lexically unaccentable words in NBB (cf. section 3.4.4)

114
(17a) [Liburu LODIA]F irakurri du Jonek
book thick read AUX Jon
“Jon read [the thick book]F”

(17b) [Bere LIBURUA]F irakurri du Jonek

his book read AUX Jon
“Jon read [his book]F”

(17c) [Bere liburu LODIA]F irakurri du Jonek

his book thick read AUX Jon
“Jon read [his thick book]F”

In all these cases the whole DP object is focused, and nuclear stress appears in
the immediately preverbal element within the DP.
However, there are some cases of narrow focus that do not follow this pattern.
In particular, nuclear stress can be assigned to a word that is not immediately
preverbal. This is the case, for instance, when we have focus on the specifier of a DP,
which, due to the opaque/island nature of the DP, cannot appear singled out from
the rest of the DP. Consider e.g., the example of (18) with narrow focus on the
possessor:

(18) [[BERE]F liburu lodia]DP irakurri du Jonek

his book thick read AUX Jon
“Jon read [his]F thick book”

So, the descriptive generalization that we arrive at is the following one:

(19) Generalization on Nuclear Stress Placement in Basque: In out-of-the-blue statements

nuclear stress falls on the last constituent before the verb and in contextualized
utterances it appears on the last constituent within focus.

115
As we just saw, in out-of-the-blue SOV sentences in Basque, nuclear stress falls
on the direct object. This pattern, however, is not specific to Basque and many
languages show similar data assigning nuclear stress to the direct object in out-of-the-
blue statements. For instance, in English nuclear stress is assigned to the direct object
in an out-of-the-blue S-V-O sentence like (20), just like in Basque (21):

(20) [John bought WATER]F

(21) [Jonek URA erosi du]F

Jon water buy AUX
“[John bought water]F”

The pattern is quite general and there is a strong tendency for both SOV
languages like Basque and SVO languages like English to assign nuclear stress to the
direct object in out-of-the-blue statements (cf. i.a., Cinque (1993)). So, my goal now is to
analyze the distribution of nuclear stress as deriving from a general Nuclear Stress
Rule (NSR). Thus, in the next section I provide an overview of the theories that try to
account for the relation between sentence stress and focus and then in section 3.4.3 I
provide my own analysis based on the DAFS I presented in sections 3.1 and 3.2.

3.4.2 Focus and the Nuclear Stress Rule

Although the most widely accepted version of the NSR stems from Cinque’s
(1993) seminal work, the analysis of the relations between sentence syntax and
sentence phonology has a long tradition. Since the pioneering work on sentence
stress and juncture in Chomsky, Halle & Lukoff (1956), it is widely accepted that
sentence structure influences the distribution of sentence accents. Thus, in order to
account for the generalizations regarding the accentuation pattern of English,
Chomsky & Halle (1968: 90) propose a rule that will assign a sentence-level stress to
the rightmost accented word. This is what they call the Nuclear Stress Rule (NSR):

116
(22) The Nuclear Stress Rule I (Chomsky & Halle (1968)):

V → [1 stress] / [# # X Y # #]
1 stress
where Y contains no vowel with the feature [1 stress]

Basically, Chomsky & Halle’s (1968) NSR assigns primary stress to the last
accented syllable of a string. This rule describes nicely the basic pattern of
accentuation in English, but as is clear, it makes no reference to focus and it cannot
explain why, for instance, focalized subjects get nuclear stress.
In the light of more complex data like (23b), which constitutes a minimal pair
with (23a), Bresnan (1971) suggested an ingenuous revision of the rule: to make the
phonological cycle apply upon each syntactic transformational cycle.

(23a) George has plans to LEAVE.

(i.e., that George is planning to leave)

(23b) George has PLANS to leave.

(i.e., that George has some plans which he intends to leave)

According to Bresnan (1971) the difference between (23a) and (23b) is that the
verb to leave in (23a) does not have any direct object but the verb to leave in (23b) does.
Assuming that at D-Structure the direct object of the verb in (23b) (i.e., plans) was in
its canonical postverbal position, we can capture the position of nuclear stress by
assuming that the NSR applies both before and after the “object shift” that provides
the relative construction in (23b).
Bresnan’s solution is elegant and simple, and it allows having accent on the
object in both preverbal and postverbal positions but still, the NSR is designed to
account for “unmarked cases”, i.e, out-of-the-blue sentences with no effect of focusing

117
(as Bolinger (1972) pointed out)7.
A different approach to sentence-level stress assignment, which partially
captures the focal phenomena, is proposed in Gussenhoven (1983a): the Sentence
Accent Assignment Rule (SAAR). The SAAR is an algorithm that operates over focus
domains (one or more constituents whose [+focus] status can be signaled by a single
accent) and assigns an accent to every domain. The rule can be formulated as
involving two steps: one forms focus domains (24a), and the other one assigns an
accent to every domain (24b). In the SAAR representation in (24) A, P and C stand for
Argument, Predicate and Condition (=adverbials), respectively. Y and X stand for
any of these. Underlining of an element denotes its [+focus] status and the square
brackets delimit the focus domain formed by the DOMAIN ASSIGNMENT algorithm in
(24a). The star in front of an element indicates sentence accent as assigned by (24b)8.

(24) Sentence Accent Assignment Rule (Gussenhoven (1983a)):

a) DOMAIN ASSIGNMENT: P(X)A Ö [P(X)A]
A(X)P Ö [A(X)P]
Y Ö [Y]
b) ACCENT ASSIGNMENT: [ ] Ö [ * ]. In AP/PA, accent A.

Thus, the SAAR allows us to correctly derive the following accent

assignments, and their correspondence to focus domains (cf. Gussenhoven (1984:
28)):

7 Recently these ideas of a cyclic computation of stress have been renewed by Kahnemuyipour (2004)
and Wagner (2005).
8 In Gussenhoven’s (1984) formulation (cf. Gussenhoven (1984: 28)) the star is placed over the element,

something that I was unable to do due to typographical reasons. Thus, I was forced to locate the star
left-adjacent to the element intended to bear the stress, that is, [‫٭‬AP] denotes an accented argument
followed by an unaccented predicate. In the very same way, [P‫٭‬A] will be an unaccented predicate
followed by an accented argument.

118
(25)
AP Ö [‫٭‬AP] Our dóg’s disappeared
ACP Ö [‫٭‬A][‫٭‬C][‫٭‬P] Our dóg’s mystériously disappéared
ACP Ö [‫٭‬ACP] (talking about mysteries) Our dóg’s
mysteriously disappeared
APC Ö [‫٭‬AP][‫٭‬C] Jáne’s had an accident in Lóndon
APAA Ö [‫٭‬A][PAA] (any news about Jane?) Jóhn’s
promised Jane a bike.
APA Ö [‫٭‬A][P‫٭‬A] Jóhn beats Máry!
APA Ö [‫٭‬APA] Her húsband beats her
APA Ö [A‫٭‬PA] He béats her
ACPCC Ö [A‫٭‬C][‫٭‬P][‫٭‬C] [‫٭‬C] Truman was quíetly búried in the
Indépendence in nineteen seventy-twó

The SAAR algorithm describes correctly the facts with respect to focus
structure and accent placement in English. But it does not go beyond that, i.e., it is a
good description with no explanation of why with a given F-Structure the accent falls
on a given element.
On quite different grounds, Halle & Vergnaud (1987) take again the NSR of
Chomsky & Halle (1968) and adapt it to the Metrical Phonology framework. This
theory purports that a segmental structure projects a two-dimensional array of points
called a ‘metrical grid’. Within a word, first, we mark each syllable in the word and
thus obtain the line called line 0. Then, a constituent structure is imposed on this line
and the heads of each constituent are marked on a line above it (line 1). This
produces a basic grid where the heads of constituents have two marks, whereas other
syllables in line 1 have just one grid mark. Additional rows of asterisks are obtained
imposing additional layers of structure. For instance, Halle & Vergnaud (1978) argue
that in Maranungku main stress falls on the initial syllable whereas in Warao it falls
on the penultimate. Thus, in order to derive the location of stress in both languages
we have to add a left-headed (for Maranungku) or right-headed (for Warao) grid-line
above line 1. This is represented in the metrical grids in (26a) for Maranungku and

119
(26b) for Warao, whose last syllable is extrametrical. (adapted from Halle &
Vergnaud (1987: 41):

(26a) Maranungku (26b) Warao

* · · · · line 2 · · · · * · line 2
* · * · * line 1 * · * · * · line 1
((* *) (* *)(*)) line 0 ((*) (* *) (* *)) <*> line 0

Concerning the NSR, the problem Halle & Vergnaud (1987) want to solve is
how to derive the sentence stress placement of English that in ‘neutral’ (i.e., out-of-the-
blue) sentences falls on the rightmost position. Their basic proposal for English is the
one given in (27), where “HT” stands for “head terminal” (“whether or not the head
of the constituent is adjacent to one of the constituent boundaries” (Halle &
Vergnaud (1987: 9)), and “BND” for “boundedness” (“whether or not the head of the
constituent is separated from its constituent boundaries by no more than one
intervening element” (Halle & Vergnaud (1987: 10)). The nuclear stress these authors
formulate is the following one:

(27) The Nuclear Stress Rule II (Halle & Vergnaud (1987: 264))
(a) The Parameter settings on line N (N≥3) of the Metrical Grid are [-BND,
+HT, right].
(b) Interpret boundaries of syntactic constituents composed of two or more
stressed words as metrical boundaries.
(c) Locate the heads of line N constituents on line N+1.

We can distinguish two types of components in Halle & Vergnaud’s NSR in

(27). On the one hand, (27c) provides the principles of the rule that are immutable
from language to language. On the other hand, (27a) and (27b) provide the
parameters to account for language variation. Thus, given the parametric settings of
English, the sentence in (28a) would have the metrical grid in (28b) (ex. 83, p. 265 of
Halle & Vergnaud (1987)):

120
(28a) Jesus preached to the people of Judéa.
(28b)
. . . * Line 6
(. . . * ) Line 5
. ( . . * ) Line 4
* * ( * * ) Line 3
[Jesus [preached to the [people of Judea]]]

Given this basic proposal and the parametric settings in (27) we can predict
English nuclear stress to fall on the rightmost element. In the same way, by assuming
different value-settings for the parameters of [+/- BND], [+/- HT], [RIGHT/LEFT],
and the limit on metrical boundaries, Halle & Vergnaud can account for the different
nuclear stress placements found across natural languages.
The NSR of Halle & Vergnaud (1987) is a very powerful algorithm. However,
the parametric segment of the rule seems idiosyncratic; i.e. there seems not to be any
principled reason for English to be parameterized as [-BND, +HT, right] instead of
with any other parameter value. It is precisely this idiosyncrasy problem that Cinque
(1993) tries to solve in a groundbreaking work on the syntax-phonology interface.
Departing from this metrical-parametric version of the NSR, Cinque (1993) develops
a more principled theory of stress placement that does not have to postulate different
parameter settings for different languages. Instead, the theory that Cinque develops
derives the crosslinguistic typological variability on stress placement from deeper
principles of grammatical variation. More explicitly, he derives the nuclear stress
placement from the head parameter. In this way, his proposal can account for the
variability of nuclear stress placement in typologically different languages such as
English or Japanese in a simple and unified way. Cinque notes that besides the
different placement of the nuclear stress in out-of-the-blue sentences, the head
parameter of these languages is also different (English being head-first, and Japanese
being head-final, as represented in (29)). Given the basic abstract architecture in (29),
he proposes that the NSR has to be computed cyclically, each syntactic node
constituting a cycle (where the complement branch is assumed to involve one more
grid mark) That is, the NSR that Cinque (1993) proposes will assign the nuclear stress

121
to the most embedded element in the metrical grid:

(29)
English: Japanese:
(A** B*** C****) (A** C*** B****)
3 3
(A*) (B** C***) (A*) (C*** B**)
3 3
(B*) (C**) (C**) (B*)
g g
(C*) (C*)

Cinque’s algorithm of the NSR is, then, as follows:

(30) The Nuclear Stress Rule III (Cinque (1993)):

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.

In comparison to Halle & Vergnaud’s (1987) proposal (cf. (27)), this proposal
has two main advantages: (i) on the one hand, the algorithm applies to all syntactic
phrases (and not only to those with more than one accented word); (ii) on the other
hand —and this is a major improvement of the NSR theory— there is no need to
stipulate which one is the head in each phrase, since phrase structure and depth of
embedding will account for that. The fact that nuclear stress is assigned to the most
embedded element in out-of-the-blue sentences can be derived from the metrical grid
construction, since the most deeply embedded element will be the first being
computed when building a metrical grid bottom-up. Thus, it will be the element with
most grid lines. As a result, the most deeply embedded element in syntactic structure
will also be the most deeply embedded one in the phonological metrical grid
construction.
This version of the NSR theory that Cinque (1993) calls “null theory” permits
to account for the crosslinguistic variation in a very elegant way. As Cinque argues

122
“only two general types of phrase stress systems should exist according to whether
the language is left or right branching” (Cinque (1993: 243)).
Cinque’s (1993) theory is remarkably simple and powerful, and it allows
explaining the basic pattern of nuclear stress placement in typologically different
languages. Furthermore, it captures the fact that a single accent placement can
correlate with different F-Structures (cf. (31)-(33)).

(31) John bought [water]F

(32) John [bought water]F
(33) [John bought water]F

According to Cinque (1993), this is due to the fact that grammar interprets
focus regarding where the nuclear stress is, and nuclear stress can project its
“focusness” all the way up. However, the theory falls short when predicting the
behavior of nuclear stress in other contextualized utterances, for the nuclear stress is
assigned to focus in other positions than the absolutely most embedded one, cf. (34)-
(35):

(34) [John]F bought water.

(35) John [bought]F water.

In order to make these data amenable to an explanation, Cinque (1993)

postulates an independent rule of “marked” stress placement for contextualized
utterances.
Given its simplicity and its nice predictions Cinque’s theory of stress
assignment has been extremely influential and many scholars have applied Cinque’s
NSR when analyzing focus (cf. chapter 4 and specially chapter 5). I think that
Cinque’s NSR is essentially correct but that the approach to focus derived from it is
incorrect. Thus, in the next section I will present my analysis of nuclear stress
assignment combining Cinque’s (1993) NSR with the DAFS I presented previously in
this chapter.

123
3.4.3 A Revision of the Nuclear Stress Rule: The Focal NSR
In this section I will present a proposal for reanalyzing the NSR. In particular,
I will propose a reformulation that I will call Focal Nuclear Stress Rule, or FNSR. The
idea behind this reformulation is very simple: In order to explain the fact that in all
the cases the PF component assigns nuclear stress to the focal XP it will be sufficient
to make the Cinquean-like NSR focus-sensitive, applying it within the F-Structure
already built in narrow syntax9:

(36) Focal NSR (FNSR): Assign nuclear stress to the most embedded element within
the F-Structure.

That is, the DAFS allows us to have the F-Structure fixed in narrow syntax.
Then, interface components can read it as the structural description for the
phonological and semantic rules of accentuation, phrasing, association with
operators etc. Thus, the FNSR will predict correctly and without any further
stipulation the nuclear stress placement in different positions, given that the
differences in the F-Structures are already set in the narrow syntax. For instance,
according to the DAFS, the sentences in (31)-(35) repeated here as (37)-(41) are
different sentences (i.e., different derivations constructed from different
Numerations). Thus, applying the focus-sensitive FNSR of (36) to each of the
derivations in (37)-(41), we predict the correct actual placement of nuclear stress:

(37) John bought [water]F

(38) John [bought water]F
(39) [John bought water]F
(40) [John]F bought water
(41) John [bought]F water

9It goes without saying, I am not the first one to formulate a rule of accent assignment applying to F-
Structure. For instance, Ladd (1980: 85) formulates the following “Focus Rule”: Accent goes on the
most accentable syllable of the focus constituent. See also Selkirk (1984, 1995a), among others.

124
In (37) the F-Structure is just the object ‘water’, so the assignment of nuclear
stress is trivial; the FNSR will assign nuclear stress to ‘water’. This is represented in
(37’)10:

(37’)
vP
3
John v’
3
v VP Nuclear Stress
3
buy water[+F] F-Structure

In (38), we have a VP focus (a vP, excluding the subject), and the with the
FNSR nuclear stress is predicted to be assigned to the most deeply embedded
element (the element with most grid marks) within the F-Structure, i.e., the direct
object, as represented in (38’):

(38’) F-Structure

vP
3
John v’
3 Nuclear
v[+F] VP Stress F-Structure
3
buy[+F] water[+F]

In sentence (39) the whole vP conforms the F-Structure since all the lexical
items that compose it have been merged with a [+F] feature. Then, the FNSR will
correctly predict nuclear stress assignment to the direct object. This is illustrated in
(39’):

10For clarity of exposition I will only represent the vP. I also assume that in cases in which the verb
forms part of the F-Structure, both V and v get the [+F] feature.

125
(39’)
vP
3
John[+F] v’
3 Nuclear F-Structure
v[+F] VP Stress
3
buy[+F] water[+F]

In (40), on the other hand, we have subject focus, and the FNSR will trivially
assign nuclear stress to it, without having to resort to ad hoc ‘marked’ stress shift
operations that NSR-based theories of F-Structure have to postulate (cf. chapter 5 for
discussion). With the DAFS and the FNSR nuclear stress is assigned to the most
deeply embedded element within the F-Structure. In this case, the F-Structure is the
singleton {John[+F]}, so nuclear stress is assigned to ‘John’, as represented in (40’):

(40’) Nuclear Stress

vP
3
F-Structure John[+F] v’
3
v VP
3
buy water

Finally, the case of (41) will be analogous to the previous ones, but instead of
having the subject as the focus we have focus on the verb (for ease of exposition, I
represent the verb after V-to-v movement):

(41’)
vP
3
John v’
3
F-Structure [V+v] VP
3
tV water
Nuclear
Stress

126
As we saw, nuclear stress placement is no mystery once we adopt the DAFS
that sets F-Structures in the narrow syntax and the FNSR that assigns nuclear stress
within them.
The data from Basque are explained in the same way; the F-Structure is set in
narrow syntax and PF assigns nuclear stress to the most deeply embedded element
within it. The Basque counterpart of the out-of-the-blue sentence of English in (39)
would be as in (42):

(42) [Jonek ura erosi du]F

Jon water buy AUX
“[Jon bought water]F”

Then, assuming a syntactic structure like (42’) whereby the vP is right-headed

(cf. i.a., Ortiz de Urbina (1989)), nuclear stress placement to the object is easily
explained by the FNSR: nuclear stress is assigned to the most deeply embedded
element within the F-Structure:

(42’)
vP
3
Jonek[+F] v’
3 F-Structure
VP erosi[+F]
3
ura[+F] tV

Nuclear
Stress

Likewise, a sentence with VP focus would be (43), whose F-Structure

derivation and nuclear stress assignment is represented in (43’).

127
(43) Jonek [ura erosi du]F
Jon water buy AUX
“Jon [bought water]F”

(43’)
vP
3
Jonek v’
3
VP erosi[+F] F-Structure
3
ura[+F] tV

Nuclear
Stress

In the same vein, nuclear stress placement on focal objects like (44) is trivially
explained if we assume with the FNSR that nuclear stress is assigned to the most
deeply embedded element within the F-Structure (in this case, the object alone).

(44) Jonek [ura]F erosi du

Jon water buy AUX
“Jon bought [water]F”

(44’)
vP
3
Jonek v’
3
VP erosi
3
F-Structure ura[+F] tV

Nuclear
Stress

128
Furthermore, notice that since the FNSR only computes the metrical
constituency of the F-Structure, we can also explain the placement of nuclear stress in
sentences where the focal element appears clause-initially. For the FNSR it will not be
problematic to have the focal phrase in a left peripheric position because the FNSR
will invariably assign an accent to it. Thus, the examples of (45)-(47) are explained as
naturally as the preceding ones11:

(45) [Ura]F erosi du Jonek

water buy AUX Jon
“Jon bought [water]F”

(46) [Jonek]F erosi du ura

Jon water buy AUX
“[Jon]F bought water”

(47) [Erosi]F egin du Jonek ura12

buy DO AUX Jon water
“Jon [bought]F water”

The analysis extends to complex DP data. Assuming that the structure of a

complex DP with a possessor is as in (48) (cf. chapter 2), the FNSR I proposed in (36)
correctly derives the nuclear stress placement in each of the sentences that we saw in
(16) to (18) above.

11 I will not enter here in the discussion on the syntax of focus in Basque. The reader is referred to
chapter 4 for an in-depth discussion on that issue.
12 In some varieties of Basque focus on the verb is expressed with a do-support strategy; something I

will not analyze here (cf. Haddican (2005, to appear)).

129
(48) DP
ei
Possessor D’
ei
NP Det
3
Noun Adj

All the PF component has to know is what the F-Structure is. Since the F-
Structure is fixed in the narrow syntax via the DAFS, PF will be able to assign nuclear
stress to the most deeply embedded element within it. As an illustration, the data in
(17) (repeated here as (49)) are explained naturally with the DAFS and the FNSR;
nuclear stress will always be assigned to the most deeply embedded constituent of
the focal XP that was constructed in the narrow syntax:

(49a) [Liburu LODI-A]F irakurri du Jonek

book thick-the read AUX Jon
“Jon read [the thick book]F”

(49b) [Bere LIBURU-A]F irakurri du Jonek

his book-the read AUX Jon
“Jon read [his book]F”

(49c) [Bere liburu LODI-A]F irakurri du Jonek

his book thick-the read AUX Jon
“Jon read [his thick book]F”

As an illustration of how the system I have proposed accounts for all these
cases in a unified way, let us look at the derivation of (49c) in detail. The F-Structure I
propose for (49c) is given in (50a); the DP is composed by a [+F] noun, a [+F]
adjective and a [+F] determiner13. The simplified metrical grid construction
represented in (50b) correctly predicts nuclear stress assignment to the adjective

13 The article ‘-a’ in Basque is a suffix and attaches to its preceding word in PF.

130
(again, assuming, like in Cinque (1993), that the non-projecting branch is added an
extra grid mark):

(50a) (50b)
DPF (bere* liburu** lodia****)
ei ei
PossessorF D’F (bere*) (liburu** lodia****)
ei ei
NPF DetF (liburu** lodi***) (-a)
3 3
NounF AdjF (liburu*) (lodi**)
g g
AdjF (lodi*)

Likewise, the example that we saw in (18), where the narrowly focused
possessor did not appear left-adjacent to the verb, will be explained in a
straightforward way. In this example, repeated here as (51a), the nuclear stress
assignment to the possessor is predicted correctly, for it is the most deeply embedded
element within the F-Structure, as illustrated in (51b):

(51a) [[BERE]F liburu lodia]DP irakurri du Jonek

his book thick read AUX Jon
“Jon read [his]F thick book”

(51b) Nuclear
Stress
DP
ei
F-Structure PossessorF D’
ei
NP Det
3
Noun Adj
g
Adj

Summarizing, under the Derivational Approach to F-Structure I have proposed

in this chapter (DAFS), F-Structure is fixed in narrow syntax. Provided with the

131
DAFS, the reformulated version of the NSR I have proposed (the FNSR) predicts the
basic pattern of nuclear stress placement in a straightforward way. In this section I
have limited myself to the discussion of the basic patterns of nuclear stress
assignment. Then, a relocation of nuclear stress might take place at a later
derivational stage depending on additional —purely phonological— phenomena (cf.
Selkirk’s (1986) ‘phonosyntactic’ and ‘phonological’ subcomponents). In the next
section I analyze one of these problematic constructions: the so-called “Schmerling
Examples”.

3.4.4. ‘Schmerling Examples’ and Other Difficulties

I have argued that the DAFS, combined with the modified FNSR of (36) can
account for the basic facts of nuclear stress placement within focus: the FNSR assigns
nuclear stress to an already syntactically built-up F-Structure.
Recall that the DAFS separates F-Structure composition from nuclear stress
placement. First, F-Structure is constructed in the narrow syntax and then a
phonological rule of accentuation assigns nuclear stress (this is just the opposite
order of Cinque’s (1993) theory). I think that the derivation of F-Structure
independently of nuclear stress assignment is a welcome result, since, as we will see
immediately, it allows us to address cases where focus on a given XP does not
correlate with nuclear stress placement on the most embedded element within that
XP. One such case is the so-called ‘Schmerling examples’. Schmerling (1976: 41 & ff.)
discusses some cases where nuclear stress placement on the subject of an
unaccusative verb correlates with focus on the whole clause. Observe the minimal
pair of out-of-the-blue statements in (52), referring to the actual deaths of two former
presidents of the USA; Harry S. Truman and Lyndon B. Johnson:

(52a) [Truman DIED]F

(52b) [JOHNSON died]F

According to Schmerling (1976) there is a pragmatic difference between both

sentences. In the context (52a) was uttered, even if uttered out-of-the-blue, people from

132
the USA was already expecting the death of Truman to happen at any time for it was
of public domain that he was seriously ill. On the other hand, in the case of (52b) the
death of Johnson was a surprise, so it was remarkable news. The puzzling feature of
(52b) is that nuclear stress falls on the subject although the whole clause is focal. In
the light of this type of phenomena, scholars like Selkirk (1984), Diesing (1992),
Rochemont (1998) or Godjevac (2000) have proposed that in some pragmatically
marked situations a “derived subject” (the subject of stage level, unaccusative and
passive verbs) can get nuclear stress and mark the whole clause as focused. In
particular, Diesing (1992) proposes that those subjects can lower in LF to their VP
internal base-generated position and project their focus nature higher up from there.
I would like to point out, however, that this account is problematic and does
not fit into the standard model of the architecture of grammar since LF is too late in
derivational terms to project focus from an accented item given that accentuation
takes place in PF (cf. chapter 5 for discussion on these issues).
Rochemont (1998), on the other hand, proposes an alternative account
suggesting that it is the trace left by the A-movement of the subject in its base-
generated object position that licenses the focus projection. This theory is not without
problems either, notably if Epstein et al. (1998), Lasnik (1999) and Epstein & Seely
(2006) are right in claiming that A-movement does not leave a trace. But more
importantly, the theory would predict that all traces of A-movements should behave
similarly, which, as Rochemont himself discusses, is not the case (cf. Rochemont
(1998)). Notice that any theory that purports a direct and causal relation between the
place of nuclear stress and focus will have important problems with this type of
cases. This is the case, for instance, with NSR-based theories of F-Structure like
Reinhart (1995, 2006), Zubizarreta (1998) and others (cf. chapter 5 for extensive
discussion).
Within the Derivational Approach to the F-Structure I presented here, the
analysis avoids all these problems: following mainstream assumptions, if the subjects
of these constructions are first merged in object position, the derivation of the
sentence-whole focus is straightforward: if both the object and the verb bear the [+F]
feature, when merged together a new set of [+F] featured elements will be created. So

133
the F-Structure of the sentence is built-up independently of any accent. Then the A-
movement applies. At PF the FNSR assigns nuclear stress to the most deeply
embedded element within the F-Structure. This explains examples like (52a) where
the verb gets the nuclear stress.
Recall that there are pragmatic differences between (52a) and (52b): it was
unexpected, a surprise that Johnson died. Surprise is usually expressed via a
particular exclamatory intonation. If so, it is possible that the initial location of the
nuclear stress (derived by the FNSR) gets affected by other factors, in particular, the
intonation associated with surprise. Then, the fact that in some marked cases like
(52b) it is the subject, rather than the lower verb that gets the nuclear stress could be
due to the somehow exclamatory nature of the utterance. This is consistent with
Schmerling (1976) who explicitly points out that sentences like (52b) need a highly
marked surprise context to be felicitous, for the unmarked case is that of (52a). I want
to argue that in both examples of (52) the sentence-whole F-Structure is derived
similarly. Then, my claim is that in (52b) the nuclear stress is assigned to the subject
due to its exclamatory nature; it was surprising that Johnson, of whom no one
expected to die, died14.
Summarizing, the DAFS I have defended in this chapter proposes that F-
Structure construal is prior to and independent from nuclear stress assignment.
Simply, an F-Structure construal does not entail any accent. Then, F-Structures are
sent to PF to be interpreted there by the FNSR15. The stress placement derived from
the FNSR can then be altered at PF by other factors. And this is what seems to be
going on, as a matter of fact, in cases like (52b). That is why it is not a problem to
have cases where the nuclear stress falls on an element other that the most deeply
embedded one (they might be due to some other process of stress relocation,
emphasis or so). Stress relocations will not alter the F-Structure of a sentence. These
cases can also be put on a par with other unaccented F-Structures, like the cases of
Lekeitio Basque that I discussed in section 2.2.1.2.1 and be explained in purely
phonological terms (cf. G. Elordieta (2007b) for such an account).

14
See Irurtzun (2006b) where I propose an analysis along these lines for Etxepare’s (1998) ‘contrastive foci’.
15
Only in languages that have such a rule, which is not universal (see section 5.2.1.1.3).

134
3.5. Conclusions
In this chapter I have developed a theory of F-Structure composition within
the Bare Phrase Structure theory of Chomsky (1995a). I have shown that the Bare
Phrase Structure approach to syntax allows us to analyze F-Structure composition as a
“syntactocentric” and derivational process; it is within this framework that I develop
the Derivational Approach to the Focus Structure (DAFS). With the DAFS, where
different items in the numeration are allowed to bear the [+F] feature, F-Structure
will be the derivational outcome of their syntactic composition. Hence, F-Structure is
fixed already in the narrow syntax, and then both phonology and semantics can take
F-Structure as the structural description of their operations, in particular the Focal
Nuclear Stress Rule on the PF side of the derivation. As I argued, the DAFS allows us
to capture in a very simple and natural way the nuclear stress assignment related to
focus, while permitting at the same time to circumvent the potential problems in
which their relation is not direct.

135
CHAPTER 4. FOCUS AT THE SYNTAX-SEMANTICS INTERFACE

In this chapter I explore the syntax, semantics, and the syntax-semantics

interface of focus. I analyze the syntactic displacements of focalizations as deriving
from a leftward movement of the focal phrase within a cartographic approach to
clausal architectures. Further, I propose that these overt syntactic displacements
provide a direct mapping from syntax to logical form whereby the focalized element
falls within the scope of a restricted quantification over events.
My analysis is based on the semantic approach to focus initially outlined in
Herburger (2000), although, as will be clear during the discussion, it differs from it in
important respects. I also take advantage of some of the most recent developments
within the Minimalist Program with regards to a derivational approach to the
relabeling of syntactic categories (cf. Hornstein & Uriagereka (2002)).
The theory of the syntax-semantics interface of focus I present here is totally
innovative and different from other analyses which adopt a cartographic approach to
focus in Basque in that the mapping from overt syntax to logical form is obtained via
the syntactic reprojection mechanism. It also differs from other syntactic approaches to
focus in that it provides a direct mapping from syntax to semantics and presents an
analysis of focus which is consistent with other proposals involving the syntax-
semantics interface of quantificational elements in natural languages. Although the
analysis I will propose is illustrated on the basis of Basque data, the proposal I make
here can be extended to focus in all languages.
This chapter is organized as follows: In the first section I present my analysis of
the syntax of focus in Basque. After comparing the two main approaches to this topic
(the traditional cartographic approach and the Nuclear Stress Rule (NSR)-based
approach), I discuss the data presented in chapter 2 and show the superiority of the
cartographic (split CP-type) approach over the NSR-based approach regarding the
data under analysis.

136
Section 4.2 is devoted to the logical form of focus. I propose a novel dual Neo-
Davidsonian analysis for the patterns of contextualized utterances and out-of-the-blue
utterances that I discussed in section 2.3. My proposal is that contextualized utterances
have the logical form of a definite description of events whereas out-of-the-blue
utterances are simple existential quantifications over events. I will argue that these
two different types of quantifiers capture naturally the properties of the semantics of
focus I discussed in section 2.3.
Section 4.3 provides a dynamic analysis of the syntax-semantics interface and
shows how the Neo-Davidsonian logical form representations proposed in section 4.2
are obtained derivationally from the left peripheric syntax proposed in section 4.1. In
particular, I will demonstrate that the properties of (i) displacement, (ii)
presuppositionality, and (iii) exhaustivity of focus are intimately related, and I will
propose an analysis whereby all these properties (or the lack of them) follow naturally
from the nature of the quantification over events.
Finally, in section 4.4, I summarize the main arguments put forward in this
chapter and discuss the main advantages of the proposal in this thesis. In particular I
show that the DAFS I proposed in chapter 3 together with the derivation to logical form
presented in this chapter allow us to explain the properties of focus in a strictly
derivational fashion, without incurring in any violation of the core principles of the
grammar. In short, we have a locus for focus in the architecture of grammar.

4.1. The Syntax of Focus in Basque: The Data and Discussion of Previous Literature
In this section I provide an analysis of the syntax of focus in Basque. First I
present the two main approaches proposed in the literature to account for focus in
Basque: the cartographic approach (cf. i.a., Ortiz de Urbina (1989)) and the NSR-based
approach (developed mainly in A. Elordieta (2001) and Arregi (2003)). I show that the
NSR-based approach has to face a series of difficulties and makes several wrong
predictions. In contrast the cartographic approach derives all the relevant facts
directly. I conclude by arguing that the cartographic approach is to be preferred over
the NSR-based approach.

137
4.1.1. Introduction: The Cartographic and NSR-based Approaches to Focus
In chapter 2, I showed that in Basque the focalized element appears left-
adjacent to the verb (cf. i.a. Azkue (1891, 1923), Altube (1920, 1929), Mitxelena (1981),
Ortiz de Urbina (1983, 1986, 1999), Eguzkitza (1986), Saltarelli et al. (1988), Osa (1990),
Aske (1997), A. Elordieta (2001), Arregi (2003), or Etxepare & Ortiz de Urbina (2003)).
The grammaticality contrast in (1a-c) illustrates this property. Example (1a) with a
focalized subject is ungrammatical because the focalized subject and the verb are not
adjacent to each other. On the other hand (1b-c), where the focalized subject is left-
adjacent to the verb, are perfectly grammatical:

(1a) *[Jonek]F mahaia hautsi du.

Jon.erg table broke AUX
“[Jon]F broke the table”

(1b) Mahaia [Jonek]F hautsi du.

table.det Jon.erg break AUX
“[Jon]F broke the table”

(1c) [Jonek]F hautsi du mahaia.

Jon.erg break aux table.det
“[Jon]F broke the table”

In the next two sections I will review the two main approaches that have been
proposed to account for the syntactic properties of focus in Basque: the cartographic
approach (4.1.1.1), and the NSR-based approach (4.1.1.2). Next, in section 4.1.2, I will
confront both approaches to the empirical evidence introduced so far and discuss the
theoretical advantages of each of them. The conclusion is that the cartographic
approach is to be preferred over the NSR-based approach.

