Phase theory: an Introduction

{DESKTOP}NEW FOLDER//9781107040847HTL.3D

i [1–1] 23.11.2013 10:53AM

Phase Theory Phase Theory is the latest empirical and conceptual innovation in syntactic theory within the Chomskyan generative tradition. Adopting a crosslinguistic perspective, this book provides an introduction to Phase Theory, tracing the development of phases in minimalist syntax. It reviews both empirical and theoretical arguments in favor of phases, and examines the role phases play at the interface with semantics and phonology. Analyzing current phasehood diagnostics, it applies them in a systematic fashion to a broad range of syntactic categories, both phases and non-phases. It concludes with a discussion of some of the more contentious issues in Phase Theory, involving crosslinguistic variation with respect to phasehood and the dynamic versus static nature of phases. Barbara Citko is Associate Professor of Linguistics at the University of Washington.

{DESKTOP}NEW FOLDER//9781107040847SER.3D

ii

[2–2] 23.11.2013 10:44AM

Research Surveys in Linguistics In large domains of theoretical and empirical linguistics, the needs of scholarly communication are directly comparable to those in analytical and natural sciences. Conspicuously lacking in the inventory of publications for linguists, compared to those in the sciences, are concise, single-authored, non-textbook reviews of rapidly evolving areas of inquiry. The series Research Surveys in Linguistics is intended to fill this gap. It consists of well-indexed volumes that survey topics of significant theoretical interest on which there has been a proliferation of research in the last two decades. The goal is to provide an efficient overview of, and entry into, the primary literature for linguists – both advanced students and researchers – who wish to move into, or stay literate in, the areas covered. Series authors are recognized authorities on the subject matter, as well as clear, highly organized writers. Each book offers the reader relatively tight structuring in sections and subsections, and a detailed index for ease of orientation. Previously published in this series A Thematic Guide to Optimality Theory, John J. McCarthy The Phonology of Tone and Intonation, Carlos Gussenhoven Argument Realization, Beth Levin and Malka Rappaport Hovav Lexicalization and Language Change, Laurel J. Brinton and Elizabeth Closs Traugott Defining Pragmatics, Mira Ariel Quantification, Anna Szabolcsi Word Order, Jae Jung Song

{DESKTOP}NEW FOLDER//9781107040847TTL.3D

iii [3–3] 23.11.2013 10:36AM

Phase Theory An Introduction

BARBARA CITKO University of Washington

{DESKTOP}NEW FOLDER//9781107040847IMP.3D

iv

[4–4] 23.11.2013 10:52AM

University Printing House, Cambridge CB2 8BS, United Kingdom Published in the United States of America by Cambridge University Press, New York Cambridge University Press is part of the University of Cambridge. It furthers the University’s mission by disseminating knowledge in the pursuit of education, learning and research at the highest international levels of excellence. www.cambridge.org Information on this title: www.cambridge.org/9781107040847 © Barbara Citko 2014 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2014 Printed in the United Kingdom by MPG Printgroup Ltd, Cambridge A catalogue record for this publication is available from the British Library Library of Congress Cataloguing in Publication data Citko, Barbara, 1970– Phase theory : an introduction / Barbara Citko. pages cm. – (Research surveys in linguistics) ISBN 978-1-107-04084-7 (hardback) 1. Phraseology 2. Minimalist theory (Linguistics) 3. Grammar, Comparative and general – Syntax. 4. Generative grammar. I. Title. P326.5.P45C48 2014 415–dc23 2013040516 ISBN 978-1-107-04084-7 Hardback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.

{DESKTOP}NEW FOLDER//9781107040847TOC.3D

v [5–6] 23.11.2013 10:44AM

Contents

Acknowledgments List of abbreviations

page vii viii 1

Introduction 1

1.1 1.2 1.3 1.4 2

3

4

7

The Minimalist Program

General architecture External and Internal Merge Features Agree

7 10 14 20

Introducing phases

23

2.1 2.2 2.3 2.4 2.5 2.6

23 27 29 31 41 46

Merge over Move preference Motivating phases Phasehood properties Phase Impenetrability Condition Multiple Spell-Out Feature Inheritance

Phasehood diagnostics

58

3.1 3.2 3.3 3.4

58 60 66 67

Anatomy of a phasehood diagnostic PF diagnostics LF diagnostics Syntactic diagnostics

71

Classic phases

4.1 CPs as phases 4.2 vPs as phases 4.3 DPs as phases 5

71 91 108

Other ph(r)ases

124

5.1 PrPs as phases 5.2 PPs as phases 5.3 AppIPs as phases

124 137 151 v

{DESKTOP}NEW FOLDER//9781107040847TOC.3D

vi

[5–6] 23.11.2013 10:44AM

vi

6

7

8

Contents

Variation in phasehood

161

6.1 Dynamic phases 6.2 Crosslinguistic variation 6.3 Non-simultaneous phases

161 168 177

Phases and the interfaces

185

7.1 Phases and PF 7.2 Phases and LF

185 199

Summary

205

References Index

207 223

{DESKTOP}NEW FOLDER//9781107040847ACK.3D

vii [7–7] 23.11.2013 11:51AM

Acknowledgments

First, I would like to thank the many cohorts of students who took my graduate syntax classes at the University of Washington. The idea for this book came from you and all the questions you asked (that I may or may not have had answers to at the time). I hope next time I will! I also want to thank all my colleagues in the Department of Linguistics for creating an intellectually stimulating work environment, and especially my fellow syntacticians, Edith Aldridge, Karen Zagona and Julia Herschensohn, for encouragement and support, especially in the final stages of the project; and Toshi Ogihara for very kindly answering all my last-minute semantic questions. Portions of this book were written at the Whiteley Writing Center, a wonderful writing oasis (on an island!), which I thank for the good writing vibes. Some parts were written when I was a visiting scholar at the Department of Linguistics and Philosophy at MIT in the Fall of 2011. I thank the department for the welcoming atmosphere it creates for all its visitors, and for all the inspiring classes, talks and conversations I had while I was there. In particular, I would like to thank David Pesetsky, Sabine Iatridou and Shigeru Miyagawa for their openness and interest in my research. I also want to thank Andrew Winnard and the whole editorial team at Cambridge University Press for their patience and guidance throughout the project, Alexander Sugar for help proofing the manuscript and Brent Woo for help with the index. Given that the literature on phases in syntactic theory is vast and evergrowing, it is impossible to do justice here to all the phenomena that bear on all the issues surrounding phases, to all the literature on phases and to all the researchers working on phases. I alone remain responsible for any omissions or misinterpretations. Last but not least, I thank my husband Randy for his support and perspective. I dedicate this book to the memory of Icarus.

vii

{DESKTOP}NEW FOLDER//9781107040847LOA.3D

viii

[8–10] 23.11.2013 10:46AM

Abbreviations

The following list includes not only abbreviations used in this book but also the abbreviations often used in the primary literature on phases. φ Φ Σ ACC ACD ATB AUGB AUX BPS C CHL CAUS C-I CP DNS DP ECM EM EPP ERG FEM FI FIFA FIFB FL FP G H

phi-features phonological component semantic component Accusative Antecedent-Contained Deletion across the board approaching UG from below Auxiliary Bare Phrase Structure Case Computation of Human Language Causative conceptual–intentional Complementizer Phrase Derivation of Narrow Syntax Determiner Phrase Exceptional Case Marking External Merge Extended Projection Principle Ergative Feminine Feature Inheritance Feature Inheritance from Above Feature Inheritance from Below language faculty Functional Projection Goal Head

viii

{DESKTOP}NEW FOLDER//9781107040847LOA.3D

ix [8–10] 23.11.2013 10:46AM

List of abbreviations

HTA IC iF IM INTER INDEF iwh LA LBE LCA LDA Lex LF LI MASC MOM MSO N NEUT NMLZ NOM NPI NS NSL NSR NUM P PART PF P-feature PH PIC PL PLA PP PrP PST Q QP QR RNR RP SA

High Tone Anticipation Interface Condition(s) or the Inclusiveness Condition Interpretable Feature Internal Merge Interrogative Indefinite interpretable wh-feature Lexical Array Left Branch Extraction Linear Correspondence Axiom Long-Distance Agree Lexicon Logical Form Lexical Item Masculine Merge over Move Multiple Spell-Out Numeration Neuter Nominalizer Nominative Negative Polarity Item Narrow Syntax Null Subject Language Nuclear Stress Rule Number Probe Participle Phonological Form Periphery Feature Phase Phase Impenetrability Condition Pair List or Plural Properly Local Agreement Prepositional Phrase Predication Phrase Past Question or Interrogative Quantifier Phrase Quantifier Raising Right Node Raising Relator Phrase Single Answer

ix

{DESKTOP}NEW FOLDER//9781107040847LOA.3D

x

SG SM SMT S-O SOR SOT T TNP TP uF uQ val VIR vP VPE

x [8–10] 23.11.2013 10:46AM

List of abbreviations

Singular Sensorimotor Strong Minimalist Thesis Spell-Out Subject to Object Raising Sequence of Tense Tense Traditional Noun Phrase Tense Phrase uninterpretable or unvalued feature uninterpretable Q feature Feature Value Virile verb Phrase VP Ellipsis

{DESKTOP}NEW FOLDER//9781107040847INT.3D

1 [1–6] 23.11.2013 10:49AM

Introduction

This research survey combines an introduction to Phase Theory with an assessment of the state of the art in Phase Theory. The term Phase Theory refers to a set of theoretical innovations in post-2000 minimalism (Chomsky 2000, 2001, 2004, 2005, 2008).1 One of the core ideas in minimalism is the idea that the language faculty is an optimal solution to the constraints imposed on it by the two cognitive systems with which it interacts: the system of thought and the articulatory– perceptual system. What Phase Theory adds to this picture is the idea that the language faculty interacts with these two cognitive systems at very specific points during the syntactic derivation, and, consequently, that syntactic derivations are constructed in chunks referred to as phases.2 In most general terms, phases cannot be accessed by the narrow syntax once they are transferred to the interfaces. My goal in this survey is to combine an introduction to a given issue within Phase Theory with an overview of the existing research on this issue (and an assessment thereof), giving the reader a sense of what is fairly settled upon and what is still under debate.3 The fact that there is lot of research that relies on phases shows a need for a survey that situates phases in current syntactic theory, introduces the technical details of Phase Theory, synthesizes the existing research

1

2

3

Chomsky in his writings is very careful about distinguishing a program from a theory, emphasizing the programmatic nature of minimalism (see in particular Boeckx 2006 for a more detailed discussion of this distinction). Chomsky does not use the term Phase Theory in his early writings on phases but does so more recently: ‘One goal of Phase Theory is to provide the mechanisms to distinguish copies from repetitions, as straightforwardly as possible’ (Chomsky 2012: 3). The resulting model is also sometimes referred to as a Multiple Spell-Out (MSO) model. The idea of Multiple Spell-Out goes back to Uriagereka (1999) (see also Uriagereka 2012 for a fuller, booklength exposition). There are a number of monographs, volumes and collections that focus on various aspects of Phase Theory, which I build on and am intellectually indebted to (see, among others, Frascarelli 2006, Gallego 2010, 2012, Grohmann 2009a, 2009b, 2009c, McGinnis & Richards 2005, and the individual contributions in the two issues of Linguistic Analysis 33, guest-edited by Kleanthes Grohmann).

1

{DESKTOP}NEW FOLDER//9781107040847INT.3D

2

2 [1–6] 23.11.2013 10:49AM

Introduction

on phases (pointing out issues that might be still contentious), outlines directions for future research, and, last but not least, standardizes the notation. Even though many (though not all) syntacticians (explicitly or implicitly) assume the concept of a phase, there seems to be less of a consensus regarding many of the most fundamental properties of phases, such as those listed in (1). (1) a. How do we define phases? b. What categories count as phases? c. Do the same categories count as phases with respect to semantic and phonological considerations? d. Are phases dynamic or static? e. Is there any crosslinguistic variation with respect to phasehood? f. How do phases interact with the interfaces?

The fundamental question to answer before we can even begin to address some of the questions listed above is in what sense a syntactic theory that assumes phases is more adequate (in a descriptive, explanatory, or beyond-explanatory adequacy kind of sense) than a syntactic theory that does not assume phases.4 This is the question we will be coming back to throughout the book. In the remainder of this introduction, I provide a brief summary of each chapter. Chapter 1 ‘The Minimalist Program’ provides an overview of the core aspects of the Minimalist Program. It outlines the general architecture of the minimalist grammar, and lays the groundwork for the discussion in the following chapters by focusing on the concepts that will be crucial to the understanding of phases, such as the distinction between interpretable and uninterpretable features and the concept of Spell-Out. This chapter is not meant as an introduction to (or survey of ) minimalist syntax; however, readers less familiar with minimalism will find all the necessary concepts, terms and mechanisms introduced in this chapter. Chapter 2 ‘Motivating phases’ turns to phases themselves. It introduces the concept of a phase, situating phases in the context of current minimalist approaches to syntactic dependencies, and asking if syntactic theory with phases is more adequate than a theory without phases, or a theory in which all phrases are phases. This chapter also gives a historical perspective on phases, and addresses some of the criticisms that have been levied in the literature against them, such as Boeckx & Grohmann’s (2007) critique of phases as ‘barriers in disguise’. Even though the idea of a cycle, which is conceptually related to a phase, goes back to the early days of generative grammar, the current concept of a phase first appeared in Chomsky’s (2000) ‘Minimalist Inquiries’, where phases (to be more specific, lexical subarrays associated with phases) were introduced as a solution to a problem arising from the so-called Merge over Move principle. Since then, 4

Following Chomsky (2004), I take beyond explanatory adequacy to refer to the why-questions about language, captured by the following quote: ‘In principle, then, we can seek a level of explanation deeper than explanatory adequacy, asking not only what the properties of language are but also why they are that way.’ (Chomsky 2004: 105).

{DESKTOP}NEW FOLDER//9781107040847INT.3D

3 [1–6] 23.11.2013 10:49AM

Introduction

3

much research has focused on defining phases and formulating independent phasehood diagnostics. The definitions of phases I survey in this chapter are listed in (2).5 (2) a. Phases are propositional objects. b. Phases are convergent objects. c. Phases are objects interacting with the interfaces. d. Phase heads are loci of uninterpretable features. e. Phases are predication structures. f. Phases are phrases.

From a diagnostic perspective, perhaps the most important aspect of Phase Theory is the so-called Phase Impenetrability Condition, which deems a portion of a phase impenetrable or inaccessible to operations from the outside. This chapter also surveys the various versions of the Phase Impenetrability Condition proposed in the literature, focusing on the empirical predictions they make, and ways to unify them (see Müller 2004, Richards 2004, 2011, among others). The Phase Impenetrability Condition is tightly linked to the concept of Multiple Spell-Out, which I also elaborate on in this chapter, sorting through the logical possibilities of how Multiple Spell-Out can proceed, i.e. spelling out to the two interfaces at different points in the derivation, for example. Finally, this chapter introduces the concept of Feature Inheritance, as developed by Chomsky (2008) and Richards (2008), which is a logical consequence of defining phase heads as hosts of uninterpretable features. If uninterpretable features are a property of phase heads, the only way non-phase heads can get them is via Feature Inheritance. Chapter 3 ‘Phasehood diagnostics’ turns to the many diagnostics that have been proposed in the literature, a subset of which will serve as the basis for the discussion of specific phases (CPs, vPs, DPs, PPs etc.) in the chapters that follow, and the arguments in favor of (or against) these categories being phases. A common thread in many existing characterizations of phases is that they should exhibit a certain amount of independence and coherence at the interfaces. This, however, only raises the question of what it means for a given category (a candidate for a phase) to be semantically or phonologically independent and coherent. Furthermore, are there any phasehood diagnostics that do not fall neatly into either of the two groups (PF versus LF diagnostics): purely syntactic or purely morphological diagnostics? Given such rather vague existing characterizations of phases, this chapter focuses on the more tangible questions that can be (and have been) asked to establish the phasehood of a given category, which I list in (3) below. It examines these questions with a critical eye towards establishing genuine phasehood diagnostics, and avoiding those that might instead be diagnosing something other than phasehood (such as constituency or phrasal status). 5

See also Boeckx (2006), Boeckx & Grohmann (2007), Den Dikken (2007) and Gallego (2010) for a discussion of these different views of phases, and of the problems some of them raise.

{DESKTOP}NEW FOLDER//9781107040847INT.3D

4

4 [1–6] 23.11.2013 10:49AM

Introduction

(3) a. Is XP a domain for feature valuation? b. Is X the locus of uninterpretable features? c. Does X trigger Spell-Out? d. Is XP a phonological domain? e. Can the complement of X be elided? f. Can XP be moved?

The Phase Impenetrability Condition also gives rise to a number of tangible phasehood diagnostics, coming mostly from the realm of successive cyclic movement through the edge of the phase, which in turn can be diagnosed by affirmative answers to the following questions: (4) a. Can the moved element be interpreted at the edge of the phase? b. Can the moved element be pronounced at the edge of the phase? c. Can the moved element leave something behind at the edge of the phase?

The discussion of phasehood diagnostics also raises the question of whether there is any crosslinguistic variation with respect to phasehood. This is the issue which I come back to in Chapter 6. While variation with respect to whether a given language has phases or not seems highly unlikely and implausible, given the conceptual arguments in favor of having phases to begin with (such as reducing computational load and being independently motivated by the interfaces), it is certainly possible for languages to differ with respect to what categories count as phases. Chapter 4 ‘Classic phases’ discusses in detail three categories that are commonly assumed to be phases – CPs, vPs and DPs – and applies the diagnostics established in Chapter 3 to these categories.6 The phasehood status of CP is relatively easy to establish: the evidence in favor of successive cyclicity from the literature on A-bar dependencies is typically taken as evidence for CPs being phases (see, for example, Lahne 2008 for an illuminating overview). The evidence includes phenomena like wh-copying (Felser 2004, Manetta 2010, McDaniel 1989, among many others), scope marking (Dayal 1996, Lutz Müller & von Stechow 2000, Stepanov 2000, among others), complementizer agreement (Carstens 2003, Carstens & Diercks 2011, Haegeman 1992, Haegeman & Van Koppen 2011, Zwart 1993, 1997), wh-quantifier stranding (McCloskey 2000, 2001), reconstruction (Barss 1986, 2001) and left branch extraction (Wiland 2010). The evidence in favor of vPs being phases is similar in spirit. This chapter also reviews the debate on whether unaccusative and passive vPs constitute phases, as argued for by Legate (2003), and against by Den Dikken (2006a). While many of the facts that are typically deemed to bear on the issue of C or v being a phase head might be given 6

The discussion in Chapters 4 and 5 is a sequence of case studies. There are other categories that are conspicuously absent from the discussion (APs, AdvPs, various functional projections in the left periphery of a clause) whose phasehood we might wonder about. I thank Kleanthes Grohmann for raising this issue.

{DESKTOP}NEW FOLDER//9781107040847INT.3D

5 [1–6] 23.11.2013 10:49AM

Introduction

5

alternative explanations that do not necessarily rely on movement through the specifier of CP or vP, such accounts typically still posit a relationship between C (endowed with uninterpretable features of the requisite sort) and the wh-pronoun in its domain. This also points towards C being a phase head, given that only phase heads are assumed to be the loci of uninterpretable features. More generally, I hope to show in this chapter that phase-theoretical accounts have the advantage of establishing connections between sets of facts that otherwise remain isolated and require independent explanations. For example, why should complementizer agreement phenomena and locality restrictions on movement involve C? Or why would Austronesian extraction restriction and constraints on parasitic gap formation be sensitive to the properties of little v? Granting these projections a privileged syntactic status (namely, the status of a phase) brings us closer towards understanding why syntactic phenomena should cluster around them. The idea that DP might be a phase as well, explored by Matushansky (2005), Hiraiwa (2005) and Svenonius (2004), among others, should not come as a surprise, given the many structural and interpretive parallels between CPs and DPs, discussed in the literature going back to the very early days of generative grammar. However, since CPs contain other phases (namely vPs), an interesting question is whether DPs contain other phases as well. In order to tackle this question, this chapter also addresses the internal structure of DPs, motivating the need for DP-internal projections such as NumberP, PersonP or ClassifierP, and asking which of them, if any, might be phases as well. Chapter 5 ‘Other ph(r)ases’ turns to categories whose phasehood status is somewhat more controversial, and still debated in the literature: Predication Phrases, Prepositional Phrases and Applicative Phrases. All of them have been argued to constitute phases (see, for example, Abels 2003, Radkevich 2010 on PPs as phases and McGinnis 2001 on Applicative Phrases as phases); yet they are not considered to fall into the widely accepted phasehood canon. What makes these categories somewhat more controversial is that many other questions about them have to be answered first before their phasehood can be entertained. For example, if phasehood is a property of functional categories, the question that needs to be resolved for prepositions is whether they are functional or lexical (or both or neither, depending on the preposition). Chapter 6 ‘Variation in phasehood’ takes up the issue of whether there is any crosslinguistic variation with respect to phasehood. There are two questions to consider here: the question of whether non-phase heads can acquire phasehood status in the course of the derivation (and conversely, whether phase heads can lose their phasehood status in the course of the derivation), and the question of whether different categories can count as phases with respect to phonological and semantic considerations. The former scenario (a head becoming a phase or ceasing to be a phase) has been argued to arise as a result of head movement (Phase Extension of Den Dikken 2007 or Phase Sliding of Gallego 2010). The latter scenario (a category being a PF phase but not an LF phase or vice versa) has been explored by Marušič (2005) as a way to handle total reconstruction and

{DESKTOP}NEW FOLDER//9781107040847INT.3D

6

6 [1–6] 23.11.2013 10:49AM

Introduction

covert movement and by Felser (2004) to handle wh-copying. This chapter also addresses crosslinguistic variation with respect to phasehood: if phases are dynamic and head movement can extend phasehood, a certain amount of variation will come from independent considerations (such as the presence or absence of certain types of head movement). Variation in phasehood can also follow from variation in lexical inventories. Chapter 7 ‘Phases at the interfaces’ examines the roles phases play at the interfaces, putting them in the more general context of the syntax–phonology and syntax–semantics interface. With respect to the PF interface, it focuses on the questions of whether phases (or Spell-Out domains) are relevant and substantive phonological units, and how these phasal or Spell-Out units are manipulated by phonology. This chapter examines the role phases play in determining linear order (see Fox & Pesetsky’s (2005) Cyclic Linearization) and nuclear stress (see Adger 2007, Kahnemuyipour 2003, 2004, 2005, Kratzer & Selkirk 2007, among others). The potential evidence for the significance of phases at the syntax–semantics interface comes not only from phenomena like scope ambiguities (on the assumption that Quantifier Raising is constrained by phasehood) and the propositional status of phases – both of which feature prominently as phasehood diagnostics – but also from the idea that the boundary between vP and CP phasal domains corresponds to the distinction between nuclear scope and restrictive domain in the tripartite quantificational structure, as proposed explicitly by Biskup (2009a). Chapter 8 ‘Summary’ provides a brief summary and conclusion.