4.1.1.1. The Cartographic Approach to Focus in Basque

Since the seminal work of Ortiz de Urbina (1986, 1989), the traditional way of
analyzing focus in Basque has been in terms of the ‘Cartographic’ or ‘Left Peripheric’

138
approach. This approach postulates that there is a functional projection in the left
periphery of the clause that attracts focus to its specifier. This projection has been
taken to be CP (cf. Ortiz de Urbina (1986, 1989)), or FocP (cf. Ortiz de Urbina (1997,
1999)). Following Ortiz de Urbina (1999), let us assume that it is FocP, with a
strong/uninterpretable focal feature in Focº. According to this theory Focº attracts the
focus phrase to its specifier and the verb raises via inflection to the head of this
projection to meet a ‘focus-criterion’. The picture then, is the following one:

(2) FocP
qp
XP[+F] Foc’
qp
V [ tXP … tV]

Following this approach, the example in (1c), repeated here as (3a), would be
derived as in (3b):

(3a) [Jonek]F hautsi du mahaia. (=1c)

Jon.erg break aux table.det
“[Jon]F broke the table”

(3b)
FocP
qp
JonekF Foc’
qp
[hautsii duj] InflP
qp
DP Infl’
qp
VP Infl
qp tj
mahaia ti

139
The focal XP (in this case, the subject ‘Jonek’) is attracted by the Focº to its
specifier, and the verb moves to Focº to fulfill a ‘focus criterion’ in Rizzi’s terms (see
Ortiz de Urbina (1999)). This analysis gives an immediate account of the word order
facts observed in sentences like (3a=1c)1.
A narrow syntactic conception of focal movements like this has many virtues.
On the one hand, it provides a unitary treatment of focus in all types of constructions
(‘normal’ focus constructions, clefts, pseudoclefts…). Under the left peripheric
movement analysis all of them are, somehow, syntactically triggered. Furthermore, it
provides a natural explanation for the morphological correlates to focus observed in a
number of languages, like the use of left peripheric focal particles of languages like
Tuki (cf. Biloa (1995)) or the agreement triggered by the focal element in languages like
Kinande (cf. Baker (2003)).
Summarizing, the main features of this syntactic approach are the following
ones:
(i) The focalized element undergoes movement.
(ii) Focus-movement takes place leftwards (to a specifier position).
(iii) The verb moves overtly to a functional head in the left periphery
which results in the phonological adjacency between the focus
and the inflected verb.
With this much background, we are in a position to compare the cartographic
approach to the NSR-based approach.

4.1.1.2. The NSR-based Approach to Focus in Basque

In recent years, another type of proposal has been pushed forward: what I will
refer to as the ‘NSR-based’ approach (cf. Cinque (1993), Reinhart (1995, 2006),
Neeleman & Reinhart (1998), Zubizarreta (1998), Szendröi (2001), and A. Elordieta
(2001) and Arregi (2003) for Basque, among others). The core idea of this approach is

1 Recall that we have mentioned before that any element preceding the focus in narrow focus sentences
is topicalized. This would be the case of (1b), repeated here as (i):
(i) [Mahaia]Top [Jonek]Foc hautsi du.
The analysis of (1b) would be exactly the same one proposed for (3b=1c) in the text, followed by
movement of ‘mahaia’ to a topic position above focus. Again, this analysis derives all the word order
facts and explains why the elements that precede focus are interpreted as topics in a straightforward
way.

140
that it is the nuclear stress placement that determines the F-Structure of a sentence in
PF.
This theory stems from Cinque’s seminal work on the NSR (cf. section 3.4.2). As
a development of his NSR theory, Cinque links the nuclear stress placement and the F-
Structure of the sentence by formulating a focus-to-nuclear stress correlation: interpret
focus where nuclear stress is. This accounts for the facts of focus projection.

(4a) Jesus preached to the people of [Judea]F

(4b) Jesus preached to the people [of Judea]F

(4c) Jesus preached to the [people of Judea]F

(4d) Jesus preached to [the people of Judea]

F

(4e) Jesus preached [to the people of Judea]

F

(4f) Jesus [preached to the people of Judea]F

(4g) [Jesus preached to the people of Judea]F

Nuclear stress is assigned to the most deeply embedded element ‘Judea’ in all
the cases in (4a-g). The variability observed in (4a-g), where focus extends from ‘Judea’
up to the whole clause is taken to be due to ‘F(ocus)-Projection’, by which the element
that bears the accent is able to ‘project’ its focal status to higher nodes that dominate it.
That is, the nuclear stress in English marks the bottom boundary of the F-Structure.
However, as we saw in chapter 3, Cinque argues that the NSR will apply
differently in out-of-the-blue and ‘marked’ environments: even if the accent placement
in the most deeply embedded element can denote different F-Structures as in (4), if we
want to mark focus on a different element than those in (4a-g) nuclear stress will have
to appear on that element (5). If this is not the case, the sentence is ungrammatical:

(5a) *[Jesus]F preached to the people of Judea.

(5b) [Jesus]F preached to the people of Judea.

(5c) *[Jesus preached to the people of Judea] .

F

141
This explains why sentences like (5c) are completely ungrammatical; nuclear
stress on the subject cannot mark sentence-focus (cf. section 3.4). Likewise, sentence
(5a) is ungrammatical given that nuclear stress does not fall in the focus.
According to the NSR based theory of F-Structure, a sentence will not have an
actual focus but ‘a set of possible foci’, the set of nodes that contain the nuclear stress
(cf. (4)). An instantiation of this notion is given in Reinhart (1995, 2006):

(6) The Focus Rule (Reinhart (1995, 2006))

The focus of IP is a(ny) constituent containing the main stress of IP, as
determined by the stress-rule.

Cinque’s proposal has become extremely influential in the studies of focus. In

recent years a very productive program of research has been developed refining his
theory as well as applying it to a wide variety of understudied languages.
With this much background, let me now analyze how this type of approach
derives the properties of the displacements triggered by focus in ‘discourse-
configurational’ languages like Basque or Hungarian. There is a crucial difference
between the cartographic approach and the NSR-based approach. Under the
cartographic approach it is focus that moves. Under the NSR-based approach if focus
is not in the most deeply embedded position, then other elements must move to
ensure that focus ends up in the most deeply embedded position.
The NSR-based approach to the F-Structure conceives the word order changes
involved in focus constructions as driven by a PF requirement. Sentences with narrow
focus involve movements that take place in order to fulfill a purported legibility
condition on derivations, which requires that focused elements bear the nuclear stress
at PF. When a non-focal element is in the most embedded position in the clause, a
‘nuclear stress avoiding’ movement of this element must take place so that the focused
element becomes the most deeply embedded element within the clause, hence getting
nuclear stress via the NSR. This is what has been called p(rosodic)-movement (see
Zubizarreta (1998) among others). In the case of the analyses that apply the NSR-based

142
approach to Basque, the movements observed in focalizations have been characterized
as ‘scrambling’ operations (cf. A. Elordieta (2001)) or left and right dislocations (cf.
Arregi (2003)).
The account of focus in A. Elordieta’s (2001) proposal is a hybrid one,
combining features of the NSR-based approach and the left peripheric approach. For
this reason, when discussing NSR-based approaches to compare them to cartographic
approaches, I will mainly discuss Arregi’s (2003) proposal, for it is a “pure” NSR-
based approach.
Arregi (2003) proposes the following explanation to account for the
displacements triggered by focus. He assumes that Basque is consistently a right-
headed language with an aspectual phrase sandwiched between TP and vP. Following
his proposal, in order to derive the surface word order of (1b) (O-SF-V), the most
deeply embedded object ‘mahaia’ is left dislocated over the subject and adjoined to
TP, as represented in (1b’). After the left dislocation of the direct object, the subject is
rendered in the most deeply embedded position and it gets nuclear stress there via the
NSR. Having the nuclear stress, it can be interpreted as focused in PF.

(1b) Mahaia [Jonek]F hautsi du

table Jon broken AUX
“Jon broke the table”

143
(1b’) TP
qp
mahaia TP
3
AspP T
3
vP Asp
3 hautsi du
Jonek v’
3
VP tv
3
t tv

Recall that in addition to (1b), (1c) is also an example of narrow focus on the
subject. The derivation Arregi proposes for (1c) is given in (1c’):

(1c) [Jonek]F hautsi du mahaia

Jon broken AUX table
“Jon broke the table”

(1c’) TP
qp
TP mahaia
3
AspP T
3
vP Asp
3 hautsi du
Jonek v’
3
VP tv
3
t tv

As shown in the tree in (1c’), in order to get the focus on the subject and the
word order in (1c), Arregi (2003) proposes a rightward dislocation of the object, which
escapes from the most deeply embedded position and right adjoins to TP. Once the

144
object has been right dislocated, the subject is the most deeply embedded element in
the clause, and it gets the nuclear stress. Hence, it can get interpreted as focus at PF.
According to this proposal, then, the surface Focus-Verb adjacency observed in
languages like Basque is just accidental. It is the result of the ‘stress avoiding
movements’ of the non-focused elements and the configuration of the lowest part of
the clause whereby the most deeply embedded element happens to be left-adjacent to
the verb. Reference to the verb or such an adjacency is not mentioned in the theory.
The basic features of this approach are then the following ones:
(i) The focal element stays in situ. Other material moves (if required by
PF).
(ii) Movements of nonfocal material can take place leftwards or
rightwards.
(iii) Focus-Verb adjacency is accidental; a byproduct of the fact that the
focal XP is the most embedded one and the verb is the next element
to its right.
Having reviewed the main features of the cartographic and NSR-based
approaches, in the next section (4.1.2.) I provide empirical evidence that favors the
cartographic approach over the NSR-based one. The discussion will be restricted to
the empirical grounds of focus in Basque, since, as I said, chapter 5 will be devoted to
a thorough discussion and criticism of the NSR-based approach.

4.1.2. Evidence in Favor of the Cartographic Approach

In this section I will compare the empirical coverage of NSR-based approaches
and Cartographic Approaches to the F-Structure regarding Basque data. I will give
ample evidence in favor of the Cartographic Approach.
To start with, the main criticisms that NSR-based approaches to the F-Structure
have against Cartographic Approaches are two:
(i) It does not account for nuclear stress placement.
(ii) It does not capture “focus projection”
Although I will put forth an analysis of the syntax of focus in the spirit of the
cartographic approach, as I argued in chapter 3 these objections raise no problem for

145
the DAFS I have proposed there. On the one hand, nuclear stress placement is
predicted correctly (even better than in NSR-based theories2). On the other hand, as I
mentioned in chapter 3, we can get away with focus projection: “focus projection” is
just an illusion (cf. chapter 5 for a development on this). Furthermore, as I will show
next on the basis of Basque, the explanatory power of the cartographic approach is
much higher than the power of the NSR-based approach.
Recall that there is a sharp difference in the conception of movement in the
cartographic and the NSR-based theories. The cartographic approach postulates a
syntactically driven movement of the focal element that is triggered by a greedy feature
checking/valuation (cf. Chomsky (1995b)). In contrast the NSR-based approach
postulates generous movements of all types of nonfocal material in order to obtain a
certain PF representation. The conceptual difference is extreme, for both systems
assume completely different motivations and operations and a completely different
architecture of grammar (cf. chapter 5). The common assumption in generative
grammar is that movement is greedy; this assumption, in and of itself, would favor a
cartographic approach over a NSR-based one, if only for its conceptual parsimony. For
instance, it treats the same way Wh- and focus-movements, movements that display
very similar patterns with regards to extractability, locality, path, and landing site3.
In the next eight sections I will compare both approaches on purely empirical
grounds testing them against empirical data. The comparison will reveal that besides
being theoretically more attractive than the NSR-based approach, the cartographic
approach is also better suited to account for the properties of focus in Basque.

4.1.2.1. Movement Patterns

As we saw in chapter 2, focalized constituents in Basque appear left-adjacent to
the verb, and the rest of the arguments or adjuncts can appear before them, or
following the complex unit [focus + inflected verb]. When non-focalized elements

2 For instance, the DAFS combined with the focus-sensitive FNSR predicts immediately nuclear stress
placement on focalized subjects in English, something that the classic NSR cannot do (cf. chapter 5 for
discussion).
3 Despite there are NSR-based theories of the F-Structure (like Arregi’s (2003)) that analyze Wh-

movements as triggered by nuclear stress there is ample evidence that Wh-phrases undergo movement
to the left periphery (see among others Cheng & Corver (2006) for a recent overview of Wh-movement).

146
appear preceding the focus they have a topic semantics and topic intonation in Central
Basque (a continuation rise at their right edge). In contrast, when they surface after the
focus in postverbal position like (7c), they just have a GIVEN interpretation and are
accompanied by the characteristic postfocal pitch compression (cf. section 2.2). This is
illustrated in the sentences in (7a-c), partially repeated from (1):

(7a) [Mahaia]F hautsi du Jonek.

table.det break AUX Jon.erg
“Jon broke [the table]F”

(7b) [Mahaia]T [Jonek]F hautsi du. (=1b)

table.det Jon.erg break AUX
“[Jon]F broke the table”

(7c) [Jonek]F hautsi du mahaia. (=1c)

Jon.erg break aux table.det
“[Jon]F broke the table”

As shown in 4.1.1.1 and 4.1.1.2, both analyses have mechanisms to derive the
word order facts in (7b). As was illustrated in (3), within the cartographic approach
the analysis would postulate a left-peripheric movement of the focus to the specifier of
FocP followed by a movement of the verb to Focº (via Inflº). This derives the structure
under (7c). Then a topicalization movement of the DP object ‘mahaia’ over the focus
sitting in Spec-FocP yields the structure in (8), which corresponds to the sentence in
(7b):

147
(8)
TopP
3
[Mahaiai]T Top’
3
Topº FocP
3
[Jonekj]F Foc’
3
[hautsi du]k TP
6
tj ti tk

Within the NSR-based approach, the analysis of the sentence in (7b) would be
the one we saw above in (1b’): left-dislocation of the most deeply embedded direct
object and everything else stays in situ, as repeated here in (9):

(9)
TP
qp
mahaia TP
3
AspP T
3
vP Asp
3 hautsi du
Jonek v’
3
VP tv
3
t tv

Likewise, the derivation of (7c) would be along the lines we saw in (1c’), just
allowing for the rightward adjunction of the direct object ‘mahaia’ to the TP.
Now, the crucial question is the following one: what motivates the movements
that lead to the [O]FVS configuration of (7a)? Within a cartographic analysis this has a
natural explanation: the focal object is attracted to Spec-FocP followed by the verb,

148
and the subject stays in situ (just like with the case of the object in (7c)). This is
represented in (10):

(10)
FocP
3
[Mahaiai]F Foc’
3
[hautsi du]j […]
6
Jonek ti tj

Within a NSR-based theory the issue becomes tricky: assuming that the object
stays in situ, the only way to account for the right-edge position of the subject would
be to posit a rightward movement of it, to right adjoin to TP; as in (11):

(11)
TP
qp
TP Jonek
3
AspP T
3
vP Asp
3 hautsi du
ti v’
3
VP tv
3
mahaia tv

Leaving aside the rightward dislocations that this analysis has to assume, the
main problem that this theory has is its ad hoc nature4. The fact is that there is no

4 Arregi (2003) assumes that leftward displacements alter scopal relations whereas rightward
displacements do not (i.e., they always reconstruct). This allows him to account for variable binding and
scopal facts. Unfortunately Arregi (2003) offers no explanation about why leftward displacements do
not reconstruct and rightward displacements have to reconstruct. Note that with the cartographic

149
trigger whatsoever for the subject’s right adjunction to TP. Here I am not talking about
the inadequacy of positing a phonologically-triggered movement (the global look ahead
of a theory that posits that some syntactic operations take place in order to provide an
intended representation in PF); see chapter 5 for an extensive discussion. Rather, the
puzzle here is that the purported phonological motivation for the displacement is
missing.
This point is crucial; according to NSR-based theories, in (11) there should be no
reason for the displacement of the subject, for it is not in the most deeply embedded
position (rather, the object ‘mahaia’ is). Hence, the movement of the subject is vacuous
in its metrical consequences, and then, it should not take place. However, sentences
like (7a) are perfectly grammatical.

4.1.2.2. Long-Distance Focus Movements

As we saw in chapter 2, in cases in which the focal phrase originates in an
embedded CP, it typically appears at the left edge of the matrix clause. These cases are
illustrated in examples like (12a)5:

(12a) [Jon]F pentsatzen dut [CP ikusi zuela Mirenek]

Jon think AUX seen AUX-that Miren
“(I) think Miren saw [Jon]F”

As I advanced in chapter 2, this phenomenon can be easily captured by positing a long

distance movement of the focal XP up to the matrix CP, just the same as with an
embedded Wh-phrase (cf. Eguzkitza (1986), Ortiz de Urbina (1989) among others). A
major virtue of this analysis is that it accounts for the Subject-Verb inversion of the
embedded clause as deriving from the successive cyclic movement of the focal XP that

approach all the facts are explained naturally for prefocal material outscope focus and postfocal
material is outscoped by focus.
5 In embedded contexts the auxiliary bears the embedding complementizer ‘-ela’ as a suffix. I will

abstract away from this issue since it is not relevant for our discussion here.

150
is accompanied by the movement of the verb. The structure of this sentence, then,
would be derived as follows6:

(12b) CP1
ei
[Jon]F C’
3
pentsatzen dut IP
3
(pro) I’
3
VP t
3
CP2 t
ei
t C’
3
ikusi zuela IP
3
Mirenek I’
3
VP t
3
t t

On the other hand, Arregi (2003) discusses some cases where the focal XP
originates in the embedded clause. In particular, he discusses cases like (13a):

(13a) [Jon]F pentsatzen dut [CP Mirenek t ikusi zuela]

Jon think AUX Miren seen AUX
“I think Miren saw [Jon]F”

According to Arregi’s (2003) analysis, these patterns are explained as follows:

contrary to the examples of simple sentences that we saw before, in embedded
sentences the focal element undergoes movement. Arregi (2003), proposes a new type
of operation of adjunction for that. This adjunction will not be like the adjunctions of

6For the clarity of exposition, throughout this discussion I use subscript numbers to mark the status of
the phrases; main (XP1) or embedded (XP2).

151
nonfocal material to TP that were posited to explain the patterns of simple sentences.
In this case, the adjunction operation takes the focus and adjoins it to the vP of the
matrix clause (and always to the left). So, the explanation, according to Arregi (2003) is
as follows: the focus is extracted from its base-generated position, adjoining it to the
left of the matrix vP1 as represented in (13b). Then, the embedded CP is right-
dislocated, adjoining it to the right of the TP1 (13b’):

(13b)
TP1
qp
AspP T
qp
vP1 [pentsatzen dut]v
wo
JonF vP1
ei
pro vP1
ei
VP1 tv
qp
[Mirenek tDP ikusi zuela]CP tV

(13b’)
TP1
wo
TP1 [Mirenek tDP ikusi zuela]CP
wo
AspP1 T
wo
vP1 pentsatzen dut
wo
JonF vP1
3
pro vP1
3
VP1 tv
3
tCP tV

152
This provides the word order of (13a) at the same time that places the subject of
the embedded clause on the most deeply embedded position given it was adjoined to
vP (setting aside the subject pro of the matrix clause, see below). If it were adjoined to
TP like the unfocussed object in (9), the verb would be the most deeply embedded
element, contradicting the main premise of NSR-based theories.
To begin with, note that assuming the postulates of NSR-based theories, there
should no reason for the existence of these sentences. The direct object of the
embedded clause is first-merged in the most deeply embedded position; then, why
should the grammar trigger all those displacement operations? In any event, setting
aside the shaky foundations of these operations, I do not share the judgments reported
by Arregi (2003). This sentence is deviant to me as well as other speakers I have
consulted. My judgment is consistent with that in earlier works like Ortiz de Urbina
(1983, 1989, 1999) and Laka & Uriagereka (1987), which mark constructions like (13a)
with a *. I feel that the deviance might not be that strong, maybe, it is enough to be
marked as a ??, as in Ortiz de Urbina (1986). At any rate, the important fact here is that
for most speakers, the optimal realization of a sentence with focus on one of the
elements in the embedded clause is the one in (12a), with verb-movement in the
embedded clause, yielding the word order [inflected verb + subject], and not [subject
+ inflected verb], as the one in (13a). And crucially, unless one more ad hoc movement
is posited to derive the Subject-Verb inversion of the embedded clause, (12a) cannot be
derived under Arregi’s (2003) analysis.
On the other hand, note that in (13a) the subject is silent (it is a pro). Being silent,
there would be no reason for it to move for PF reasons. However, the correct word
order of a sentence analogous to (13a) with an overt subject instead of a pro in the
matrix clause would be as in (14)7:

7 As I said before, sentences like (14) without the subject-verb inversion in the embedded clause are
deviant for me, but let me use (14) for the sake of the argument. To my ear, as well as for other speakers
I have consulted, an optimal version of (14) would be (i):
(i) [Jon]F pentsatzen du Aitorrek [CP ikusi zuela Mirenek]
Jon think AUX Aitor seen AUX Miren
“Aitor thinks that Miren saw [Jon]F”

153
(14) [Jon]F pentsatzen du Aitorrek [CP Mirenek t ikusi zuela]
Jon think AUX Aitor Miren seen AUX
“Aitor thinks that Miren saw [Jon]F”

In order to explain (14) the NSR-based approach will have to posit that the
subject of the matrix clause is also right-dislocated and adjoined to TP, as represented
in (14’):

(14’)
TP1
wo
TP1 [Mirenek tDP ikusi zuela]CP
wo
TP1 Aitorrek
wo
AspP1 T
wo
vP1 pentsatzen du
wo
JonF vP1
3
tDP vP1
3
VP1 tv
3
tCP tV

This set of operations would provide the word order of (14) at the same time
that it leaves the object ‘Jon’ of the embedded clause in the most embedded position.
Note however that a NSR-based theory of F-Structure will have to impose an arbitrary
order in the rightward dislocations so that the dislocation of the subject applies before
the embedded clause adjoins to TP, otherwise the word order would be the one
represented in (15), which is ungrammatical:

154
(15) *[Jon]F pentsatzen du [CP Mirenek t ikusi zuela] Aitorrek
Jon think AUX Miren seen AUX Aitor
“Aitor thinks that Miren saw [Jon]F”

A NSR-based theory of F-Structure cannot explain this restriction on word

order. On the other hand, the cartographic approach derives these facts
straightforwardly; sentence (14) corresponds to the structure in (12) but with an overt
subject instead of pro8. The unavailability of sentences like (15) is as well easily
explained given that we would have to posit ad hoc movements to derive them. Since
(15) is ungrammatical, those movements need not be posited.
As we saw, the explanatory power of the cartographic approach is bigger than
the explanatory power of the NSR-based approach when confronted to both sinple
sentences and sentences with long distance movements. In the next section I address
the clausal pied-piping constructions.

4.1.2.3. Pied-piping Phenomena

As already explained in chapter 2, a focal element within an embedded clause
can also trigger an operation of “clausal pied-piping”. (16) would be an example of
this type of construction9:

(16) [[Mirenek]F edan duela ura]CP esan du Jonek.

Miren drink AUX.that water say AUX Jon
“Jon said that [Miren]F drank water”

As we saw in chapter 2, the derivation of a sentence like (16) within a

cartographic analysis would be along the lines in (17): first, the focus phrase is
extracted to the Specifier of a left peripheric projection, depending on the analysis it is

8 As I said before, the appropriate sentence would involve head movement of the verb in both matrix
and embedded clauses.
9 In chapter 2, footnote 3 I noted that these sentences aredegraded with material from the pied-piped

clause in postverbal position. Nevertheless I will continue using these examples for the ease of
exposition.

155
either CP or FocP (although here I will use CP for ease of exposition). This movement
triggers the V2 effect within CP2, cf. (17a). Then the whole CP2 is pied-pipied to Spec-
CP1, triggering I-to-C movement in CP1, as represented in (17b):

(17a) CP2
3
[Mirenek]F C’
3
edan duela IP
3
t I’
3
VP t
3
ura t

(17b) CP1
qp
[[Mirenek]F edan duela ura]CP2 C’
3
esan du IP
3
Jonek I’
3
VP t
3
t t

In contrast to this simple cartographic proposal, Arregi (2003) proposes an

alternative analysis. Within his NSR-based theory of the F-Structure, the same
movements as those proposed for the long distance movements would have to be
posited for these constructions. The only difference is that in pied-piping constructions
it is not only the focal XP of the embedded clause that adjoins to the matrix vP, but the
entire subordinate clause containing the focus.
According to Arregi’s (2003) analysis, example (16) is derived as in (18), with a
a right dislocation of the embedded object adjoining it to TP2, just like in matrix
clauses. This renders the subject as the most embedded element within CP2. Further,

156
in order to explain the word order in (16), whereby the subject of the matrix clause
appears in sentence-final position, a right-dislocation of this subject will have to be
posited:

(18)
TP1
3
TP1 Jonek
3
AspP T
3
vP Asp
3 esan du
tDP v’
3
VP tv
3
CP2 tv
3
TP2 Cº
3
TP2 ura
3
AspP T
3
vP Asp
3 edan duela
Mirenek v’
3
VP tv
3
tDP tv

But again, we face the same puzzle as in (11) above. From the point of view of
phonology movement of the subject of the matrix clause is unmotivated. This
movement is sterile in its metrical consequences.
Again, we saw that the cartographic approach is much more attractive than the
NSR-based approach when confronted to pied-piping data. In the next section I
analyze the island restrictions that focal movements observe.

157
4.1.2.4. Islandhood
As explained before, one of the most striking characteristics of syntactic
displacements resides on the restrictions that these context-sensitive dependencies
have to observe (cf. Ortiz de Urbina (1989)). As discussed in section 1.3.1.3.2., some
syntactic domains are opaque in that they do not allow for extraction of Wh-phrases.
This is illustrated with the data in (19) (adjunct island), (20) (coordinate structure), and
(21) (left branch extraction):

(19) Adjunct Island:

*Zer atera zara etxetik [zzeerr galdu duzulako]Adj.
what extract AUX house.from loose AUX
“What did you go out of home because you lost?”

(20a) Coordinate Structure Island (first conjunct):

*Zer jango du lehengusinak [zzeerr eta txuleta]CoordP.
what eat.fut AUX cousin and cutlet
“What will the cousin eat and cutlet?”

(20b) Coordinate Structure Island (second conjunct):

*Zer jango du lehengusinak [bakailaoa eta zzeerr]CoordP.
what eat.fut AUX cousin cod and
“What will the cousin eat cod and?”

(21) Left Branch Extraction Island:

*Nor da [nnoorr etorriko dela]Subj. berri handia.
who BE come aux-comp news big
“Who is big news that will come?”

As Ortiz de Urbina (1989) notes, the displacements triggered by focus are also
sensitive to these barriers. As illustrated below, focus cannot be extracted out of the
classic island constructions. Below are some examples of adjunct (22), coordinate

158
structure (23a-b) and left branch (subject) islands (24) (potentially, the answers to the
questions in (22)-(24)):

(22) *[Giltza]F atera naiz etxetik [ggiillttzzaa galdu duelako]Adj.

key.det extract AUX house.from loose AUX
“I went out of home because he/she lost the key”

(23a) *[Bakailaoa]F jango du lehengusinak [bbaakkaaiillaaooaa eta txuleta]CoordP.

cod.det eat.fut aux cousin and cutlet
“The cousin will eat cod and cutlet”

(23b) *[Txuleta]F jango du lehengusinak [bakailaoa eta ttxxuulleettaa]CoordP.

cutlet eat.fut aux cousin cod and
“The cousin will eat cod and cutlet”

(24) *[Aitor]F da [A
Aiittoorr etorriko dela]Subj. berri handia.
Aitor BE come aux-comp news big
“That Aitor will come is big news”

As discussed in section 1.3.1.3.2 syntactic islands are very mysterious within the
minimalist theory. From a descriptive point of view, however, the facts regarding
focus are clear: a focal element first-merged within one of these “island phrases”
cannot be extracted. In this regard, the left peripheric approach to focus provides a
neat way of analyzing the data: a focalized phrase, just like a Wh-phrase, has to
undergo a leftward movement that is blocked by the island status of the phrase it
departs from. A NSR-based approach cannot account for this data in such a way since
within that analysis the focus phrase is taken to undergo no movement (unless the
violation is put in terms of extraction of nonfocused elements due to p-movements).
Besides, a NSR-based analysis of focus that does not analyze Wh-movement as driven
by nuclear stress would force a different analysis for the Wh-movement patterns of
(19)-(21) and the focus patterns of (22)-(24), which at first sight does not seem

159
conceptually desirable given the similar restrictions they have to obey (see Ortiz de
Urbina (1989) and Etxepare & Ortiz de Urbina (2003), among others). It is true that we
do not have a good understanding of the nature of the Coordination Structure
Constraint. However, whichever the ultimate explanation, given their syntactic,
semantic, and phonological similarities, a unitary account of islands for focus and Wh-
phrases should be preferred instead of different explanations for data that vary just in
the interrogative vs. focal interpretation of the moved element. The island effects
displayed by focus fit perfectly in what we expect from a left-peripheric movement,
whereas in an NSR-based approach the coincidence in sensitivity to islands of Wh- and
focus constructions is left unexplained.
Furthermore, as I will argue in the next sections, other properties show as well
the plausibility of the idea that a left peripheric movement is involved in focus
constructions.

4.1.2.5. Dialectal Variation

In all cases we have seen so far the focal phrase (or a phrase containing it in
cases of pied-piping) appears immediately left-adjacent to the verb. Interestingly, a
look at Eastern dialects of Basque shows that this adjacency is not accidental, since
what we find in these dialects is not Focus-Verb adjacency but Focus-Tense adjacency.
Thus, alongside the Central and Western Basque (25), we can find the Eastern variants
of (26). This suggests that what is relevant is the adjacency between focus and tense.
The Western dialect would include exactly the same operation, with the difference
that the verb is pied-piped together with the tense when the latter moves to Focº cf.
Ortiz de Urbina (1989, 1995), and, more generally, Etxepare & Ortiz de Urbina (2003)).

(25) Central and Western Basque:

(a) [Mahaia]T [Jonek]F hautsi du .

table Jon broke AUX
“[Jon]F broke the table”

160
(b) [Jonek]F hautsi du mahaia.
Jon broke AUX table
“[Jon]F broke the table”

(26) Eastern Basque:

(a) [Mahaia]T [Jonek]F du hautsi.

table Jon AUX broke
“[Jon]F broke the table”

(b) [Jonek]F du mahaia hautsi.

Jon AUX table broke
“[Jon]F broke the table”

Consider the example in (26b). Its analysis within a cartographic approach

would be the same one that we saw in (3b), but just raising Tº to Focº instead of the
whole verbal complex:

(27)
FocP
qp
JonekF Foc’
qp
du TP
qp
tDP T’
qp
VP tT
qp
mahaia hautsi

On the contrary, there is no way of analyzing such a construction within a NSR-

based approach. Even if we right-dislocate the object, to make the subject the most
embedded element no accidental adjacency could arise between the focused element

161
and the inflected auxiliary. In order to get such a configuration a new movement
would have to be posited to dislocate the verb over Tº/Inflº to the right of the object10:

(28)
TP
qp
TP hautsi
qp
TP mahaia
3
AspP T
3
vP Asp
3 du
Jonek v’
3
VP tv
3
t tv

As we have seen so far, the cartographic approach makes use of a small number
of operations and accounts for all the data in a unified way, the NSR-based analysis,
on the other hand, has difficulties in explaining a number of patterns and it has to
postulate new ad hoc operations for every new construction under analysis. In the next
section I analyze the patterns of infinitival constructions.

4.1.2.6. Focus in Infinitival Constructions

This point is a corollary to the previous one. In order to account for the
displacement of the inflected verb as well as the adjacency between the focalized
element and the inflected verb the cartographic approach assumes that focus moves to
Spec-FocP and the inflected verb head-moves to a position in the left periphery.
There is no agreement as to what formal feature triggers the displacement of
the verb, whether it should be stated as a feature that the focus has to check (for
instance, a T feature in Focº), a formal feature on other heads of the left periphery (like

10 Note that in classical terms this would amount to a violation of the Head Movement Constraint.

162
Finº) or a formal feature on Tº. Suppose that we assume the third hypothesis, and
assume that Tº has to raise to Focº; this would explain why the finite verb has to move.
The prediction of this analysis would be that whenever we do not have an inflected
auxiliary, the verb will not necessarily be left-adjacent to focus. This is precisely the
pattern that we find in Basque: when we construct examples with infinitival
subordinate clauses, the verb in the infinitival clause does not appear left-adjacent to
the focal element (providing focus is on an element different than the object, of
course). Compare in the following dialogue the out-of-the-blue sentence uttered by A
with the correction of B11:

(29) A: [Kepak ardoa edate-a] gauza arraroa da.

Kepa wine drink-INF thing strange BE
“It is a strange thing for Kepa to drink wine”
B: Ez!, [[Julenek]F ardoa edate-a] da gauza arraroa!
No Julen wine drink-INF BE thing strange
“No!, It is a strange thing for [Julen]F to drink wine!”

In B’s response the basic word order of the embedded clause is maintained
([S]F OV) whereas in the main clause the verb is fronted to be left-adjacent to the
embedded clause.
These examples confirm that the reason for the adjacency between the focus
and the verb lies on Tense, since in its absence, the verb in the subordinate clause
does not move, and the sentence is still grammatical. Note that an analysis of (29B)
within a cartographic approach would be straightforward; along the lines described
for the pied-piping cases in section 4.1.2.3, the subject ‘Julenek’ moves to the Spec-Foc
of the embedded clause without triggering any movement of the verb (30). Then, it
pied-pipes the embedded clause to Spec-FocP1 triggering T-to-C movement there, as
represented in (31).

11This is not the only possibility, since some speakers accept also the variant of (29B) with focus-verb
adjacency (cf. also Ortiz de Urbina (1983)).

163
(30)
FocP2
3
JulenekF Foc’
3
Focº [tDP ardoa edatea]

(31)
FocP1
qp
[[Julenek]F ardoa edatea]CP2 Foc’
wo
da TP
wo
[ tCP2 gauza arraroa] T’

…tT…

Alternatively, we could posit that the verb in the embedded clause does move
to Focº, but that this movement gets later ‘masked’ by a rightward movement of the
verb to adjoin to the clitic determiner ‘–a’, as represented in (32)12:

(32)
DP
3
[…] Dº
g edate-a
FocP
3
JulenekF Foc’
ei
Focº [tDP ardoa tV]

12This would be similar to what happens in embedded sentences when the complementizer is
morphophonologically a clitic (cf. Laka (1990)).

164
On the other hand, a NSR-based approach cannot derive these sentences unless
it is allowed to move anything anywhere and freely. A possible analysis would be to
posit rightward dislocations of the direct object ‘mahaia’ and the verb ‘edatea’ of the
embedded clause adjoining to TP2 (assuming infinitivals always project a TP (which is
not important for our discussion here)). Furthermore these dislocations would have to
take place in this order so that we get the desired word order. Such a derivation is
provided in the tree in (33):

(33a) TP2
3
TP2 edatea
3
TP2 ardoa
3
AspP T
3
vP Asp
3
Julenek v’
3
VP tv
3
tDP tV

Then the attributive DP ‘gauza arraroa’ would have to right-dislocate over the
copula and adjoin to TP1:

165
(33b) TP2
wo
TP2 [gauza arraroa]DP
3
AspP T
eo
vP Asp
qp da
[[Julenek]F ardoa edatea]CP2 v’
(=33a) 3
VP tv
3
TDP tV

So, the explanation becomes completely ad hoc, for elements like the verb that
were not posited to move in previous cases now have to move. Furthermore, none of
these movements would give a grammatical result if they were not posited as taking
place rightwards and in that specific order. But recall that the NSR theory conceives
these movements as triggered by a PF legibility condition.

4.1.2.7. Rightward Foci

So far I have discussed constructions where the focal XP appears on the left
edge of the clause, claiming that this surface word order is evidence of a leftward
dislocation of the focal XP. In this section I will present some ‘marked’ constructions
where the focus appears at the right edge of the clause and claim that these
constructions as well are evidence in favor of the cartographic analysis for focus.
These are cases where the focalized element appears at the right edge of the clause
preceded by the rest of the sentence where the order of constituents is the same as in
out-of-the-blue sentences. For instance, the example in (34) could be a ‘marked’ variant
of (1c), with a reinforced degree of exhaustivity and a clear topic-focus intonation:

(34) Mahaia hautsi du [Jonek]F.