{DESKTOP}NEW FOLDER//9781107040847C01.3D

7 [7–22] 23.11.2013 11:50AM

1 The Minimalist Program

1.1 General architecture The current chapter offers a bird’s eye view of the Minimalist Program. It is not meant as a comprehensive introduction (or a thorough overview) of minimalism. Rather, its goal is to give readers less familiar with minimalism the necessary and sufficient background to follow the discussion of phases in the rest of this book. For the sake of clarity, the technical terms that I will be referring to throughout the book will be given in bold when they are first introduced. For a more thorough textbook-style introduction to minimalism, I refer the interested reader to Adger (2003) and Hornstein, Nunes and Grohmann (2005) and the references therein. What came to be known as the Minimalist Program was articulated explicitly in the early nineties with the publication of works such as Chomsky’s (1991) ‘Some Notes on Economy of Derivation and Representation’, and his (1993) ‘A Minimalist Program for Linguistic Theory’, both of which later became two of the four chapters of Chomsky’s (1995) The Minimalist Program. As Chomsky emphasizes in his writings, minimalism is grounded in the Principles and Parameters model, which gave us the beginnings of an understanding of which properties of language are universal (and perhaps unique to it), and which ones are subject to crosslinguistic variation. This, in turn, led to deeper questions, which are at the core of minimalist theorizing nowadays. These are questions that go beyond explanatory adequacy, alluded to above, such as the question of why language is the way it is. Computational efficiency and interface conditions play a central role, as stated succinctly in the following quote from ‘Beyond Explanatory Adequacy’. (1) Its [the Minimalist Program’s, B.C.] task is to examine every device (principle, idea, etc.) that is employed in characterizing languages to determine to what extent it can be eliminated in favor of a principled account in terms of general conditions of computational efficiency and the interface condition [emphasis mine, B.C.] that the organ must satisfy for it to function at all. (Chomsky 2004: 106)

7

{DESKTOP}NEW FOLDER//9781107040847C01.3D

8

8 [7–22] 23.11.2013 11:50AM

The Minimalist Program

This is also clear in Chomsky’s discussion of the so-called three factors in language design and three types of conditions in language acquisition (see Chomsky 2005 in particular). These are listed in (2a–c), and they help determine how a child gets from the initial state (S0) of linguistic competence to the final state: the fully formed adult state of linguistic competence. (2b) are the interface conditions, and (2c) are the general properties of efficient computation. (2) a. unexplained elements of S0 b. IC (the principled part of S0) c. general properties (Chomsky 2004: 106)

The minimum that the language faculty (FL) has to accomplish is to interface with language-external systems. The two external systems in question are the sensorimotor (SM) system and the conceptual-intentional (C-I) system. The conditions imposed by these two external systems are referred to as Legibility Conditions, Bare Output Conditions or Interface Conditions (IC).1 The Strong Minimalist Thesis (SMT), given in (3), states that language is designed to interface with the external systems in an optimum way.2 (3) The substantive thesis is that language design may really be optimal in some respects, approaching a “perfect solution” to minimal design specifications. (Chomsky 2000: 93)

The general architecture of the language faculty is as follows. Language has three components: Narrow Syntax (NS), the phonological component Φ and the semantic component Σ.3 For the most part, we will be concerned here with Narrow Syntax and its computational processes. Each derivation starts with a set of ‘lexical items’ which are manipulated in the course of the derivation by the syntactic operations Merge and Agree. I will discuss these two operations in more detail in Sections 1.2 and 1.4, respectively. The ‘lexical item’ is, strictly speaking, a bundle of features, not a primitive syntactic object (hence the quotes).4 This set of lexical items is called a Lexical Array (LA) and is represented as an unordered set. A Lexical Array augmented by information 1 2

To the best of my knowledge, these terms are used interchangeably. The formulation of the Strong Minimalist Thesis in ‘Minimalist Inquiries’ (which is the one given in (3)) is slightly different from the one Chomsky gives in ‘Beyond Explanatory Adequacy’, where he formulates it as in (i): (i) The set of unexplained elements of S0 is empty. (Chomsky 2004: 106)

3

4

S0 refers to the genetically determined initial state in the process of language acquisition, which is what UG provides. There is no PF or LF cycle and thus there are no PF or LF operations. The terms are used to refer to PF or LF representations. The terms PF interface and LF interface are used to refer to the interface with SM or C-I systems, respectively. This assumes something like the Late Vocabulary Insertion model of Distributed Morphology of Halle & Marantz (1993) and much later work.

{DESKTOP}NEW FOLDER//9781107040847C01.3D

9 [7–22] 23.11.2013 11:50AM

General architecture

9

on how many times each lexical item is selected from the lexicon is called a Numeration (N). This information is represented by the subscripts. In simple cases, Numerations and Lexical Arrays are equivalent, as shown in (4b–c). (4) a. Icarus likes nuts. b. LA = {Icarus, likes, nuts, v, T, C} c. N = {Icarus1, likes1, nuts1, v1, T1, C1}

The two diverge when a single item is used more than once in a given derivation, as in the infamous example given in (5a).5 (5) a. Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo. (Pinker 1994: 2006, crediting Annie Senghas) b. LA = {Buffalo, buffalo, buffalo, C, v, T} c. N = {Buffalo3, buffalo2, buffalo2, C2, v2, T2}

However, one does not need to resort to exotic examples of this sort to show how Lexical Arrays and Numerations differ; relatively simple sentences with any level of embedding make the same point: (6) a. Icarus thinks he likes nuts. b. LA = {Icarus, thinks, he, likes, nuts, v, T, C} c. N = {Icarus1, thinks1, he1, likes1, nuts1, v2, T2, C2}

The output of a Narrow Syntax derivation is a pair of representations , which is accessed by the two interfaces defined above: the SM and the C-I interface. The derivation converges if the two representations satisfy the conditions imposed by the two interfaces; otherwise it crashes.6 For a given representation to meet the Interface Conditions simply means it has to be legible to the external systems; hence the term Legibility Conditions. The question of what it means to be legible at a given interface is not trivial. A common and intuitively correct understanding of this concept of legibility is the following: an expression is legible at a given interface level (PF or LF) only if it consists of features that can be interpreted by the language-external systems: the SM and C-I system, respectively. But, of course, making convergence contingent on the presence of features that can only be interpreted at the interfaces raises an obvious question of what features the two interfaces can interpret. It seems quite plausible to assume that 5

Its notoriety comes from lexical ambiguity, of which it provides a very extreme illustration, not from the distinction between a lexical array, which in this case contains distinct lexical items ‘buffalo’ (i.e. the proper name Buffalo, the common noun buffalo, and the less commonly used transitive verb to buffalo meaning to bully) and the Numeration that includes multiple occurrences of each of them. The following paraphrase helps distinguish the different meanings of buffalo: (i) The buffalo from Buffalo that (another) buffalo from Buffalo bullies himself bullies (yet another) buffalo from Buffalo.

6

See, however, Frampton & Gutmann (2002) for a proposal that syntax only generates convergent derivations, and Preminger (2011) for a proposal that having unvalued features does not always lead to a crash.

{DESKTOP}NEW FOLDER//9781107040847C01.3D

10

[7–22] 23.11.2013 11:50AM

10

The Minimalist Program

the SM interface can interpret features having to do with linear order (if such exist), syllable structure, prosodic structure or intonation. The C-I interface, on the other hand, should be able to interpret features having to do with scope, quantification, referentiality, specificity, propositional status etc. Neither interface can interpret formal features, such as structural case features or categorial features. Since features play such a major role in minimalism and there are quite a few contentious issues surrounding them, we will devote an entire section to them (Section 1.3). The three basic operations that manipulate lexical items selected from the lexicon are External Merge, Internal Merge and Agree. These three, in conjunction with a more detailed discussion of features, are the focus of the next three sections.

1.2 External and Internal Merge Recursion, the property of language that allows smaller units to combine iteratively to form larger units forming hierarchically structured objects, and displacement, the property that gives us the intuition that syntactic objects can surface in one position but be understood as belonging in another position, are two very fundamental (perhaps the most fundamental) properties of language. Chomsky (2004) distinguishes two kinds of Merge, External Merge (EM) and Internal Merge (IM), to capture these two fundamental phenomena. External Merge is the basic concatenation operation responsible for recursion in language. It takes two objects (such as X and Y in (7a), which have been first selected from the Numeration), and combines them into one bigger object, as shown in (7b).7 External Merge is a recursive operation; one of these two objects could 7

Elsewhere, I have argued that Merge can also create structures in which a single object can end up shared between two objects, referring to this type of Merge as Parallel Merge (see Citko 2011b and the references therein), illustrated in (i–ii). Parallel Merge combines the properties of External Merge and Internal Merge. Before Merge takes place, Z and YP are disjoint (as in External Merge) but Z merges with a subpart of YP (as in Internal Merge). (i) Merge X and Y, Project Y

YP Y

XP

(ii) (Parallel) Merge X and Z, Project Z

YP Y

ZP XP

Z

Chomsky (2007) excludes such a possibility on the grounds that ‘it requires new operations and conditions on what counts as a copy, hence additional properties of UG’ (Chomsky 2007: 8, note 10).

{DESKTOP}NEW FOLDER//9781107040847C01.3D

11

[7–22] 23.11.2013 11:50AM

External and Internal Merge

11

itself be the output of a prior Merge operation, as shown in (7d), where the complex object that is the output of (7b) merges with Z.8 (7) a. Select X and Y from the Numeration b. Merge X and Y

X X

Y

c. Select Z from the Numeration d. Merge Z with X

Z Z

X Y

X

8

Following Chomsky (1994), I assume a bare-bones approach to phrase structure, referred to as Bare Phrase Structure (BPS), in which the X-bar status of a given projection (in particular, whether it is a head (X0), a phrase (XP) or an intermediate projection (X0 )) is not a primitive of the theory, as it can always be derived from the structure in which it appears. Thus a head is an element that is not a projection of another element of the same type. It can (but it does not have to) project further. A phrase is an element that does not project any further. Again, it can, but it does not have to, be a projection of another element of the same type. The representations in (7a–d) are BPS representations; the ones in (i) and (ii) are their more traditional X-bar theoretical counterparts, with the ones in (ib) and (iib) simplified in that they lack vacuous intermediate projections. (i)

a. XP

b. X′

X (ii)

a.

XP X

YP

YP

ZP

b. Z′

Z

ZP Z

XP

XP X

YP

X′ X

YP

Even though I will be using representations of the kind given in (ib)–(iib) throughout this book, mostly for familiarity’s sake, I do assume a BPS approach throughout. The advantages of such an approach are twofold. First, it allows a given element to be simultaneously a head and a phrase, something that the behavior of clitics suggests has to be allowed in the grammar. Y in (7b) and (7d) is such an element; it is a head because it is not a projection of another Y and it is a phrase because it does not project any further. Second, BPS dispenses with vacuous intermediate projections, a welcome move from the perspective of the Strong Minimalist Thesis (SMT).