Table break AUX Jon
“[Jon]F broke the table”

166
Ortiz de Urbina (2002) argues that in this type of constructions we have the
regular focus movement of the [+F] marked XP illustrated in (35a), followed by a
remnant movement of the rest of the clause to a higher TopP (35b)13:

(35a) FocP[ XP[+F]i [Focº CP[...ti...]]]]

(35b) TopP[ CPj [Topº FocP[ XP[+F] [Focº tj]]]]

This type of analysis immediately explains the sentence-final position of the

focal phrase (it follows from the fact that the rest of the clause is remnant-moved to a
position higher than Spec-Foc)14. The topic-comment intonational contour that
accompanies them and the highly contrastive nature of their interpretation are
explained as deriving from the Topic position of the clause. This analysis can also
explain some scopal relations that would otherwise be hard to explain. Consider the
rightward focus construction of (36), where the focal XP takes scope over negation,
linear order notwithstanding:

(36) Ez da etorri [horregatik]F.

not AUX come because.of.that
“He has not come because of that”
(=That, and not (the) other, is the reason why he has not arrived).

Under the NSR-based approach Negº should take scope over ‘horregatik’ since
‘horregatik’ has to be the most deeply embedded element to be focalized, so Negº
would c-command it. On the contrary, on the cartographic approach we expect
negation not to take scope over ‘horregatik’, for Negº is within the remnant-moved
phrase and it cannot c-command ‘horregatik’. In opposition to the NSR-based

13 See Uribe-Etxebarria’s (2002) and Etxepare & Uribe-Etxebarria’s (2005) proposal for a similar analysis
of in situ-like Wh-questions of Spanish.
14 Notice that since the verb is contained in the remnant-moved phrase T-to-Foc movement will not take

place given that in the remnant-moved phrase there is no Focº. This is similar to what happens in other
languages like Spanish, see Uribe-Etxebarria (2002) for discussion.

167
approach, the scope facts show that the prediction of the cartographic approach is
correct and the NSR-based approach is wrong.
Again, these facts suggest that the cartographic approach is well suited to
account for the data, whereas an in situ NSR-based approach is highly problematic.

4.1.2.8. Fragmental Answers

The last piece of evidence in favor of the left peripheric approach comes from
fragmental answers. Throughout this dissertation I have always provided full
sentences as answers to Wh-questions. While full answers are grammatical, they are
odd in ordinary language use, where we tend to use subsentential or fragmental
answers providing just the focus and omitting the part of information that is GIVEN.
For instance, the fragment in (38) is more natural than the complete sentence (39) as an
answer to the question in (37):

(37) Zer irakurtzen du Urtzik?

what read AUX Urtzi
“What is Urtzi reading?“
(38) [Obabakoak]F.
(39) [Obabakoak]F irakurtzen du Urtzik.

Merchant (2004) has proposed an ellipsis analysis of these fragmental

constructions. According to Merchant’s (2004) analysis in (38) we have focus
movement of the object ‘Obabakoak’ to a left peripheric FocP, followed by ellipsis of
the remnant GIVEN material, as represented in (40) (cf. as well Kim (1997) for a similar
analysis):

168
(40)
FocP
3
[Obabakoak]F Foc’
3 ELLIPSIS
Focº [……]

This type of analysis has many advantages: the subsentential fragment is but
the tip of the iceberg and the whole clause is beneath15. Then, under this approach, all
case, agreement, and argument-structure relations are established as usual, the only
difference would be that the GIVEN information, which forms a constituent, is
phonologically elided (avoiding scattered deletion). Furthermore, this analysis helps
explaining many (anti-)connectivity, and island effects that we observe in these
constructions (cf. Merchant (2004) for discussion).
What is interesting for the purposes of our discussion is that in order to be able
to get Merchant’s (2004) elegant analysis and its nice predictions, we need the focus to
undergo leftward movement to a higher projection, escaping the ellipsis site. This is
something that is impossible to capture with a NSR-based approach to focus that
postulates that focus rests in situ.

4.1.3. Summary and Conclusions

To sum up this section on the syntax of focus in Basque, I showed that
focalizations in Basque are better analyzed as instances of syntactic movements of the
focal phrase to some left peripheric position. Thus, we observe the locality restrictions
that we expect from these movements. An analysis based on the NSR would not
predict this behavior. Furthermore, I showed that a cartographic approach has a big
explanatory power regarding the data; it can account for all the patterns observed so
far in a restricted and unified way just by positing a small number of operations that
are independently needed. The NSR-based approach, on the other hand, is forced to
provide different analyses in a case by case basis. It needs to postulate different types

15 See Valmala (2007) for a different analysis of these constructions.

169
of movements involving a variety of elements to different positions —movements
which are not motivated independently. Also, all these operations will have to take
place in a specific and different order depending on the pattern involved. There is no
explanation for these ordering restrictions.
In the next section I will present my analysis of the semantic contribution of
focus. I will argue that it provides content to the nuclear scope of a (restricted)
quantification over events.

4.2. The logical form of Focus

In this section I develop my analysis of the semantic contribution of focus.
Starting out from the analysis in Herburger (2000), I will propose that a focalized
element falls under the scope of a quantifier over events. In particular, I would like to
propose a modification of Herburger’s (2000) proposal and posit two different types of
quantifiers: (i) a weak existential quantifier over events giving rise to out-of-the-blue
readings, and (ii) a strong, definite description of events leading to contextualized
readings. Furthermore, I will argue that the clustered interpretive properties of these
two types of constructions that I introduced and discussed in chapter 2 follow in a
natural way from their logical form representation.

4.2.1. A Neo-Davidsonian Semantics for Focus

In order to analyze the semantic contribution of focus, I will adopt Herburger’s

(2000) Neo-Davidsonian analysis of event semantics (cf. section 1.2.2.2). The basic idea
of Herburger’s (2000) proposal is that a focalized constituent will be mapped into the
scope of an existential quantification over events. According to her analysis, the
sentence in (41a), with a focalized object, has the logical form in (41b), where the non-
focused chunk is the restrictor of the existential quantification (i.e., the sentence’s
‘aboutness’) and the focus falls within the scope of the quantificational structure (cf.
Herburger (1998, 2000)):

(41a) Milan bought [cider]F.

(41b) ∃e[Agent(e, Milan) & Buy(e) & Past(e)] Theme(e, cider)

170
This semantic representation captures in a very neat way the semantic relation
between the focus and the rest of the sentence. There is no need to add notions such as
‘novelty of information’ or ‘not given information’ to focus as primitives. Under this
approach, the novelty of focus will be its relation with the rest of the sentence (which
constitutes the restrictor of the existential quantification). The notion of ‘relative
novelty’ is much more accurate than the notion of ‘absolute novelty’ of focus proposed
in many of the works that analyze focus as ‘new information’ vs. ‘old information’.
This is illustrated in the question/answer pair in (42a-b). As this example shows, the
actual focus of the sentence does not have to be absolutely new or non-given in the
discourse (ex. from Rooth (1996: 271)):

(42a) Does Ede want tea or coffee?

(42b) Ede wants [coffee]F
(42c) ∃e [Agent(e, ede) & Want(e) & Present(e)] Theme(e, coffee)

As (42) clearly illustrates, the focal element in (42b) is not absolute new
information in the discourse, but rather relative new information in the event
description: it is new relative to the non-focused part (the restrictor of the event
description). Here it contrasts with the other alternative in the discourse, ‘tea’ (which
in this dialogue is explicit but need not be so).
Herburger’s analysis is simple and elegant, and it allows us to explain a wide
range of phenomena, such as the interactions between focus and negation or the
interactions between focus and other quantifiers. Recall that, as we saw in chapter 2,
the main reason that has been adduced against a presuppositional analysis of focus
was its failure in negative contexts like (43) (ex. (29) in chapter 2):

(43) John bought [nothing]F

According to Herburger (2000) there is now an easy solution for this puzzle if
we allow negation to outscope the existential quantification over events, as illustrated
in (44):

171
(44) Nothing(x)[∃e [Buying(e) & Past(e) & Agent(e, John)] Theme(e, x)]

What the formula in (44) means is, roughly, that nothing is such that there was
a buying by John that had it as a theme16.
Herburger, observes that focus might have non-local effects; thus, it will not be
enough to have just the [+F] featured material within the scope of the quantification.
Otherwise we would not be able to capture some of the possible meanings of
sentences like (45) (ex. 57, p. 41 of Herburger (2000)):

(45) [CP1 Mary told me about [DP the rumor [CP2 that Bill had said [CP3 that SUE was
going to India].

According to the judgments reported in Herburger (2000), sentence (45) can have at
least the following three readings:

(46a) “Mary told me about the rumor that Bill said that it was SUE who was going to
India”.
(46b) “Mary told me about the rumor that it was SUE who was such that Bill had said
that she was going to India”
(46c) “It was SUE who was such that Mary told me about the rumor that Bill had said
that she was going to India”

Herburger suggests that the reading in (46b) is obtained by having the focus
affecting the event denoted by the verb ‘say’ in CP2. The reading in (46c), on the other
hand, is one where the focus affects the event of the CP1.
Given these non-local effects of focus, we cannot just have the focused phrase at
the scope of the event quantifier but rather we need to take into account the entire
sentence. If this is correct, the conclusion that follows from there is that the specific
import of focus will be just not being present in the restriction of the existential
quantifier; that is, not being present in the sentence’s ‘aboutness’.

16 See Herburger (2000 §2) for a fully developed analysis of the relations between focus and negation.

172
My proposal is that if we combine Herburger’s (2000) basic logical form
representation with the DAFS I proposed in chapter 3 of this dissertation, we can
account now for the semantic import of the focus phrase in a very straightforward,
bottom-up fashion. As an illustration of how my analysis works, consider sentence
(47a). Sentence (47a) is an answer to the question in (47b) and has the logical form in
(47c). The syntactic F-Structure construal I propose for (47c) is represented in the
derivation in (47d) (tense is omitted for simplicity):17, 18

(47a) Milan [bought cider]F

(47b) What did Milan do?

(47c) ∃e[Agent(e, Milan)] Theme(e, cider) & Agent(e, Milan) & Buy(e)

(47d)
vP
qp
Milan v’
qp
v[+F] VP
qp
buy[+F] cider[+F]
e: e is a buying

FOCUS

So, with the DAFS I proposed in chapter 3, the size and interpretation of the F-
Structure of a sentence will be predictable from the lexical items that bear the [+F]
feature and the syntactic composition of the clause. The syntactic F-Structure will be
the derivational outcome of the merger of [+F] featured lexical items. Likewise, the
semantic interpretation of the focus of a sentence will be predictable regarding the
elements that are marked [+F] in the numeration and their mode of syntactic
composition. Finally, as I argued in chapter 3, by fixing the F-Structure in narrow
syntax we allow the PF component to apply some operations like phonological
phrasing or nuclear stress assignment (via the FNSR I proposed in section 3.4.3). All in

17I will leave the issue of where the existential quantifier comes from for section 4.3.1.
18For the ease of exposition, I am assuming here that as an answer to the question of (47b), the ‘v’ in
(47d) is marked as being [+F], nothing crucial for the analysis to be developed.

173
all, the theory of F-Structure composition I am developing is ‘syntactocentric’ in
essence, F-Structure is set derivationally in the narrow syntax, and then interpretive
modules can target it applying some phonological rules in PF or mapping it into the
scope of the restricted quantification over events at logical form.
But now, a puzzle arises: Herburger’s (2000) semantic account of focus assumes
a binary quantifier taking its restriction and its scope as the logical form representation
of a sentence, but this configuration does not come for free. If we want to keep with
the minimalist assumption that the syntax-semantics mapping is transparent and a
direct function of the syntactic structure and the lexical meaning of the words it
contains, then the intended logical form representation will have to be derived directly
from a syntactic structure that mirrors the postulated semantics (without involving ad
hoc mechanisms of any type). In order to accommodate the syntax to the tripartite
semantic structure she defends for focus, Herburger proposes a LF process of “focal
mapping”, an operation by which the nonfocused material in the c-command of the
existential quantifier rises to its restriction. This is represented in (48) (cf. Herburger
(2000: 43)):

(48)
XP
qp
QP ZP
3
Q WP
Syntax: internal argument external argument
Semantics: restriction scope

This is the intended result, but at first sight, it is not obvious that the semantic
representation follows from its surface structure or even from its Logical Form. To
obtain the intended representation in (48) Herburger’s ‘focal mapping’ would involve
a number of syntactic operations that are not possible under standard analyses.
In section 4.3, I shall argue for a novel account that provides a transparent
syntax-semantics mapping adapting the reprojection analysis of binary quantifiers in
Hornstein & Uriagereka (2002). But before doing that, in the next section I will
propose a slight modification of Herburger’s (2000) system in order to account for the

174
semantic-pragmatic effects of focus in both out-of-the-blue and contextualized
environments.

4.2.2. Syntactic Complexity and logical form Complexity

In this section I propose a modification of Herburger’s (2000) analysis of focus
semantics in order to capture the difference in the interpretation of contextualized
utterances and out-of-the-blue utterances that I discussed in chapter 2. According to the
proposal I will make, out-of-the-blue utterances correspond to existential descriptions of
events (hence, their new information reading and their lack of focal presuppositions);
contextualized utterances, in turn, correspond to definite descriptions of events (hence,
their contrastive reading and their focal presuppositions).

4.2.2.1. Two Types of Expressions

As we saw in chapter 2, in Basque (as well as in many other languages)
contextualized utterances and out-of-the-blue utterances show different syntactic,
semantic, and phonological structures. Let us recapitulate the conclusion we reached
in chapter 2 regarding these questions:

(49) Generalization on the Possible Interpretations of Focus: (section 2.3): In contextualized

utterances the focus has a contrastive interpretation, there is a focal
presupposition, and the focal XP is displaced to the left periphery of the clause.
In contrast, in out-of-the-blue statements the focus has a non-contrastive
interpretation, there is no focal presupposition, and there is no focus movement.

These differences are illustrated in the examples (50a), with a contextualized

utterance, and (50b), with an out-of-the-blue utterance:

(50a) Contextualizad utterance:

[Patatak]F erosi ditu Jonek. F-movement, Contrastive & Focal Presupposition

Potatoes buy aux Jon

“Jon bought [potatoes]F”

175
(50b) Out-of-the-blue utterance:

[Jonek patatak erosi ditu]F. In-Situ, Informative & No Focal Presupposition

Jon potatoes buy aux

“[Jon bought potatoes]F”

Herburger’s (2000) original analysis, whereby focus is mapped into the nuclear
scope of a restricted existential quantification over events, cannot capture those
differences. She assumes that it is a pragmatic issue. I will take a different route and
in order to explain these differences I propose that the question is that propositional
focus (when out-of-the-blue) cannot give a strongly exhaustive answer to the Question
Under Discussion (QUD), the question to which a given sentence can provide an
answer (cf. Roberts (1996)). In fact, it is definitional of out-of-the-blue utterances that
there is no QUD when they are uttered (although intuitively right, a definition does
not suffice to account for the facts). What I want to propose is that the impossibility of
an exhaustive reading with out-of-the-blue utterances follows from their logical form and
the discourse appropriateness conditions that it entails. Let me explain this in detail.
First, in order to set the background of what we are talking about when we talk
about contrastive and informative interpretations, I will assume the idea of the
denotation of a question given by the theory of ‘Partition Semantics’ (cf.
Higginbotham & May (1981), Groenendijk & Stokhof (1982)). These authors argue that
a question like (51) demands two pieces of information in order to be completely
answered, (51-i) and (51-ii):

(51) Who is from San Sebastian?

(i) Who is from San Sebastian?
(ii) Who is not from San Sebastian?

That is, in order to know the answer to question (51) it will not be enough to
know what satisfies the variable in the question; rather, the answer also has to provide

176
in some way the negative information of (51ii) in order to be a satisfactory answer. I
will argue below that normally this is done via an exhaustiveness implicature.
Notably, the need of this second demand of the question is clear if we consider
questions in embedded contexts like (52):

(52) Ainhoa knows who is from San Sebastian.

In order for the sentence in (52) to be truthfully uttered Ainhoa has to know
(within a restricted domain of discourse) who is from San Sebastian, but she also has
to know who is not; otherwise, despite the fact that she might know of some people
that they are from San Sebastian, she will not know that she knows the answer to the
question in (52) (even if she knows the origin of all the people that are from San
Sebastian). In other words, she will have to know how to tease apart the people from
San Sebastian from the people that are not from San Sebastian, something that you
cannot do if you only happen to know about the origin of the people from San
Sebastian. Basically, you have to know that the people you know that are from San
Sebastian are all the people that are from San Sebastian.
Thus, a way of formulating this double requirement of a question in set-
theoretic terms is to assume that a question creates a partition of the world into
mutually exclusive states of nature, where a “partition” is defined as in (53), taken
from Lahiri (2002):

(53) Partition: X is a partition on a set S iff X is a set of non-empty sets such that (i) ∪X
= S, and (ii) for any Y, Z ∈ X, if Y≠ Z, then Y∩ Z = ∅

Under this approach, then, a question like (54a) would create a partition like
(54b), a set of propositions where each of them is a complete specification of possible
states of nature:

177
(54a) Who’s from San Sebastian?
(54b) If p1, …, pn are the people in the world, then
[[who is from San Sebastian]] = {

that p1, …, pn are from San Sebastian

that p1 is from San Sebastian, p2, …, pn are not from San Sebastian
that p2 is from San Sebastian, p1, …, pn are not from San Sebastian
that p1, p2 is from San Sebastian, p3, …, pn are not from San Sebastian
…
that p1, …, pn are not from San Sebastian }

Assuming that this is the correct characterization of the import of a question,

we can now formulate two different types of answer: a ‘complete answer’ (55) and a
‘partial answer’ (56). Following Lahiri (2002), we can define them as follows:

(55) Complete Answer: A proposition p is a complete answer to a question Q iff p is

compatible with exactly one q ∈ Q, i.e., iff there is exactly one
p ∈ Q such that p ∩ q ≠ ∅.

(56) Partial Answer: A proposition p is a partial answer to a question Q iff p is

incompatible with at least one q ∈ Q, i.e., iff there is at least one
p∈ Q such that p ∩ q = ∅.

I think that this is what is at stake when we talk about informative vs.
contrastive interpretations of focus: whether the focus provides a partial or a complete
answer to the Question Under Discussion or not (cf. Roberts (1996)). In other words,
whether the proposition expressed by the utterance is interpreted exhaustively or not.
At this point, the question that arises is the following one: why can a contextualized
utterance be interpreted exhaustively but an out-of-the-blue cannot? Assuming a
partition-like import of questions like the one described above, I would like to propose
that, pragmatically, question interpretations are two-sided processes of gricean

178
reasoning (cf. Grice (1975), Horn (2004)): (i) A first step of a partial answer
interpretation that provides the informativity of the content or ‘what is said’. This is
obtained via the standard compositional semantics, and (ii) a second step that provides
the complete answer interpretation (i.e. exhaustivity or ‘what is implicated’) obtained as
a conventional implicature (cf. Lambrecht (1994) and Spector (2006) and section
4.2.2.2.2). Contextualized utterances would involve both steps while out-of-the-blue ones
would provide just the first one, i.e., the partial information of ‘what is said’ in the
proposition.
Under the theoretical approach I adopt in this dissertation now my goal is to
trace this difference between contextualized utterances and out-of-the-blue utterances
back to their different logical forms. I will argue that the semantic differences derive
directly from their different syntax, which derives from the different type of elements
that constitute the clause. Thus, in the next section I will propose two different types
of quantifiers that carry with them the (im)possibility of these readings.

4.2.2.2. Two Types of Quantifiers for Events

Having stated the formal and interpretive differences between out-of-the-blue
and contextualized utterances, now I would like to propose that the differences
observed in these constructions do not derive from two different types of foci (that is,
they will not be two foci of different nature). Rather the differences observed in both
types of constructions will derive from differences in the types of quantifiers involved
in their logical forms, and, ultimately, from differences in their CP systems (see section
4.3).
Let me sum up a little bit the background I will adopt. As explained in section
4.2, Herburger (2000) proposes that at logical form focus falls within the scope of a
restricted existential quantification over events. For instance, the logical form of the
sentence (57a), a proper answer to the question in (57b), would be as represented in
(57c):

(57a) Ainhoa drank [wine]F.

(57b) What did Ainhoa drink?

179
(57c) ∃e[Agent(e, Ainhoa) & Drink(e) & Past(e)] Theme(e, wine) & Agent(e, Ainhoa) &
Drink(e) & Past(e)

Elegant and powerful as it is, Herburger’s representation falls short in

explaining focus-triggered presuppositions (cf. Eckardt (1998)). Furthermore, it can
neither capture the different properties of out-of-the-blue and contextualized utterances
discussed in chapter 2 and reviewed in the previous section.
Thus, in order to explain the cluster of properties of the different focus
constructions, I would like to propose that in Basque out-of-the-blue and contextualized
sentences involve different event-quantifiers: out-of-the-blue environments involve an
existential quantifier, and contextualized utterances involve a δ-quantifier that creates
definite descriptions of events. This proposal is explained in detail in the next sections.

4.2.2.2.1. ∃ in out-of-the-blue utterances

The semantic properties of an out-of-the-blue statement like (58) can neatly be
accounted for with an analysis based on an existential quantifier over events à la
Herburger (2000). This logical form is sketched in (59):

(58) [Jonek Miren musukatu du]F.

Jon Miren kiss aux
“[Jon kissed Miren]F”

(59) ∃e [FOCUS]

In the schema of (59) we have an existential quantification over events with the
event predicate and its participants in the scope of the quantification, as in (60):

(60) ∃e [Agent(e, Jon) & Kiss(e) & Past(e) & Theme(e, Miren)]

As I said in chapter 2, out-of-the-blue sentences like (58) have a cluster of

semantic properties that distinguish them from contextualized utterances. In particular

180
out-of-the-blue sentences have a nonexhaustive interpretation and they do not trigger
any focal presupposition (contrary to contextualized utterances). The logical form in (60)
predicts all these propoerties of out-of-the-blue sentences: first, we expect these
constructions to have no focally induced presupposition, since there is no contentful
restriction in the existential quantification; i.e. nothing under discussion. Furthermore,
the information focus reading of these constructions is also to be expected given that it
is characteristic of existential quantifiers that they introduce new elements into the
discourse. If we look at the discourse-contribution of these constructions from a
dynamic point of view (cf. Kamp (1981), Kamp & Reyle (1993), as well as Asher (1993),
Vallduví (1993), von Heusinger (1999), Erteschik-Shir (1997) and Gómez Txurruka
(2002)), the existential quantifier over events will create an index card for an event that
is introduced in the discourse. For instance, for the sentence in (58), we would have
the simplified Discourse Representation Structure (DRS) in (61) (cf. Kamp & Reyle
(1993) for a fully developed analysis):

(61) Discourse Representation Structure (simplified)

e
Kissing(e)
Theme(e, Miren)
Past(e)
Agent(e, Jon)

This is a purely presentational DRS where all the predicates of the event are
introduced at the same level. Hence, the informativity and nonpresuppositionality of
these constructions are just natural and expected properties of the quantifications they
involve.
My proposal is that contextualized utterances differ from out-of-the-blue
statements in that they are introduced by a different quantifier.

181
4.2.2.2.2. δ in contextualized utterances
We just saw that a unary ∃ quantifier makes the correct predictions for out-of-
the-blue statements: the informativity and non-presuppositionality of these constructions.
Now, in order to capture the interpretation of contextualized utterances like (63b)
below, I would like to propose that their logical form is as roughly represented in (62),
where δ (taken and adapted from Hegarty (1992)) is a quantifier employed to
construct definite descriptions of events that will anaphorically anchor into the
discourse19. That is, δ will select event cards from the discourse frame and predicate
over them (just like the ι quantifier does with individuals). Thus, in its logical form δ
takes all the nonfocal material in its restriction, and the focal material (as well as all the
rest of the clause within its nuclear scope:

(62) δe [BACKGROUND] [FOCUS]

Following the proposal I have just made according to which contextualized

utterances involve a δ-quantifier, a sentence like (63b) (as an answer to the question in
(63a)) will have the logical form in (63c):

(63a) What did Jon buy?

(63b) [Patatak] F erosi ditu Jonek.

potatoes buy aux Jon
“Jon bought [potatoes]F”

(63c) δe [Buy(e) & Past(e) & Agent(e, Jon)] Theme(e, potatoes) & Buy(e) & Past(e) &
Agent(e, Jon)

This representation roughly says that “the past event of buying by John was an
event of buying potatoes”. Notice that the semantic properties of a contextualized

19 Hegarty’s (1992) original formulation is framed within Heim’s (1982) File Change Semantics; thus δ
selects file cards within the discourse frame. Thanks to Javi Ormazabal for bringing this work to my
attention.

182
utterance like (63b) are naturally predicted from the logical form representation I have
proposed in (63c):
(i) Presuppositionality: First, the focally induced presupposition derives directly
from the availability of a restricted quantification in the definite description; the
restriction provides the presuppositional background material (the information on the
anaphoricity of the event description). For the sake of simplicity, let me adopt here an
ontologically naïve and pragmatic view of presuppositions (cf. Karttunen (1974)):

(64) Presupposition: Surface sentence A pragmatically presupposes a logical form L, if

and only if it is the case that A can be felicitously uttered only in
contexts which entail L.

In dynamic terms, we would have the event introduced in the question of

(63a) as the proper antecedent for the focal presupposition of (63b), i.e., the restriction
of the quantification in (63c). Thus, uttering (63b) we would have an update of the
discourse frame: having introduced the question of (63a) we have the DRS in (65a),
then, the answer provides a thematic predicate to that event description (65b):

(65) Discourse Representation Structure (simplified)

a. b.
e e
Buying(e) Buying(e)
Agent(e, Jon) Agent(e, Jon)
Past(e) Past(e)
Theme(e, potatoes)

The definite description of the event of buying is used to update the description
of an event that was previously introduced in the discourse.
Furthermore, with the analysis just presented, we can account not only for the
presuppositionality triggered by focus, but also for the lack of it in the contexts where

183
it disappears. In chapter 2 I showed that the focal presupposition disappears in a
presupposition filter environment like the apodosis of a conditional. This behavior
was used to support the presuppositional analysis of focus:

(66) If anybody cheered, then [John]F cheered.

In (66), we have a focused sentence in the apodosis (“[John]F cheered”) whose

focally induced presupposition (“that someone cheered”) is not projected up to the
whole sentence. The content of the protasis provides a contextual antecedent for
anchoring the definite description of events but crucially, the lack of an existential
assertion for the protasis, due to the conditional nature of the construction, forces the
focal presupposition triggered by the apodosis to be cancelled:

(67) ∃e[cheer(e) & Past(e) & ∃x[Agent(e, x)] ⇒ δe[cheer(e) & Past(e) & ∃x[Agent(e, x)]]
cheer(e) & Past(e) & ∃x[Agent(e, x)] & John(x)

(ii) Exhaustivity: Moreover, under the analysis I propose here the other major
characteristic of these constructions, that of their contrastivity, will not be any more
mysterious than their presuppositionality. The contrastive/exhaustive essence of
contextualized utterances follows naturally from the uniqueness restriction of the
definite quantification, that, I assume, is lexically encoded in δ as a conventional
implicature. Having this restriction, there can be no other event that satisfies the
restriction and thus, we can capture focal contrast without postulating extraneous
elements like a contextually built set of focal alternatives.
Summing up: the dual analysis of focal constructions I defend, which postulates
different types of quantificational structures, allows us to account for the clustered
semantic properties of out-of-the-blue and contextualized utterances in a straightforward
way.
Nevertheless, there was one more property that came in the cluster: the
syntactic movement observed in contextualized sentences that is lacking in out-of-the-
blue ones. This constitutes the topic of next section. There I will provide an analysis of

184
this distinction as deriving from differences in the CP system of each type of
construction. I will argue (i) that the lack of movement in out-of-the-blue sentences is
due to a CP system which has no left peripheric FocP (a ‘simple’ CP) and (ii) that the
mapping from that syntactic structure to its logical form is transparent. On the other
hand, the movements observed in contextualized sentences are due to a more
articulated CP structure involving FocP and TopP projections à la Rizzi (1997) (a
‘complex’ CP). As I will argue, the focus movement will be an essential operation in
order to get the logical form representation of the binary quantification that I just
presented.

4.3. The Road to logical form

In this section I will put forward my analysis of the syntax-semantics interface
of focus. The goal is to get to the logical forms just proposed in section 4.2 from the
syntax defended in section 4.1 in a direct and dynamic way. First, in section 4.3.1, I
will propose that the event quantifier is introduced in the Finº head (adapting Rizzi’s
(1997) split-CP system). Then, in section 4.3.2, I provide my analysis of the syntax-
semantics interface of contextualied and out-of-the-blue utterances. I will argue that the
focus movement observed in Basque is an essential syntactic step towards the logical
form representations I just discussed in the previous section20.

4.3.1. The Introduction of the Event Quantifier

Assuming a direct mapping from syntax to logical form and minimalist

desiderata, the first question that we have to solve is the following one: Where does
the (implicit) Davidsonian quantifier come from if we are not going to violate the
Inclusiveness Condition?
Let us look at the sentence in (68a), whose partial parsetree would be as
represented in (68b)21:

20 Even though my analysis of the syntax-semantics interface will be presented on the basis of Basque
data, it is meant to be crosslinguistically valid.
21 This is a simplified representation for the ease of exposition. See Pietroski (2003a) for an analysis of

the semantic contribution of ‘v’ as relating two eventualities.

185
(68a) Milan bought [cider]F
(68b)
VP e[Agent(e, Milan) & Buy(e) & Theme(e, cider)]
qp
Milan V’ e[Buy(e) & Theme(e, cider)]
qp
buy e: e is a buying cider

Now, since the vP denotes a property of events and not yet a truth value,
scholars like Higginbotham (1985) or Kratzer (1996) have proposed that a higher
functional head introduces the existential closure of the event denoted by the verb.
According to their proposals it is Inflº. The idea is appealing but we would run into a
problem whenever we had Tº (or Inflº) marked as [+F] for the same head would
introduce the Q and its nuclear scope22. Furthermore, there is convincing evidence
against treating Tense as an operator and arguments in favor of analyzing it as a
predicate of events that orders an interval of the event with respect to the utterance
time (cf. Stowell (1996), Demirdache & Uribe-Etxebarria (2000, 2007) and
Higginbotham (2002) among others). Given this state of affairs, another syntactic node
will have to introduce the quantifier to the event denoted by the verb.
We saw in section 4.1 that the best way to analyze the syntax of focus in Basque
was to postulate a left peripheric head attracting the focal element to its specifier.
Thus, let me assume the ‘split CP’ proposed by Rizzi (1997) with four projections
‘ForceP’, ‘TopP’, ‘FocP’ and ‘FinP’, where Rizzi describes ‘TopP’ and ‘FocP’ as
optional, as roughly illustrated in (69):

(69) Force .... (Topic) ... (Focus) ... Fin ... TP

Adopting this system, I would like to propose that it is the Finiteness head that
introduces the quantifier to the event denoted by the verb, whereas Tenseº (or Inflº)

22 Examples of focused tenses might not be very common but as I argued in section 3.2.3, this might be a
matter of information flow (‘informatics’ in the sense of Vallduví (1993)). Not a restriction in the
computational system. In fact, relevant examples can be constructed in contrastive environments like
(i):
(i) A: I heard that John is married.
B: No, he WAS married! (i.e., he is not married anymore)

186
introduces a temporal predicate of events that combines with the vP via predicate
conjunction23. This is represented in (70), where Q stands for the corresponding
quantifier (δ or ∃)24:

(70)
FinP Qe[…& T(e)...] [ ... ]
qp
Fin TP e[... & T(e)...]
Q 3
T vP
T(e) e[...]

With this assumption, I would want to trace the syntactic differences between
contextualized and out-of-the-blue utterances back to the CP systems of the sentences.
Thus, my proposal would be that we have two types of CP systems, a full CP for
contextualized utterances with the four projections above IP in (69) and a δ quantifier
sitting in Finº, as represented in (71a). Out-of-the-blue utterances, on the other hand
would have a simpler CP with just the obligatory ForceP and FinP, as in (71b):

(71a) Contextualized Utterances (71b) Out-of-the-blue utterance

ForceP ForceP
2 2
Spec. Force’ Spec. Force’
2 2
Forceº TopP Forceº FinP
2 2
Spec. Top’ Spec. Fin’
2 2
Topº FocP Finº [...]
2 (=∃)
Spec. FocP
2
Focº FinP
2
Finº [...]
(= δ)

23 This is a simplified version (see references above).

24 For clarity of exposition I will represent left-headed projections.

187
Looking at it from a selectional point of view, we can say that different Forceº
heads will select for different complements: a complex declarative Forceº that relates
the assertion to the previous discourse will select a TopP that will serve as a link in
Vallduví’s (1993) sense. This anchors the proposition in the discourse (i.e., with
contextualized utterances like (71a)). On the other hand a simple declarative Forceº will
select a bare existential Finº providing no anaphoric discourse anchoring (i.e., out-of-
the-blue utterances like (71b)).
So, the idea I want to put forward is that the logical form of each type of
construction mirrors directly their syntax; that is, a higher syntactic complexity is used
for a higher semantico-discoursive complexity. In this respect, sentence form and
sentence function would go hand in hand, as I explain in the next section.

4.3.2. The Syntax-Semantics Interface

In this section I present my analysis of the syntax-semantics interface of focus. I
propose that the mapping from syntax to logical form in out-of-the-blue sentences is a
direct one, whereas in contextualized utterances the syntax-logical form mapping is more
complex: it involves more structure and more operations; as a result, the logical form
representation of these sentences will be more articulated.

4.3.2.1. Out-of-the-blue Sentences

Assuming the logical form representation proposed in 4.2.2.2.1, and the syntactic
structure of (71b), the syntax-logical form interface of out-of-the-blue sentences will be
straightforward: the ∃ placed in Finº selects a completely focal sister (the out-of-the-blue
phrase) taking scope over it. This, directly, provides an ‘everything in situ’ sentence
like (72a) and a transparent mapping to logical form as in (72b).

(72a) [Jonek patatak erosi ditu]F.

Jon potatoes buy aux
“[Jon bought potatoes]F”

188
(72b)
FinP
qp
Fin TP
∃ 3
T[+F] vP
3
Jonek[+F] v’ SCOPE
3
v[+F] VP
3
patatak[+F] erosi[+F]

∃e [Buy(e) & Past(e) & Agent(e, John) & Theme(e, potatoes)]

That is, the derivation takes place without involving any focal movement given
that there is no [+F] feature attractor (no Focº) in the left periphery. The clause
structure in (72b) suffices to provide the existential logical form as a direct mapping
from its simple syntax25.

As I show next, the syntax-logical form mapping of contextualized utterances is

more complex.

4.3.2.2. Contextualized Sentences

In this section I present my analysis of the syntax-logical form mapping for
contextualized utterances. The analysis will be based on Hornstein & Uriagereka’s
(2002) label reprojection mechanism26. In order to familiarize the reader with that
proposal, I first summarize their analysis in section 4.3.2.2.1. Then, in section 4.3.2.2.2 I
present a novel proposal for the syntax-semantics interface of contextualized sentences.
My proposal unifies the logical form I proposed in 4.2.2.2.2 with the clause structure of
contextualized utterances I put forward in 4.3.1 via Hornstein & Uriagereka’s (2002)
reprojection mechanism.

25 As opposed to these out-of-the-blue sentences we will see in next section that contextualized sentences
involve movements to the left periphery. As I will argue, these movements provide the specificity
reading of the definite description of events; in this vein, the approach I am developing here threads
within a whole tradition relating specificity to left peripheric positions (cf. i.a. Diesing (1992)).
26 Thanks to Juan Uriagereka for suggesting me this route.

189
4.3.2.2.1. Dynamic Label Reprojections
In this section I present one of the key steps for the syntax-semantics interface
of focus: the label reprojections proposed by Hornstein & Uriagereka (1999, 2002). As I
will argue, it is this mechanism that will provide the syntax-semantics interface of
focus constructions (cf. section 4.3.2.2.2).
Hornstein & Uriagereka (1999, 2002) implement in a dynamic fashion the
relational view of binary quantifiers proposed in Larson (1991) and Larson & Segal
(1995), which I briefly reviewed in section 3.1.3.2.2 of this dissertation. Their point of
departure is the observation that binary quantifiers take their restriction and scope in a
syntactically ordered way (first the restriction, then the nuclear scope) and propose a
Bare Phrase Structure analysis of the syntax of these quantifiers.
Hornstein & Uriagereka (1999, 2002) propose that in Bare Phrase Structure terms
(cf. chapter 3) a given phrase marker could have different labelings at different stages
of the derivation. This would arise via a “reprojection” mechanism. A reprojection
turns a phrase marker {Y, {Y, X}} into a phrase marker {X, {Y, X}}. Formally, this
mechanism will be simply a label ‘projection’ that takes place for a second time in the
same geometrical object:

(73)
{Y, {Y, X}} reprojection {X, {Y, X}}
2 2
Y X Y X

Within these authors’ analysis, and applying the idea that quantifiers are
predicative, reprojection allows us to ensure the locality of quantifiers’ argument-
taking without the stipulation of a null pro in their specifier (cf. Larson’s (1991)
analysis in section 1.3.1.3.7). In Hornstein & Uriagereka’s proposal, binary quantifiers
like ‘most’ in (74), acquire their ‘internal argument’ (i.e., the restrictor ‘people’) just by
being first-merged with them. Then they get the ‘external argument’ (i.e., the nuclear
scope) in the course of the derivation. Following Hornstein & Uriagereka’s proposal,
we can ensure the locality of the semantic composition of the Q and its arguments if
we allow the Q to reproject higher up in the phrase, turning the phrase marker that at
some derivational time Dt was {I, {Q, I}} into {Q, {Q, I}} at a later derivational time Dt’.

190
This is illustrated in (75). Reprojection, by extending the Q’s projection space, allows
the Q to take the external argument ‘t love children’ in its extended local domain:

(74) Most people love children.

(75)
Derivational Time Dt:
IP
qp
QPi I’
2 3
most people I VP
3
ti love children

Q restriction

Derivational Time Dt’ (after reprojection of the Q):

QP
qp
Q’i IP
2 3
most people I VP
3
ti love children

Q restriction scope

Besides providing a neat syntax-logical form mapping for binary quantifiers, the
analysis has several welcome empirical consequences. Among others, it provides, for
instance, a natural explanation for the appearance of some logical form islands (see
Hornstein & Uriagereka (2002) and Hornstein, Lasnik & Uriagereka (2007) for details).
With this much background, let me address in the next section the syntax-semantics
interface of focus constructions.

191
4.3.2.2.2. A New Proposal for the Syntax-Semantics Interface of Contextualized
Utterances
Assuming the DAFS presented in chapter 3 and the phrase structure for
contextualized sentences I proposed in (71a), I will now present my analysis of the
derivation of the syntax-semantics interface of sentences like (76):

(76) [Ogia]F jan du Urtzik.

bread eat AUX Urtzi
“Urtzi ate [bread]F”

As we saw in section 4.1, the best analysis of a sentence with narrow focus on
the object like that in (76) is by positing that the focal element is attracted to the
specifier position of a FocP. Within the clausal architecture I adopted from Rizzi (1997)
and Ortiz de Urbina (1999), this projection is above Finº, which is the introductor of
the event quantifier. Under this analysis then, sentence (76) involves a derivational
stage like (77) where the Finº (i.e. the lexically binary δ) has been merged with TP. This
gives to δ its internal argument by first-merge:

(77)
FinP
qp
Fin TP
δ 3
T vP
3
Urtzi v’ RESTRICTION
3
v VP
3
ogia[+F] jan

192
Then the next head of the articulated CP is inserted: Focº. As argued in section
4.1 this head has an uniterpretable [-F] feature that has to be checked so it attracts the
[+F] marked element to its specifier:27,28
(78)
FocP
qp
ogia[+F] Foc'
qp
Focº FinP
qp
Fin TP
δ 3
T vP
[Urtzi ogia[+F] jan]

The Topº head is merged next:

(79)
TopP
qp
Topº FocP
qp
ogia[+F] Foc'
qp
Focº FinP
qp
Fin TP
δ 3
T vP
[Urtzi jan]

27 Given the DAFS presented in chapter 3 I will assume that the focal head attracts all the [+F] featured

material (cf. Bošković (1999) and Jeong (2003, 2007)). Since in (78) there is only one element marked as
[+F] that will be the only one moved to Spec-FocP
28 This movement will have to leave no copy behind (or the copy will have to be necessarily deleted at

least before we arrive at a logical form). This argument, stipulative as it is, seems to be necessary given
that focus movements in Basque show scope-freezing effects where reconstruction is impossible. For
instance, the sentence in (i) with a focalized object lacks the reading where the universal quantifier takes
scope over the existential one:
(i) [Txakur bat]F maite dute haur guztiek. √ ∃>∀, *∀>∃
dog one love AUX child all
‘All the children love one dog’

193
Now, following the relational view of quantifiers of Larson (1991) and Larson &
Segal (1995) (cf. section 1.3.1.3.7), and the reprojection mechanism proposed in
Hornstein & Uriagereka (1999, 2002) that I just presented, I want to suggest that the
lexically binary δ quantifier in Finº that obtained its “internal argument” (TP) when
first merged with it, is moved to Spec-TopP pied-piping its restriction. As we saw in
section 4.2.2.2.2.2, δ creates anaphoric expressions, events that have discursive
antecedents, and we can assume that δ is inherently topical. So, the δP/FinP moves to
TopicP given its topical/anaphoric feature, an operation that amounts to a Quantifier
Raising of the binary δ. Then, the FinP sitting under TopP reprojects, thus gaining the
“external argument” in its extended local domain (80):29

(80)
TopP
qp
FinP Top’
2 qp
Fin TP Topº FocP
qp
ogia[+F] Foc'
qp
Focº FinP
qp
Fin TP
QR δ 3
T vP
[Urtzi jan]

29 Note that this QR movement, if overt, could be the trigger for the remnant movement operations
argued for in Uribe-Etxebarria (2002), Etxepare & Uribe-Etxebarria (2005) and Ortiz de Urbina (2002)
when analyzing “in situ”-like Wh constructions and corrective focalizations in Spanish and Basque
respectively (cf. section 4.1.2.7). Thus, we could derive both unmarked (XP fronted) and marked (XP “in
situ”) constructions in a unitary way (pace Dominguez (2004)). The asymmetries between both
constructions would derive simply from timing differences between the reprojection and Spell Out,
something I will not explore here.

194
(81)

FinP
qp
FinP TopP
2 qp
Fin TP Topº FocP
qp
ogia[+F] Foc'
qp
Focº FinP
qp
Fin TP
δ 3
T vP
[Urtzi jan]

Q Restriction Scope
δ [[Eat(e) & Past(e) & Agent(e, Urtzi)] Theme(e, bread) & Eat(e) & Past(e) & Agent(e, Urtzi)

That is, by way of the reprojection, we get the quantificational phrase

(FinP/ δP) at the top node dominating a quantifier (δ), its restriction (TP) and its
nuclear scope (TopP).
A more general point is that with the adoption of this system, the postulation of
a left peripheric head attracting the focal material will not be just a way of coding a
movement that we observe in syntax (as reasonably denounced in Reinhart (2006:
126)). The focal movement will be an essential operation for the syntax-semantics
interface, since it is this movement that will allow the focal XP to escape from the
restriction of the δ quantifier.30
Finally, a possible puzzle arises concerning the timing of the reprojection: that a
reprojection could take place overtly is seen as problematic by Hornstein & Uriagereka
(2002), who argue that it would change the checking relations (notably for case and
agreement). Hence, they claim that reprojections would have to take place covertly
and very late in the derivation, once all these relations have already been exploited in
overt syntax. I have no qualms for this move. However, in principle, there should be
no problem in having a system where reprojection can take place after all agreement

30 Obviously, after QR, some ‘trace conversion’ will have to take place deleting the lower δ quantifier.

195
relations are set, and just before Spell Out (as potentially in the proposal just sketched
for focus, where the reprojection of Finº takes place in the ‘C-domain’).31

4.4. Summary
In this chapter I have offered a dynamic analysis of the syntax-logical form
interface for Basque focalizations. First, in section 4.1 I showed that in Basque
focalizations a leftward displacement of the focal XP is involved, and that we observe
a number of expected locality restrictions for this movement. Thus, I concluded that a
cartographic approach to focus provides a good understanding of the phenomenon
where a NSR-based approach falls short. Then, in section 4.2 I analyzed the logical form
for focus constructions, proposing a dual analysis for out-of-the-blue and contextualized
utterances: out-of-the-blue sentences would have a simple logical form characterized by
an existential quantification over events whereas contextualized sentences would have
the logical form of a definite description of events. I showed that with this proposal the
semantico-pragmatic properties of contrastivity and presuppositionality that I showed
in chapter 2 derive directly from these logical form rerpresentations. Finally, in section
4.3 I proposed a dynamic analysis of the syntax-logical form interface, getting the logical
form representations of section 4.2 from the syntactic analysis of section 4.1 in a direct
and natural way.

See Grohmann (2000) and Platzack (2001) for discussion on the different functions of the vP, TP and
31

CP domains.

196
CHAPTER 5. INTERFACE PROPERTIES AND THE

ARCHITECTURE OF GRAMMAR

In this chapter I analyze the basic postulates of the mainstream NSR-based

approaches to F-Structure. I show that this type of approach has many empirical and
conceptual drawbacks and that the approach I have presented in chapters 3 and 4
overcomes them in a straightforward way.
The chapter is organized as follows: in section 5.1 I offer a summary of the
core ideas of the NSR-based approaches to F-Structure. Section 5.2 addresses some
empirical evidence against them. To be more specific, in 5.2.1 I discuss empirical
evidence; namely: (i) the categorial difference in the accents conveying narrow and
broad foci, (ii) the focus-induced phonological phrasing facts, (iii) the lack of any
accent in some focus constructions, and (iv) the evidence based on the syntactic
pattern presented in chapter 4. Section 5.2.2 is a detailed discussion of several
conceptual drawbacks that NSR-based approaches to F-Structure face from a
minimalist point of view. Finally, in section 5.3 I summarize the major conclusions of
the discussion.

5.1. Theories of F-Structure based on the Nuclear Stress Rule

As mentioned in the previous chapter, a common assumption in the literature
is that we can understand the fact that focus tends to bear the nuclear stress in
several languages as deriving from a representational legibility condition on
derivations that requires focus to have nuclear stress at PF. This is the line of research
adopted by some referential works such as Selkirk (1995), Reinhart (1995, 2006),
Neeleman & Reinhart (1998), Zubizarreta (1998), Zubizarreta & Vergnaud (2000),
Szendröi (2001), Arregi (2003a), Elordieta (2001), and Ishihara (2003), which adopt the
original ideas of Cinque (1993) proposing some readjustments and refinements.
While there are some differences in the implementation of their analyses, what all
these approaches have in common is that they propose that the different F-Structures

197
observed in the sentences in (1a-g) would be amenable to an explanation in terms of
‘focus projection’; the element that bears the nuclear stress ‘projects’ its focal status to
higher nodes that dominate it. Thus, with a single accent placement we can have
different F-Structures:

(1a) Jesus preached to the people of [Judea]F

(1b) Jesus preached to the people [of Judea]F
(1c) Jesus preached to the [people of Judea]F
(1d) Jesus preached to [the people of Judea]F
(1e) Jesus preached [to the people of Judea]F
(1f) Jesus [preached to the people of Judea]F
(1g) [Jesus preached to the people of Judea]F

As explained in chapter 4, in NSR-based approaches there is no F-Structure as

such in narrow syntax; rather, the focus set has just to be inferred from the placement
of the nuclear stress in PF. Any element containing the nuclear stress might be
interpreted as focus, and hence, a single nuclear stress placement might mark
different F-Structures. This view implies as a natural consequence the notion of focus
projection. According to this view, if nuclear stress falls on the most deeply embedded
element in (1), this stress placement will be able to convey many different F-
Structures (all the constituents containing the nuclear stress up to the whole sentence
(cf. (1a-g)).
The main consequence that follows from this interrelation between phonology
(via the nuclear stress), syntax (the properties if clausal architecture), and semantics
(the focus of a sentence) is that a sentence won’t have an ‘actual focus’ per se but
rather ‘a set of possible foci’, that is, the set of nodes that an actual nuclear stress
placement can mark as focused. This idea is explicitly stated, for instance, in Reinhart
(2006: 158)1:

1 Selkirk (1995) proposes an analogous rule of focus projection from heads to phrases that has also

been very influential (cf. i.a. Schwarzschild (1999)). The main difference between the proposals in
Selkirk (1995) and Reinhart (1995, 2006) is that for the former the PF rule of projection applies
optionally and for the latter it applies automatically (creating focus sets).

198
(2) The focus set: The focus set of a derivation D includes all and only the constituents
that contain the main stress of D.

This focus rule accounts for the basic facts regarding focus projection in the
different sentences in (1). Under this approach, in a SVO sentence of a language like
English this rule allows the nuclear stress on the object to mark as focused either the
direct object itself, the VP, or the whole clause. According to Reinhart (1995, 2006),
the default case is that where NSR assigns nuclear stress to the most embedded
position2. The ‘focus projection’ is seen as automatic, and then the discourse will
decide what the actual focus is from among the elements in the focus set.
However, as observed by Reinhart (1995, 2006), this strategy will not serve to
mark focus on certain elements like the subject or the verb since, clearly, they don’t
contain the nuclear stressed element (providing that, as said before, by default the
NSR assigns the nuclear stress to the most deeply embedded position in absolute
terms (i.e., the direct object)). Thus, according to Reinhart (1995, 2006), in order to
mark focus on a phrase that cannot be focused by the focus projection of the object
(i.e., a phrase which is not in the original focus set), some marked strategies must be
employed. For instance, in English-like languages —where focus doesn’t affect word
order—, a deaccentuation rule will deaccent the object and a marked stress rule will
assign nuclear stress to whichever element has to be interpreted as focused (3)3:

(3) Accent Shift:

Step1 (base) Ö SVO
Step2 (NSR) Ö SVO
Step3 (deaccentuation) Ö SVO
Step4 (marked stress placement) Ö SVO

2 Likewise for sentences with oblique objects.

3 Here I mark nuclear stress with boldface and with an underline for clarity.

199
On the other hand, as we saw in chapter 4, in languages like Basque, where
focus affects the basic order of constituents, it is posited that those elements that are
most deeply embedded must scramble to a position higher up in the structure. As a
result of this scrambling operation, the element that has to be interpreted as focus
becomes the most deeply embedded one and thus it can receive nuclear stress as
roughly illustrated in (4):

(4) Displacement:
Step1 (base) Ö SOV
Step2 (NSR) Ö SOV
Step3 (deaccentuation) Ö SOV
Step4 (scrambling of O over S) Ö OSV
Step5 (NSR) Ö OSV

Having the nuclear stress on the subject, a new focus set is created and the
elements in this set (in particular the subject) can be interpreted as focus (see below).
So, these two strategies (namely, (i) stress shifts in languages like English, and
(ii) scrambling of unfocused constituents in languages like Basque) provide the
intended focus-nuclear stress correlation at PF.4
There is still a question for this type of approach, though. According to the
focus rule in (2), nuclear stress placement on a subject should also be able to denote
sentence-focus. This is so because the CP, dominating all material, contains the
nuclear stressed item when it is on the subject. Likewise for TP5. Thus, if we apply
any of the marked operations proposed by NSR-based approaches that I have
illustrated in (3) and (4), then, the focus set of a sentence with nuclear stress on the
subject is {Subject, TP, CP}. However, Reinhart (1995, 2006) argues that this is not an
available option because it is anti-economical: the sentence-focus interpretation could
have been obtained via the regular projection algorithm from the nuclear stress on
the object in the base configuration (just with the steps 1 and 2 in (3) and (4)) and

4 cf. Reinhart (2006) for the specificities of these operations and chapter 4 of this dissertation for

discussion of Basque focus constructions.

5 Let me assume for the sake of the argument that the subject is out of the vP in SpecTP.

200
without having to incur into marked operations. That is, Reinhart (1995, 2006) argues
is that it is more economical to interpret sentence focus from the projection of the
NSR that would apply to the object by default than to interpret it as the outcome of
three operations: (i) default NSR to the object first, (ii) deaccenting it, and (iii)
applying the ‘marked’ stress shift of (3) to assign nuclear stress to the subject, then
interpreting sentence-focus from the focus set projected from the accent on the subject.
Therefore, Reinhart (1995, 2006) concludes that an economy principle would prevent
that choice. Under Reinhart’s theory, economy is calculated by comparison of both
derivational possibilities for interpreting focus on the whole clause (what Reinhart
(2006) calls a ‘reference-set computation’). Thus, even if both focus sets of the
sentences in (5) and (6) (with nuclear stress on the object and subject respectively)
contain TP and CP, the choice of TP or CP as the actual focus in a sentence with
nuclear stress on the subject would be antieconomical, as those options would also be
available with nuclear stress on the object6:

(5) John likes wine.

Focus set: {Obj, VP, TP, CP}

(6) John likes wine.

Focus set: {Subj, TP, CP}

Based on this view of the NSR and F-Structure, a very productive program of
research has been recently developed refining the details of the theory and applying
it to previously unstudied languages (cf. i.a., Reinhart (1995, 2006) and Neeleman &
Reinhart (1998) for English and Dutch, Zubizarreta (1998) and Dominguez (2004) for
Spanish, A. Elordieta (2001) and Arregi (2003) for Basque, Szendröi (2003) for
Hungarian, English and Italian, Ishihara (2003) for Japanese, Arnaudova (2003) for
Bulgarian and Georgiafentis (2004) for Greek). I will refer to this program of research
as the ‘NSR-based theories of F-Structure’, given that what all these proposals have

6Following Reinhart’s notation, potential foci that are not available due to their ‘antieconomical’
nature are marked with italics.

201
in common is that, according to them, there is no focus phrase per se during the
syntactic computation to the interfaces. The gist of all these approaches is that there
is a legibility condition on derivations that requires focus to have nuclear stress at PF.
Focus is calculated taking into account the phrases that contain the nuclear stress in
the PF representation. Under this view, nuclear stress placement is prior to the
setting of F-Structure. As a result, the F-Structure of a sentence is computationally
ambiguous (since the accented element denotes, via focus projection, a set of different
F-Structures).
In the following section I provide empirical and conceptual arguments that
show that this type of theory cannot be maintained.

5.2. NSR-based Theories and the Architecture of Grammar

In this section I present both empirical and conceptual arguments against the
NSR-based theories of F-Structure. The evidence I will introduce shows that this type
of theory is untenable, not only because there are counterexamples to it, but because
it cannot be operative in computational terms. In the discussion section (5.3) I will
argue that the proposal in this thesis is to be preferred for it overcomes all these
problems.
Here, I will first discuss empirical evidence against NSR-based theories of F-
Structure. In section 5.2.1.1 I will provide phonological evidence while section 5.2.1.2
will be devoted to syntactic evidence. Then, in section 5.2.2 I will turn to more
conceptual arguments which pose theoretical problems for any NSR-based theory of
F-Structure, showing that this type of theory cannot be maintained from a minimalist
point of view because of the architectural paradoxes it carries.

5.2.1. Empirical Evidence Against NSR-based Approaches

In this section I will critically review empirical evidence against NSR-based
approaches to the F-Structure. My goal is not just to provide a list of
counterexamples for this theory. Rather, I want to discuss a set of empirical
phenomena that demonstrate that these theories are conceptually ill formed.
Ultimately I will try to show that NSR-based approaches to F-Structure cannot be

202
modified so that the evidence I will discuss in the next sections falls under their
scope (cf. as well chapter 6 of this dissertation).

5.2.1.1 Phonological Evidence

In this section I will present some of the phonological evidence against NSR-
based theories of F-Structure. The evidence I will discuss in sections 5.2.1.1.1 and
5.2.1.1.2 shows that in many cases the phonological structure of a sentence shows
unambiguously a specific F-Structure, and hence, that F-Structure is fixed. This fact
constitutes evidence against the existence of focus sets. In section 5.2.1.1.3 I will
review the facts of some languages that don’t show any accent on the focus phrase,
which contradicts the main axiom of NSR-based theories of F-Structure.

5.2.1.1.1 Categorially Different Nuclear Pitch Accents

As explained in the previous section, according to NSR-based theories of F-
Structure there is a Bare Output Condition that requires focused material to bear the
nuclear stress. This stress placement, via projection, is able to mark as focused any
phrase that contains the nuclear stress at PF (cf. (2)). Then, economy conditions will
dictate what the actual focus possibilities are, and the actual focus will be chosen
from the focus set at the discourse level.
If I understand this approach correctly, nuclear accents are taken by these
theories to be some sort of [+Accented] diacritics (i.e., suprasegmental features with
no internal tonal structure nor autosegmental alignment information). However, at
least since Liberman (1975) and Pierrehumbert’s (1980) seminal works on English
intonation, pitch-accents with different tune-structures have been described
associated with the expression of different meanings (see, among others, Ward &
Hirschberg (1989), Pierrehumbert & Hirschberg (1990), Hirschberg & Ward (1992),
Ladd (1996)). Moreover, in many of the best studied languages within the
Autosegmental-Metrical framework of intonational phonology we observe
systematic differences between the nuclear accents used to convey narrow focus and
those used to convey broad focus. Crucially, the different tonal compositions and
alignments of pitch-accents are categorial distinctive features of accents; that is, H*,

203
L+H* and L*+H accents are different phonological entities stored as such in the
‘intonational lexicon’ of a given language7. Among the languages or varieties that
make use of categorially different pitch-accents to denote broad and narrow foci we
can mention the following ones8:

(7)
-English (cf. Selkirk (2002a)): Broad Focus H*, Narrow Focus L+H*.
-Bengali (cf. Hayes & Lahiri (1991)): Broad Focus H*/L+H*, Narrow Focus:
L* HP LI.
-Italian (cf. D’Imperio (2002)): Broad Focus H+L*, Narrow Focus L+H*.
-European Portuguese (variety of Lisbon) (cf. Frota (2000)): Broad Focus
H+L*, Narrow Focus H*+L.
-Greek (cf. Baltazani (2002)): Broad Focus H*/H*+L, Narrow Focus L+H*.
-Madrid Spanish (cf. Face (2002)): Broad Focus L*+H, Narrow Focus L+H*
-Errenteria Basque (cf. Irurtzun (2003)): Broad Focus H*, Narrow Focus
H*+L.

Hence, if we pay attention to these facts, a stressed direct object in a SVO

configuration like (8) (for a head-first language like English) doesn’t convey an
ambiguous F-Structure. That is, there is a correlation between type of focus and type
of accent and an accent cannot mark broad and narrow foci at the same time9, 10:

7 As I have shown in detail in chapters 2 and 4, broad focus and narrow focus have different
interpretations regarding exhaustiveness and presupposition. These differences correlate with the
choice of a narrow focus accent or broad focus accent. But what is really relevant for the matter at
hand is that the difference in pitch accents implies that the actual F-Structure has to be already set for
the phonological component to assign an ‘appropriate’ nuclear accent to each type of focus or F-
Structure.
8 See as well the Hirst & Di Cristo (1998) for a wide variety of languages that display these

phonological differences.
9 Obviously, there might be some cases of real ambiguity, but the important issue here is that the fact

that there are unambiguous structures invalidates the hypothesis of the focus set.
10 The subscripts 'NF' and 'BF' stand for 'Narrow Focus' and 'Broad Focus' respectively.

204
(8) ?
AA g
*[S [V [O](NF)](BF)](BF)

If the assigned accent is the one used to convey broad focus, then this implies
that the direct object is not the focus of the sentence. Alternatively, if the accent that
the object bears is the one used to convey narrow focus, then the sentence is not an
out-of-the-blue utterance. Both ways, the focus set is just an illusion.
As a consequence, the systematic distinction of categorially different pitch
accents for narrow and broad foci shows that there can be no focus projection. A
determinate accent marks a determinate type of focus (broad or narrow). Clearly, the
correlation between the type of accent and the type of focus requires that F-Structure
has to be set derivationally before the phonological component assigns a specific type
of accent to the metrical structure built up from the syntactic structure11. The key idea
here is that the purported F-Structure ambiguity is not such an ambiguity12. And
without that ambiguity, the focus projection and the availability of a focus set are
untenable. Even if in both out-of-the-blue and contextualized utterances nuclear stress
falls on the direct object focus sets are just an illusion, for the intonational facts show
unambiguously whether it is a broad focus or a narrow focus. Hence, it cannot be the
case that the F-Structure is computed at the discourse level, after all syntactic,
semantic, and phonological computations are over.

5.2.1.1.2. Focus-induced Phonological Phrasing

A similar conclusion can be obtained from the fact that some languages use
phonological phrasing to convey focus. In many languages, there is a systematic
correlation between focusing and phonological phrasing. In these languages, the left
(9) or right edge (10) of the focal XP is aligned with the edge of a phonological
phrase, which is marked by prosodic or segmental cues. These are cues such as
postfocal pitch compression, glottalizations in phrase-initial positions, final

11 Tonal languages like Mandarin also pattern alike, showing unambiguous F-Structures (cf. Peng et al.

(2005))
12 A fact that, by the way, was also noted by Cinque himself (cf. Cinque (1993: 259 fn 23)).

205
lengthening, the presence of boundary tones before the focalized constituent, and
perceptible breaks or sense of disjuncture with material preceding it:

(9)13
Left Alignment: PHONOLOGY: [φP
SYNTAX: ZP [XP]F YP

(10)
Right Alignment: PHONOLOGY: ]φP
SYNTAX: ZP [XP]F YP

These patterns of phonological phrasing associated to focus clearly contravene

the premises of NSR-based theories of F-Structure; basically, if a phonological
boundary aligns with the focus phrase, then there has to be a specific focus phrase to
align with.
Among the languages in which there is a correlation between F-Structure and
phonological phrase alignment we can mention the following:

(11)
-Left Alignment: Tokyo Japanese (cf. Pierrehumbert & Beckman (1988),
Bengali (cf. Hayes & Lahiri (1991), Selkirk (2006)), Korean
(cf. Jun (1993)), Greek (cf. Condoravdi (1990)), Lekeitio
Basque (cf. Elordieta (1997, 2007b))…

-Right Alignment: Swedish (cf. Bruce (1977)), Chicheŵa (cf. Kannerva (1990),
Truckenbrodt (1995, 1999)), English (cf. Selkirk (2000),
Brazilian Portuguese (cf. Sandalo & Truckenbrodt
(2001))…

13The subscript ‘φP’ is used to denote whichever prosodic phrase (accentual phrase, minor phrase…) of
the prosodic hierarchy is aligned with focus.

206
Thus it cannot be the case that the F-Structure of a sentence is undetermined or
ambiguous, nor that the ‘actual focus’ is chosen from a set of ‘possible foci’ at the
discourse level (once all syntactic, semantic, and phonological computations are
over). If the phonological phrasing of the sentence is affected by the nature of the
actual focus, then the actual focus has to be set derivationally before the phonological
component deals with prosodic phrasing14.
From this discussion we can preliminarily conclude that, as the intonational
evidence shows, sentences don’t have ambiguous F-Structures. This means that F-
Structure has to be set at some point in the derivation from the numeration to PF.
This prosodic non-ambiguity of F-Structure is correlated with recent discoveries
regarding the non-ambiguous nature of prosodic phenomena (cf. Hirchsberg &
Avesani (2000), Speer (2003), Kang & Speer (2003), Schafer, Speer & Warren (2003)).
These studies demonstrate that even in the most contextually unambiguous
environments the prosodic structure of a sentence shows that the sentence is not
ambiguous (hence, blurring any functionalist explanation in terms of emphatic
disambiguation (pace Ladd (1996)). And in fact, this empirical scenario is something
to be expected given minimalist assumptions; a sentence is a grammatical object
construed from a set of lexical items (that are no more that arrangements of features)
via a derivation that undergoes Full Interpretation (FI). As a grammatical (hence
mental) object that undergoes FI, a sentence cannot be structurally ambiguous,
although the orthographically written representation of a sentence might be so.

5.2.1.1.3. Lack of Focal Accent

As we saw before, the main premise of NSR-based theories of F-Structure is
the purported necessary correlation between focus and nuclear stress at PF.
However, it has been discussed in the literature that some languages and varieties
make use of no accentual cue to mark focus. Among them we find Hyxkariana (cf.
Derbyshire (1985)), Yucatec Maya (cf. Gussenhoven (2006)), Q’anjob’al (cf. Parafita
(2005)), Kaqchikel (cf. Parafita (2005)), West Greenlandic (cf. Arnhold (2007)), Navajo

14I am agnostic as to what is the correct characterization of the phrasing itself, be it rule-based or
harmonic candidate computation. The fact is that both types of architecture of the grammar require
unambiguous F-Structures to align them with the relevant phonological phrase.

207
(cf. Hale et al. (2003)), Guyanese English Creole (cf. Bickerton (1993)), French (cf. Féry
(2001)), Egyptian Arabic (cf. Hellmuth (2007)), Aghem (cf. Watters (1979)), Wolof (cf.
Rialland & Robert (2001)) as well as some constructions of Russian (cf. King (1995))
and Northern Bizkaian Basque (cf. G. Elordieta (1997, 2007a) and chapter 2 of this
dissertation). This is as well the case in Ewe (cf. Jannedy & Fiedler (2007)), and
Chichewa, Chitumbuka and Durban Zulu (cf. Downing (2007)). In fact, as Hyman
(1999: 166) puts it, it is not clear at all that any tonal language of the Bantu family
displays any accentual or tonal effect directly linked to [+F] marking: “In no case […]
have we seen what can be called a “direct mapping” from focus to tone. That is, I am
unaware of a “pure” example where semantic focus (and only semantic focus)
unambiguously conditions a [+focus] tonal effect, or where the absence of semantic
focus (and only its absence) conditions a [-focus] tonal effect”.
Crucially, then, a focalization mechanism different from the NSR would have
to be posited for these languages so as to set the F-Structure. Recall here that the
issue is not about a parametrically different realization of F-Structure in different
languages (something that, in fact, any look at crosslinguistic data reveals). The
crucial point is that if the NSR-based theories of F-Structure were right in postulating
a legibility condition (i.e. a Bare Output Condition) that requires focused phrases to
bear the nuclear stress, the languages mentioned above would have to have a
different architecture of the grammar with different Bare Output Conditions from, for
instance, those of English speakers’. Clearly, this view is incoherent given minimalist
assumptions, since interface conditions on representations cannot be subject to
parametric variation. In fact, by assumption, “conditions on representations […] hold
only at the interface, and are motivated by properties of the interface, perhaps
properly understood as modes of interpretation by performance systems” (Chomsky
(1995b: 170)). And, of course, performance systems are the same for, say, Hyxkariana
speakers and Dutch speakers, and, likewise, they are also the same for (potential)
Hyxkariana-Dutch bilinguals.
Note that it could be argued that the purported legibility condition is more
abstract than the explicit proposal of the NSR-based theories of F-Structure, that is,
something like “focus has to be overtly realized in some way”. But an argument of

208
this sort would carry back the question I mentioned above. From a minimalist point
of view, legibility conditions are just Bare Output Conditions; constraints or filters on
the type of representations that are readable by the interfaces (like the need for linear
order of terminals for the A-M interface and, maybe, the avoidance of vacuous
quantification for the C-I interface15). Hence, by definition, they are not subject to
parametric variation (see 5.4. below). Furthermore, this postulated legibility
condition is at the same time too specific and too unspecific. It is hard to believe that
our Articulatory-Motor system (AM) is designed in such a way that it requires
anything about focus. The most widely accepted Bare Output Conditions have a
natural flavor regarding what we know about the interfaces. For instance, the
linearity requirement for terminals in PF is understandable due to the fact that
spelling out sentences necessarily takes time. Thus, phonological words have to be
arranged linearly so that the articulatory apparatus can spell them out sequentially.
But a specific constraint on the shape that a focus phrase must have is another issue.
It follows from nothing that we know about the A-M system; i.e., the design of the A-
M system does not require such a constraint.
But more importantly, notice that such a constraint is at the same time too
unspecific, and that it doesn’t offer any explanation for these facts. Suppose that in
order to account for the wide crosslinguistic variation in surface effects of focus we
assume that there is a legibility constraint requiring focus to be “somehow realized”
(phonologically, morphologically and/or syntactically). We could equally say the
same thing for any other aspect of the grammar. For instance, given that
crosslinguistically conditionals and subjunctives are realized in some way or another,
nothing keeps us from applying the argument of the unspecified focus constraint and
claiming that there are other constraints requiring that [+conditional] and
[+subjunctive] have to be “somehow realized”. But in the end, the explanatory power
of such an unspecified Bare Output Condition is null, for, besides of being

15 See, e.g., Chomsky (1982: 11): “[…] a general principle of LF to the effect that each operator must
bind a distinct variable”, or Kratzer (1995: 131): Prohibition Against Vacuous Quantification: For every
quantifier Q, there must be a variable x such that Q binds an occurrence of x in both its restrictive
clause and its nuclear scope.

209
unfalsifiable, it doesn’t help us to understand the phenomenon of conditionals or
subjunctives. Likewise for focus.
Summarizing the discussion so far, in this section I have provided three
different types of arguments against NSR-based theories of F-Structure: arguments
from (i) the categorically different accents used to convey different F-Structures, (ii)
the alignment of the focus phrase and phonological phrase boundaries, and (iii) the
lack of any focal accent in different languages. They show that, on the one hand,
focus is encoded unambiguously in the grammar, and on the other hand, that focus is
not necessarily linked with nuclear stress placement. What all these problems have in
common is that they show that the architecture of the grammar the NSR-based
approaches need to postulate is untenable. In the next section I will review the
syntactic evidence that shows that this type of approach to F-Structure cannot be
maintained.

5.2.1.2. Syntactic Evidence Against NSR-based Theories of F-Structure

In chapter 4 I have provided a wide range of evidence in favour of a
movement analysis of focalization. Of a special relevance are the sensitivity to islands
and the pied-piping phenomena, which are to be expected under a cartographic
approach that postulates movements to a left peripheric position (since they show
the same restrictions as Wh-movements). Those patterns become mysterious in a
NSR-based in situ approach. These movements would lack any motivation and we
would loose the generalization obtained with Wh-movements. Similarly, the patterns
of Focus-Auxiliary adjacency analyzed in chapter 4 also suggest that NSR-based
theories of F-Structure are wrong in postulating an accidental adjacency between the
verb and the focal XP (cf. the Appendix). In this section I will analyze other types of
syntactic phenomena that show the inadequacy of NSR-based theories of F-Structure.
The evidence I want to discuss is based on two predictions of those approaches: (i)
the prediction that nuclear stress is assigned to the most embedded position (section
5.2.1.2.1), and (ii) the prediction that there is focus projection from the element that
bears the nuclear accent (section 5.2.1.2.2).