{DESKTOP}NEW FOLDER//9781107040847C01.3D

12

[7–22] 23.11.2013 11:50AM

12

The Minimalist Program

The output of External Merge could also be represented in set notation. (8a) is equivalent to (7b) and (8b) to (7d). (8) a. {X, {X, Y}} b. {Z, {Z, {X, {X, Y}}}}

It is standard to assume that the result of Merge also has to have a label, bolded in (8a–b) above.9 The label is determined by one of the two merged elements; other options would violate the Inclusiveness Condition (IC), which prohibits the introduction of new elements in the course of the derivation. Any label other than X or Y in (8a) (and other than Z and X in (8b)) would be a violation of this condition. Chomsky (1995) discusses two other logical possibilities. In a set consisting of two elements (such as X and Y in (8a)), the label could also be the intersection of X and Y or the union of X and Y. However, he excludes the first option on the grounds that the intersection in many cases is a null set. The second option is excluded because the union of X and Y is contradictory if X and Y differ in features, as is the case when one is N and the other V, for example. This leaves us with only one option: either for X or Y to project as the label. The question of which of the two it is, or the even more fundamental question of whether labels are necessary at all, is far from being settled. A common and fairly intuitive idea is that the object which has its selectional feature checked as a result of Merge is the object that determines the label. For example, when a verb or a preposition merges with a noun, this verb (or preposition) has its selectional feature satisfied via the Merge operation. This means that the label is going to be determined by this verb or preposition.10,11 Many existing labeling algorithms make reference to selection; the one from Cecchetto & Donati (2010), given in (9) below, is a representative example:12

9

10

11

An interesting issue is whether all the features of X determine the label, or only a subset thereof. While it seems clear (and commonly assumed) that categorial features (N, V, D etc.) are part of the label, what other features need to project is somewhat less clear (see Citko 2008a, 2011b, Cecchetto & Donati 2010, Donati & Cecchetto 2011, for a discussion of this and related issues). Things get a little more complicated if c-selection is removed from our syntactic toolkit, as Chomsky suggests in ‘Beyond Explanatory Adequacy’ (2004). Chomsky (2008) proposes a different labeling algorithm, given in (i) below, which does not make reference to selection. (i) In {H, α}, H a lexical item (LI), H is the label. If α is internally merged to β forming {α, β}then the label of β is the label of {α, β} (Chomsky 2008: 145)

12

Other examples similar in spirit, as also noted by Cecchetto & Donati (2010) and Gallego (2010) include those proposed by Adger (2003), Pesetsky & Torrego (2006) and Boeckx (2008b): (i) The head is the syntactic object which selects in any Merge operation. (Adger 2003: 91) (ii) If α and β merge, some feature F of α must probe F on β. (Pesetsky & Torrego 2006) (ii) The label of {α, β} is whichever of α or β probes the other, where the Probe = Lexical Item whose uF gets valued. (Boeckx 2008b: 96)

{DESKTOP}NEW FOLDER//9781107040847C01.3D

13

[7–22] 23.11.2013 11:50AM

External and Internal Merge

13

(9) The label of a syntactic object {α, β} is the feature(s) that act(s) as a Probe of the merging operation creating {α, β}. (Cecchetto & Donati 2010: 245)

Chomsky (2004) considers the possibility that the grammar requires no labels whatsoever (see also Collins 2002, Gallego 2010, Seely 2006, among others).13 While intuitively appealing from the perspective of the Strong Minimalist Thesis, eliminating labels raises some non-trivial issues, such as how selection or relativized minimality works. I will not discuss these issues here (as they strike me as tangential to the issue of phases. Instead, I refer the interested reader to Citko (2008a, 2011b) and the references therein for relevant discussion, pointing to the conclusion that labels are not dispensable. One argument often adduced against the existence of labels comes from the Inclusiveness Condition. However, if a label is simply a copy of the features of one of the two merged elements (or more likely, of a subset of its features), inclusiveness is not violated. Internal Merge, which as Chomsky suggests in ‘Beyond Explanatory Adequacy’ (2004) is a very natural way to capture movement, is similar to External Merge in that it takes two objects and combines them to form one larger object. The only difference is that one of these two objects is already part of the other one (hence the term Internal Merge). (10) below provides an illustration; Y undergoes Internal Merge with X or, to use more traditional terminology, moves to the specifier of X. (10) Internal Merge of X and Y14

a.

X Y

b. X

X EPP Y

X X X EPP

Y

The label in Internal Merge structures is determined by the element that drives movement. In (10a–b), for example, it is some feature of X that drives movement of Y to its specifier. I assume that the feature in question is the EPP feature, which is checked by this movement. Checking is marked by a strikethrough. In what 13

14

However, Chomsky (2013) seems to depart from this conclusion, and points out a number of cases in which the lack of labels forces movement. This suggests that in other (non-movement) cases, labels have to be in principle available. (10a) represents movement (Internal Merge in our terms) in a familiar Copy and Merge notation, whereas (10b) does so in a way that reflects the term ‘Internal Merge’ perhaps a little more directly. Both are BPS-style representations. Strictly speaking, the movement illustrated here (i.e. movement of a complement to a specifier of the same head) might be banned by independent anti-locality principles, which I discuss in more detail in Chapters 4 and 5.

{DESKTOP}NEW FOLDER//9781107040847C01.3D

14

14

[7–22] 23.11.2013 11:50AM

The Minimalist Program

follows, I use the term valuation to refer what used to be known as checking and reserve the term checking for Extended Projection Principle (EPP) feature checking, the only remnant of checking in the current system. The third operation that plays a crucial role in minimalist syntax is Agree. Since it is an operation that affects features, let me digress briefly and discuss features first. This is the focus of the next section, which dicusses syntactic features and the role they play in Narrow Syntax.

1.3 Features Examples of syntactic features that populate syntactic literature include categorial features (such as N features or V features), case features, φ-features (gender, number, person), EPP features, wh-features, and Q features, among others. Features play a crucial role in minimalism in that they drive syntactic operations. Features come in two guises: Interpretable Features (iF) and Uninterpretable Features (uF). Chomsky (2001, 2004) reduces the distinction between these two types to feature valuation. Uninterpretable features are simply features that enter the derivation unvalued, whereas interpretable ones are the ones that come with inherently specified feature values: (11) An uninterpretable feature F must be distinguished somehow in LEX from interpretable features. The simplest way, introducing no new devices, is to enter F without value: for example, [uNumber]. That is particularly natural because the value is redundant, determined by Agree. (Chomsky 2004: 116)

I will use the following notation to distinguish these two types, where val stands for feature value. (12) a. Interpretable feature: iF[val] (e.g. iT[past], iφ[3sg, masc]) b. Uninterpretable feature: uF[ ] (e.g. uCase[ ], uφ[ ])

In ‘Minimalist Inquiries’ (2000), Chomsky takes the presence of uninterpretable features to be an imperfection from the perspective of the Strong Minimalist Thesis, since the existence of such features is not forced by Bare Output Conditions. Interestingly, though, he takes this imperfection to be linked to another imperfection in the system: movement. Uninterpretable features drive movement. ‘Perhaps these devices [uninterpretable features, B.C.] are used to yield the dislocation property’ (Chomsky 2000: 121). The fact that these two imperfections are related is somewhat ‘suspect’, and it opens up the possibility that they might not be imperfections after all. This is indeed the view Chomsky takes in ‘Beyond Explanatory Adequacy’, where he takes movement to be Internal Merge. The concept of interpretability assumed here is a relative notion in that it refers to interpretability with respect to one of the two (or perhaps both) interfaces: the C-I interface and the SM interface. Being interpretable at a given interface means containing features that this interface can interpret. For example,

{DESKTOP}NEW FOLDER//9781107040847C01.3D

15

[7–22] 23.11.2013 11:50AM

Features

15

it seems plausible to assume the C-I interface can interpret features that have some semantic import; examples of such features might be features involving tense (past, present, future (and null, on some accounts)), aspect, quantification or φ-features (person, number and gender). The SM interface, on the other hand, can interpret features that are relevant for articulation. This allows us to distinguish features that are interpretable at a given interface from features which the interface cannot interpret: a feature is interpretable at a given interface if it makes a distinction between two representations at this interface.15,16 Svenonius (2007a) (see also Adger & Svenonius 2011 for relevant discussion) puts it succinctly as follows: (13) A feature F is an X feature iff F can constitute a distinction between two different X representations.(Svenonius 2007a: 375)

Crucially, a given feature is not inherently interpretable or uninterpretable. It can be interpretable in one position but not in another, or at one interface but not the other. This, in turn, depends on what lexical item the feature in question is a feature of. Perhaps the most straightforward illustration comes from the domain of φ-features. These features are typically taken to be interpretable on nouns but not interpretable on verbs, which reflects the intuition that verbs agree with nouns and not vice versa. One way to think about it is that nouns are selected from the Numeration inherently specified for these features. Verbs, however, are selected from the Numeration unspecified. The following examples from Polish, a language with grammatical gender distinctions, provide an illustration.17 In (14), the noun papuga ‘parrot’ is lexically specified as a

15

16

This view of interpretable features bears some resemblance to distinctive features, not an unwelcome result. Distinctive features in phonology are features that distinguish between two phonological representations. By analogy, we could think of features that are interpretable at the C-I interface as distinctive semantic features. Svenonius also makes the following distinction between module-internal features (for example, syntax-internal features) and interface features (for example, syntax–semantics features). Features that have no semantic import at all, such as class features, or arguably EPP features, provide good examples of syntax-internal features. (i) For any F, and any modules X and Y, a. F is an X-internal feature iff F is an X feature and not a feature of any other module. b. F is an X-Y interface feature iff F is an X feature and a Y feature. (Svenonius 2007a: 375)

17

On one view, any syntax-internal feature is an uninterpretable feature. Svenonius, however, gives uninterpretable features a narrower definition: uninterpretable features for him are those features that are not translated or interpreted in the mapping from one module to another. I use English glosses for the sake of clarity. Also, the use of actual lexical items in the Numeration is misleading. Since, in this case, the verb enters the derivation with its inflectional features unvalued, it is more accurate to think of the Numeration as consisting of feature bundles and the derivation as manipulating these feature bundles. The use of small caps is meant to capture this intuition.

{DESKTOP}NEW FOLDER//9781107040847C01.3D

16

[7–22] 23.11.2013 11:50AM

16

The Minimalist Program

third-person singular feminine noun; the presence of third-person singular feminine agreement on the verb, however, is a result of run-of-the-mill subject–verb agreement.18 In (15), on the other hand, the noun ptak ‘bird’ is specified as a thirdperson singular masculine noun and the verb agrees with it. (14) a. Papuga

siedziała parrot.3SG.FEM sit.PST.3SG.FEM ‘A parrot sat on a tree.’

na on

drzewie. tree

na on

drzewie. tree

[Polish]

b. N = {PARROTiφ[3SG.FEM], . . ., SITuφ[ ], } (15) a. Ptak

siedział bird.3SG.MASC sit.PST.3SG.MASC ‘A bird sat on a tree.’

b. N = {BIRDiφ[3SG.MASC], . . ., SITuφ[ ], }

The idea that φ-features are always interpretable on nouns has sometimes been questioned in the relevant literature. We have seen above (note 18) that number might be different from other features in this respect. Legate (2002, 2012) also raises some issues for the way Chomsky characterizes the distinction between interpretable and uninterpretable features, pointing out that it is not clear in what sense gender features in languages with non-natural gender are semantically

18

Number is different from person and gender. For most nouns, gender specification is inherent and invariant; the Polish noun książka ‘book’ is always third person feminine but it can be either singular or plural. This suggests that number might not be inherent to the noun itself in the same way person and gender are. One way to implement it is to think of Number as being uninterpretable on the noun but interpretable on a higher Number head, as shown in (i) (see Danon 2011 for a more concrete proposal).