210
5.2.1.2.1. Failure to Meet the Most Embeddedness Condition
Recall that the crucial assumption of the NSR-based analysis of the F-Structure
is that the nuclear stress is invariably assigned to the most deeply embedded position
of the whole clause. Thus, if we want to focalize an element that is not contained in
the focus set created by focus projection from the direct object (the absolutely most
deeply embedded element), then an alternative mechanism is necessary. Different
languages employ different strategies:

(12)
(i) English-like languages allow for the assignment of the nuclear stress in a
position different than the most deeply embedded one. This would be a
‘marked’ stress shift.
(ii) Basque-like languages employ movements of nonfocal material to
positions higher up in the structure so that the element to be focused
results in the most deeply embedded position and, thus, nuclear stress is
assigned to it.

But this raises several questions. In particular my concern here is the following
one: if we allow for these two different strategies, then what is the nature of the NSR?
Regarding languages like English, then the conclusion should be that the syntax-PF
interface cannot be that blind. That is, the position of the nuclear stress cannot be said
to be derived just from the mapping of syntactic embedding into prosodic (metrical)
embedding; something else is needed if we are going to allow for stress shifts (for
instance in the case of subject focalizations, which are not obtainable with the
unmarked pattern). Something has to tell PF where to put the nuclear stress. And I
think that what is needed is the availability of the F-Structure at the structural
description of the NSR (cf. my reanalysis of the cinquean NSR in chapter 3). In a
nutshell, we cannot capture the English facts without the availability of a specific F-
Structure in the structural description of the NSR. But obviously, the moment we
accept the availability of an F-Structure during the computation the whole focus set
approach disappears (cf. 5.2.2 for further discussion). However, what I want to

211
discuss here relies on the patterns of Basque-like languages. These languages are said
to be different from English-like languages in that they do not recur to marked shifts
of the nuclear stress to higher positions. Rather, these languages are analyzed as
allowing some sort of displacements of nonfocal material so that the element to be
focused results in the absolutely most deeply embedded position (cf. i.a., Reinhart
(1995, 2006), Zubizarreta (1998), A. Elordieta (2001), Arregi (2003)). But this
argumentation is built on the assumption that the nuclear stress is assigned to the
most deeply embedded position in the whole clause. That is, at PF we just turn
syntactic trees into metrical grids and whatever gets the most grid marks, that will
receive the nuclear stress. So, nuclear stress is assigned to an element not because of
any specific ‘focusness’ of this element, but just because that element happens to be
the one with the most grid marks:

(13) Syntax-Prosody Interface

NSR

( Z** ( X*** Y****))

3
( Z* ( X** Y***))
3
( X* Y**)
g
( Y* )

I will address later on (cf. section 5.2.2) the problems raised by the lack of any
motivation for the displacement of nonfocal material. For now, we can observe that
the idea that nuclear stress is assigned just computing a metrical grid is too strict to
account for the empirical facts. Remember that the NSR-based analysis requires any
element that gets the nuclear stress to be the absolutely most deeply embedded one.
This assumption is, nevertheless, highly problematic since nuclear stress can be
assigned, for instance, to specifiers of full XPs like the genitive in (14), whose

212
underlying structure would be the one in (15), with a right-headed DP (cf. section
3.4.3)16, 17:

(14) [Amaiaren]F lagunak maite du Jon.

Amaia.of friend love AUX Jon
“[Amaia’s]F friend loves Jon”

(15)
DP
qp
Spec. D’
Amaiaren 3
NP Dº
lagun -a

Following the premises of Arregi (2003a)’s proposal, if the base-structure of

the sentence in (14) (before the dislocation of nonfocal material) is the one in (16),
then, we would have to conclude that the derivation of the sentence in (14) is along
the lines in (17):

(16)
TP
3
AspP T
qp
vP Asp
qp
DP v’
qu 3
Amaiaren D’ VP v
3 3
lagun -ak Jon maite

16 For the sake of the argument I am just following Arregi’s (2003) analysis of a right-headed DP. If
genitives were adjuncts instead of specifiers the argument would still be the same.
17 In this tree-diagram I have omitted the ergative case-marking ‘-k’, that appears attached to the

determiner in (14).

213
(17)
TP
qp
TP Joni
3
AspP T
qp
vP Asp
qp maite du
DP v’
3 3
DP D’ VP tv
3 2 2
Amaiaren tj lagun -ak ti tv
j

That is, in order to get the genitive ‘Amaiaren’ to be the most deeply
embedded element we would have to postulate (i) the right dislocation of the direct
object adjoining it to the TP (like for any [S]FVO sentence (cf. chapter 4) as well as (ii)
a right dislocation of the D’ inside the subject DP, adjoining to its DP projection (in
order to obtain the intended linear order). But this is a strange type of movement for
it constitutes a displacement of an intermediate projection (D’) to adjoin it to a leave
of the tree. The problem is that if we allow for this type of derivations in a case by
case basis, then it is hard to evaluate the theory. There is simply no way to predict
where an element will move, and then, the explanation becomes completely ad hoc.
Here I have discussed just one instance of the type of problems that derive
from the assumption that focus is assigned through nuclear stress in the most
embedded position. Note, however, that NSR-based theories of focus will have to
posit equally baroque accounts for any other construction where the focal element is
not predicted to be in the absolutely most embedded position. In particular for
focusing constructions like Heavy-NP Shift, Gapping, Pseudogapping, Clefts or
Pseudoclefts. Not only in “stress shifting” languages like English, that allow for
stress assignment wherever necessary, but also for “scrambling” languages like
Basque, where the focal XP is supposed to be always in the most deeply embedded
position (cf. Saltarelli et al. (1988) for a presentation of these constructions). Likewise

214
for the ‘Schmerling examples’ I discussed in section 3.4.4, nuclear stress on the
subject should not be able to mark focus on the whole clause.

5.2.1.2.2. Lack of ‘Focus Projection’

There is a further empirical problem for NSR-based theories of F-Structure.
These theories make a strong prediction concerning the focus projection. F-Structure
is computed from the accented item and any node that contains the accented item
becomes a member of the focus set; it is just a PF issue of containment. Then we
should expect this to happen independently of the syntactic constituency of the
phrases in the metrical grid. But that seems to be wrong. Recall the basic pattern
already discussed: in both SVO and SOV languages, nuclear accent on the deepmost
embedded object can mark the entire sentence as well as the VP as focused, as shown
in (18)-(19):

(18) [S [V [Ó]F]F]F

(19) [S [[Ó]F V]F]F

However, this claim implies that in VSO languages, nuclear accent on the object
should be able to mark as focal the phrase containing both the subject and the object
and excluding the verb; after all, all that the calculation of the F-Structure cares about
is surface embedding. That is, the prediction is that something like (20) should be
natural (see (22), below):

(20) [V [S [Ó]F]F]F

But this is a too strong conclusion. A quick look at VSO languages like Valley
Zapotec shows that such a pattern is ungrammatical. Observe the data in (21), taken
from Rosembaum (1977):

215
(21a) Gudo Xwain biza (Out-of-the-blue) >>> [VSO]F
ate Juan beans
“[Juan ate the beans]F”

(21b) #Gudo Xwain biza (Who ate what?) >>> *V[SO]F

ate Juan beans
“[Juan]F ate [the beans]F”

(21c) Xwain biza gudo (Who ate what?) >>> [S]F[O]F V

Juan beans ate
“[Juan]F ate [the beans]F”

Suppose that, as predicted by NSR-based theories of F-Structure, the nuclear

stress of (21a) is on the most deeply embedded object ‘biza’18. This accent placement
would allow, via focus projection, to mark the whole sentence as focal. However, the
case is that if we want to have focus on both the object and the subject, we cannot
leave them in situ (21b). Necessarily, we have to left-dislocate them (21c). Following
NSR-based theories of F-Structure this is unexpected because if VSO can be a broad
focus, then [SO] should also be a possible focus in the VSO order. That is, if, as
defined by the NSR-based theories of F-Structure focus projection is just a matter of
which phrases contain the nuclear stress in PF, then and any phrase containing the
nuclear stress should be part of the focus set, regardless of its phrasal constituency.
But here, it seems that this is not the case, given that we cannot project the ‘focusness’
of the object to the subject in a VSO sentence:

18Rosenbaum (1977) doesn’t provide these data, but since Zapotec is a language that makes use of
accents, (cf. Chávez-Peón (2006)), let us suppose that for the sake of the argument.

216
(22)
XP
3
gudo […]
3
Xwain […] Illegitimate F-Structure
5
biza

On the other hand, within the DAFS presented in chapter 3, this is

unproblematic, since the need of a syntactic constituency for the F-Structure is
inherited from the narrow-syntactic setting of the F-Structure19.

5.2.1.3. Conclusion: Unambiguous F-Structures in the Computation

I have discussed in detail a variety of empirical evidence against NSR-based
theories of F-Structure. The phonological evidence was of two sorts: (i) the
directionality of the causal relation between focus and nuclear accent (that is, the fact
that F-Structures are assigned nuclear stress (only in some languages), and not the
other way around) and (ii) the necessity of having the F-Structure already set in the
narrow syntax in order for some PF operations (NSR/phonological phrasing) to
operate on it. This last argument extends also to the syntactic component in that the
displacement operations need to make reference to an already fixed F-Structure (see
also next section). Besides, in chapter 4 I have also argued that the postulation of an
accidental adjacency between the focal XP and the verb (necessarily so, if there is no
‘actual focus’ in the syntax) makes the wrong predictions. In many cases this
adjacency cannot be obtained without the postulation of specific ad hoc movements.
But, if we postulate these movements, then the adjacency is not accidental.
Crucially, none of these criticisms applies to the proposal I have put forth in
chapters 3 and 4 of this dissertation:

19See chapter 6 for an analysis of cases of split focus constructions where the syntactic constituency is
not maintained.

217
(i) On the one hand, severing the F-Structure from nuclear stress
placement avoids the qualms brought by the lack of nuclear accent in
focal XPs in some languages; the F-Structure is derived
independently from any accent placement.
(ii) Second, as I argued in chapter 3, setting the F-Structure in the
narrow syntax allows for its introduction as such in the structural
description of the PF operations of nuclear stress assignment and
phonological phrasing (cf. 2.3 for discussion). The F-Structure is
unambiguously encoded in the grammar and thus we can expect all
the effects of focusing that we observe at the interfaces.
(iii) As a consequence, the problems brought by having focus projection
and a focus set disappear.

In the next section I will discuss some of the conceptual arguments to reject the
NSR-based theories of F-Structure adopting minimalist assumptions and desiderata.

5.2.2. Conceptual Arguments Against NSR-based Approaches

After reviewing the empirical drawbacks of NSR-based theories of F-Structure,
I will now address the core conceptual problems that they raise. I will be focusing on
three main problems: (i) the lack of explicitness of the NSR-based approach to F-
Structure, (ii) the problem of the focal feature and the setting of F-Structure, and (iii)
the fuzzy nature of the purported focus set.

5.2.2.1. Architectural Issues

The first criticism I would like to make here concerns the lack of explicitness of
NSR-based approaches. As explained before, their main premises are based on a
purported interface legibility condition that requires focus to have the nuclear stress
at PF. This argumentation is paradoxical in its conception of F-Structure: the crucial
idea of NSR-based approaches is that there is no F-Structure per se during the
computation from the Numeration to PF, the actual focus is determined after nuclear
stress placement. But on the other hand the system needs to know what the focus

218
will be in order to trigger the operations that lead to assign nuclear stress within it.
That is, there is a paradox concerning where focus appears: on the one hand, NSR-
based theories involve the assumption that F-Structure is somehow there at the
beginning of the derivation, that is, the NSR applies either regularly or through a
marked operation, depending on the intended F-Structure. But on the other hand, it
is claimed that PF is responsible for determining the possible F-Structures (the focus
set) given the NSR and focus projection. Basically, the question is the following one: if
there is no F-Structure from the beginning, then ceteris paribus the NSR should
invariably assign nuclear stress to the most deeply embedded element, and there
would be no reasons for the scrambling movements of languages like Basque, nor for
the marked stress assignments of languages like English. On the other hand, if the F-
Structure is there from the beginning, then there is no need to postulate ambiguities,
focus sets, nor focus projections; it is there. As I will discuss next, the problem is that
the argumentation of NSR-based theories of F-Structure is necessarily circular: the
focal XP of a sentence will be inferred from the placement of the nuclear stress, but
on the other hand, the position of the nuclear stress will depend on what the actual
‘semantic’ focus is. Clearly, this type of argumentation is circular.
Recall as well that an important problem arises with the computation from the
Numeration to PF: a crucial assumption of NSR-based theories of F-Structure is that
there is a legibility condition on derivations requiring that convergent derivations
have nuclear stress on the focused XP. Otherwise the derivation crashes. The
problem arises with the trigger of the operations that lead to the intended
configuration. As explained before (cf. chapter 4), A. Elordieta (2001) or Arregi
(2003a) have proposed some displacement operations (‘scrambling’ and ‘left and
right dislocations’ respectively) in order to account for the Verb-Focus adjacency
observed in Basque. Under this view, the adjacency between the verb and the focal
XP would be just a by-product of these phonologically triggered movements. It
would just be the result of the need to assign nuclear stress to the most deeply
embedded element (i.e., the one that happens to be to the left of the verb in a right-
headed SOV language like Basque (cf. chapter 4)). But this type of approach is highly
problematic. First, there is a problem with this adjacency, since, as discussed in

219
chapter 4, it cannot be that accidental. But, more importantly, for these theories the
major problem is that, according to basic assumptions in the minimalist program,
movement does not come for free and displacement operations in syntax take place
as the result of feature-checking under probe-goal relations. However, the
movements that would provide the correct configurations for the postulates of NSR-
based theories of F-Structure lack any other motivation than just providing that
configuration.
Furthermore, if we assume a Y-model of the architecture of the grammar, then,
a condition imposed on PF will be unable to drive narrow syntactic operations, that
is, there is no look ahead. And, unless specified further, the displacements that the
NSR-based theories of F-Structure postulate are look ahead operations. Obviously this
issue becomes problematic provided that we accept the classical Y-model of the
architecture of the grammar. And in fact, many of the theorists adopting NSR-based
theories of F-Structure postulate that the conclusion of their work is that the
architecture of the grammar has to be changed radically in order to accommodate for
focus phenomena. Thus, for instance, Zubizarreta (1998) proposes a novel
architecture of the grammar like (23) where the derivation unfolds creating sets of
phrase markers until one single phrase marker is obtained (Σ-Structure). Then some
operations like F-marking, NSR and prosodic movements take place until we reach
the level of LF. At this stage the derivation branches in two branches, one that
derives in a PF representation and the other one that provides what Zubizarreta
(1998) calls the “Assertion Structure”, the information structure of the sentence
where the focus-presupposition partition of the sentence is encoded20:

20 Cf. Vallduví’s (1995) architecture with a similar level of “Information Structure”.

220
(23) Lexicon

Σ-Structure

LF

PF Assertion Structure

A more radical change of the model of the grammar is proposed by authors

like Szendröi (2003) or Dominguez (2005), who, following ideas of Jackendoff (1987,
1998, 2002), postulate a “Parallel Architecture” of the grammar where all modules
create their own derivations in a parallel fashion. Then, there will be some
correspondence rules between the different derivations (24):

(24)
Phonological Syntactic Conceptual
Formation Rules Formation Rules Formation Rules

Phonological Syntactic Conceptual

Structures Structures Structures

PS-SS SS-CS
Correspondence Rules Correspondence Rules

PS-CS
Correspondence Rules

An in-depth analysis of these alternative models would take us too far away
from the goals of this dissertation but I would like to point some general problems I
see with these proposals:

221
(i) Zubizarreta’s (1998) model assumes that all the focally relevant operations
of NSR and “prosodic movements” take place in the stretch between Σ-Structure and
LF, but at this stage there is still no phonological representation, so even changing
the architecture of the grammar all these operations take place looking ahead.
Furthermore, it is assumed that the phonological representation derives from the LF
representation (which, I assume, provides also the interface with the C-I system). But
then, what type of variability do we expect in PF if its input is the output of the
logically unambiguous and universal representation in LF? The LF→PF stretch
should be very powerful in order to allow for all the variability we observe in the
surface of natural languages. Basically, the LF→PF stretch would be equivalent to the
D-Structure→PF strecht of the P&P model, so instead of simplifying the grammar we
would complicate it exponentially. All word order, phononological phrasing and
morphological differences (among others) would be just idiosyncratic properties of
languages (non-traceable to their syntax).
(ii) Parallel architectures where any module is allowed to affect any other are
much more debated, and they are often invoked whenever an interface phenomenon
escapes from a straightforward explanation, especially in the syntax-phonology
interface (cf. i.a. Zec & Inkelas (1990)). In fact, one of the most popular arguments
used for arguing in favor of the jackendoffian parallel architecture is the fact that
focus constructions display phonological, syntactic, semantic and/or morphological
effects (cf. the references cited above and specially Jackendoff (2002, §12)). This issue
deserves a closer inspection: as I argued in chapters 3 and 4, the phonological and
semantic effects of focus can be naturally explained in a “syntactocentric” model,
building the F-Structure in the narrow syntax. Indeed, that seems to me to be the
only way to explain such effects: given that in the jackendoffian parallel arquitecture
“correspondence rules” are introduced to link the independent derivations, this
model as well incurs in a global look ahead. The purported derivational operations
that take place to provide the intended representations are unmotivated otherwise;
they are posited in order to arrive to a certain PF representation. However, the
system is supposed to be derivational in essence, there is nothing like ‘move α‘, so no
way to obtain the intended representations. It is, in my view, like an Optimality

222
Theoretic grammar but without the GEN function to provide the infinite set of
candidates to be evaluated21.
But more importantly, the idea of independent derivations is extremely
dubious: what type of phonological representation are we going to build
independently of syntax? And how are they going to affect syntactic structure
building? For instance, given a syntactically unordered set of words in a numeration
like {these, words, are, not, syntactically, combined}, how does the computational
system build phonological structures if it doesn’t take as input the output of syntax?
Because, if it does, then it is syntactocentric in essence (as Jackendoff somehow
recognizes (cf. Jackendoff 1998: 23)). Jackendoff (1997: 28-29) postulates that the
syntax-phonology interface is but a list of correspondence rules between
independently built syntactic phrases and phonological phrases, but we are never
told how to build a phonological phrase independent of syntax. All the jackendoffian
correspondence rules do for us is to claim that “if a syntactic constituent X1
corresponds to a phonological constituent Y1, and a syntactic constituent X2
corresponds to a phonological constituent Y2, then the linear order of X1 and X2
preferably corresponds to the linear order of Y1 and Y2” (cf. Jackendoff (1997: 28)).
Jackendoff denies any mapping operation from syntax to phonology; there are just
correspondences between two independent representations. The question is; how are
those representations obtained? A customary assumption is that the processes of
syntactic structure building and linear ordering are obtained from selection (i.e., the
locus principle (cf. chapter 3)) and c-command. But how does phonology work to
create structure from the lexical items? How do we get the phonological
representations that are to be corresponded to the syntactic representations?
Jackendoff doesn’t offer an answer. In fact, whenever Jackendoff presents the syntax-
phonology correspondence, he departs from already built structures like (25) and
then formulates the optimal correspondence rules like (26), where ‘IntP’ stands for
Intonational Phrase (cf. Jackendoff 2002: 119-120):

21
See Szendröi (2001) for a proposal mixing Brody’s (1995) Lexico-Logical Form, a jackendoffian
parallel architecture, and an OT grammar for the interfaces.

223
(25)
Syntax: [NP this] [VP is [NP the cat [CP that [VP caught [NP the rat [CP that [VP stole [NP the
cheese]]]]]]]]
Phonology: [IntP this is the cat][IntP that caught the rat][IntP that stole the cheese]

(26) a. An utterance consists of a series of one or more concatenated IntPs forming

a flat structure. Each IntP is a sequence of Words.
b. Preferably, the IntPs are of equal length.
c. Preferably, the longest IntP is at the end.
d. (Possibly, some strong preferences on maximum duration of IntPs, e.g. try
not to go over three seconds).

The interface constraint relating IntPs to syntactic structure would be as

roughly stated in (27):

(27) An IntP corresponds to all of a syntactic constituent C, except that a

subconstituent at the right-hand end of C can be omitted.

Thus, according to Jackendoff, the example of (25) is explained as follows (cf.

Jackendoff 2002: 120): “The first IntP corresponds to the whole sentence, but the large
relative clause that caught the rat that stole the cheese is omitted. In turn, this relative
clause is further carved: the final relative clause is taken out to form the third IntP.
The result is perfectly balanced prosodically.”
The theory sketched describes the facts nicely. But as can be observed, there is
no syntax-independent process of phonological phrase construction. On the contrary,
the process in (27) is defined as applying upon the different syntactic cycles. There
were no phonological phrases before, for the number of IntPs depends on the
number of CPs. So, the alleged correspondence rule is more like the classic syntax-
phonology mapping. That is, there are some operations that take as input syntactic

224
phrases and give as output phonological phrases; however complex this process
might be22, 23.
In any event, notice that having a phonological restriction or correspondence
rule on the type of representations that are acceptable wouldn’t entail the assumption
of a parallel architecture. Let alone a syntax-phonology correspondence like the one
observed with focus and nuclear stress.
The type of evidence that would show that phonology does affect syntax and
hence that syntactocentrism is wrong would be to observe the application of a purely
phonological rule triggering the application of a purely syntactic operation24. This is
something that, up to my knowledge, has not been attested in any language. I believe

22 Obviously, syntactic phrases are not isomorphic to phonological phrases (which, in no case proofs the
mistake of syntactocentrism). There will always be some readjustments due to phonological reasons
like length, accent placements and so on (that is the role of Jackendoff’s “preferably” clause in (26)).
The fact that those mismatches are due to readjustements (in derivational prosodic phrasing
algorithms) or unfaithful mappings (in OT grammars) shows us that syntactocentrism is on the right
track.
23 Note that, furthermore, in various syntactocentric models phonology even can access the output of

syntax several times in the derivation, as in Uriagereka’s (1999) Multiple Spell Out or Chomsky’s (2000
et seq.) phases models (cf. Dobashi (2003) for an analysis of phonological phrasing in these terms).
24 For instance something like the invented rule depicted in (i):

(i) [+ANT, COR] → [-ANT, COR] / [+HIGH, -ROUND]+__X]v ] →V1

This is a contextually restricted rule of palatalization of a dental stop within a verb (the structural
description), triggering a syntactic rule of verb fronting (the structural change): A hypothetical
paradigm corresponding to this sort of rule would be to have in Basque a paradigm like (ii), where
depending on the aspectual affix that the verb ‘lodi’ (to thicken) takes, the syntax of an out-of-the-blue
sentence changes. Imagine that if it takes the suffix ‘-tzen’ (punctual/habitual, where ‘tz’ is a dorsal
affricate) no phonological rule applies and the syntax doesn’t change (the rule in (i) doesn’t apply).
Thus we have the regular SOV word order we saw so far (cf. (iib-c)). But imagine that it takes the
suffix ‘-tu’ (perfective). That would trigger a phonological palatalization (T→C) and, as a consequence
rule (i) applies raising the verb and ending in a VSO order (iid-e):

(iia) Base: Jonek zopa lodi[affix] du >>> SOV

Jon soup thicken AUX
(iib) Jonek zopa loditzen du >>> SOV
Jon soup thicken AUX
“Jon thickens the soup”
(iic) *Loditzen Jonek zopa du >>> *VSO
thicken Jon soup AUX
“Jon thickens the soup”
(iid) *Jonek zopa lodittu du >>> *SOV
Jon soup thicken AUX
“Jon thickened the soup”
(iie) Lodittu Jonek zopa du >>> VSO
thicken Jon soup AUX
“Jon thickened the soup”

225
that this indicates that the classical “syntactocentric” model is essentially correct and
syntax is “phonology-free” (cf. Miller, Pullum & Zwicky (1997)), as we assumed at
the outset of this dissertation. Syntax creates structure that is then interpreted by the
interfaces:

Lexicon

Language

[Phonological Form] [logical form]

A-P System C-I System

I will leave architectural issues aside and in the next section I address another
core problem of the NSR-based approaches: the issue of the focal feature.

5.2.2.2. The Role and Nature of the Focal Feature in Building the F-Structure
The paradox of circularity of argumentation that I have unveiled brings forth
another issue concerning the identification of the focal XP. The question can be
formulated in the following way: if there is a [+F] feature from the beginning of the
derivation, how is it assigned? If there is no such feature, how is focus interpreted?
Within the framework of NSR-based theories of F-Structure two main types of
proposals have been made with respect to the issue of the [+F] feature:
(a) Theories which do not postulate any [+F] feature (cf. i.a. Szendröi
(2001), Reinhart (2006))
(b) Theories that postulate one [+F] feature (cf. i.a. Zubizarreta (1998)).
I believe both of them are wrong for the reasons I will expose in what follows.

226
5.2.2.2.1. Theories that Do Not Postulate Any [+F] Feature
Some authors like Szendröi (2001) or Reinhart (2006) have proposed that we
can dispense with any [+F] feature. The focalized constituents in a sentence would be
just inferred from the nuclear stress placement in PF.
I believe that this view cannot be operative from a minimalist perspective. If in
those theories it is the nuclear accent placement what will trigger the setting of a focus
set, then the computational system will not be able to recognise an element as focused
or unfocused since, assuming Bare Phrase Structure (cf. chapter 3), lexical items are
just bundles of features, and nothing distinguishes a ‘discursively’ focused XP from
an unfocused XP if there is no feature difference among them. To put it in a plain
way: given the core minimalist assumptions, if there is no feature difference, there is
no difference. So, according to these theories without any [+F] feature, the DP
subjects in (28b) and (29b) are just formally identical. If there is no feature difference
among the subjects in (28b) and (29b), there should be no distinction between them
for the computational system, and ceteris paribus, they should behave in the very
same way with respect to all operations (accentuation, displacement, phonological
phrasing, etc.).

(28a) Who likes cod?

(28b) Ibon likes cod.

(29a) What does Ibon like?

(29b) Ibon likes cod.

Likewise for the examples of Basque, where the syntax of both sentences should be
the very same (contrary to fact):

(28b’) Ibonek maite du bakailua.

Ibon love AUX cod
“[Ibon]F likes cod”

227
(29b’) Bakailua maite du Ibonek.
cod love AUX Ibon
“Ibon likes [cod]F”

Thus, in the absence of any feature distinction, in a language like English the
phonological component shouldn’t be able to ‘know’ where to assign the nuclear
stress or how to phrase the structure, since a notion such as ‘discursively focused’
cannot be available during the syntactic and phonological computation. Actually, if
there were no F-Structure fixed at PF, we should have invariably nuclear stress on
the most deeply embedded element and consequently, the focus set should always be
{CP, TP, vP, VP, Obj}. Thus, there should be no reason to apply the marked
operations that NSR-based theories of F-Structure postulate. On the other hand,
proposing marked operations to allow for the focalization of elements that were not
originally contained in the focus set requires syntactically fixed F-Structures.
Targeting a constituent not in the focus set by some marked operation is tantamount
to having it specified with a [+F] feature, which refutes the hypothesis that focus is
just computed in PF over the XPs that contain the element that has the nuclear stress.
The case is that if we want to focalize an element that is not in the ‘unmarked’ focus
set (say, the subject) we have to give precise instructions to the computational system
so that the subject gets the nuclear stress. These instructions necessarily will make
reference to the intended focussing of the subject. Hence, specific representations of
the F-Structure are needed. The paradoxical nature of these theories is even clearer in
the light of languages like Basque. In the absence of any [+F] feature the
computational system shouldn’t be able to “know” which phrase to move and where
to move it (and recall that, in cases of embedded foci NSR-based theories have to
posit that the focal phrase moves first (cf. sections 4.1.2.2 and 4.1.2.3)).
As can be observed, the theory of F-Structure I developed in chapter 3
circumvents these issues straightforwardly, because according to my proposal F-
Structures are set in the narrow syntax, so focus constituents can be affected by
syntactic transformations and interpreted at the interfaces.

228
5.2.2.2.2. Theories that Assume One [+F] Feature
F-Structure theories that assume the early assignment of just one [+F] feature
and its possible percolation or projection also have an important drawback. For
instance, there is no inherent reason for the N in the following configuration (30),

(30)
N[+F]

that will form an NP,

NP[+F] (projection)
g
N[+F]

that will form the DP Object,

DP[+F] (projection)
3
D NP[+F] (projection)
g
N[+F]

that will be embedded in a VP,

VP[+F] (projection)
3
V DP[+F] (projection)
3
D NP[+F] (projection)
g
N[+F]

229
that will form a vP,
vP[+F] (projection)
3
DP v’[+F] (projection)
3
v VP[+F] (projection)
3
V DP[+F] (projection)
3
D NP[+F] (projection)
g
N[+F]

that will end up forming a CP,

CP[+F] (projection)
3
[…][+F] (projection)
3
vP[+F] (projection)
3
DP v’[+F] (projection)
3
v VP[+F] (projection)
3
V DP[+F] (projection)
3
D NP[+F] (projection)
g
N[+F]

to receive the [+F] feature when the derivation is going to end with a sentence-whole
focus (an out-of-the-blue utterance). To have this N marked as [+F] and get it to project
its feature in order to get a CP focus is just a theory-convenient stipulation. That is,
the theory has a teleological flavor: all the system works towards the correct
focalization of an element by the satisfaction of a PF legibility constraint. If we want
to mark a sentence as focus, then we assign the most deeply embedded element
within the CP a [+F] feature, so that it projects it higher up. On the other hand, if we
want to have focus on, say, the subject, the [+F] feature will have to be assigned to the

230
most deeply embedded element within it and the system adopts marked operations.
That is, the system depicted by this type of theory is somehow ‘intelligent’, knowing
what it has to do so that the nuclear stress falls in a specific item and it becomes
interpreted as focused. Thus, if the element that ultimately has to be focused is not on
the most embedded position, some marked operations are posited.
Note here that this doesn’t constitute a problem for a purely derivational
theory like the one described in chapter 3, [+F] feature assignment is free and then, as
the output of Merge what you see is what you get: the F-Structure will be the
set/phrase constituted by just [+F] featured material.
Given this, we can conclude that ‘one [+F] featured theories’ that postulate the
early insertion of this feature are unable to explain their ad hoc axiom of feature
assignment.
A different alternative has been suggested by theories which postulate the late
insertion of one [+F] feature like Zubizarreta (1998). Again, a theoretical puzzle
arises; if the [+F] feature is assigned to a built up XP, and hence, the [+F] feature was
not in the original numeration, we incur in a violation of one of the most basic
assumptions of the minimalist program; the Inclusiveness Condition25:

(31)
“A ‘perfect language’ should meet the condition of inclusiveness: any structure formed
by the computation (in particular, π and λ) is constituted of elements already present
in the lexical items selected for N; no new objects are added in the course of
computation apart than rearrangements of lexical properties…”.
[Chomsky (1995b: 228)]

If we want to maintain the Inclusiveness Condition, then an operation inserting

a feature after the syntactic computation part cannot be posited.

As I explained earlier, Zubizarreta (1998) proposes a different architecture of the language faculty
25

which dispenses with the ‘Inclusiveness Condition’ in order to allow for this late insertion.

231
5.2.2.2.3. The Problematic Nature of the Focus set
As explained before, the notion of focus set is crucial to this theory in which
there is no ‘actual focus’ in the grammar, but just a set of ‘possible foci’ from which
discourse will choose the ‘actual focus’. The implications of such a view are
problematic: according to these theories, the ‘actual focus’ is set after the derivation
has undergone Spell Out. Thus, for these theories, in the examples of the purported
focus projection in (1) –repeated here as (32) for convenience– we don’t have
different grammatical sentences (understood as different derivations with the same
word order that undergone Full Interpretation) but just one sentence whose F-
Structure is ambiguous:

(32a) Jesus preached to the people of [Judea]F

(32b) Jesus preached to the people [of Judea]F
(32c) Jesus preached to the [people of Judea]F
(32d) Jesus preached to [the people of Judea]F
(32e) Jesus preached [to the people of Judea]F
(32f) Jesus [preached to the people of Judea]F
(32g) [Jesus preached to the people of Judea]F

According to these theories, then, the ‘actual focus’ will not be a grammatical
notion but a discursive one, and hence, unable to have any effect at logical form26.
However, the effects of focus in logical form are widely attested in the literature (cf.
i.a. Rooth (1985, 1992), Herburger (2000), Krifka (2006) and chapter 4 of this
dissertation).
In a nutshell, the theoretical corollary of this section is the following one: if
there is no [+F] feature and focus is just a configurational interpretation, then there
should be no focally-induced displacements, and nuclear stress should invariably fall
on the most deeply embedded element. Basically, there should be no reason to apply
marked operations. On the contrary, if there is a [+F] feature marking the focal XP,

26 Notably, it should also be unable to have the PF effects that it has (as we saw in sections 3.2 and
5.2.1.1 of this dissertation).

232
then it is not PF embedding and nuclear stress placement that sets the F-Structure.
Both ways then, the NSR-based theories of F-Structure are incoherent.

5.3. Summary
I have shown that the NSR-based theory of F-Structure has both empirical and
conceptual drawbacks. The reasons I adduced to claim this can be summarized as
follows:
(i) Some languages make use of categorially different pitch accents to
convey broad and narrow foci, showing that the actual F-Structure is
present at the phonological component. Hence, there can be no focus set
nor ‘focus projection’.
(ii) The strict alignment between the F-Structure and a phonological phrase
boundary observed in many languages reinforces the impossibility of
computationally ambiguous F-Structures.
(iii) The lack of any accent to convey focus in some languages makes dubious
the existence of the Bare Output Condition (by definition, universal) that
would require focused elements to bear nuclear stress.
(v) I have also argued how theories that purport the assignment of one [+F]
feature are stipulative and conceptually ill formed, and [+F] featureless
theories are computationally inoperative.
On the other hand, the Derivational Approach to the F-Structure presented in
chapter 3 and the syntax-semantics interface analysis of chapter 4 overcome all these
qualms in a natural way: the F-Structure of the sentence is built up from [+F]
featured lexical items in narrow syntax and independently of any nuclear stress.
Then, it can be interpreted at both interfaces without any violation of the Inclusiveness
Condition and without incurring in any look ahead operation. In a nutshell: syntax
proposes and the interfaces dispose.
The central proposal in this dissertation, thus, is an allegation in favor of the
classical ‘syntactocentric’ architecture of the grammar I presented in section 1.2.1. As
I argued in chapters 3 and 4, we can account for the interface effects of focus

233
assuming a syntactically set F-Structure. And as I argued in this chapter, that is the
only way to do it (pace Jackendoff (1987, 1997, 2002, 2007)).

234
CHAPTER 6. THE DERIVATIONAL APPROACH TO SPLIT
FOCUS CONSTRUCTIONS

In this chapter I analyze the patterns and information-packaging status of

answers to multiple Wh-questions. I argue that they are instances of split F-
Structures. A prediction of the Derivational Approach to F-Structure (DAFS) I presented
in chapter 3 is that split focus constructions should be possible. Note that this type of
constructions is unapproachable from a NSR-based theory of F-Structure. In NSR-
based theories an F-Structure has to be a continuum (i.e. the projection of the focal
status of the element that bears the nuclear accent). And obviously, NSR-based
approaches cannot accept but the introduction of just one nuclear stress since,
otherwise, they would not be able to account for focus projection (or the lack of it).
Here I will argue that split focus constructions do indeed exist, and that they
give rise to the patterns observed in the typical answers to multiple-Wh questions (as
proposed by Steedman (2000), Kadmon (2001) and van Hoof (2003) among others
and pace de Rijk (1969), Rebuschi (1983), Eguzkitza (1986)). In order to do that, I will
discuss the intonation, morphosyntax and semantic properties of these constructions
in different languages.