DP

(i)

D

NumP NP

NumiNum[val]

NiPers[val] iGen[val] uNum[ ]

However, there do exist nouns that can be either feminine or masculine (such as natural gender marked nouns in (ii)), or nouns that are always singular or plural (such as so-called pluralia tantum nouns in (iii)). (ii) aktor/aktorka

‘actor/actress’

lekarz/lekarka

‘male doctor/female doctor’

pilot/pilotka

‘male pilot/female pilot’

(iii) spodnie noz·yczki

‘pants’ ‘scissors’

Since little hinges on this distinction as far as the specifics of Phase Theory are concerned, I will abstract away from it in what follows.

{DESKTOP}NEW FOLDER//9781107040847C01.3D

17

[7–22] 23.11.2013 11:50AM

Features

17

interpretable. Consider, for example, the noun book: in Russian it is feminine, in French masculine, and in German neuter, as shown in (16a–c). (16) a. eta this.3SG.FEM b. un a.3SG.MASC c. ein a.3sg.NEUT

kniga book3SG.FEM livre book.3SG.MASC Buch book.3SG.NEUT

[Russian] [French] [German]

The mismatches between syntactic number and semantic number (discussed by Munn 1999, for example) raise similar issues. A noun like ‘group’ or ‘committee’ is syntactically singular but semantically plural. Does this mean that its interpretable number feature is singular or plural? Given that interpretability in this case refers to the semantic interface, it seems that it should be plural, but then the question is why (at least in American English) it determines singular verb agreement. So perhaps a better distinction is between inherently/lexically specified features (which may or may not map into semantically relevant features but which can provide values to other features) and those that receive their specification in the course of a syntactic derivation. The idea that the same feature can be interpretable in one location but uninterpretable in another one raises the question of whether every interpretable feature has an uninterpretable counterpart and vice versa: whether every uninterpretable feature has an interpretable counterpart. For features such as φ-features, the answer to both questions seems to be yes. φ-features, as we saw above, are interpretable on nouns but uninterpretable on verbs. What about other features? Tense features (i.e. past, present, future) have to be interpretable somewhere; the question is where and what their uninterpretable counterparts are. It is generally assumed that the Complementizer–Tense complex provides tense information; this suggests that tense features are interpretable on these heads in spite of the fact that in many languages these features are realized on verbs, as pointed out by Pesetsky & Torrego (2007). If tense is interpreted on T and it ‘needs’ an uninterpretable counterpart, the tense feature on the verb comes to mind as its uninterpretable counterpart, as shown in (17). (17) TiT[PST]

walkeduT[PST]

A slightly different explanation (but not one that is incompatible with tense feature on the verb being uninterpretable) comes from Pesetsky and Torrego’s (2001) work on case. In their view, Nominative case is an instance of an uninterpretable T feature on the subject: (18) TiT[PST]

DPuT[NOM]

In a similar spirit, Svenonius (2002) reanalyzes the Accusative case feature as an uninterpretable Aspect feature (uAsp). Support for such a reanalysis comes from languages like Turkish or Finnish (see Enç 1991, Kiparsky 1998, Megerdoomian

{DESKTOP}NEW FOLDER//9781107040847C01.3D

18

[7–22] 23.11.2013 11:50AM

18

The Minimalist Program

2000, among others), in which Accusative case marks aspectual notions such as delimitedness or specificity: (19) a. Ali

[Turkish]

(20) a. Matti

[Finnish]

bir kitabı aldı. Ali one book.ACC bought ‘A book is such that Ali bought it.’ b. Ali bir kitap aldı. Ali one book bought ‘Ali bought some book or other.’ luk-i kirja-t (tunni-ssa). Matti-SG/NOM read-PST/3SG book-PL-ACC (hour-INESS) ‘Matti read the books (in an hour).’ b. Matti luk-i kirjo-j-a (tunni-n). Matti-SG/NOM read-PST/3SG book-PL-PART (hour-ACC) ‘Matti read books (for an hour).’ (Megerdoomian 2000: 316–17)

The mechanism responsible for valuing unvalued features is called Agree. Since Agree is the subject of the next section, for now all we need to know is that Agree values uninterpretable features. I use the notation in (21) to capture this distinction. Val stands for any feature value complex, empty brackets signify the lack of value, and filled brackets signify valued features. The only two possibilities are valued interpretable features and unvalued uninterpretable features; other combinations, to which I turn shortly, would be contradictions in terms. (21) a. valued/interpretable features: iF[val] b. unvalued/uninterpretable features: uF[ ]

Uninterpretable features have to be valued in the course of the derivation. If there are only two types of features, uninterpretable/unvalued and interpretable/valued, providing a value to an uninterpretable feature will make it indistinguishable from a feature that was interpretable (i.e. it had a value) to begin with. Furthermore, if unvalued features need to be deleted by the time of Transfer to the interfaces, there is no way to know which features need to be deleted, as they are all valued by the time of Transfer. This gives rise to the following paradox, discussed by Epstein & Seely (2002), Richards (2008) and Chomsky (2008), among others. Uninterpretable features cannot be interpreted by the interfaces, so they have to be deleted by the time they reach the interfaces. However, since they might have a phonetic reflex (such as the presence of subject–verb agreement as a reflex of an uninterpretable φfeature on T or overt case morphology as a reflex of an uninterpretable case feature), they must be transferred to the phonological component before they get deleted. This is not a problem, however, if feature valuation, Transfer and deletion happen at the same point during the derivation. This point in the derivation is determined by phase heads. Thus the solution to the timing paradox provides an indirect argument in favor of phases. Since we have not said much about phases yet, I will defer a more complete discussion of the

{DESKTOP}NEW FOLDER//9781107040847C01.3D

19

[7–22] 23.11.2013 11:50AM

Features

19

interaction between feature valuation, Transfer and deletion till the next chapter. To preview, we will see there that feature valuation and deletion have to happen simultaneously at the point of Transfer to the interfaces. There is thus no need for the derivation to keep track of which features entered the derivation valued and which ones did not. As pointed out by Pesetsky & Torrego, Chomsky’s reasoning for equating interpretability with valuation has to do with the fact that Narrow Syntax lacks the look-ahead property to determine whether a given feature is going to be interpretable at the interface or not. It is important to note that the timing issue alluded to here only arises if we equate uninterpretability with lack of value; if uninterpretable features remain uninterpretable throughout the derivation, the issue does not arise. Indeed, this equivalence has also been questioned in the literature. Pesetsky & Torrego (2007) and Bošković (2011), for example, argue that valuation and interpretability should be taken as independ ent of each other. This allows feature combinations that were contradictory if feature interpretability amounts to valuation. The ‘new’ possibilities are bolded in (22). (22) Types of features uF[val] iF[val] uF[ ] iF[ ]

an uninterpretable and valued feature an interpretable and valued feature an uninterpretable and unvalued feature an interpretable and unvalued feature (cf. Pesetsky & Torrego 2007)

For Pesetsky & Torrego (2007), tense on the verb is valued in the lexicon (which, they argue, is plausible given the fact that quite often tense surfaces on the verb).19 It is nevertheless uninterpretable and remains so throughout the derivation. The tense feature on the T head, on the other hand, is interpretable but unvalued: (23) a. TiT[ ] b. TiT[PST]

walkeduT[PST] walkeduT[PST]

There are thus two plausible candidates for what an uninterpretable T feature might be: a tense feature on the verb or a Nominative case feature on the noun. Furthermore, Pesetsky & Torrego argue that such a dissociation of interpretability and valuation also removes the need for two types of features on interrogative complementizer heads and wh-phrases. Complementizer heads in wh-questions are typically thought to require two types of features: the feature that marks the clause as interrogative (essentially Cheng’s (1991) typing

19

I am simplifying Pesetsky and Torrego’s system somewhat. For them, feature valuation is feature sharing of the kind proposed by Frampton and Gutmann (2000), for example. Instead of (23), they suggest (i), in which [2] is the value. The number 2 has no significance, it just marks the same feature value in both locations: (i) TenseiT[2]

walkeduT+past[2]

{DESKTOP}NEW FOLDER//9781107040847C01.3D

20

[7–22] 23.11.2013 11:50AM

20

The Minimalist Program

feature) and the feature that drives a wh-element to the specifier of the interrogative clause. The two are typically referred to as a Q-feature and a wh-feature, respectively; the complementizer has an interpretable Q feature (iQ feature) and an uninterpretable wh-feature (uwh-feature). A wh-phrase in its scope, on the other hand, has an interpretable wh-feature feature (iwh) and an uninterpretable Q feature (uQ).20 Agree, as shown in (24), ensures valuation of both unvalued features, the uninterpretable wh-feature of C and the uninterpretable Q feature of the wh-phrase. (24) a. CiQ, uwh[ ] b. C iQ, uwh[val]

WHiwh,uQ[ ] WHiwh,uQ[val]

This is a proliferation of interrogative-type features. For Pesetsky and Torrego, C has an interpretable unvalued Q feature and the wh-phrase has its uninterpretable valued counterpart. As shown in (25), there is only one type of feature (Q feature), and wh is its value. (25) a. CiQ[ ] b. CiQ[wh]

WHuQ[wh] WHuQ[wh]

1.4 Agree The discussion in the previous section relied on the assumption that Agree is an operation responsible for valuing unvalued features.21 In this section, we turn to the details of the Agree mechanism. Following the general consensus in the field, I will capitalize Agree when I use it in this technical sense and distinguish it from the descriptive terms ‘agreement’ or ‘agrees with’, used quite broadly in the literature. For Agree to be possible, the following conditions have to hold (cf. Chomsky 2000: 122–3). (26) a. The Probe and the Goal have to be active, where being active means having uninterpretable/unvalued features. THE ACTIVITY CONDITION

20

21

Some sources take C to have an interpretable wh-feature and an uninterpretable Q feature, and the wh-phrase to have an uninterpretable wh-feature and an uninterpretable Q feature. This seems to be a terminological rather than a substantive difference. Chomsky describes Agree as an operation responsible for deleting uninterpretable features under matching: (i) The erasure of uninterpretable features of probe and goal is the operation we called Agree. (Chomsky 2000: 122)

Feature deletion is distinct from feature valuation, which raises the question of whether we have been mistaken in taking Agree to be responsible for feature valuation. I do not believe so. As we will see in Section 2.6, if valuation and deletion happen simultaneously (at the phase level), there is no incompatibility here.

{DESKTOP}NEW FOLDER//9781107040847C01.3D

21

[7–22] 23.11.2013 11:50AM

Agree

21

b. The features of the Probe and Goal have to match, where matching refers to feature identity. THE MATCHING CONDITION c. The Goal has to be inside the domain of the Probe, where the domain of the Probe is its sister. THE DOMAIN CONDITION d. The Goal has to be in a local relationship, where locality is closest ccommand. THE LOCALITY CONDITION

When all these conditions are met, the unvalued features get valued (and deleted). A typical configuration satisfying all four conditions is given in (27). Agree is indicated by a dotted line (a convention I am going to employ throughout the book). Agree is typically assumed to be impossible if both the Probe and the Goal have only unvalued features.22 (27)

a. PuF[

b. AGREE

]

GiF[val], uF[

]

PuF[val] GiF[val], uF[val]

A very straightforward illustration comes from Nominative case licensing, which in current terms reduces to valuation of the uninterpretable Case feature (uC[ ], or uT[ ] in Pesetsky & Torrego’s system). Nominative case is assumed to be a reflex of the Agree relationship between a finite T (which, in current terms, means it has a valued/interpretable tense feature and an unvalued set of φfeatures) and a subject in the domain of T, which has a valued set of φ-features and an unvalued Case feature, as represented schematically in (28) for a past tense T and a third-person singular subject. Agree provides values to the uninterpretable features on both; in (28b) there are no unvalued features left. (28) a. Tuφ[ ], iT[pst] b. Tuφ[3sg.fem] iT[pst]

DPiφ[3sg.fem], uC[ ] DPiφ[3sg.fem], uC[Nom]

For Chomsky, Agree is a binary relationship, involving a single Probe and a single Goal. However, others have argued that Agree does not have to be binary and that it also can involve one Probe undergoing Agree with multiple Goals or multiple Probes undergoing Agree with a single Goal. Hiraiwa (2001) dubs this type of Agree Multiple Agree, focusing specifically on the scenarios in which a single Probe (T in this case) agrees simultaneously with two (or more) Goals (two DPs in the case at hand): (29) a. Tuφ[ ], iT[val] DPiφ[val], uC[ ] DPiφ[val], uC[ ] b. Tuφ[val] iT[val] DPiφ[val], uC[val] DPiφ[val], uC[val]

22

Some researchers allow such non-standard Agree (see Adger 2003: 169, for example).