6.1. Split F-Structures

In this section I analyze one of the predictions that arise if we adopt the DAFS
I advocated for in chapter 3: the possibility of having syntactically split F-Structures.
The question is the following one: as presented in chapter 3, my proposal is
that F-Structure is built up in the narrow syntax with the dynamics of the derivation:
within Bare Phrase Structure, when two [+F] marked elements (α, β) are merged
together, the new syntactic object created will also be focal (a set/phrase composed
only of [+F] featured lexical items):

235
(1)
{αF, {αF, βF}}
3
αF βF

This theory has an interesting prediction: whenever two elements enter the
derivation bearing each of them a [+F] feature but are not merged together, two
isolated F-Structures will arise. This is depicted in (2), where γ and β bear a [+F]
feature whereas α does not bear it:

(2)
{α,{γF,{α, {α, βF}}}}
3
γF {α, {α, βF}}
3
α βF

For instance, a case of this sort of construction could arise when a DP subject
and a DP object enter the derivation being [+F] marked but the verb does not bear
any [+F] feature; something like the structure depicted in (3):

(3) Lexical Array (simplified): {{Mary}F, {John}F, {kiss}, {v}}

vP
qp
MaryF v’
3
kissed VP
2
tv JohnF

NO MERGE

At first sight, this structure might seem problematic, given the commonly
assumed uniqueness of focus; that is, the standard assumption that there is one and
just one focus per sentence. However, the idea that I would want to propose is that
indeed this type of construction is fully grammatical. My proposal is that one such

236
case is constituted by typical answers to multiple-Wh questions where the only focus
is a pair constituted by the subject and the object. Thus, the idea I would want to
push forward is that underlying in the sentence in (4b) we would have the structure
in (3), a proper (partial) answer to the question in (4a):

(4a) Who kissed whom?

(4b) [Mary]F kissed [John]F.

As I explained in chapter 4, at the syntax-semantic interface all the [+F]

material will be mapped into the scope of a quantification over events. This implies
that in the case of (3-4) both the subject and the object will be in the scope of the
quantification over events, having escaped from its restriction (cf. section 4.3.2.2.2):

(5)
vP
3
DPF v’
3
v VP
3 logical form: [δe [RESTRICTION] [SCOPE]]
V DPF

So, pursuing the idea of the quantificational nature of the logical form
representation of foci, in these constructions we will not have two independent foci,
nor two different information-packaging primitives like a ‘contrastive topic’ and a
focus (as I will argue in section 6.2). We will have just one focus which is a pair of
elements that happens to be derivationally split (what Herburger (2000: 45) calls a
‘discontinuous focus’).
In fact, note that the availability of these structures is just a natural conclusion
of the DAFS I argued for in chapter 3. As I argued there, being marked [+F] in the
numeration will not entitle a lexical item to be the actual focus of the utterance per se;
rather, being marked [+F] only entitles a lexical item to take part in the construction
of the F-Structure. And, trivially, this can be done in a strictly compositional way

237
(merging all the [+F] featured lexical items with each other, as in the cases discussed
in chapter 3), or in split focus constructions like the one depicted in (3)-(5), where
[+F] marked elements do not merge together.
Obviously, a question that arises is whether the [+F] feature assignment to
lexical items is completely free, or whether we can have any potentially possible
assignment of [+F] features and therefore, any possible F-Structure. This question
amounts to asking how we can prevent, for instance, a determiner in the subject and
a verb to be assigned a [+F] feature, like in the structure represented in (6):

(6) Lexical Array (simplified): {{The}F, {boy}, {John}, {kiss} F, {v}}

vP
qp
DP v’
2 3
theF boy kissed F VP
2
tv John

NO MERGE

The potential problem with this numeration is that the type of sentence with
such an F-Structure would be rather odd; something like (7):

(7) [The]F boy [kissed]F John.

However, for me, the oddity of the statement in (7) would not derive from any
computational anomaly in its derivation, but rather from independent discursive
factors: factually, there is no question whose answer could be a sentence like (7)1. The
system presented here would predict no ungrammaticality for (7), and in fact
sentences like (7) may be fine in contrastive environments. Therefore, I am assuming
—as I advanced in section 3.3— no a priori restriction on the assignment of [+F]
features. According to the DAFS I proposed in chapter 3, in computational terms,

1
In fact, sentence (7) might not be that odd. (7’), on the other hand, is highly degraded.
(7’) [The]F boy [that]F John [and]F Mary [saw]F yesterday [kissed]F John.

238
[+F] feature assignment might be free. Then, the sentence derived from this feature
assignment will have to fulfill the independently needed discourse congruence
requirements in order to be appropriate. But this will be a matter of discourse; not a
matter of syntax.
Note, finally, that these constructions of split foci that I am interested in are
very different from the so-called ‘multiple foci’ constructions analyzed in works like
Rooth (1985), Krifka (1991) or Wold (1998). These authors study constructions of
secondary focus like (8) and (9):

(8) John only introduced BILL to SUE.

(9) Even JOHN drank only WATER.

Clearly, (8) and (9) are very different from the patterns of answers to multiple-
Wh questions I am interested in. These secondary focus constructions need a marked
context to be felicitous, where one of the foci has an ‘echoic’ flavor. That is, a sentence
like (9) is only felicitous in a context where it clarifies or emphasizes a previous
statement like (9’), which has a focal object associated with the focus-sensitive
operator ‘only’:

(9’) Almost everyone drank only WATER.

The association of ‘water’ with ‘only’ in the case of (9), then, is a property
inherited from the previous utterance (9’), and hence does not constitute a real focus
in (9). That is, sentences like (10b) (=9) cannot constitute a proper (partial) answer to
multiple-Wh questions like (10a). Rather, in order to be acceptable they have to be
corrections of previous utterances like (10c) (in which case, they are better introduced
with something like “You’re wrong…”):

(10a) Who drank what?

(10b) Even John drank only water.
(10c) I think that just Izaskun drank only water.

239
Thus, due to their different character, the analysis of these structures falls out
of the scope of this dissertation (cf. the analysis of Rooth (1985), Krifka (1991) and
Wold (1998), among others).
Having advanced the theoretical argument that I am going to put forward in
this chapter, let me review now some of the properties of split foci constructions. I
will first review the mainstream analysis of these sentences and then discuss some of
their phonological, semantic and morphosyntactic properties. The conclusion I will
arrive at will be that:
(i) In these constructions we don’t have two different information packaging
elements like a ‘contrastive topic’ and a ‘focus’; rather, both elements will
be focal in nature.
(ii) These focal elements are not independent foci from each other; rather,
together they constitute the focus of the sentence (that is, we won’t have n
foci per sentence but only one, which might be a pair of elements).

With the analysis I will propose, the uniqueness property of focus will not only
be preserved, but it will also be predicted due to the logical form representation that
these sentences have (cf. section 6.4)).

6.2. An Earlier Analysis: Split F-Structures as Contrastive Topic + Focus

The mainstream analysis of answers to multiple-Wh questions is the one
advanced in Jackendoff (1972) and recently pushed forward in Büring (1999, 2003).
Büring’s (2003) analysis is more elaborated, so I will restrict my comments to this
study.
Büring (2003) takes the semantics of a question to denote a set of propositions
(cf. Hamblin (1973) and chapter 2 of this dissertation). For instance, the denotation of
the question in (11a) would be the set of propositions in (11b), where the Wh-phrase
in the question has been replaced by different alternative values that match the Wh-
phrase in semantic type and that are available in the context. And according to
Rooth’s (1985, 1992) theory of focus semantics, an appropriate answer to the question

240
in (11a) will be any of the propositions in this set, for instance (11c):

(11a) Who got the flu?

(11b) [[Who got the flu]] = {[[Kepa got the flu]], [[Eider got the flu]], [[Adam got the
flu]], [[Ibon got the flu]], …}
(11c) Kepa got the flu.

Furthermore, following this approach, a multiple Wh-question like (12a) will

denote a set of questions; that is, a set of sets of propositions like (12b). This question
could be answered by the sentence in (12c):

(12a) Who cooked what?

(12b) [[Who cooked what]] = {{[[Adam cooked cod]], [[Adam cooked rice]], [[Adam
cooked eggplants]]…}, {[[Julen cooked rice]], [[Julen
cooked pasta]], [[Julen cooked tuna]]…} …}
(12c) Adam cooked eggplants and Julen cooked pasta.

This type of approach to the semantics of questions is adopted by Büring

(2003) in his analysis of discourse structuration and answerhood, proposing that in
an answer to a multiple-Wh question we have different possible answer strategies.
Following Kuppevelt (1991, 1995, 1996) and Roberts (1996), Büring (2003) proposes
that the question in (12a) is split up, as it were, into sub-questions such as ‘Who
cooked cod?, Who cooked rice?...’ or into sub-questions such as ‘What did Adam
cook?, What did Julen cook?...’. Sets of questions would be hierarchically structured
in the discourse, as abstractly represented in (13) (taken from Büring (2003: 516)):

241
(13)
Discourse

Question Question
g
Sub-Q Sub-Q Sub-Q …
g ei g
Answer SubSubQ SubSubQ Answer
g g
Answer Answer

A complex question is related to a set of contextually given questions that

constitute different strategies. The different answer strategies are represented in the
discourse trees (D-Trees) in (14b) and (15b), for questions (14a) and (15a),
respectively. In this case, the choice of answer strategy would ivolve whether to start
answering by the agents of the event of buying or by its themes. Crucially, according
to Büring (2003), this will determine the information-packaging nature of the
elements that are not given in the question (the elements that stand for each of the
Wh-phrases). Büring’s (2003) analysis makes reference to two independent discourse-
configurational primitives: (i) the ‘contrastive topic’ (CT) that would indicate the
answer strategy to follow, and (ii) the ‘focus’ (F) that would provide the answer
within the strategy indicated by the contrastive topic:

(14a) Who bought what?

(14b) [[Who bought what?]]ct = {{x bought y │ y ∈ De}│x ∈ De}

Discourse-Tree

What did John buy? What did Mary buy? What did ...?

JOHNCT bought BEERF MARYCT bought WINEF

242
(15a) Who bought what?
(15b) [[Who bought what?]]ct = {{x bought y │ x ∈ De}│y ∈ De}

Discourse-Tree

Who bought beer? Who bought wine? Who bought ...?

JOHNF bought BEERCT MARYF bought WINECT

I agree with Büring’s (2003) analysis of the contribution of each of the elements
in that both elements have a contrastive nature. But I find the taxonomy of
information-packaging elements proposed by Büring (2003) puzzling. According to
Büring (2003), both information-packaging elements are analysed as having the very
same semantic import; i.e., a contrastive nature raising alternative values à la Rooth
(1985). Recall that according to Rooth’s ‘Alternative Semantics’ approach, a sentence
with focus would have two denotations: the ‘Ordinary Semantic Value’, that will be
the proposition obtained compositionally by montagovian function application (this
proposition won’t be affected by the focus), and the ‘Focus Semantic Value’, a set of
propositions obtained by the substitution of the focused phrase with alternatives
available in the discourse that match the focus in semantic type (i.e., roughly, the
semantic value of the question being answered in a Hamblin-Karttunen-type
semantics of questions). My concern is that although the theory of answer strategies
of Büring (2003) captures in an elegant way the semantics of multiple Wh-questions,
the ‘topicness’ of the ‘contrastive topics’ proposed is not very well established; after
all, both the ‘focus’ and the ‘contrastive topic’ are analyzed as having the very same
semantic import, the only difference being that the contrastive topic is the first
variable whose value is decided.
Furthermore, as Büring himself notes (Büring (2003: 512)), the so-called
‘contrastive topic’ does not behave in some relevant respects like other topics. For
instance, its presence is mandatory and not optional (hence, they cannot be elided or
silent), and they constitute the answer (in part), instead of stating necessarily

243
old/given information. Clearly, the answer function that ‘contrastive topics’ display
is definitional of focus. Thus, I would want to suggest that we don’t need the
theoretical primitive of ‘contrastive topic’ in order to capture the semantics of these
sentences2. These elements would be just foci that by coming first, indicate the
answer strategy3. Therefore, the proposal to be developed in this chapter is that in
these constructions we have a pair of elements as the focus. For instance, in the case
of the discourses of (14) and (15), the focal elements can be regarded as taking part in
a relation denoted by the verb; the first element that stands for a Wh-word of the
question sets the domain of the relation and the second element sets its range.
In a nutshell, then, my claim is that the semantic representation proposed by
Büring (2003) captures correctly the discourse-contribution of multiple-Wh questions,
but that the two (or n) elements that stand for a Wh-phrase in the question have a
focal nature. Thus, we could dispense with the theoretical primitive of ‘Contrastive
Topic’ for these constructions.
In the next section I will discuss some of the intonational, semantic and
morphosyntactic properties observed in these constructions.

6.3. The Properties of Split Foci Revisited

In this section I discuss the phonological, semantic, and morphosyntactic

properties of split foci in a variety of languages. I show that an analysis in terms of
’Contrastive Topic + Focus’ is unsustainable and that the patterns that they display
strongly suggest that they are split (pairing) foci.

6.3.1 Intonational Properties of Split Foci

It is a widely known fact that in many languages the elements that stand for a
Wh-phrase in a multiple Wh-question bear a pitch accent (cf. Bolinger (1958),
Jackendoff (1972), Liberman & Pierrehumbert (1984) and Büring (2003)) for English,

2 However, probably something like this will be necessary to analyze the ‘additional topic’
constructions analyzed in Umbach (2001). See as well the analysis of ‘implicit subquestions’ in Büring
(2003).
3 Similar points of view with different theoretical points of departure can be found among others in

Jackendoff (1972), Williams (1997), Steedman (2000), Kadmon (2001) and van Hoof (2003).

244
Büring (1999) for German, Godjevać (2000) for Serbo-Croatian and Aske (1997), A.
Elordieta (2001) and Irurtzun (2003) for Basque, among others).
However, even if it is true that each of the elements that stand for a Wh-phrase
bears a pitch accent, the tunes associated to each of the elements are quite different.
For instance, Jackendoff (1972) analyzes an answer to a multiple-Wh question as
having two different pitch-accents which he calls ‘A’ and ‘B’ accents. This is
represented in (16):

(16)

FRED ate the BEANS

B A

As Jackendoff (1972) describes them, the ‘B accent’ is characterized by a ‘fall-

rise’ tune contour and the ‘A accent’ is characterized by a simple ‘fall’. According to
Jackendoff’s (1972) analysis, the B tune is associated with a ‘free’ variable and the A
tune to a ‘dependent’ variable, where the identification of the second variable will
depend on the identification of the first one (cf. as well Liberman & Pierrehumbert
(1984)). These differences in tune-structure and ‘freedom’ of the variables have been
taken by Büring (2003) and others as implying that these are two different
informational-packaging primitives: a focus (characterized by the A accent), and a
contrastive topic (characterized by the B accent; cf. section 6.2). However, I would
want to remark that notwithstanding the idiosyncrasies of focus-marking tunes in
different languages, there is some regularity in the tunes for the so-called ‘contrastive
topics’ across languages: like the ‘B accents’ of English, in other languages like
German, Basque or Serbo-Croatian the so-called ‘contrastive topics’ are characterized
by a final pitch rise.
If we focus on the Basque case, for Central Basque (cf. section 2.2.1.1) I have
analyzed this type of constructions as involving a tune composed by a H* pitch
accent and a H- boundary tone for the element answering the first Wh-phrase,
followed by a H*+L L- in the second phrase standing for a Wh-phrase in the question

245
(cf. Irurtzun (2003)). That is, there are two different pitch accents; the first one is H*
and it is followed by a rise (a H-), whereas the second one is H*+L and it is followed
by a L- tone, as illustrated in (18):

(18) Mariak liburua, anaiak mandarina, Amagoiak mangera eta Enarak ordenadorea
eraman dituzte. (Maria (took) the book, the brother (took) the tangerine and Enara took the
computer).

[from Irurtzun (2003: 75)]

First Part of the F-Structure: H* H- Second Part of the F-Structure: H*+L L-

%L H* H- %L H*+L L-L%
| | | | | |
[[Ma ri ak]ι [li bu ru a]ι]υ

In the pitch accent varieties of NBB (cf. section 2.2.1.2) A. Elordieta (2001) notes
that the first element in split focus constructions can bear an accent, as in (17):

(17) Nire nebâk Amaiâri utzi dio kotxe berria.

my brother Amaia.to lend AUX car new

‘[My brother]F lent the new car [to Amaia]F’

246
G. Elordieta (p.c.) does not find unambiguous evidence for an accent in the
first focalized constituent. However, he confirms that this constituent ends in a rising
pitch (contrary to topics, where the pitch level is maintained in a plateau, without a
rise or a fall, until the right boundary of the phrase (cf. G. Elordieta (1998, 2003)).
However, if we consider the properties exhibited by other languages with
respect to the intonational properties of answers to multiple Wh-questions, Serbo-
Croatian reveals a particularly interesting fact (cf. Godjevac (2000)). In an answer to a
multiple-Wh question, each of the elements bears a L*+H pitch accent in Serbo-
Croatian; further, akin to English or Basque, the so-called ‘contrastive topic’ phrase
ends in a H- phrase accent and the ‘focus’ in a L- tone. However, the interesting point
is that there is one additional tonal gesture involved in these constructions: an initial
high boundary tone in the ‘focus’ (i.e., %H cf. Godjevac (2000)). This is shown in (19),
a possible answer to a question like ‘Who gave a lemon to whom?’:

(19)
%L L*+H H- %H L*+H L-
| | | | | |
Je le na je Ma ri ji dala.
“[Jelena]F gave it [to Mary]F”

The high F0 of the first syllable of ‘Mariji’ in (19) is not caused by the
proximity of the H- phrase accent of ‘Jelena’. In fact looking at (20) (where both
elements are not close to each other), it seems that the initial %H is a categorial
property of these constructions, since in normal/single focus utterances there is no
%H at the left edge of the focus phrase:

(20)
%L L*+H H- %H L*+H L-
| | | | | |
JE LE NA je dala ravan MA RI JI.
“[Jelena]F gave the flat one [to Mary]F”

247
In my view, these facts constitute evidence that, on the one hand, in answers
to multiple-Wh questions, both elements standing for a Wh-phrase bear a pitch
accent, which might be indicative of their focal nature. On the other hand, it is also
evidence that the tune differences between both elements are usually phrasal, and
that there is a striking regularity across languages in that the tunes associated to the
so-called ‘contrastive topics’ end in a high tone. Furthermore, as observed in Serbo-
Croatian, the so-called ‘foci’ of the answers to multiple-Wh questions are not the
same elements as the foci of answers of single-Wh questions (which lack any initial
%H). This suggests that the focus of answers to multiple-Wh questions are global
entities (i.e., the pair of elements that stand for the Wh-phrases in the question). Thus,
the conclusion that follows from the phonological properties of split focus
constructions we discussed so far is that in answers to multiple-Wh questions we
don’t have a ‘contrastive topic’ and a ‘focus’ (as proposed by Büring (2003)), given
that the intonational patterns associated to them are not the same as those in
sentences with bona fide topics and single foci. In these multiple Wh-constructions,
what we have is a single focus that is the pair of both elements, and the H- phrase
accents at the right edge of so-called ‘contrastive topics’ could be analyzed as
‘continuation rise’ contours, something that would not be surprising under the
analysis defended in this dissertation, whereby the focus structure is split among
both elements bearing the [+F] features. That is, the continuation rise at the end of the
first focalized element (i.e., the first part of the F-Structure) indicates that the second
part of the F-Structure is to come.

6.3.2. Semantic Properties of Split Foci

As is widely acknowledged (cf. among others Comorovski (1996), Bošković
(2002), Büring (2003)), in languages like English (21) or Basque (22) that show overt
movement of (one of) the Wh-phrases to Spec-CP, sentences like (21b) and (22b)
constitute partial answers of multiple-Wh questions like (21a) and (22a) respectively45:

4 Basque allows for single Wh/Focus fronting, with the rest remaining in situ, or multiple Wh/focus
fronting with no superiority effects (see Reglero (2005), Jeong (2005)).

248
(21a) Who broke what?
(21b) John broke the door, Mary broke the computer… (pair list)

(22a) Zeinek erosi du zer?

which buy AUX what
“Who bought what?”
(22b) Jonek esnea erosi du, Maiderrek ardoa erosi du… (pair list)
Jon milk buy AUX, Maider wine buy AUX
“John bought milk, Maider bought wine…”

In fact, in English, a question like (16a) is incongruent in a scenario that

demands a single-pair answer. Scholars like Wachowicz (1974, 1975) or Bolinger
(1978) make a distinction between two types of multiple-Wh questions: Matching
Questions and Conjoined Questions. Matching questions are the “usual” multiple-Wh
questions, those questions like (16a) that demand request about how to pair some
elements; then they demand a pair list answer (cf. Boskovic (2002) and references
therein for a possible analysis). Thus, the example in (16a), which demands a pair list
answer, is a good example of matching questions, but examples like (23) and (24) are
not (examples from Wachowicz (1974, 1975)):

(23) *Who killed Robert Kennedy when?

(24) *Who is keeping the silver dollar in which museum?

Arguably, the oddness of these questions is due to the impossibility of having

several events of killing Robert Kennedy (23) or keeping the silver dollar at the same
time in various museums (24). The oddity of these sentences is strictly related to the
incompatibility between one single event and a pair-list answer; compare (23) and
(24) to similar examples that allow the multiplicity of events like those in (25) (for
(23)) and (26) (for (24)):

5
Bošković (2002), based on Hagstrom (1998), associates the possibility of having single-pair answers to
multiple-Wh questions in a given language to the possibility of moving the Wh-phrase to FocP.
However, this conclusion is too strong and the typological split is not that clear (see below).

249
(25a) Who saw Robert Kennedy when?
(25b) Who killed which Kennedy?

(26) Who kept the silver dollar in which museum?

On the other hand, conjoined questions are characterized as a conjunction of

questions that demands for the independent identification of two variables. In many
languages, this type of questions can only be asked with a conjunction of questions.
An example of a conjoined question in English could be the one in (27), a
grammatical variant of (23):

(27) Who killed Robert Kennedy, and when did he do it?

Here the question is perfectly natural because it demands independently and

in two single-Wh questions for two pieces of information.
Leaving aside the conjoined questions, what I want to argue is that the pairing
pattern of matching multiple-Wh questions is explained straightforwardly with the
analysis of the derivation and interpretation of focus structure I argue for in this
dissertation. Chomsky (1973), Higginbotham & May (1981), Gutiérrez-Rexach (1999)
and others have argued that in a multiple-Wh question like (28a), an operator
absorption takes place creating a compound polyadic operator that quantifies over
pairs of variables (in the specific case of (28a), people and eatable things). This is
represented in the logical form representation in (28b):

(28a) Who ate what?

(28b) [WH x, WHy: person(x) & eatable thing(y)] x ate y

According to these authors, this logical form representation for multiple-Wh

questions is what will provide us with the bijective interpretation these questions

250
typically show. Thus, the most natural assumption is to take both elements that stand
for the pairs of variables in the question to be focal: there is one question (which is
complex), and one answer (which is complex)6. As a result, the uniqueness of focus,
i.e., the idea that each sentence has just one focus is kept given the logical form
representation proposed in chapter 4 whereby all the [+F] material will fall in the
scope of a δ-quantifier over events. We do not have to postulate it as a principle. In
the cases at hand, the actual focus will be split; it will be the pair of elements being
marked [+F], for they are mapped into the scope of the quantification over events.
This was illustrated in (5) and is repeated here as (29) for convenience:

(29)
vP
3
DPF v’
3
v VP
3 LF: [δe [RESTRICTION] [SCOPE]]
V DPF

So, for instance, for the sentence in (30b) (as a partial answer to (30a), cf.
chapters 2 and 4), we would have the logical form in (30c)7:

(30a) Who at what?

(30b) [John]F ate [cod]F...
(30c) δe [Eating(e) & Past(e)][Eating(e) & Past(e) & Agent(e, John) & Theme(e, cod)]

As I said before, the corollary of such a proposal is that there will be just one
focus per sentence, even if it has the form of a pair. An important advantage of this

6 See Pesetsky (1982) for a previous analysis that relates the Wh-absorption with a Focus-absorption for
gapping constructions. My analysis differs from Pesetsky’s (1982) in that even if questions and
answers are isomorphic, having one focal element standing for each Wh-phrase, in multiple foci
constructions I will not postulate the absorption mechanism, for the pair-list reading is obtained
directly from the quantificational logical form representation proposed and the necessity of a complex
event implied in the multiple-Wh question.
7 Obviously, the moment we acknowledge the need of a pair-list answer we have the problem of the

plurality of events and its mereology. This extremely interesting and complicated issue falls out of the
scope of this dissertation (cf. Schein (1993) for an analysis).

251
proposal from a theoretical point of view is that instead of introducing the uniqueness
of focus as a principle, it will be just derivative of the nature of the logical form
representation of sentences with focus: in the logical form of these sentences we have a
quantification over events, and this quantification has just has one scope. There is no
room for more. I think this is a nice prediction, and one of the advantages of this
proposal compared to previous approaches.

6.3.3. Morphosyntactic Properties of Split Foci

In this section I will review morphosyntactic evidence that reinforces the
previous points that reveal that (i) both elements that stand for a Wh-phrase in the
question are focal, and (ii) there aren’t two foci; rather, the pair formed by the two
[+F] marked phrases constitutes the single focus of the utterance.
From the point of view of word order, it is difficult to make a distinction
between a syntactic structure with ‘multiple foci’ and a syntactic structure with a
focus and one or more topic phrases as the schematic syntactic trees in (31a) and
(31b) show:

(31a) Multiple [+F] XPs (31b) One [+F] XP, and several [+Top] XPs
FocP TopP
3 3
XP1 Foc’ XP1 Top’
3 3
XP2 Foc’ XP2 Top’
3 3
XP3 Foc’ Topº FocP
3 3
Focº […] XP3 Foc’
3
Focº […]

There is little difference in syntactic (word order) terms, just a Topº between
XP2 and XP3 in (31b)).
In the literature, there are a number of studies on multiple Wh-movement
questions that propose that one Wh-phrase undergoes focus movement and the rest

252
undergo topic movement (cf. i.a. Bošković (1999), Lambova (2001) and Reglero

(2002)). However, these analyses pose several puzzles:

(i) As far as I can tell, there is no explanation of what it means for a Wh-phrase
to be topical.
(i) Moreover, some of these studies propose that both types of elements are
displaced to the same projection, for instance, a ΔP as in Lambova (2001) and Reglero
(2002) (‘Δ’ for ‘discourse; cf. as well Uriagereka (1995)).
(ii) Furthermore, there is no way to predict which element is going to be the
‘contrastive topic’ and which one the ‘focus’ in the answer. Consider (32b), where
there is a change of strategy in the answer. In the first two sentences the word
occupying the putative contrastive topic position is the subject and the focal position
is occupied by the direct object. However, in the third sentence the pair is reversed
and the direct object occupies the putative “contrastive topic” position and the
subject occupies the putative “focus” position.

(32a) Zeinek ekarri du zer?

which bring AUX what?
“Who brought what?”

(32b) Nik atuna ekarri dut, Pablok txuletak ekarri ditu, eta ardoa denok ekarri dugu.
I tuna bring AUX Pablo cutlet bring AUX and wine all bring AUX
“[I]F brought [the tuna]F, [Pablo]F brought [the cutlets]F, and [all of us]F
brought [the wine]F”

The objects of the first two sentences in (32b) don’t appear to have anything
special in them that make them discursively different from the subject in the third
sentence. Likewise for the direct objects; we could change the strategy and the
discourse-congruence would be kept.
Similar evidence can be found in Russian. According to Grebenyova (2006), a
multiple-Wh question like “What did Ded Moroz give to whom?”, could be answered as

253
in (33a), with a series of sentences with fronted foci, or as in (33b), where in the first
sentence the two elements that stand for a Wh-phrase in the question appear in situ:

(33a) [Ivanu]F [velosiped]F Ded Moroz podaril, [Lene]F [knigu]F on podaril,

Ivan.to bike Ded Moroz gave Lena.to book he gave
a [Saše]F [časy]F on podaril.
and Saša.to watch he gave.
“Ded Moroz gave [Ivan]F [a bike]F, he gave [Lena]F [a book]F, and he gave
[Saša]F [a watch]F”

(33b) Ded Moroz podaril [Ivanu]F [velosiped]F, [Lene]F [knigu]F on podaril,

Ded Moroz gave Ivan.to bike Lena.to book he gave
a [Saše]F [časy]F on podaril.
and Saša.to watch he gave.
“Ded Moroz gave [Ivan]F [a bike]F, he gave [Lena]F [a book]F, and he gave [Saša]F
[a watch]F”

According to Grebenyova (2006) a variant of (33a-b) with all the elements that
stand for a Wh-phrase in the question appearing in situ would be ungrammatical. She
speculates that it might be because in the series of answers to a multiple-Wh question
we need a real contrast between the elements, and that is obtained with contrastive
focalization to the left periphery. Nevertheless, the important data for me appears in
the first conjunct of (33b), where the elements that stand for a Wh-phrase in the
question appear in situ. In this position, they can hardly be contrastive topics,
whereas their focal status is well established.
This strongly suggests that these elements are not topics. As a conclusion, I
will rely on the semantic interpretation and compositional parsimony for claiming
that all the XPs that stand for a Wh-phrase in the question are focal. It is their
definitional property. In the next section I will present some morphosyntactic
properties that give plausibility to this claim. Split focus constructions are not very
well described syntactically in the literature, unfortunately, so I will use evidence
from the multiple-Wh questions they answer, and the morphological attachment of

254
particles to show that we don’t have a topic and a focus, nor independent foci, but a
single complex focus.

6.3.3.1 ‘Emphatic’ Morphemes and Information Structure

Some dialects of Basque like the dialect of Lekeitio (G. Elordieta (p.c.)) make
optional use of a set of ‘strong’ pronouns in focus positions. Thus, an answer to a Wh-
question like (34a) might be answered with a regular 1st person singular pronoun
(34b) or with a strong variant of it (34c):

(34a) Nork ikusi du Mikel?

Who see AUX Mikel
“Who saw Mikel?”

(34b) [Nik]F ikusi dut Mikel.

I see AUX Mikel
“[I]F saw Mikel“

(34c) [Neuk]F ikusi dut Mikel.

I see AUX Mikel
“[I]F saw Mikel“

Interestingly, strong pronouns can also be used in putative ‘contrastive topic’

positions of answers to multiple Wh-questions like (35b). However, these strong
pronouns create ungrammaticality in topicalizations (36):

(35a) Nork ikusi du nor?

Who see AUX who
“Who saw whom?”

(35b) [Neuk]F [Mikel]F ikusi dut…

I Mikel see AUX

255
“[I]F saw [Mikel]F”

(36) *Neuk, denbora banu, gehiago bidaiatuko nuke.

I time if.have more travel.fut AUX
“Me, if I had time, I would travel more”

In these respects all phrases that stand for a Wh-phrase of the question pattern
like foci, and unlike topic phrases. The possibility of having a strong pronoun, then,
seems to be a characteristic of the focal nature of the item.
Furthermore, with the analysis just sketched, we can account for the
usage/lack of usage of contrast particles of Wh-in-situ languages like Japanese or
Korean, where multiple-Wh questions can be answered with either a single-pair or
pair-list answer (cf. Hagstrom (1998), Bošković (2002) for discussion). Bošković (2002)
gives the following scenario for triggering single-pair answers:

(37) SCENARIO: John is in a store and in the distance sees somebody buying a piece of
clothing, but does not see who it is and does not see what the person is
buying.

With this scenario, in a language like English, that has overt movement of Wh-
phrases a question like (38) is incoherent (since, as said earlier, it is inherently a
matching question). Its counterpart in a ‘Wh-in-situ language’ like Japanese, on the
other hand, is fine, as represented in (39):

(38) Who bought what?

(39) Dare-ga nani-o katta no?

who-nom what-acc bought Q
‘Who bought what?’

256
Whichever the explanation for the lack of single-pair reading in Wh-movement
languages (see Hagstrom (1998) and Boskovic (2002) for a possible analysis), the case
is that this reading is available in Wh-in-situ languages; i.e., that in both Japanese and
Korean multiple-Wh questions can equally be interpreted as matching questions and
as conjoined questions (cf. Bošković (2002) for an analysis). The crucial fact here is that
in this type of languages, an answer to a multiple-Wh question is different when it is
a single-pair or a pair-list answer. In languages like Japanese or Korean that allow for
the single-pair reading, the usage of particles like ‘-wa’ for Japanese or ‘-nun’ for
Korean varies with the type of answer: the appearance of those particles is
mandatory after the first element in cases of a pair-list answer but, remarkably, when
the question demands a single pair, the first element cannot bear such a particle. In
that case the subject must appear with the particle ‘-ga’ and ‘-wa’ is not allowed (cf.
(40a-b) for Japanese and (41a-b) for Korean)8:

(40a) Takako-wa wain-o kaimashita... (pair list)

Takako-WA wine-ACC bought
“Takako bought wine…”
(40b) Takako-ga wain-o kaimashita (single pair)
Takako-GA wine-ACC bought
“Takako bought wine…”

(41a) Yenghui-nun wain-ul sassta.... (pair list)

Yenghui-NUN wine-ACC bought
“Yenghui bought wine…”
(41b) Yenghui-ga wain-ul sassta. (single pair)
Yenghui-GA wine-ACC bought
“Yenghui bought wine”

8Many thanks to T. Oda, T. Iseda, T. Sawada, H. Hoshi, S-Y Kuroda for their Japanese judgments and
to D-H. An, B. Kang, B-S. Park for their Korean judgments.

257
Again, although these particles have been analyzed as conveying the
discursive notion of ‘topic’, in these cases it is hard to maintain that the words they
are attached to are topics, since they are attached to an element that answers partially
the question and might not be mentioned in the previous discourse. In fact, as has
been argued by some scholars (cf. Kuroda (1965), Munakata (2002), Maruyama
(2003)), these particles should be better reanalyzed as marking ‘contrast’ rather than
topic, this being one of the core properties of focal elements. Hence, the appearance
of these contrast particles in pair-list answers but not in single-pair ones would
follow from the matching type of the former ones and the conjoined type of the latter
ones9. This evidence suggests that we are not dealing with ‘contrastive topics’.
Further evidence for this claim comes from “no da constructions”; single pair
questions in Japanese can typically be answered with some prototypical focus
constructions: the so-called “no da construction” or cleft constructions. In “no da
constructions”, the entire matrix clause is headed by the nominalizer -no followed by
the copula -da. (cf. Kuno (1973), Hiraiwa & Ishihara (2002)). If there is no element
within the nominalised phrase bearing a nuclear stress, then the whole phrase is
interpreted as focus. On the other hand, all XPs within the nominalised phrase that
bear a nuclear accent are interpreted as focus, thus, having several elements
highlighted provides an answer to a multiple-Wh question:

(43) [TARO-ga KONO-RINGO-o tabeta no] da

Taro-NOM this-apple-ACC ate C BE
“(Lit.) It is Taro, this apple that (he) ate”

Besides, answers to multiple-Wh questions can also be formed with a multiple

cleft construction like (44) (cf. Kuroda (1999), Hiraiwa & Ishihara (2002)):

(44) [Taro-ga ageta no]-wa Hanako-ni ringo-o (3-tu) da

Taro-NOM gave C wa Hanako-DAT apple-ACC 3-Class. BE
“(Lit.) It is (three) apples to Hanako that Taro gave”

9 cf. van Hoof (2003) for a similar analysis of German and Dutch data.