{DESKTOP}NEW FOLDER//9781107040847C01.3D

22

22

[7–22] 23.11.2013 11:50AM

The Minimalist Program

In my own work on Parallel Merge structures (see Citko 2011b for the most complete exposition), I argued in favor of the other possibility: the possibility of a single Goal agreeing simultaneously with multiple Probes: (30) a. Tuφ[ ], iT[val] Tuφ[ ], iT[val] b. Tuφ[val], iT[val] Tuφ[val], iT[val]

DPiφ[val], uC[ ] DPiφ[val], uC[val]

Since the issue of whether Agree is a binary operation or not does not bear directly on any aspect of Phase Theory, I will not elaborate on it further here, and I refer the interested reader to the works cited above (and the references therein) for data, evidence and further discussion. The one thing about Agree that is relevant to Phase Theory concerns the locality conditions on Agree; in particular the issue of whether Agree is possible across phase boundaries. This is the issue we come back to in Section 2.4, which focuses on the Phase Impenetrability Condition.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

23

[23–57] 23.11.2013 11:49AM

2 Introducing phases

2.1

Merge over Move preference1

This chapter introduces the concept of a phase, providing both theoretical and empirical motivation for it. To the best of my knowledge, the term first appeared in Chomsky’s (2000) ‘Minimalist Inquiries’, where phases (to be more specific, lexical subarrays associated with phases) were introduced as a solution to a problem arising from the Merge over Move (MOM) principle. The concept of a phase, however, builds on many prior principles involving locality domains: cycles, barriers, islands, to name a few. Let us start by looking at the Merge over Move Principle in more detail. What it says is that, all things being equal, Merge is preferred over Move.2 Consider the following contrast involving a raising predicate be likely, modeled upon Chomsky’s (2000) (non-parrot) examples: (1) a. There are likely to be many parrots at the clay lick right now. b. *There are likely many parrots to be at the clay lick right now.

Without the MOM principle, it is impossible to explain the ungrammaticality of (1b). The crucial step in its derivation is the one illustrated in (2b), where the next step has to involve checking the EPP feature of T. (2) a. N = {there, T, are, likely, to, be, many, parrots, at, the, clay lick, right, now} b. [TP toEPP be many parrots iφ[3pl],uC[ ] at the clay lick right now]

1

2

See also Hornstein Nunes & Grohmann (2005) for a very lucid overview of the rationale behind the Merge over Move principle. The logic of the presentation here mirrors theirs. See, however, Castillo, Drury & Gohmann (2004) for a discussion of some issues MOM raises and ways to resolve these issues. Chomsky points out that MOM could be derived from more general economy principles, if Move is a more complex operation than Merge, consisting of Copy and Merge. It is not clear, however, if the logic survives when Move is treated as a variant of Merge, i.e. Internal Merge. There is no sense in which Internal Merge is more complex than External Merge.

23

{DESKTOP}NEW FOLDER//9781107040847C02.3D

24

[23–57] 23.11.2013 11:49AM

24

Introducing phases

We have a choice here: the EPP feature of T can be checked either by merging the expletive there in [Spec,TP] or by moving the DP many parrots there. However, only one of these choices leads to a convergent derivation. If we move many parrots to [Spec,TP], the derivation proceeds as follows: (3) a. Move [DP many parrots] to [Spec,TP], checking the EPP feature of T

TP DPiϕ[3pl],uC[ ] T′ many parrots T to EPP

VP

be many parrots at the clay lick right now b. Merge [TP many parrots to be at the clay lick right now] with the adjective likely

AdjP Adj likely

TP DP iϕ[3pl],uC[

many parrots

T′

]

T to EPP

VP be many parrots at the clay lick right now

c. Merge [AP likely many parrots to be at the clay lick right now] with the verb are VP V are

AdjP Adj likely

TP DPiϕ[3pl],uC[

many parrots

]

T to EPP

T′ VP

be many parrots at the clay lick right now

d. Merge [VP are likely many parrots to be at the clay lick right now] with T Move are to T Agree between T and [DP many parrots]

{DESKTOP}NEW FOLDER//9781107040847C02.3D

25

[23–57] 23.11.2013 11:49AM

Merge over Move preference

25

TP T EPP,uϕ[3pl] VP are V AdjP are Adj TP likely DPiϕ[3pl],uC[Nom] T′ many parrots T to EPP

VP

be many parrots at the clay lick right now e. Merge the expletive there with [TP are likely many parrots to be at the clay lick right now], checking the EPP feature of T

TP T′

there

TEPP,uϕ[3pl] VP are V AdjP are Adj TP likely DPiϕ[3sg],uC[Nom] T′

VP many parrots T to EPP be many parrots at the clay lick right now f. There are likely many parrots to be at the clay lick right now.

Interestingly, nothing seems to go wrong with this derivation, yet the result (example (3f)) is ungrammatical. The uninterpretable features are all checked and/or valued, the Numeration is exhausted, and locality is respected. This is where the Merge over Move (MOM) principle comes in: the step in (3a) violates MOM, as we are moving the DP many parrots to [Spec,TP] rather than merging the expletive there. If we merge the expletive there first, the derivation proceeds as schematized in (4) instead, leading to the grammatical (4f). (4) a. Merge there with [TP to be many parrots at the clay lick right now], checking the EPP feature of T

TP there

T′ toEPP

VP

be many parrotsiϕ[3pl],uC[ ] at the clay lick right now

{DESKTOP}NEW FOLDER//9781107040847C02.3D

26

[23–57] 23.11.2013 11:49AM

26

Introducing phases

b. Merge [TP there to be many parrots at the clay lick right now] with the adjective likely

AdjP Adj likely

TP there

T′

T toEPP

VP

be many parrotsiϕ[3pl],uC[ ] at the clay lick right now c. Merge [AdjP likely there to be many people present] with the verb are

VP V are

AdjP Adj likely

TP T′

there T toEPP

VP be many parrotsϕ[3pl],uC[ ] at the clay lick right now

d. Merge the [VP are likely many parrots to be at the clay lick right now] with the matrix T Agree between T and many parrots Move are to T TP Tuϕ[pl],EPP VP are V are

AdjP Adj TP likely T′ there T to EPP

VP

be [DP many parrots iϕ[3pl],uC[Nom]] at the clay lick right now

{DESKTOP}NEW FOLDER//9781107040847C02.3D

27

[23–57] 23.11.2013 11:49AM

Motivating phases

27

e. Move the expletive there to [Spec,TP], checking the EPP feature of T

TP there

T′

Tuj[3pl],EPP VP are V AdjP are Adj TP likely there T′ T to EPP

VP

be many parrots[3pl],uC[Nom] at the clay lick right now f. There are likely to be many parrots at the clay lick right now.

2.2 Motivating phases Now, with a basic grasp of the Merge over Move principle, let us look at slightly more complex examples, such as (5). (5) There is a strong likelihood that many parrots will be at the clay lick right now.

Since its Numeration, given in (6a) below, contains an expletive, the prediction is that (5) should be ungrammatical, since at the stage given in (6b), according to MOM, the Merge of there should win out over the Move of many parrots. (6) a. N = {there, is, a, strong, likelihood, that, many, parrots, will, be, at, the, clay lick, right, now}

b.

TP T′ T VP willuϕ[3sg]EPP be many parrots iϕ[3pl],uC[Nom]t be at the clay lick right now

c.

TP

there

T′

T willuϕ[3sg]EPP

VP be many parrotsiϕ[3pl],uC[Nom] be at the clay lick right now

{DESKTOP}NEW FOLDER//9781107040847C02.3D

28

[23–57] 23.11.2013 11:49AM

28

Introducing phases

If that happens, however, the derivation is going to run into problems when the matrix T is merged. Raising of there from the subject position of the embedded finite clause to the matrix subject position, which results in (7), is unmotivated; hence ruled out by economy. (7) * Therei is a strong likelihood that ti will be many parrots at the clay lick right now.

The solution to this problem relies on the intuition that at the point in the derivation when the EPP feature of the embedded T needs to get checked (which is immediately upon the Merge of T), the expletive there is not available. In ‘Minimalist Inquiries’, Chomsky implements it in terms of so-called subarrays. (8) Suppose we select LA as before . . . Suppose further that at each stage of the derivation a subset LAi is extracted, placed in active memory (the ‘work space’), and submitted to the procedure L. When LAi is exhausted, the computation may proceed if possible. Or it may return to LA and extract LAj, proceeding as before. (Chomsky 2000: 106)

Crucially, only when a given subarray has been exhausted can the next subarray be accessed, which can be thought of as a form of cyclicity in lexical access. With lexical arrays in place, the Numeration becomes a more structured object, which instead of being a set (as in (9a)), is a set of sets (as in (9b)). These subsets are the subarrays we have just learned about. (9) a. N = { a, b, c, d, e, f } b. N = { {a, b},{c, d},{e, f} }

Now, the choice between Merge and Move (and the preference for Merge) arises only if the relevant subarray contains both the expletive and the potentially movable DP. What this means for our problematic example (5) above is that the Merge over Move principle does not apply since the expletive there and the DP people are in two different subarrays, as shown in (10). (10) N = { {there, is, a, strong, likelihood}, {that, many, parrots, will, be, at, the, clay lick, right, now } }

The only way to check the EPP feature of T is thus to move many people to [Spec, TP], which will eventually lead to the grammatical string (11b). (11) a.

TP DP iϕ[3pl],uC[Nom] T′ many parrots

T EPP uϕ[3pl] VP will be many parrots at the clay lick right now

b. There is a strong likelihood that many parrots will be at the clay lick right now.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

29

[23–57] 23.11.2013 11:49AM

Phasehood properties

29

Given the focus of this book on phases, the question is: what is the relationship between subarrays (the need for which arises from problems relating to the MOM principle) and phases? In ‘Minimalist Inquiries’, this relationship is fairly straightforward. Chomsky’s characterization of phases is simply in terms of subarrays: (12) A phase of a derivation is a syntactic object derived . . . by choice of LAi. (Chomsky 2000: 106)

Such a characterization of phases only raises the question of how lexical subarrays (LAi) are chosen. Otherwise, the argument is circular: phases are determined relative to subarrays, and subarrays are determined relative to phases. To avoid such circularity, we need an independent way to characterize phases (or subarrays). This is the focus of the next section.

2.3 Phasehood properties In ‘Minimalist Inquiries’, Chomsky characterizes phases as ‘natural syntactic objects’, ‘relatively independent in terms of interface properties’ (2000: 106).3 The C-I interface requires phases to be complete from a semantic perspective. In Chomsky’s words, a phase is ‘the closest syntactic counterpart to a proposition: either a verb phrase in which all theta roles are assigned or a full clause including tense and force’ (2000: 106). Chomsky thus assumes that CPs are phases, as are transitive and unergative vPs, but TPs, unaccusative and passive vPs are not.4 What distinguishes verb phrases that are not phases from the ones that are is the fact that they lack external arguments. It is not clear, however, in what sense the argument structure of unaccusative or passive verbs is less complete than the argument structure of transitive verbs; since the external argument is not selected, there is no real sense in which it is missing. We could only make this case by comparing the two, but such cross-derivational comparisons introduce non-significant complications into the grammar (as pointed out by Epstein 2007, for example). Furthermore, such a comparison is not always available, as shown by the following contrasts: (13) a. The ship sank. b. The captain sank the ship.

3

4

Chomsky actually describes subarrays in those terms in ‘Minimalist Inquiries’, but this seems to be a better characterization of phases. A subarray is a set, so it cannot be a syntactic object that can have the semantic type of a proposition or be movable, for example. A phase, on the other hand, being a hierarchically organized object, can have these properties. Even though Chomsky does not explicitly address the phase status of verb phrases involving ditransitive verbs, it seems reasonable to group them with transitive ones. I will come back to them in Chapter 5.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

30

[23–57] 23.11.2013 11:49AM

30

Introducing phases

(14) a. The sailor died. b. *The captain died the sailor.

By analogy, one could also make a case that unergative verbs are not complete in terms of their argument structure (because they lack an internal argument which can sometimes be realized (as in (15b)), or that many transitive verbs are not complete in terms of their argument structure because they have double object counterparts, as shown in (16b). (15) a. John ran. b. John ran a race. (16) a. John baked a cake. b. John baked Mary a cake.

There are other characterizations of phases that do not rely on propositional status or completeness. I list them in (17). (17) a. Phases are convergent objects. b. Phases are objects that determine points of Transfer. c. Phase heads are the loci of uninterpretable features.

Chomsky (2000) rejects the idea of characterizing phases in terms of convergence on the grounds that it would require a considerable look-ahead, thus negating whatever advantages of reducing computational complexity phases were meant to afford. We have seen above that defining phases in terms of the interfaces is appealing. One way to make it more concrete is to take phases to be objects that determine points of Transfer to the two interfaces. As we will see in the next section, seeing phases in this light provides a very natural way to understand the so-called Phase Impenetrability Condition (PIC), a powerful locality constraint. A slightly different view to think of phases is by looking at the properties of phase heads. On this view, articulated as (17c), phase heads are more ‘powerful’, syntactically speaking, than non-phase heads. Their syntactic prowess comes from the fact that they are the loci of uninterpretable features, and as such, they trigger syntactic operations. In simple terms, they are the syntactic engines of every derivation. Chomsky is not very explicit on this point in ‘Minimalist Inquiries’, but (18) below does imply that phase heads trigger syntactic operations (and are subject to strong cyclicity). From here, it is not a huge leap to assume that phase heads are the only triggers.5

5

This is different from the issue of whether the phase head itself can be a target of operations from outside the phase (as opposed to the issue of whether the phase head can trigger any operations). The Phase Impenetrability Condition, the focus of the next section, determines how much of a given phase is accessible from the outside.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

31

[23–57] 23.11.2013 11:49AM

Phase Impenetrability Condition

31

(18) The head of a phase is inert after the phase is completed, triggering no further operations. (Chomsky 2000: 107)

The idea of phase heads being the loci of uninterpretable features is often viewed as a definitional property of phase heads. This is the view articulated explicitly by Gallego (2010), Legate (2012) and Miyagawa (2011), among others, whose formulations are given in (19a–c). (19) a. Uninterpretable features signal phase boundaries. (Gallego 2010: 151) b. A C/v possessing an unvalued φ-feature in the numeration is a phase head. (Legate 2012: 239) c. Case identifies phase heads. (Miyagawa 2011: 1273)

2.4 Phase Impenetrability Condition From an empirical perspective, perhaps one of the most important aspects of Phase Theory is the so-called Phase Impenetrability Condition (PIC), which deems a portion of the phase impenetrable (inaccessible) from outside of the phase. In this section, we first examine the various versions of PIC that have been proposed in the literature, focusing on the different empirical predictions they make. Next, we turn to the empirical coverage of PIC, particularly the issue of whether it governs all syntactic operations or only a subset thereof (i.e. only Move/Internal Merge or both Move and Agree). What counts as phase-external can vary depending on the specific formulation of PIC, and, as we will see in this section, different versions of PIC differ with respect to how long in the derivational history the interior of the phase is accessible. All of them refer to the same general configuration, given in (20a) below, in which Z and H, given in bold, are phase heads, and there is a non-phase head X between them. This configuration corresponds to a clausal structure in (20b) in which C and v are phase heads and T is not. (20) a. [ZP Z . . . [XP X [HP α [ H YP ] ] ] ] (cf. Chomsky 2001: 13) b. [CP C . . .[TP T [vP DP [v VP ] ] ] ]

Let us start with the original version of PIC, from Chomsky’s (2000) ‘Minimalist Inquiries’. (21) Phase Impenetrability Condition In phase α with head H, the domain of H is not accessible to operations outside α, only H and its edge are accessible to such operations. (Chomsky 2000: 108)

According to (21), as soon as HP is complete, the complement of H is spelled out. HP is complete when H no longer projects. Being inaccessible in this context

{DESKTOP}NEW FOLDER//9781107040847C02.3D

32

[23–57] 23.11.2013 11:49AM

32

Introducing phases

means if there are any features that are unvalued (or unchecked), they are going to remain so. The general configuration PIC refers to is shown in (22a), and applied to a more concrete clausal configuration in (22b); HP/vP is the phase (with H/v being a phase head), YP/VP is the Spell-Out domain, and H/v and its specifier constitute the phase edge. In (22a–b), as well as in the tree diagrams that follow, the spelled-out portions are faded.6 (22) a.

PHASE XP X

HP α

H′ H

YP

SPELL-OUT DOMAIN

EDGE

b.

PHASE

TP T

vP DP

v′

v

VP

SPELL-OUT DOMAIN

EDGE

Thus the only way something can move out of the Spell-Out domain is if it gets to the phase edge first. This is how the Phase Impenetrability Condition, in conjunction with (23) – which states that phase heads have the requisite features to trigger movement – forces movement out of the phase to proceed through the phase edge.7 (23) The head H of phase Ph may be assigned an EPP-feature. (Chomsky 2000: 109)

We will review the evidence in favor of movement proceeding through phase edges in Chapter 4 and, to a lesser extent, Chapter 5. For now, suffice it to note that such evidence exists, and it involves well-known diagnostics of successive cyclic movement (i.e. reconstruction, partial wh-movement, agreement, quantifier

6

7

I borrow the convention from Hornstein, Nunes Grohmann (2005). They, however, use it to mark checked features rather than spelled-out domains. The EPP feature is also sometimes referred to as the P-feature (Periphery feature) (see Chomsky 2000: 144, note 5).

{DESKTOP}NEW FOLDER//9781107040847C02.3D

33

[23–57] 23.11.2013 11:49AM

Phase Impenetrability Condition

33

float, to name a few). If, for example, a wh-phrase in a long-distance question passes successive-cyclically through the edge of vP or CP (the edge being there because of the ‘extra’ EPP (or P) feature, as per (23)), we should be able to see evidence of it having passed through that edge. This is illustrated schematically in (24). (24) [CP WH [Cʹ C . . . [vP WH [vʹ v . . . [CP WH [Cʹ C . . . [vP WH v [VP V WH ] ] ] ] ] ] ] ]

As briefly mentioned above, there are at least two major versions of the Phase Impenetrability Condition: one from Chomsky’s (2000) ‘Minimalist Inquiries’ and the other one from Chomsky’s (2001) ‘Derivation by Phase’. The two are juxtaposed in (25a–b). I will refer to the two versions as a Strong PIC and a Weak PIC, or PIC1 and PIC2, respectively (following Chomsky 2001, Müller 2004, Richards 2011, among others, who also provide a more detailed discussion of the differences between them).8 (25) a. The domain of H is not accessible to operations outside HP [emphasis mine, B.C.]; only H and its edge are accessible to such operations. STRONG PIC/PIC1 b. The domain of H is not accessible to operations at ZP [emphasis mine, B.C.]; only H and its edge are accessible to such operations. WEAK PIC/PIC2 (Chomsky 2001: 13–14)

The two definitions differ with respect to when the domain of the phase head H becomes inaccessible: as soon as HP is complete versus at the point the next phase head (Z) is merged. Thus the two versions impose different restrictions on how long (in derivational terms) the elements inside the complement domain (complement of H) can be probed. If the phase HP is directly dominated by Z (another phase head), the two versions of PIC are empirically equivalent. Both state that at the level of ZP, only the edge of H (i.e. H itself, its specifiers and adjuncts) are accessible to Z. However, the two versions of PIC diverge in their empirical consequences in configurations of the kind given in (26a) below, where there is a non-phase head between the two phase heads Z and H (and corresponding phrases), given in bold. 8

Müller (2004), however, argues in favor of a yet different type of PIC, given in (i) below, which he dubs PIC3, in which movement proceeds through the edge of every phrase (and in which every phrase is thus a phase). I take here the more restrictive (and the more standardly assumed) approach to phasehood, in which only some phrases are phases. I also refrain from digressing into a detailed comparison of PIC3 with its ‘predecessors’, I refer the interested reader to Müller’s work instead. (i) Phrase Impenetrability Condition (PIC3) The domain of a head X of a phrase XP is not accessible to operations outside XP; only X and its edge are accessible to such operations.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

34

[23–57] 23.11.2013 11:49AM

34

Introducing phases

(26) a.

XP X

HP α

PHASE 1

H′ H

YP

SPELL-OUT DOMAIN

EDGE

b.

ZP Z

PHASE 2

XP X

HP α

PHASE 1

H′ H

SPELL-OUT DOMAIN

YP

The two versions of PIC also make different predictions with respect to what X can agree with in (26a–b). Under PIC1, X cannot agree with YP (or anything inside YP), since YP is spelled out as soon as X is merged, as shown in (26a). Under PIC2, however, X can agree with YP since YP is not spelled out until Z is merged, as shown in (26b). A concrete illustration comes from the relationship between T and a direct object. According to PIC1, T cannot agree with DP, as shown in (27). According to PIC2, T can undergo such an Agree relationship, since the complement of v becomes inaccessible only at the next phase level, the CP level, as shown in (28b). The two versions of PIC do not differ in their predictions regarding Agree between T and the elements at the edge of vP (the external argument, shifted object if present, and the v head itself ). (27)

TP Tuϕ[ ]

v Agree between T and DP blocked

vP VP

SPELL-OUT DOMAIN

...DPiϕ[val],uC[

]...

{DESKTOP}NEW FOLDER//9781107040847C02.3D

35

[23–57] 23.11.2013 11:49AM

Phase Impenetrability Condition

(28) a.

35

TP Tuϕ[val]

vP

v

VP

Agree between T and DP possible

b.

… DPiϕ[val],uC[val]…

CP C

TP Tuϕ[val]

vP

v

VP

SPELL-OUT DOMAIN

...DPiϕ[val],uC[val]...