258
According to Hiraiwa & Ishihara (2002), these constructions are derived from
the no da constructions in two steps: (i) movement of the elements that stand for the
Wh-phrases in the question to SpecFoc, and (ii) remnant topicalization of the rest of
the clause10. The copula –da of these constructions is a grammaticalized focus particle
that heads FocP. If Hiraiwa & Ishihara’s (2002) analysis is correct, in these
constructions we are in front of focal elements, for both the no da construction and
the cleft construction are bona fide focalization strategies. An analysis based on the
notion of contrastive topic would be problematic.
This idea is reinforced with the patterns of usage of ‘focus particles’. In some
languages that mark focus with ‘focal particles’ the observation is that these particles
are not assigned to each element that stands for the Wh-phrase. For instance,
Manding is a language that marks morphologically the focus with the attachment of
the suffix ‘-le’ to the focal XP. However, as an answer to a multiple-Wh construction,
a sentence like (45) results in ungrammaticality (cf. Bamba & Liberman (1999)):

(45) *Músà-lè yé kú sán Búlàmá-lè má…

Musa-FOC AUX yam buy Ibrahim-FOC from
‘[Musa]F bought yam(s) from [Ibrahim]F…’

In the same line, languages that don’t allow for multiple-Wh questions like
Italian don’t allow either for multiple foci (cf. Calabrese (1984), Rizzi (1997), Belletti
(1999)). Observe the ungrammaticality of the question and answer in (46):

(46a) *Chi ha scritto che cosa?

who AUX write what thing
“Who wrote what?”

10 Very similar, then, to Ortiz de Urbina’s analysis of rightward foci that we saw in section 4.1.2.7.

259
(46b) *Mario ha scritto la lettera
Mario AUX write the letter
“[Mario]F wrote [he letter]F

The fact that both question and answer pattern together is somehow
predictable from the analysis proposed in this chapter. Proposing that ‘Mario’ in the
answer of (46b) is a ‘contrastive topic’ would complicate the pattern, since there
would have to propose two independent reasons for the ungrammaticality of the
question of (46a) and the answer in (46b), unless, obviously, we consider ‘chi’ in the
question of (46a) a ‘contrastive topic’ itself, which does not seem very explanatory.

6.4. Summary and Conclusions

In this chapter, I have analyzed the properties of the answers to multiple-Wh
questions. I have argued that in these constructions, we have a split focal structure
and that at logical form, it leads towards having a pair of elements as the actual focus.
This analysis provides us with a natural understanding of the question-answer
pairings since all the material that stands for a variable in the question is taken to be
focal in nature. Thus, treating these answers as instances of split foci, we can
dispense with the theoretical primitive of ‘contrastive topic’ used previously in the
literature to avoid having two foci. Thus, we gain in our understanding of the
interface phenomena observed crosslinguistically.

260
CHAPTER 7. SUMMARY AND GENERAL CONCLUSIONS

In this dissertation I have explored the nature of focus and its locus within the
overall architecture of grammar.
The main conclusion that derives from this work is that the only way to
capture the effects of focus at the interfaces is with a ‘syntactocentric’ architecture of
grammar and a derivational analysis of F-Structure composition, as the one I
developed here. As I showed throughout the dissertation, the effects of focus give
rise to a wide variety of syntactic, phonological, morphological and semantic
phenomena in different languages. I have offered a detailed analysis of Basque, a
language that displays several of these phonological and syntactic phenomena
overtly. For instance, in Basque an out-of-the-blue sentence like (1) has the form
[SOV]F, with nuclear stress on the object, whereas a contextualized utterance with
focus on the subject has the word order [S]FVO with nuclear stress on the subject, as
represented in (2a). Alternatively, it can also have the word order [O]T[S]FV with the
object topicalized over the focal subject, cf. (2b):

(1) [Bidasoak maila galdu du]F [out-of-the-blue utterance]

Bidasoa category loose AUX Broad Focus
“[Bidasoa lost the category]F” S-Ó-V

(2a) [Bidasoak]F galdu du maila

Bidasoa loose AUX category
“[Bidasoa]F lost the category” [contextualized utterances]
Narrow Focus on the Subject

(2b) Maila [Bidasoak]F galdu du (OT)-Ś-V-(O)

category Bidasoa loose AUX

“The category, [Bidasoa]F lost it”

In order to explain these facts, previous analyses of the grammar of focus

purported a causal relation between the nuclear stress and F(ocus)-Structure. I have

261
referred to those analyses as ‘NSR-based theories of F-Structure’. As an instance of
that type of approach, works like Reinhart’s (1995, 2006) propose that the F-Structure
of a sentence is calculated from the placement of nuclear stress in PF. Basically, the
idea is that phrases containing nuclear stress in PF can be interpreted as focused.
Reinhart’s precise formulation is illustrated in (3), taken from Reinhart (2006: 158):

(3) The focus set: The focus set of a derivation D includes all and only the constituents
that contain the main stress of D.

Following this approach, then, the focus of a sentence will be computationally

ambiguous and the actual focus will be chosen at the discourse level from among the
elements within the focus set.
Assuming that syntactic trees are mapped into metrical grids and that nuclear
stress is assigned to the most deeply embedded position (à la Cinque (1993)), A.
Elordieta (2001) and Arregi (2003) have analysed Basque examples like (1) as
instances of focus projection from the element that bears nuclear stress. In order to
account for the word order changes exhibited by some sentences (i.e. contextualized
utterances like (2a) and (2b)) these authors propose that nonfocal material must
dislocate to ensure that the focal element is the most deeply embedded. There it gets
nuclear stress and, by the principle in (3), it can be interpreted as focus.
These NSR-based approaches to F-Structure carry two strong implications for
the theory of grammar:
(i) The grammatical architecture is one where the PF component ‘invades’
syntax, triggering some displacement operations in order to obtain a certain
configuration at a later derivational stage of PF. Thus the interface between syntax
and PF is bidirectional.
(ii) On the other hand, given that nuclear stress is assigned in PF, advocates of
NSR-based theories of F-Structure suppose a direct connection between both
interpretive modules, PF and LF. According to those approaches there is no F-
Structure per se during the computation from the numeration to the interfaces; F-
Structure ‘emerges’ in PF deriving from the computation of the phrases that contain

262
nuclear stress. This view necessarily leads to the conclusion that PF (nuclear stress
placement) and LF (focus interpretation) are linked to each other.
An architecture of this sort, for instance, is provided by Jackendoff’s (1997)
parallel model illustrated in (4):

(4)
Phonological Syntactic Conceptual
Formation Rules Formation Rules Formation Rules

Phonological Syntactic Conceptual

Structures Structures Structures

PS-SS SS-CS
Correspondence Rules Correspondence Rules

PS-CS
Correspondence Rules

I have shown that this cannot be the case, and I have discussed a wide variety
of phenomena in different languages that show that this conception of the
architecture of grammar is misguided. I have brought into the discussion the patterns
of nuclear stress placement in languages like Basque, English, Bengali, Italian, Greek
and European Portuguese, the phonological phrasing associated with focus in
languages like Japanese, Swedish, Korean, Greek and Chicheŵa, the use of syntactic
dislocation in languages like Dutch, Hungarian, Russian and Zapotec or the specific
morphological particles used in Korean, Tuki, Manding or Japanese among others.
My position has been that by exploring those grammaticalizations of focus in
different languages we can gain a better understanding of the interactions of
different modules of the grammar and over all, about the general design of the
architecture of grammar. In particular, I have argued that there is ample evidence
that shows that parallel architectures cannot work.

263
In particular, have shown that the F-Structure of a sentence is unambiguously
encoded in the derivation and I have argued that the effects of focus are better
captured in a ‘syntactocentric’ and derivational model of the architecture of
grammar. I have proposed that F-Structure is fixed in the narrow syntax via Merge,
and that interface modules can interpret it, hence maintaining the overall inverted-T
model of the architecture of grammar. The F-Structure of a sentence is fixed and
unambiguous in computational terms. Otherwise, it wouldn’t be able, for instance, to
have the interface effects that it has (association with specific pitch accents,
phonological phrase boundaries, quantifiers etc.). Furthermore, I have shown that
any derivational theory that purports that some syntactic operations take place in
order to provide a certain PF representation cannot work; they necessarily imply
global look ahead operations that crash with the derivational conception of the
architecture of grammar and computationally ambiguous F-Structures, contrary to
fact.
As an alternative, I have offered an innovative analysis that locates F-Structure
composition within narrow syntax and which allows us to unify all effects of
focussing in a parsimonious way. My proposal is that the F-Structure of a sentence is
built up derivationally from the elements that bear a [+F] formal feature, and that
then it is interpreted by both interface components. F-Structure, then, will be the
outcome of phrase structure building in Bare Phrase Structure theory.
The virtues of this proposal are evident; on the one hand, the F-Structure of a
sentence will be composed and fixed in narrow syntax since the F-Structure of a
sentence will be the outcome of the derivation of [+F] featured lexical items. Then,
we can maintain the classic inverted-T model of the architecture of grammar while
accounting for F-Structure composition in Bare Phrase Structure:

264
(5)

Lexicon

Language

[Phonological Form] [logical form]

A-P System C-I System

So, according the proposal I have put forth narrow syntax creates a well
specified F-Structure and interface components can ‘read’ it and apply some
operations upon it. By constructing F-Structure in a derivational fashion, my
proposal avoids all the problems of previous approaches that violated the
‘inclusiveness condition’ with the assignment of a [+F] feature to built-up XPs.
Furthermore, I have also shown that with the Derivational Approach to F-Structure
(DAFS) that I proposed, all the problems derived from the introduction of focus sets
disappear; the F-Structure is unambiguously fixed in the narrow syntax and thus, the
F-Structure can get interpreted at interface components:

(i) On the phonology side, depending on the language the F-Structure

will be aligned with a phonological phrase boundary and assigned a
nuclear accent.
(ii) On the semantics side, focus will be mapped into the nuclear scope
of a quantification over events.

In PF, the syntactically fixed F-Structure will be placed at the structural

description of phonological rules; namely prosodic phrasing and nuclear stress
assignment operations. In particular, the nuclear stress assignment rule I proposed
applies within the F-Structure (cf. (6)),

265
(6) Focal NSR (FNSR): Assign nuclear stress to the most embedded element within the
F-Structure.

With the FNSR I have shown that we circumvent all the circularity problems
of alternative NSR-based approaches that, paradoxically, need to ‘know’ where to
assign the nuclear stress (i.e., which constituent is the focus) while maintaining that
the focus is just ‘inferred’ from the PF representation. As I argued in chapter 3, the
FNSR captures in a straightforward way the stress assignments associated with
focus: the FNSR interprets the F-Structure which has been built on the basis of the
elements that are marked [+F] regardless of the position they occupy in the clause.
Likewise, my proposal of the DAFS also provides a coherent way to interpret
F-Structure on the semantic side. Following Herburger’s (2000) analysis I have
proposed that the core semantic contribution of focus can be captured by assuming
that the F-Structure is mapped into the nuclear scope of a quantification over events.
I have proposed two such quantification structures:

a) A unary existential quantification over events for out-of-the-blue utterances.

b) A binary quantification over events creating definite descriptions for
contextualized utterances.

As I showed, all the presupposition and exhaustivity features of foci derive

directly from those logical form representations. Further, I have argued that those
semantic representations correspond to the syntactic patterns of each type of
construction: On the one hand, out-of-the-blue sentences have a simple left periphery
area involving no FocP projection; this provides a direct syntax-semantics mapping
for these sentences (i.e, the existential quantifier over events is merged in Finº with its
restriction). On the other hand, contextualized utterances will have a more complex
CP area involving FocP and TopP projections. As I argued, the richness of the CP
area accounts at the same time for the syntactic focus displacements and the
semantically presuppositional and exhaustive interpretation of contextualized

266
utterances. The semantics of these sentences is obtained when a binary quantifier like
δ reprojects to gain the nuclear scope in its extended local domain.
Finally, I have also shown that the DAFS provides an appealing analysis of
answers to multiple-Wh questions as derivationally split foci. The prediction of the
system I argued for is that we can, in principle, have several items marked [+F] that
do not merge together. Those numerations, I argued, derive into isolated F-Structures
that provide the patterns of the answers to multiple-Wh questions. These patterns are
unexplainable in alternative theories that assumed the existence of just one pitch
accent.
All in all, the gist of this dissertation has been to provide a way to capture the
essence of one of the core functional properties of natural languages (i.e., focus)
within the highly restricted formal framework of the Minimalist Program.
The corollary is that this epistemological move has proven not only plausible
but also highly clarifying both for the study of focus and for the reinforcement of the
explanatory adequacy of the “syntactocentric” architecture of grammar and the Bare
Phrase Structure theory.

267
REFERENCES

Altube, S., 1929, Erderismos, Bermeo: Euskaltzaindia.

Altube, S., 1920, De sintaxis euskérika, Donostia: Martin Mena y Cía.
Arnaudova, O., 2003, Focus and Bulgarian clause structure, PhD. Dissertation: University
of Ottawa.
Arnhold, A., 2007, “Intonation and focus in West Greenlandic”, paper presented at the
ICPhS Workshop on “Intonational phonology: Understudied or fieldwork languages”,
Saarbrücken, Germany.
Arregi, E. & J. Ormazabal, 2003, “Aditz ditrantsitiboen barne-egitura”, in IKER, 14: 119-
136.
Arregi, K., 2007, “Stress and islands in Northern Bizkaian Basque”, in J.A. Lakarra & J.I.
Hualde (eds.), Studies in historical and Basque linguistics dedicated to the memory of
R.L. Trask, Donostia: ASJU.
Arregi, K., 2003, Focus on Basque movements, PhD. Dissertation: MIT.
Artiagoitia, X., 2006, “Basque adjectives and the functional structure of the noun
phrase”, manuscript: UPV-EHU, URL: http://artxiker.ccsd.cnrs.fr/artxibo-
00087305/fr/
Artstein R., 2004, “Focus below the word level”, in Natural Language Semantics 12, 1-22.
Arvaniti, A., D.R. Ladd & I. Mennen, 2000, “What is a starred tone? Evidence from
Greek”, in M. Broe & J. Pierrehumbert (eds.), Papers in laboratory phonology V,
Cambridge: Cambridge University Press, 119-135.
Asher, N., 1993, Reference to abstract objects in discourse, Dordrecht: Kluwer.
Aske, J., 1997, Basque word order and disorder: principles, variation, and prospects, University
of California, Berkeley.
Atterer, M. & D.R. Ladd, 2004, “On the phonetics and phonology of “segmental
anchoring” of F0: Evidence from German”, in Journal of Phonetics 32, 177-197.
Azkue, R.M., 1923-5, Morfología Vasca [2nd edition, 1969, Bilbo: La Gran Enciclopedia
Vasca].
Azkue, R.M., 1891, Euskal-izkindea/Gramática vasca, Bilbo: José de Ayastuy.
Baker, M.C., 2003, “Agreement, dislocation, and partial configurationality”, in A. Carnie,
H. Harley & M. Willie (eds.), Formal approaches to function, Amsterdam: John
Benjamins, 107-134.
Baker, M. C., 1988, Incorporation: A theory of grammatical function changing: Chicago &
London: Chicago University Press.
Baltazani, M., 2002, Quantifier scope and the role of intonation in Greek, Ph. Dissertation:
UCLA.

268
Barwise, J. & J. Cooper, 1981, “Generalized quantifiers and natural language”, in
Linguistics and Philosophy 4, 159-219 [Reprinted in P. Portner & B.H. Partee, 2002,
Formal semantics, Oxford: Blackwell]
Beckman, M.E., 1996, “The parsing of prosody”, in Language and Cognitive Processes 11,
17-67.
Beckman, M.E., 1986, Stress and non-stress languages, Foris: Dordrecht.
Bickerton, D., 1993, “Subject focus and pronouns”, in F. Byrne & D. Winford (eds.), Focus
and grammatical relations in Creole languages, Amsterdam & Philadelphia: John
Benjamins, 189-212.
Biloa, E., 1995, Functional categories and the syntax of focus in Tuki, Müenchen &
Newcastle: Lincom Europa.
Boeckx, C., 2007, Bare syntax, Oxford: Oxford University Press (in press.).
Boeckx, C., 2006, Linguistic minimalism. Origins, concepts, methods, and aims, Oxford:
Oxford University Press.
Boeckx, C. & K. Grohmann, 2006, “Putting phases into perspective”, manuscript:
Harvard University & University of Cyprus.
Bolinger, D., 1978, “Asking more than one question at a time”, in H. Hïz (ed.), Questions,
Dordrecht: Reidel, 107-150.
Bolinger, D., 1972, “Accent is predictable (if you’re a mind-reader)”, in Language 48-3,
633-644.
Booij, G., 1983, “Principles and parameters in prosodic phonology”, in Linguistics 21,
249-280.
Botinis, A. (ed.), 2000, Intonation: analysis, modeling and technology, Dordrecht: Kluwer.
Borer, H., 2005, The normal course of events. Structuring sense, vol II. Oxford: Oxford
University Press.
Bošković, Ž., 2002, “On the interpretation of Multiple Questions”, in J. Bresnan et alii
(eds.), Celebration: An Electronic Festschrift in honor of Noam Chomsky’s 70th birthday:
MIT Press. URL: http://cognet.mit.edu/library/books/chomsky/celebration/
Bošković, Ž. 1999. “On multiple feature checking: Multiple wh-fronting and multiple
head-movement”, in S. Epstein & N. Hornstein (eds), Working Minimalism,
Cambridge: MIT Press, 159-187.
Bresnan, J., 1971, “Sentence stress and syntactic transformations”, in Language 47-2, 257-
281.
Brody, M., 1998, “Projection and phrase structure”, in Linguistic Inquiry 29, 367-398.
Brody, M., 1995, Lexico-logical form: a radically minimalist theory, Cambridge: MIT Press.
Bruce, G., 1977, Swedish Word Accent in Sentence Perspective, Travaux de l’Institut de
Linguistique de Lund 12, Gleerup: Lund.
Büring, D., 2003, “On D-Trees, Beans, and B-Accents”, in Linguistics & Philosophy 26:5,
511-545.

269
Büring, D., 1999, “Topic”, in P. Bosch & R. van der Sandt (eds.), Focus: linguistic,
cognitive, and computational perspectives, Cambridge (UK): Cambridge University
Press, 142-165.
Carlson, G.N., 1984, “Thematic roles and their role in semantic interpretation”,
Linguistics 22, 259-279.
Casati, R., & A. C. Varzi (eds.), 1996, Events, Dartmouth: Aldershot.
Castañeda, H-N., 1967, “Comments on D. Davidson's 'The logical form of action
sentences'”, in N. Rescher (ed.), The logic of decision and action, Pittsburgh:
Pittsburgh University Press, 104-112.
Chávez-Peón, M.E., 2006, “Loanword Phonology in San Lucas Quiaviní Zapotec”, in
Proceedings of NWLC 22, 10-19.
Cheng, L.L. & N. Corver, 2006, Wh-movement: moving on, Cambridge: MIT Press.
Chomsky, N., 2006, “Approaching UG from below”, manuscript: MIT.
Chomsky, N. 2005, “Three factors in language design,” in Linguistic Inquiry 36, 1-22.
Chomsky, N., 2004, “Beyond explanatory adequacy,” in A. Belletti (ed.), Structures and
beyond: the cartography of syntactic structures (vol. 3), Oxford: Oxford University
Press, 104-131.
Chomsky, N., 2001, “Derivation by phase,” in M. Kenstowicz (ed.), Ken Hale: A life in
language, Cambridge: MIT Press, 1-52.
Chomsky, N., 2000, “Minimalist inquiries: the framework”, in R. Martin, D. Michaels &
J. Uriagereka (eds.), Step by step: essays on minimalist syntax in honor of Howard
Lasnik, Cambridge: MIT Press, 89-155.
Chomsky, N., 1995a, “Bare phrase structure”, in G. Webelhuth (ed.), Government and
binding theory and the minimalist program, Oxford, Blackwell, 383-439.
Chomsky, N., 1995b, The minimalist program, Cambridge: MIT Press.
Chomsky, N., 1993, “A minimalist program for linguistic theory”, in K. Hale & S.J.
Keyser (eds.), The view from Building 20: Essays in linguistics in honor of Sylvain
Bromberger, Cambridge: MIT Press, 1-52 [Reprinted in Chomsky (1995b)].
Chomsky, N., 1986, Knowledge of language: its nature, its origins, its use, New York:
Praeger.
Chomsky, N., 1982, Some concepts and consequences of the theory of government and binding,
Cambridge: MIT Press.
Chomsky, N., 1981, Lectures on government and binding, Dordrecht: Foris.
Chomsky, N. 1976, “Conditions on rules of grammar”, in Linguistic Analysis 2-4, 303-351.
Chomsky, N., 1973, “Conditions on transformations”, in S. Anderson & P. Kiparsky
(eds.), A festschrift for Morris Halle: Holt Rinehart & Winston. [Reprinted in
Chomsky, 1977, Essays on form & interpretation, New York: Elsevier North
Holland, 81-162].

270
Chomsky, N., 1970, “Deep Structure, Surface Structure and Semantic Interpretation” in
R. Jakobson & S. Kawamoto (eds.), Studies in General and Oriental Linguistics,
Tokyo: TEC Corporation for Language Research, 183-216. [Reedited in D.
Steinberg & L. Jakobovits (eds.), 1971, Semantics in Generative Grammar, New
York: Cambridge University Press, 183-216]
Chomsky, N., 1965, Aspects of the theory of syntax, Cambridge: MIT Press.
Chomsky, N., 1957, Syntactic structures, Berlin: Mouton de Gruyter.
Chomsky, N., 1955, The Logical Structure of Linguistic Theory, manuscript: Harvard/MIT
[Published in 1975 as The Logical Structure of Linguistic Theory, New
York/London: Plenum Press].
Chomsky, N. & M. Halle, 1968, The sound pattern of English, Cambridge: MIT Press.
Chomsky, M. Halle & F. Lukoff, 1956, “On accent and juncture in English”, in M. Halle
et al. (eds.), For Roman Jacobson, The Hague: Mouton & Co, 65-80.
Chomsky, N. & H. Lasnik, 1977, “Filters and control”, in Linguistic Inquiry 8, 425-504.
Chomsky, N., & G. Miller, 1963, “Introduction to the formal analysis of natural
languages”, in R. D. Luce, R. B. Bush & E. Galanter (eds.) Handbook of mathematical
psychology. New York/London: John Wiley & Sons, 271-232.
Cinque, G., 1993, “A null theory of phrase and compound stress”, in Linguistic Inquiry
24, pp. 239-298.
Clark, A., 1997, Being there: putting brain, body and world together again, Cambridge: MIT
Press.
Collins, C., 2002, “Eliminating labels”, in S. D. Epstein & T. D. Seely (eds.), Derivation and
explanation in the minimalist program, Oxford: Blackwell, 42-64.
Comorovski, I., 1996, Interrogative Phrases & the Syntax-Semantics Interface, Dordrecht:
Kluwer.
Condoravdi, C., 1990, “Sandhi rules of Greek and prosodic theory”, in S. Inkelas & D.
Zec (eds.), The phonology-syntax connection, Chicago: The University of Chicago
Press, 63-85.
Contreras, H., 1987, “Small clauses in Spanish and English” in Natural Language and
Linguistic Theory 5, 225-244.
D’Imperio, M.P., 2002, “Italian intonation: An overview and some questions”, in Probus
14, 37-69.
Davidson, D.D., 2001, Essays on actions and events, 2nd ed., Oxford: Oxford University
Press.
Davidson, D.D., 1985, “Adverbs of action”, in B. Vermazen & M. Hintikka (eds.), Essays
on Davidson: actions and events, Oxford: Oxford University Press, 230-241
[Reprinted in Davidson (2001), 293-304].
Davidson, D.D., 1967a, “The logical form of action sentences”, in N. Rescher (ed.), The
logic of decision and action, Pittsburgh: Pittsburgh University Press, 81-95.

271
Davidson, D.D., 1967b, “Reply to Castañeda on agent and patient”, in Davidson (2001),
125-126.
Demirdache, H. & M. Uribe-Etxebarria, 2007, “The syntax of time arguments”, in Lingua
177, 330-366.
Demirdache, H. & M. Uribe-Etxebarria, 2000, “The primitives of temporal relations”, in
R. Martin, D. Michaels & J. Uriagereka (eds.), Step by step: Essays on minimalist
syntax in honor of Howard Lasnik, Cambridge: MIT Press, 157-187.
Derbyshire, D. C., 1985, Hyxkariana and linguistic typology, Dallas: SIL & University of
Texas at Arlington.
Diesing, M., 1992, Indefinites, Cambridge: MIT Press.
Dobashi, Y., 2003, Phonological phrasing and syntactic derivation, PhD. dissertation: Cornell
University.
Dominguez, L., 2004, Mapping focus: the syntax and prosody of focus in Spanish, PhD.
Dissertation: Boston University.
Downing, L., 2007, “Focus prosody divorced from stress and intonation in Chichewa,
Chitumbuka and Durban Zulu”, paper presented at the ICPhS Workshop on
“Intonational phonology: Understudied or fieldwork languages”, Saarbrücken,
Germany.
Dowty, D.R., 1989, “On the semantic content of the notion of thematic role”, in G.
Chierchia, B.H. Partee & R. Turner (eds.), Properties, types and meaning. Studies in
linguistic and philosophy, 2: Semantic issues, Dordrecht: Kluver, 69-129.
Dretske, F., 1972, “Contrastive Statements”, Philosophical Review 81, 411–37.
Eckardt, R. 1998, “Review of In the event of focus, by E. Herburger”, in Glot International 3,
9/10: 12-15.
Eguzkitza, A., 1986, Topics on the syntax of Basque and Romance, PhD. Dissertation, UCLA.
Elordieta, A., 2006, “Foku azentuak eta foku semantikoak bat egiten ez duten zenbait
kasuren azalpen baterantz”, in B. Fernández & I. Laka (eds.), Andolin gogoan:
Essays in honor of Professor Eguzkitza, Bilbo: UPV-EHU, 252-268.
Elordieta, A., 2004, “Prosodiaren eta egitura konfigurazionalaren arteko eraginaz”, in P.
Albizu & B. Fernández (eds.), Euskal gramatika XXI. mendearen atarian: Arazo
zaharrak, azterbide berriak, Gasteiz: UPV-EHU & Arabako Foru Aldundia, 39-60.
Elordieta, A., 2002, “ On the (im)possibility of prosodic focus marking in embedded
contexts in Northern Bizkaian Basque”, in X. Artiagoitia, P. Goenaga & J.A.
Lakarra (eds.), Erramu boneta: Festschrift for Rudolf P.G. de Rijk, supplements of
ASJU XLIV, Donostia: Gipuzkoako Foru Aldundia & UPV-EHU, 153-178.
Elordieta, A., 2001, Verb movement and constituent permutation in Basque, PhD.
Dissertation, HIL/Leiden University, LOT Dissertation Series.
Elordieta, G., 2007a, “Constraints on intonational prominence of focalized constituents”,
in C. Lee, M. Gordon & D. Büring (eds.), Topic and focus: Cross-linguistic
perspectives on meaning and intonation, Dordrecht: Springer, 1-22.

272
Elordieta, G., 2007b, “A constraint-based analysis of the intonational realization of focus
in Northern Bizkaian Basque”, to appear in T. Riad & C. Gussenhoven (eds.),
Tones and tunes: Volume I, Typological studies in word and sentence prosody, Berlin:
Mouton de Gruyter, 201-234.
Elordieta, G., 2003, “Intonation.”, in J.I. Hualde & J. Ortiz de Urbina (eds.), A grammar of
Basque, Berlin: Mouton de Gruyter, 72-112.
Elordieta, G., 1998, “Intonation in a pitch accent variety of Basque”, in Anuario del
Seminario de Filología Vasca «Julio de Urquijo» 32, 511-569.
Elordieta, G., 1997, “Accent, tone, and intonation in Lekeitio Basque” in F. Martínez-Gil
& A. Morales-Front (eds.), Issues in the Phonology and Morphology of the Major
Iberian Languages, Washington, D.C.: Georgetown University Press, 3-78.
Elordieta, G. & N. Calleja, 2005, “Microvariation in accentual alignment in Basque
Spanish”, Language and Speech, 48, 359-396.
Elordieta, G., S. Frota & M. Vigário, 2005, “Subjects, objects and intonational phrasing in
Spanish and Portuguese”, Studia Linguistica 59(2-3), 110-143.
Elordieta, G. & J.I. Hualde, 2003, “Tonal and durational correlates of accents in contexts
of downstep in Northern Bizkaian Basque”, Journal of the International Phonetic
Association 33: 195-209.
Elordieta, G. & J.I. Hualde, 2001, “The role of duration as a correlate of accent in Lekeitio
Basque”, Proceedings of Erospeech 2001 – Scandivavia, 105-108.
Epstein, S-D., E. Groat, R. Kawashima & H. Kitahara, 1998, A derivational approach to
syntactic relations, Oxford: Oxford University Press.
Epstein, S-D. & T. D. Seely, 2006, Derivations in Minimalism, Cambridge: Cambridge
University Press.
Epstein, S-D. & T. D. Seely, 2002, “Rule applications as cycles in a level-free syntax”, in
S-D. Epstein & T. D. Seely, Derivation and explanation in the minimalist program,
Oxford: Blackwell.
Erteschik-Shir, N., 1997, The dynamics of focus structure, Cambridge: Cambridge
University Press.
Etxepare, R., 1998, “A case for two types of focus in Basque”, in E. Benedicto, M. Romero
& S. Tomioka (eds.), UMOP 21: Proceedings of the workshop on focus, Amherst:
GLSA, 65-81.
Etxepare, R. & J. Ortiz de Urbina, 2003, “Focalization”, in J. I. Hualde & J. Ortiz de
Urbina (eds.), A grammar of Basque, Berlin: Mouton de Gruyter, 459-515.
Etxepare, R. & M. Uribe-Etxebarria, 2005, “Wh-phrases in situ in Spanish: scope and
locality”, Recherches linguistiques de Vincennes 33, 9-34.
Face, T.L., 2002, Intonational marking of contrastive focus in Madrid Spanish, Munich:
Lincom Europa.
Face, T.L. & P. Prieto, 2007, “Rising accents in Castillian Spanish: A revision of
Sp_ToBI”, in Journal of Portuguese Linguistics, 117-146.

273
Féry, C., 2001, “Focus and phrasing in French”, in C. Féry & W. Sternfeld (eds.), Audiatur
vox sapientiae. A festschift for Arnim von Stechow, Berlin: Akademie-Verlag, pp. 153-
181.
Fox, D. & D. Pesetsky, 2005, “Cyclic linearization of syntactic structure”, Theoretical
Linguistics 31, 1-46.
von Fintel, K., 2004, “Would you believe it? The King of France is back! (Presuppositions
and truth-value intuitions)”, in M. Reimer & A. Bezuidenhout (eds.), Descriptions
and beyond, Oxford: Oxford University Press, 315-341.
von Fintel, K., 1994, Restrictions on quantifier domains, PhD. Dissertation, University of
Massachusetts at Amherst.
Frank, R., 2002, Phrase structure composition and syntactic dependencies, Cambridge: MIT
Press.
Fortuny, J., 2007, The emergence of order in syntax, PhD. dissertation: Universitat de
Barcelona.
Frota, S., 2000, Prosody and focus in European Portuguese, New York: Garland.
Fujisaki, H., 1983, “Dynamic characteristics of voice fundamental frequency in Speech
and singing”, in P.F. MacNeilage (ed.), The production of speech,
Gallego, A.J., 2007, Phase theory and parametric variation, PhD. dissertation: Universitat
Autònoma de Barcelona.
Georgiafentis, M., 2004, Focus and word order variation in Greek, PhD. dissertation:
Reading University.
Geurts, B. & R. van der Sandt, 2004, “Interpreting Focus”, in Theoretical Linguistics 30, 1-
44.
Godjevać, S., 2000, Intonation, word order, and focus projection in Serbo-Croatian, Ph.D
Dissertation: The Ohio State University.
Goenaga, P., 1978, Gramatika bideetan, Donostia: Erein.
Goldberg, A., 1999, “The Emergence of the semantics of argument structure
constructions”, in B. MacWhinney (ed.), Emergence of language, Hillsdale:
Lawrence Earlbaum Associates. 197-212.
Gómez Txurruka, I., 2002, Foco y tema: Una aproximación discursiva, Bilbao: EHU-UPV.
Grebenyova, L., 2006, Multiple interrogatives: Syntax, semantics and learnability, PhD.
Dissertation: University of Maryland.
Green, M. & P. J. Jaggar, 2003, “Ex-situ and in-situ focus in Hausa: syntax, semantics and
discourse”, in J. Lecarme (ed.), Research in Afroasiatic grammar II, Amsterdam:
John Benjamins, 187-213.
Grice, P., 1975, “Logic and conversation”, in P. Cole & J.L. Morgan, (eds.), Syntax and
Semantics, Vol. 3, Speech Acts, New York, Academic Press, 41-58. [Reprinted in
Grice, 1989, Studies in the Way of Words, Cambridge: Harvard University Press]

274
Grohmann, K.K., 2000, Prolific pripheries: a radical view from the left, PhD. Dissertation: U.
Maryland.
Gussenhoven, C., 2006, “Yucatec Maya tones in sentence perspective”, poster presented
at LabPhon 10, Paris.
Gussenhoven, C., 2004, The phonology of tone and intonation, Cambridge: Cambridge
University Press.
Gussenhoven, C., 2002, “Phonology of intonation”, in Glot International 6, 9/10, 271-284.
Gussenhoven, C., 1984, On the grammar and semantics of sentence accents, Foris: Dordrecht.
Gutiérrez-Rexach, J., 1999, “Interrogatives and Polyadic Quantification”, in N. Scott
(ed.), Proceedings of the International Conference in Questions, Liverpool: University
of Liverpool, 1-14 [Reprinted in J. Gutiérrez-Rexach (ed.), 2003, Semantics: Critical
Concepts vol. V, London: Routledge, 418-433].
Haddican, W., to appear, “On egin: do-support in Basque”, in Natural Language and
Linguistic Theory.
Haddican, W., 2005, “On egin: do-support and Verb-focalization in Central and Western
Basque” in Penn Working Papers in Linguistics: Proceedings of PLC29.
Hagstrom, P.A., 1998, Decomposing questions, PhD. dissertation: MIT.
Hale, K., E. Jelinek & M.A. Willie, 2003, “Topic and focus scope positions in Navajo”, in
S. Karimi (ed.), Word order and scrambling, Oxford: Blackwell, 1-21.
Hale, K. & S. J. Keyser, 2002, Prolegomenon to a theory of argument structure, Cambridge:
MIT Press.
Halle, M. & J. R. Vergnaud, 1987, An essay on stress, Cambridge: MIT Press.
Hamblin, C.L., 1973, “Questions in Montague English”, Foundations of Language 10, 41–
53.
Haspelmath, M., M.S. Dryer, D. Gil & B. Comrie, 2005, The world atlas of language
structures, Oxford: Oxford University Press.
Hauser, M.D., N. Chomsky & T. Fitch, 2002, “The faculty of language: what is it, who
has it, and how did it evolve?”, in Science 298: 1569-1579.
Hayes, B., 1989, “The prosodic hierarchy in meter”, in P. Kiparsky & G. Youmans (eds.),
Rhythm and meter, Academic Press: Orlando, 201-260.
Hayes, B. & A. Lahiri, 1991, “Bengali intonational phonology”, in Natural Language and
Linguistic Theory 9, 47-96.
Hegarty, M., 1992, Adjunct extraction and chain configurations, PhD. Dissertation: MIT.
Heim, I.R., 1982, The Semantics of definite and indefinite noun phrases, PhD. Dissertation:
MIT.
Hellmuth, S., 2007, “The (absence of) prosodic reflexes of given/new information status
in Egyptian Arabic”, manuscript: UNIVERSITY OF Potsdam.
Herburger, E., 2000, What counts: focus and quantification, Cambridge (Mass.): MIT Press.

275
Herburger, E., 1998, “Presupposition vs. assertion: a Davidsonian account”, in in E.
Benedicto, M. Romero & S. Tomioka (eds.), UMOP 21: Proceedings of the workshop
on focus, Amherst: GLSA, 125-141.
Hernández, M., 1990, La semántica de Davidson: una introducción crítica, Madrid: Visor.
von Heusinger, K., 1999, “Intonation and information structure”, manuscript: University
of Konstanz.
Higginbotham, J., 2002, “Why is sequence of tense obligatory?”, in G. Preyer & G. Peter
(eds.), Logical form and language, Oxford: Oxford University Press, 207-227.
Higginbotham, J., 2000, “On events in linguistic semantics”, in J. Higginbotham, F.
Pianesi & A. Varzi (eds.), Speaking of events, Oxford: Oxford University Press, 49-
79.
Higginbotham, J., 1986, “Linguistic Theory and Davidson's Program in Semantics”, in
Lepore (ed.), Truth and interpretation: Perspectives on the philosophy of Donald
Davidson, Oxford: Blackwell, 29-48.
Higginbotham, J., 1985, “On semantics”, Linguistic Inquiry 16, 547-593. [Reprinted in
Lepore (ed.), 1987, New directions in semantics, London: Academic Press, 1-54.
Higginbotham, J., 1983, “The logic of perceptual reports: an extensional alternative to
situation semantics”, in Journal of Philosophy 80-2, 100-127.
Higginbotham, J. & R. May, 1981, “Questions, quantifiers and crossing”, in The Linguistic
Review 1, 41-80.
Hinzen, W., 2006a, Mind design and minimal syntax, Oxford: Oxford University Press.
Hinzen, W., 2006b, “The successor function + Lexicon = Human language?”,
manuscript: Universiteit van Amsterdam & University of Durham.
Hirschberg, J. & G. Ward, 1992, “The influence of pitch range, duration, amplitude and
spectral features on the interpretation of the rise-fall-rise intonation contour in
English”, in Journal of Phonetics 20: 241-251.
Hirschberg, J., & C. Avesani, 2000, “Prosodic disambiguation in English and Italian”, in
A. Botinis (ed.), Intonation, Dordrecht: Kluwer Academic Publishers, 87-95.
Holmer, N.M., 1991, El idioma vasco hablado. Un estudio de dialectología vasca, Supplements
of ASJU V, Donostia & Bilbao: Gipuzkoako Foru Aldundia eta EHU-UPV.
van Hoof, H., 2003, “The rise in the rise-fall contour: does it evoke a contrastive topic or
a contrastive focus?”, in Linguistics 41-3, 515-563.
Horn, L.R., 2004, “Implicature”, in L.R. Horn & G. Ward (eds.), The Handbook of
Pragmatics, Oxford: Blackwell, 3-28.
Hornstein, N., 2002, “A grammatical argument for a Neo-Davidsonian semantics”, in G.
Preyer & G. Peters (eds.), 345-264.
Hornstein, N., 1995, Logical form: from GB to minimalism, Oxford: Blackwell.
Hornstein N., H. Lasnik & J. Uriagereka, 2007, “The Dynamics of Islands: Speculations
on the Locality of Movement”, Linguistic Analysis 33, 1-2, 149-175.

276
Hornstein, N., J. Nunes & K. Grohmann, 2005, Undestanding minimalism, Cambridge:
Cambridge University Press.
Hornstein, N. & J. Uriagereka, 1999, “Labels and Projections: A Note on the Syntax of
Quantifiers”, University of Maryland Working Papers in Linguistics 8, 249-270.
Hornstein, N. & J. Uriagereka, 2002, “Reprojections”, in S. Epstein & T. D. Seely (eds.),
Derivation and explanation in the minimalist program, Blackwell, 107-132.
Hualde, J.I, 2007, “Remarks on Word-Prosodic Typology”, paper presented at the
Berkeley Linguistics Society 2006.
Hualde, J.I., 2003, “El modelo métrico y autosegmental”, in P. Prieto (ed.), Teorías de la
entonación, Barcelona: Ariel, 155-184.
Hualde, J.I, 1999, “Basque Accentuation”, in van der Hulst (ed.), Word Prosodic Systems in
the Languages of Europe, Berlin eta New York: Mouton de Gruyter, 947-994.
Hualde, J.I, 1997, Euskararen azentuerak, Donostia: Gipuzkoako Foru Aldundia & UPV-
EHU..
Hualde, J.I. & G. Elordieta, 2003, “Tonal and durational correlates of accent in contexts
of downstep in Lekeitio Basque”, in Journal of the International Phonetic Association
33, 195-209.
Hualde, J.I., G. Elordieta & A. Elordieta, 1994, The Basque dialect of Lekeitio, Donostia:
Gipuzkoako Foru Aldundia & UPV-EHU.
Hualde, J.I., G. Elordieta, I. Gaminde & R. Smiljanic, 2002, “From pitch-accent to stress-
accent in Basque”. in C. Gussenhoven & N. Warner (eds.), Laboratory Phonology 7,
Berlin: Mouton de Gruyter, 547-584.
Hualde, J.I. & J. Ortiz de Urbina, 2003, A grammar of Basque, Berlin: Mouton de Gruyter.
Hurford, J., 2003, ``The neural basis of predicate-argument structure”, Behavioral and
Brain Sciences, 26(3): 261-283.
Hurford, J. R., 2001, “Protothought had no logical names”, in J. Trabant & S. Ward (eds.),
New Essays on the Origin of Language, Berlin & NY: Mouton de Gruyter, 119-132.
Hyman, L.M., 1999, “The interaction between focus and tone in Bantu”, in Rebuschi &
Tuller (eds.), 1999, 151-178.
Hyman, L.M., 1977, “On the nature of linguistic stress”, in L. Hyman (ed.), Studies on
stress and accent, Southern California Occasional Papers in Linguistics, 37-83.
Irurtzun, A., 2007, “Propositions and pro positions”, manuscript: EHU-UPV & IKER.
Irurtzun, A., 2006a, “Fonetikak Bideratutako Aldaketa Fonologikoak”, in C. Isasi & S.
Gómez (eds.), Oihenart 21: Lingüística Vasco-Románica. Euskal-Erromantze
Linguistika. Donostia: Eusko Ikaskuntza, 195-208.
Irurtzun, A., 2006b, “Galdegaia bere bakardadean”, in Lapurdum X, 95-125.
Irurtzun, A., 2003, “Stress and pitch: Errenteriako azentu eta intonazioa”, manuscript:
EHU-U. Basque Country.
Ishihara, S., 2003, Intonation and interface conditions, PhD. dissertation: MIT.

277
Ishihara, S., 2000, “Stress, focus, and scrambling in Japanese”, in MIT Working Papers in
Linguistics 39; A View from Building E-39, 142-175.
Jackendoff, R., 1972, Semantic interpretation in generative grammar, Cambridge: MIT Press.
Jackendoff, R., 1997, The architecture of the language faculty, Cambridge: MIT Press.
Jackendoff, R., 2002, Foundations of language: Brain, meaning, grammar, evolution, Oxford:
Oxford University Press.
Jackendoff, R., 2007, Language, consciousness, culture: Essays on mental structure,
Cambridge: MIT Press.
Jannedy, S. & I. Fiedler, 2007, “Qualitative exploration of Ewe focus prosody”, paper
presented at the ICPhS Workshop on “Intonational Phonology: Understudied or
Fieldwork Languages”, Saarbrücken, Germany.
Jeong, Y., 2007, “Multiple Wh-fronting in Basque”, in A. Conroy, C. Jing., C. Nagao & E.
Takahashi (eds.), University of Maryland Working Papers in Linguistics 15, 98-142.
Jeong, Y., 2003, “Multiple Wh-Fronting in Basque”, manuscript, University of Maryland.
Johnston, M., 1994, The syntax and semantics of adverbial adjunct, PhD. dissertation: UCSC.
Jun, S-A., 1993, The phonetics and phonology of Korean prosody, PhD. dissertation: The Ohio
State University.
Jun, S-A., & G. Elordieta, 1997, “Intonational structure of Lekeitio Basque”, in A.
Botinis, G. Kouroupetroglou & G. Carayiannis (eds.) Proceedings of an ESCA
Workshop on Intonation: Theory, Models and Applications, 193-196.
Kadmon, N., 2001, Formal pragmatics, Oxford: Blackwell.
Kahnemuypour, A., 2005, The syntax of sentential stress, PhD. dissertation: University of
Toronto.
Kamp, H., 1981, “A theory of truth and semantic interpretation”, in J. Groenendijk et al.
(ed.), Formal methods in the study of language, Amsterdam: Mathematical Center
Tracts, 277-322.
Kamp, H. & U. Reyle, 1993, From discourse to logic, Dordrecht: Kluwer.
Kang, S. & S. Speer, 2003, “Effects of prosodic phrasings on syntactic disambiguation”,
paper presented at the 26th GLOW Colloquium, Lund (Sweden).
Kannerva, J. M., 1990, “Focusing on Phonological Phrases in Chicheŵa”, in S. Inkelas &
D. Zec (eds.), The phonology-syntax connection, Chicago & London: University of
Chicago Press, 145-162.
Karttunen, L., 1977, “Syntax and semantics of questions”, Linguistics and Philosophy 1, 3-
44.
Karttunen, L., 1974, “Presuppositions and linguistic context”, in Theoretical Linguistics 1,
181-194.
Kawamura, T., 2006, “NPI connectivity as a focal phenomenon”, manuscript: Stony
Brook University.

278
Kearns, K., 2000, Semantics, London: Macmillan.
Kim, J-S., 1997, Syntactic focus movement and ellipsis: a minimalist approach, PhD.
dissertation, University of Connecticut.
King, T. H., 1995, Configuring topic and focus in Russian, Stanford: CSLI.
Kiss, K. É., 1998, “Identificational focus versus information focus”, in Language 74-2, 245-
273.
Kiss, K.É., 1995, “Introduction”, in K. É. Kiss (ed.), Discourse configurational languages,
New York (NY): Oxford University Press, 3-27.
Koopman, H., & D. Sportiche, 1991, “The position of subjects”, in Lingua 85, 211-258.
Kratzer, A., 1996, “Severing the external argument from its verb”, in J. Rooryck & L.
Zaring (eds.): Phrase structure and the lexicon, Dordrecht: Kluwer, 109-137.
Kratzer, A. 1995, “Stage-level and individual-level predicates as inherent generics”, in G.
N. Carlson & F. J. Pelletier (eds.), The generic book, Chicago: University of Chicago
Press, 125-75.
Kratzer, A. & E. Selkirk, 2007, “Phase theory and prosodic spellout: the case of verbs”, in
The Linguistic Review 24, 93-135.
Krifka, M. (ed.), 2004, Theoretical Linguistics 30.
Krifka, M., 1999, “At least some determiners aren’t determiners”, in K. Turner (ed.), The
Semantics/Pragmatics Interface from Different Points of View, Oxford: Elsevier, 257-
291.
Krifka, M., 1991, “A compositional semantics for multiple focus constructions”, in S.
Moore & A.Z. Wyner (eds). Proceedings of SALT 1, 17-53.
Kuno, S. & J. J. Robinson, 1972, “Multiple Wh Questions”, in Linguistic Inquiry 3-4, pp.
463-487.
Kuno, S., 1973, The Structure of the Japanese Language, Cambridge (Mass.): MIT Press.
Kuroda, S-Y., 2003, “Indeterminates and Islands”, class lectures: EHU-U. Basque
Country.
Kuroda, S.-Y., 1988, “Whether we agree or not: a comparative syntax of English and
Japanese”, in Linguisticae Investigationes 12, 1-47.
Kuroda, S-Y., 1965, Generative grammatical studies in the Japanese language, PhD.
Dissertation: MIT.
Ladd, D. R., 1996, Intonational phonology, Cambridge: Cambridge University Press.
Ladd, D.R., & A. Schepman, 2003, “Sagging transitions" between high accent peaks in
English: experimental evidence”, Journal of Phonetics 31: 81-112.
Lahiri, U., 2002, Questions and answers in embedded contexts, Oxford: Oxford University
Press.
Laka, I., 1993, “Negation in syntax: The view from Basque”, in Rivista de Linguistica 5.2,
245-273.

279
Laka, I., 1990, Negation in syntax: On the nature of functional categories and projections, PhD.
Dissertation: MIT.
Laka, I. & J. Uriagereka, 1987, “Barriers for Basque and vice-versa”, in J. McDonough &
B. Phunkett (eds.), Proceedings of NELS 17, Amherst: GLSA, 394-408.
Lambova, M. 2001. “On A-bar movements in Bulgarian and their interaction”, The
Linguistic Review18: 327-374.
Lambrecht, K., 1994, Information structure and sentence form, Cambridge: Cambridge
University Press.
Larson, R.K., 1988, “On the double object construction”, in Linguistic Inquiry 19-3, 335-
391.
Larson, R.K., 1990, “Double object constructions: reply to Jackendoff”, in Linguistic
Inquiry 21-4, 589-632.
Larson, R.K., 1991, “The projection of DP (and DegP)”, manuscript: Stony Brook
University.
Larson, R.K. & G. Segal, 1995, Knowledge of meaning: An introduction to semantic theory,
Cambridge & London: MIT Press.
Lasnik, H., 1999, “Chains of arguments”, in S.D. Epstein & N. Hornstein (eds.), Working
minimalism, Cambridge: MIT Press, 189-216.
Lasnik, H. & J. Kuppin, 1977, “A restrictive theory of transformational grammar”, in
Theoretical Linguistics 4: 173-196.
Lasnik, H., J. Uriagereka & C. Boeckx, 2005, A course in minimalist syntax: foundations and
prospects, Oxford: Blackwell.
Lecarme, J., 1999, “Focus in Somali”, in Rebuschi & Tuller (eds.), The Grammar of Focus,
Amsterdam: John Benjamins, 275-310.
Legate, J., 2003, “Some Interface Properties of the Phase”, in Linguistic Inquiry 34-3, 506-
516.
Lepore, E. & K. Ludwig, 2007, Donald Davidson’s truth-theoretic semantics, Oxford: Oxford
University Press.
Lepore, E. & K. Ludwig, 2005, Donald Davidson: meaning, truth, language, and reality,
Oxford: Oxford University Press.
Liberman, M. Y., 1975, The intonational system of English, PhD. Dissertation: MIT.
Linebarger, M., 1987, “Negative polarity and grammatical representation”, in Linguistic
and Philosophy 10, 325-287.
Longa, V. M., 2006, “Dependencia de la estructura y complejidad computacional”, in B.
Fernández & I. Laka (eds.), Andolin gogoan: Essays in Honor of Professor Eguzkitza,
Bilbao: EHU-UPV, 633-649.
Lombard, L.B., 1998, “Ontologies of Events”, in S. Laurence and C. Macdonald (eds.),
Contemporary Readings in the Foundations of Metaphysics, Oxford: Blackwell, 277-
294.

280
López, L., 2001, “Head of a projection”, in Linguistic Inquiry 32-2, 521-532.
Lorenzo, G., 2006a, El vacío sexual, la tautología natural y la promesa minimalista, Boadilla
del Monte (Madrid): Antonio Machado Libros.
Lorenzo, G., 2006b, “Some minimalist thoughts on the emergence of recursion”, in J.
Rosselló & J. Martín (eds.), The biolinguistic turn: Issues on language and biology,
PPU: Barcelona, 119-128.
Lupyan, G., 2006. “Labels facilitate learning of novel categories”, in A. Cangelosi,
A.D.M. Smith & K.R. Smith (eds.) The evolution of language: Proceedings of the 6th
international conference. Singapore: World Scientific, 190-197.
Marantz, A., 1995, “The minimalist program”, in G. Webelhuth (ed.), Government and
binding theory and the minimalist program, Oxford: Blackwell, 349-382.
Marušič, F., 2005, On non-simultaneous phases, PhD. dissertation: Stony Brook University.
Martin, R. & J. Uriagereka, 2000, “Some possible foundations of the minimalist
program”, in R. Martin, D. Michaels & H. Lasnik (eds.), Step by step: essays on
minimalist syntax in honor of Howard Lasnik, Cambridge: MIT Press, 1-30.
Maruyama, A., 2003, “Japanese wa in conversational discourse: A Contrast Marker” in
Studies in Language 27-2, pp. 245-285.
Marvin, T., 2003, Topics in the stress and syntax of words, PhD. Dissertation: MIT.
May, R., 1985, Logical form: its structure and derivation, Cambridge: MIT Press.
McCarthy, J. & A. Prince, 1993, “Generalized alignment”, in Yearbook of morphology 1993,
79–153.
Merchant, J., 2004, “Fragments and ellipsis”, Linguistics and Philosophy 27, 661-738.
Miller, P.H., G.K. Pullum & A.M. Zwicky, 1997, “The principle of phonology-free
syntax: four apparent counterexamples in French”, in Journal of Linguistics 33, 67-
90.
Mitxelena, K., 1987, Palabras y textos, Leioa: EHU-UPV.
Mitxelena, K. 1985, Fonética Histórica Vasca, 3rd ed., Supplements of ASJU IV, Donostia &
Bilbo: Gipuzkoako Foru Aldundia & EHU-UPV.
Mitxelena, K., 1981, “Galdegaia eta mintzagaia euskaraz”, in Euskal linguistika eta
literatura: Bide berriak, Bilbo: Deustuko Unibertsitatea, 57-81 [Reprinted in J. A.
Lakarra & B. Urgell (eds.), 1988, Mitxelenaren euskal idazlan guztiak VI, Zarautz:
Euskal Editoreen Elkartea, 137-167].
Mitxelena, K., 1976, “Acentuación alto-navarra”, in Fontes Linguae Vasconum, 23, 147-162.
[Reprinted in Mitxelena 1987: 245-260]
Mitxelena, K., 1972, “A note on old Labourdin accentuation”, in Anuario del Seminario de
Filología Vasca “Julio de Urquijo” 6, 110-120. [Reprinted in Mitxelena 1987: 235-244]
Molnár, V. & S. Winkler (eds.), 2006, The architecture of focus, Berlin: Mouton de Gruyter.

281
Munakata, T., 2002, “Contrastive-Topic wa as Focus Interpretation Operator”, in Y.
Endo, R. Martin & H. Yamashita (eds.), Working papers in Biolinguistics 1: Papers on
Syntax and Semantics, Yokohama: BAY, pp. 21-38.
Neeleman, A., & T. Reinhart, 1998, “Scrambling and the PF interface”, in M. Butt & W.
Geuder (eds.), The projection of arguments; Lexical and compositional factors,
Stanford: CSLI Publications, 309-353.
Nespor, M. & I. Vogel, 1986, Prosodic phonology, Foris: Dordrecht.
Ortiz de Urbina, J., 2002, “Focus of correction and remnant movement in Basque”, in J.
A. Lakarra & X. Artiagoitia (eds.), Erramu boneta: A festschrift for Rudolph P.G. De
Rijk, Supplements of ASJU XLIV, Donostia: Gipuzkoako Foru Aldundia & UPV-
EHU, 511-524.
Ortiz de Urbina, J., 2001, “Ezezko perpausetako fokoak (galdegaiak eta ukagaiak)”, in
IKER 14, Bilbo: Euskaltzaindia, 343-356.
Ortiz de Urbina, J., 1999, “Focus in Basque”, in Rebuschi & Tuller (eds.), 1999, 311-334.
Ortiz de Urbina, J., 1995, “Residual verb first and verb second in Basque”, in K. É. Kiss
(ed.), Discourse configurational languages, New York (NY): Oxford University Press,
99-121.
Ortiz de Urbina, J., 1993, “Feature percolation and clausal pied-piping”, in J. I. Hualde &
J. Ortiz de Urbina (eds.), Generative studies in Basque linguistics, Amsterdam: John
Benjamins.
Ortiz de Urbina, J., 1989, Some parameters in the grammar of Basque, Dordrecht: Foris.
Ortiz de Urbina, J., 1988, “Clausal pied-piping”, in Estudios de lengua y literatura, Bilbo:
Deustuko Unibertsitatea, 5-112.
Ortiz de Urbina, J., 1986, Parameters in the grammar of Basque, PhD. Dissertation:
University of Illinois at Urbana Champaign.
Ortiz de Urbina, J., 1983, “Empty categories and focus in Basque”, in Studies in the
Linguistic Sciences, 13-1.
Osa, E, 1990, Euskararen hitzordena komunikazio zereginaren arauera, Leioa: UPV-EHU.
Parafita, M.C., 2005, Focus at the interface, PhD. Dissertation: University of Kansas.
Parsons, T., 1990, Events in the semantics of English: a study of subatomic semantics,
Cambridge: MIT Press.
Peng, S., M.K.M. Chan, Ch. Tseng, Ts. Huang. O.J. Lee & M. Beckman, 2005, “Towards a
pan-mandarin system of prosodic transcription”, in S. Jun (ed.), Prosodic typology:
The phonology of intonation and phrasing, Oxford: Oxford University Press, 230-270.
Pesetsky, D., 1982, Paths and categories, PhD. Dissertation: MIT.
Pianesi, F. & A. Varzi, 2000, “Events and event talk: an introduction” in F. Pianesi & A.
Varzi (eds.), Speaking of events, Oxford: Oxford University Press, 3-49.
Pierrehumbert, J.B., 1980, The phonology and phonetics of English intonation, PhD.
Dissertation: MIT.

282
Pierrehumbert, J.B. & J. Hirschberg, 1990, “The meaning of intonational contours in the
interpretation of discourse”, in P. Cohen, J. Morgan & M. Pollack (eds.), Intentions
in communication, Cambridge: MIT Press, pp. 271–311.
Pierrehumbert, J.B. & M.E. Beckman, 1988, Japanese tone structure, Cambridge: MIT Press.
Pierrehumbert, J.B., M.E. Beckman & R. Ladd, 2000, “Conceptual foundations of
phonology as a laboratory science”, in N. Burton-Roberts, P. Carr & G. Docherty
(eds.), Phonological Knowledge: Oxford University Press, 273-304.
Pietroski, P.M., 2005, Events and semantic architecture, Oxford: Oxford University Press.
Pietroski, P.M., 2003a, “Small verbs, complex events: analyticity without synonymy”, in
Antony & Hornstein (eds.), Chomsky and his critics, Malden, Oxford & Melbourne:
Blackwell Publishers, 179-214.
Pietroski, P.M., 2003b, “Semantics and metaphysics of events”, in K. Ludwig (ed.),
Donald Davidson, Cambridge: Cambridge University Press, 137-162.
Pietroski, P.M., 2002, “Function and concatenation”, in G. Preyer & G. Peters (eds.), 91-
117.
Pietroski, P.M., 1999, Causing actions, Oxford: Oxford University Press.
Pietroski, P.M. & J. Uriagereka, 2002, “Dimensions of natural language”, in J.
Uriagereka, Derivations: exploring the dynamics of syntax, London: Routledge, 266-
287.
Platzack, Ch., 2001, “Multiple Interfaces”, in U. Nikkane & E. van der Zee (eds.),
Cognitive interfaces: constraints on linking cognitive information, Oxford: Oxford
University Press, 21-53.
Portner, P. & K. Yabushita, 1998, “The Semantics and Pragmatics of Topic Phrases”, in
Linguistics & Philosophy 21, pp. 117-157.
Potts, C., 2007, “The dimensions of quotation”, in C. Barker & P. Jacobson (eds.), Direct
compositionality, Oxford: Oxford University Press, 405-431.
Prieto, P., 2003a (ed.), Teorías de la entonación, Barcelona: Ariel.
Prieto, P., 2003b, “Teorías lingüísticas de la entonación”, in Prieto (2003a), 13-33.
Preyer, G. & G. Peter (eds.), 2002, Logical form and language, Oxford: Oxford University
Press.
Quine, W.V.O, 1963, Set theory and its logic, Cambridge: Harvard University Press.
Ramchand, G. C., 2006, Verb meaning and the lexicon: a first phase syntax, manuscript:
University of Tromsø.
Rebuschi, G., 1983, “A note on focalization in Basque”, in ASJU-International Journal of
Basque Language and Linguistics 4-2, 29-42
Rebuschi, G. & L. Tuller (eds.), 1999, The grammar of focus, Amsterdam & Philadelphia:
John Benjamins.
Reglero, L., 2004, A’-Dependencies in Spanish and Basque, PhD. Dissertation, University of
Connecticut.

283
Reglero, L., 2003 “Non-wh-fronting in Basque”, in C. Boeckx & K. Grohmann (eds),
Multiple Wh Fronting, Amsterdam: Benjamins 187-227.
Reglero, L., 2002, “Multiple questions in Basque”, manuscript: UConn.
Reinhart, T., 1995, “Interface strategies”, OTS Working Papers in Linguistic, TL-95-002,
Utrecht: Utrecht University.
Reinhart, T., 2006, Interface strategies: Optimal and costly computations, Cambridge: MIT
Press.
Revithiadou, A. & V. Spyropoulos, 2003, “Trapped within a phrase”: effects of syntactic
derivation on prosodic phrasing”, in A. Mettouchi & G. Ferré (eds.), Interfaces
prosodiques – Prosodic interfaces, Nantes: AAI, 167-172.
Rialland, A. & S. Robert, 2001, “The intonational system of Wolof”, in Linguistics 39-5,
pp. 893-939.
de Rijk, R. P. G., 1996, “Focus and quasifocus in Basque negative statements”, in Revista
Internacional de Estudios Vascos 41-1, 63-76. [Reprinted in R. de Rijk, 1998, De
Lingua Vasconum: Selected Writings, EHU & Gipuzkoako Foru Aldundia: Bilbo &
Donostia, 421-433].
de Rijk, R. P. G., 1969, “Is Basque an SOV Language?”, in Fontes Linguae Vasconum, 1,
319-351. [Reprinted in R. de Rijk, 1998, De Lingua Vasconum: Selected Writings,
EHU & Gipuzkoako Foru Aldundia: Bilbo & Donostia, 13-38].
Rizzi, L., 1997, “The fine structure of the left periphery”, in L. Haegeman (ed.), Elements
of grammar, Dordrecht: Kluwer, 281-337.
Roberts, C., 1996, “Information Structure: Towards an integrated formal theory of
pragmatics”, in J. Hak Yoon & A. Kathol (eds.) OSUWPL Volume 49: Papers in
Semantics, 1996, The Ohio State University Department of Linguistics.
Rochemont, M.S., 1998, “Phonological focus and structural focus”, in P. Culicover & L.
McNally (eds.), Syntax and semantics: The limits of syntax, San Diego (CA):
Academic Press, pp. 337-363.
Rooth, M., 1999, “Association with focus or association with presupposition?”, in P.
Bosch & R. Van der Sandt (arg.), Focus: linguistic, cognitive and computational
perspectives, Cambridge: Cambridge University Press, 232-244.
Rooth, M., 1985, Association with focus, PhD. Dissertation: UMass.
Rosenbaum, H., 1977, “Zapotec gapping as counterevidence to some universal
proposals”, in Linguistic Inquiry 8-2, 379-395.
Saltarelli, M., M. Azkarate, D. Farwell, J. Ortiz de Urbina & L. Oñederra, 1988, Basque,
London & New Cork.
Sandalo, F. & H. Truckenbrodt, 2002, “Some notes on phonological phrasing in Brazilian
Portuguese”, in A. Csirmaz et alii (eds.), Phonological answers (and their
corresponding questions): MIT Working Papers in Linguistics 42, pp. 285-310.
Schafer, A., Sh. Speer & P. Warren, 2003, “Intonation in sentence processing”, in
Proceedings of 15th ICPhS, Universitat Autónoma de Barcelona: Barcelona.

284
Schein, B., 2002, “Events and the semantic content of thematic relations”, in G. Preyer &
G. Peter (eds), 263-344.
Schein, B., 1993, Plurals and events, Cambridge: MIT Press.
Schmerling, S., 1976, Aspects of English sentence stress, Austin: Austin University Press.
Schwarzschild, R., 1999, “GIVENness, AVOIDF, and other constraints on the placement of
accent”, in Natural Language Semantics 7, 141-177.
Seely, T. D., 2006, “Merge, derivational c-command, and subcategorization in a label-
free syntax”, in C. Boeckx (ed.), Minimalist essays, Amsterdam: John Benjamins,
182-221.
Seidl, A., 2001, Minimal indirect reference: A theory of the syntax-phonology interface. New
York: Routledge.
Selkirk, E., 2006, “Bengali intonation revisited: An optimality theoretic analysis in which
FOCUS stress prominence drives FOCUS phrasing”, in Ch. Lee, M. Gordon & D.
Büring (eds.), Topic and focus: Intonation and meaning, Dordrecht: Kluwer
Academic Press, 215-244.
Selkirk, E., 2002, “Contrastive FOCUS vs. presentational focus: Prosodic evidence from
right node raising in English”, in B. Bel & I. Marlin (eds.), Speech prosody 2002:
Proceedings of the 1st international conference on speech prosody, Aix-en-Provence,
643-646.
Selkirk, E., 2000, “The interaction of constraints on prosodic phrasing”, in M. Horne
(ed.), Prosody: Theory and experiment: Studies presented to Gösta Bruce, Dordrecht:
Kluwer, 231-262.
Selkirk, E., 1995a, “Sentence prosody: Intonation, stress, and phrasing”, in J. A.
Goldsmith (ed.), The handbook of phonological theory, Oxford: Blackwell, 550-569.
Selkirk, E., 1995b, “The prosodic structure of function words”, in J. Beckman, L. W.
Dickey, & S. Urbanczyk (eds.), Papers in optimality theory, Amherst: GLSA
Publications, 439-70 [Reprinted in J. McCarthy (ed.), 2004, Optimality theory in
phonology – A reader, Malden & Oxford: Blackwell, 464-482.
Selkirk, E., 1986, “On derived domains in sentence phonology”, in Phonology Yearbook 3,
371-405.
Selkirk, E., 1984, Phonology and syntax, Cambridge: MIT Press.
Selkirk, E., 1980, “The role of prosodic categories in English word stress”, in Linguistic
Inquiry 11, 563-605.
Shattuck-Hufnagel, S. & A.E. Turk, 1996, “A prosody tutorial for investigators of
auditory sentence processing”, Journal of Psycholinguistic Research 25, 193-246.
Spector, B., 2006, Aspects de la pragmatique des opérateurs logiques, PhD. dissertation:
Université Denis Diderot – Paris 7.
Speer, S., 2003, “Intonational disambiguation in sentence processing”, paper presented
at the 2nd Workshop of Tone and Intonation in Europe, april 2003, Oxford (United
Kingdom).

285
Sportiche, D., 1988, “A theory of floating quantifiers and its corollaries for constituent
structure”, in Linguistic Inquiry 19, 425-449.
Steedman, M., 2000, The Syntactic Process, Cambridge (Mass.): MIT Press.

Stowell, T., 1996, “The phrase structure of tense”, in J. Rooryck & L. Zaring (eds.), Phrase
structure and the lexicon, Dordrecht: Kluwer, 277–291.
Svenonius, P., 2004, “Adpositions, particles, and the arguments they introduce”,
manuscript: U. Tromsö.
Szabolcsi, A., 1994, “The noun phrase”, in F. Kiefer & K.E. Kiss (eds.), The syntactic
structure of Hungarian, New York: Academic Press, 179-274.
Szabolcsi, A., 1981, “The semantics of topic-focus articulation”, in J.A.G. Groenendijk,
T.M.V. Janssen, M.B.J. Stokhov (eds.), Formal methods in the study of language,
Amsterdam: Matematisch Centrum, 513-541.
Szendröi, K., 2001, Focus and the syntax-phonology interface, PhD. Dissertation: UCL.
Taylor, B.,1985, Modes of occurrence: Verbs, adverbs & events, Oxford: Blackwell.
Tancredi, Ch., 1992, Deletion, Deaccenting & Presupposition, PhD. Diss: MIT.
Tomalin, M., 2007, “Reconsidering recursion in linguistic theory”, in Lingua 117: 1784-
1800.
Truckenbrodt, H., 1999 “On the relation between syntactic phrases and phonological
phrases”, in Linguistic Inquiry 90-2, 219-255.
Truckenbrodt, H., 1995, Phonological phrases: Their relation to syntax, focus, and prominence,
PhD. Dissertation: MIT.
Umbach, C., 2001, “Contrast & Contrastive Topic” in I. Kruijff-Korbayová & M.
Steedman (eds.), Proceedings of ESSLLI 2001 Workshop on Information Structure,
Discourse Structure and Discourse Semantics: University of Helsinki, pp. 175-188.
Uriagereka, J., 2006, Syntactic anchors: On the semantic interfaces, manuscript: University
of Maryland.
Uriagereka, J., 1999, “Multiple spell out” in S.D. Epstein & N. Hornstein (eds.), Working
minimalism, Cambridge: MIT Press, 251-282.
Uriagereka, J., 1998, Rhyme and reason, Cambridge: MIT Press.
Uriagereka, J., 1995, “An F position in Western Romance”, in K.É. Kiss (ed.), Discourse
configurational languages, Oxford: Oxford University Press, 153-176.
Uribe-Etxebarria, M., 2002, “In situ questions and masked movement”, in P. Pica & J.
Rooryck (eds.), Linguistic Variation Yearbook 2, 217-257.
Vallduví, E., 1995, “Structural properties of information packaging in Catalan”, in K.É.
Kiss (ed.), Discourse configurational languages, Oxford: Oxford University Press,
122-152.
Vallduví, E., 1993, The informational component, PhD. dissertation: University of
Pennsylvania.

286
Vlach, F, 1983, “On situation semantics for perception”, Synthèse 54-1, 129-152.
Wachowicz, K., 1975, “Multiple questions”, in Linguistica Silesiana 1: 155-166.
Wachowicz, K., 1974, On the syntax and semantics of multiple questions, PhD. Dissertation:
University of Texas, Austin.
Wagner, M., 2005, Prosody and recursion, PhD. Dissertation: MIT.
Ward & Hirschberg, 1989, “Intonation and propositional attitude: the pragmatics of
L*+H L H%”, in Proceedings of 5th ESCOL, The Ohio State University, 512-524.
Watters, J. R., 1979, “Focus in Aghem: a study of its formal correlates and typology”, in
L. Hyman (ed.), Aghem Grammatical Structure, USC: Los Angeles, 137-197.
Williams, E., 1997, “Blocking and anaphora”, Linguistic Inquiry, 577-628.
Williams, E., 1994, Thematic structure in syntax, Cambridge: MIT Press.
Williams, E., 1977, “Discourse and logical form”, Linguistic Inquiry 8, 692-696.
Wold, D.E., 1998, “How to interpret multiple foci without moving a focused
constituent”, in E. Benedicto, M. Romero & S. Tomioka (eds.), UMOP 21:
Proceedings of the workshop on focus, Amherst: GLSA, 277-289.
Zagona, K., 1987, Verb phrase syntax: A parametric study of Spanish and English, Dordrecht:
Kluwer.
Zec, D. & S. Inkelas, 1990, “Prosodically constrained syntax”, in S. Inkelas & D. Zec
(eds.), The phonology-syntax connection, Chicago: University of Chicago Press, 365-
378.
Zubizarreta, M. L., 1998, Prosody, focus and word order, Cambridge: MIT Press.
Zubizarreta, M. L. & J. R. Vergnaud, 2000, “Phrasal stress and syntax”, in M. Van
Oostendorp & E. Anagnastopoulou (eds.), Progress in grammar: Articles at the 20th
anniversary of the comparison of grammatical models group in tilburg, electronic
publication. URL:
http://www.roquade.nl/meertens/progressingrammar/index.html

287