The reason to believe that this type of Agree should be possible comes from languages that allow so-called quirky Nominative objects. On the assumption that Nominative case is a result of Agree between a finite T and the object in question, VP cannot be spelled out before Agree takes place. This is thus an argument in favor of PIC2 over PIC1. Illustrative examples from Icelandic and Polish are given in (29a–b), respectively.9, 10 (29) a. Henni höfðu leiðst Þeir. her.DAT had.3PL bored.at they.NOM ‘She had found them boring.’ (Sigurðsson 2002: 692) b. Marii podobała sie¸ ta ksia˛z·ka. Maria.DAT please REFL ‘Maria liked this book.’ 9

this

[Icelandic]

[Polish]

book.NOM

It is clear that T, rather than some unorthodox form of v, values Nominative case on the object, given that in Polish it remains Nominative under negation. If v were involved in valuing the case of the (Nominative) object, we would expect Genitive case instead (so-called Genitive of Negation) under negation. This is what happens obligatorily with Accusative objects: (i) Marii Maria.DAT

nie

podobaja˛

sie¸

kwiaty/*kwiatów.

not

please

REFL

flowers.NOM/*GEN

[Polish]

‘Maria is not pleased with the flowers.’ (ii) Maria Maria.NOM

nie

lubi

kwiatów/*kwiaty.

not

likes

flowers.GEN/ACC

‘Maria does not like flowers.’ 10

Many of the examples with Dative subjects and Nominative objects involve so-called psych-verbs, which is interesting in itself but does not change the fundamental insight about case valuation (see Belletti & Rizzi 1988 for a classic early account of psych-verbs, and Pesetsky 1995 and Landau 2010, among many others, for more recent ones).

{DESKTOP}NEW FOLDER//9781107040847C02.3D

36

[23–57] 23.11.2013 11:49AM

36

Introducing phases

Abstracting away from the non-trivial issue of how Agree can take place across a Dative experiencer in such a configuration, Agree between T and DP is only possible under PIC2; under PIC1,VP is spelled out as soon as T is merged. A similar issue arises in expletive constructions; if the associate of the expletive bears Nominative case and if v in this case is also a phase head (a somewhat controversial assumption, as we will see in Chapter 5), the example in (30) also requires Agree between T and the VP internal argument, schematized in (30).11,12 (30) a. There arrived a train. b. [CP C [TP there T uφ[3sg] [vP [v [VP arrived [DP a train]iφ[3sg],uC[Nom] ] ] ] ] ]

Again, if the VPs are spelled out as soon as T is merged, the case feature on the associate of the expletive cannot be valued. Richards (2011) argues that the differences between the two versions of PIC are reducible to which subarray contains T. He contrasts two general approaches to phasehood, which he dubs the ‘subarrays approach’ versus ‘all-powerful head approach’, and argues in favor of the subarray-based one (see, however, the discussion in Section 2.2 above). Since in principle T could belong either to the subarray ‘headed’ by C or to the one ‘headed’ by v, depending on which of the two it is, the effect is either PIC1 or PIC2. (31) a. N= { {C, T},{v, V} } b. N = { {C},{T,v},{V} }

STRONG PIC/PIC1 WEAK PIC/PIC2

However, given the empirical problems associated with PIC1 (as well as the inherent circularity in defining phases in terms of subarrays and subarrays in terms of phases), it is not clear why we would ever need both versions. Richards (2011) points out that PIC1 might be preferable conceptually, as PIC2 increases the search space, thus increasing computational complexity 11

The assumption that the associate is Nominative seems problematic in light of the examples involving pronouns, such as the one in (i), in which the case of the associate is Accusative rather than Nominative. This led Belletti (1988) and Lasnik (1992) to a different proposal, relying on socalled partitive case. Interestingly though, even in such cases the associate determines agreement, as shown by the contrast between (ii) and (iii). (i) There arrived him/*he. (ii) There arrives him. (iii) There arrive them.

12

The so-called transitive expletive constructions, such as the one in (i) from Icelandic, are not a problem, since the Nominative argument is at the phase edge, as shown in (ii). (i) Ϸað there

borðuðu

sennilega

margir

jólasveinar

bjúgun.

ate

probably

many

Christmas.trolls

the.sausages

[Icelandic]

‘Many Christmas trolls probably ate the sausages.’(Bobalijik & Jonas1996: 230) (ii) [CP C [TP there T uφ[3pl] [vP [DP many Christmas trolls]iφ[3pl]uC[Nom] [v [VP eaten [DP the sausages] iφ[3pl] uC[Acc]]

]]]]

{DESKTOP}NEW FOLDER//9781107040847C02.3D

37

[23–57] 23.11.2013 11:49AM

Phase Impenetrability Condition

37

(see also Müller 2004). Richards also notes, together with Bošković (2002), Abels (2003), Matushansky (2005), Boeckx & Grohmann (2007) and Epstein (2007) among others, that there is a mismatch between what is a phase and what is transferred to the interfaces. Phases are identified by their heads, but it is the complement of the phase head that gets transferred to the interfaces. Richards’ system avoids this problem. For him, the phase conceived in terms of subarrays is the Spell-Out unit. An important question about the Phase Impenetrability Condition, which has been given a fair amount of attention in the literature on phases, is whether all syntactic mechanisms are subject to it or constrained by it. The two most relevant syntactic mechanisms (arguably, the only two) are Move/ Internal Merge and Agree. The discussion so far implies that they both have to obey PIC. However, this conclusion is not uncontroversial (see, for example, Bošković 2003, 2007, Bhatt 2005 for claims to the contrary). Bošković, based on data of the following sort, argues that Agree into finite clauses is possible, and, consequently, that Agree cannot be subject to PIC. In what follows, we will refer to this type of Agree as Long Distance Agree (LDA). (32) ənan qəlɣil¸u lənərk ə-nin-et [iŋqun Ø-rətəm ŋəv-nen-at qora-t]. [Chukchee] he-INSTR regrets-3-PL that 3SG-lost-3-PL reindeer-PL(NOM) ‘He regrets that he lost the reindeers.’ (Boškovic´ 2003: 57, citing Inèenlikèj & Nedjalkov 1973)

In (32), the matrix verb is plural. Since the only plural element in the clause is the embedded object, Bošković takes this to indicate the presence of an Agree relationship between the matrix v and the embedded object. This leaves us with the question of whether such LDA is universally possible, or subject to parametric variation. Bošković takes the latter view, attributing the difference between languages like Chukchee and ones like English to the presence or absence of φ-features on CP. If CPs have φ-features, which he takes to be the case in English, LDA is impossible (presumably, due to intervention effects, as a violation of Attract Closest). The phenomenon of LDA crossing a finite CP boundary is also attested in Tsez, another language Bošković discusses. His analysis, however, departs from Polinsky & Potsdam (2001), from whom he draws his data. (33a) illustrates what Polinsky and Potsdam dub Properly Local Agreement (PLA); (33b) illustrates the LDA pattern we are interested in here. Polinsky and Potsdam, arguing against specifier–head-based accounts of agreement in general, reanalyze LDA as involving covert topic movement. The element triggering agreement moves to the edge of the embedded clause, from which it is accessible to Agree with the matrix V-v complex. In (33a), the verb agrees with the embedded clause. In (33b), which is an example of LDA, the verb agrees with the embedded object.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

38

38

[23–57] 23.11.2013 11:49AM

Introducing phases

(33) a. enir [uža¯ magalu ba¯c’rułi] r-iyxo. [Tsez] mother [boy bread.III.ABS ate].IV IV-know The mother knows [the boy ate the bread]. ‘The mother knows the boy ate the bread.’ b. enir [uža¯ magalu ba¯c’rułi] b-iyxo. III-know Mother [boy bread.III.ABS ate] The mother knows [the boy ate the bread]. ‘The mother knows the boy ate the bread.’ (Polinsky & Potsdam 2001: 584)

Another language that allows LDA is Hindi. As in Tsez, in Hindi–Urdu long-distance agreement appears to be optional, as shown by the examples in (34a–b), which Bhatt attributes to Mahajan (1989). The difference in agreement correlates with a difference in interpretation. Following Mahajan, Bhatt takes the relevant factor to be specificity for Hindi–Urdu, whereas Polinsky and Potsdam take it to be topic-hood for Tsez. Irrespective of what the difference is, the availability of LDA in languages like Hindi–Urdu or Tsez suggests that there must be some factors that (in some languages) allow Agree not to obey PIC. (34) a. Ram-ne [rotii khaa-nii] chaah-ii. Ram-ERG bread.FEM eat-INF.FEM want-PFV.FEM.SG ‘Ram wanted to eat bread.’ b. Ram-ne [rotii khaa-naa] chaah-aa. Ram-ERG bread.FEM eat-INF.MASC want-PFV.MASC.SG ‘Ram wanted to eat bread.’ (Bhatt 2005: 761)

[Hindi–Urdu]

Hindi–Urdu is slightly different from the Chukchee and Tzez data we have just seen in that LDA is only possible into infinitival clauses, which leads Bhatt to correlate LDA with restructuring. It also leads him to dissociate case from agreement; in LDA examples, the embedded object agrees with the matrix verb but does not get case from it.13 A slightly different version of the Phase Impenetrability Condition, which also allows LDA into embedded CPs, comes from Landau’s work on control (see also witko’s 2010a, b).14 In a series of works, Landau develops a theory of control, often referred to as Agree-based Calculus of Control (see Landau 1999, 2000, 2003, 2008), in which the interpretation of PRO is determined by what it undergoes Agree

13

14

In order to account for Hindi LDA, Bhatt (2005) also reformulates Agree (dubbing it AGREE). His version of AGREE differs from the standard Agree in that it allows the Goal to have no unvalued features. In standard Agree cases, having unvalued features (uCase features in most cases we have seen in this section) was a definitional characteristic of a Goal; this was the property that made it active. This is slightly different from the issue of Agree being exempt from PIC. Here we are dealing with PIC being reformulated in a way that allows certain elements to agree with outside others their phase (without being at the phase edge).

{DESKTOP}NEW FOLDER//9781107040847C02.3D

39

[23–57] 23.11.2013 11:49AM

Phase Impenetrability Condition

39

with. In Landau’s system, the relationship between PRO and its controller is mediated via Agree, but, crucially, not necessarily Agree between PRO and its controller. Rather, he proposes two mechanisms of control, which correlate with the distinction between Partial and Exhaustive control.15 The difference between the two types of control, which Landau convincingly argues are both subspecies of Obligatory Control, is illustrated by the contrast between the example in (35a), illustrating Partial Control, and the one in (36a), illustrating Exhaustive Control. The former allows a mismatch in plurality between a (singular) controller (i.e. matrix subject) and a plural embedded PRO. The plurality of PRO, which Landau indicates by the subscript 1+, is evidenced by the fact that it is possible with distributive predicates like gather, meet or do something together. Both Exhaustive and Partial Control involve a number of Agree relationships. Interestingly though, neither involves a direct Agree relationship between the controller and PRO. In Partial Control cases, such as the one in (35a), none of the three Agree relationships violates PIC. (35) a. Maria1 preferred [PRO1+ to go for a walk together].

b.

TP DP

T′ vP

T

v′

Maria v

VP

preferred

CP C

TP

T C PRO

T′ T

vP

PRO go for a walk together

However, the Agree Landau posits for Exhaustive Control is different and it does involve Agree across a CP boundary, in violation of the Phase Impenetrability Condition. This is the Agree between the matrix T and the PRO in the embedded [Spec,TP] position.

15

The distinction between Exhaustive and Partial Control is a matter of lexical selection. Some verbs are inherently specified as selecting Partial Control complements, whereas others are specified as selecting Exhaustive Control complements.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

40

[23–57] 23.11.2013 11:49AM

40

Introducing phases

(36) a. Maria1 managed [PRO1 to go for a walk].

b.

TP T′

DP T1

vP v′

Maria

VP

v managed

CP C

TP T′

PRO T

vP PRO go for a walk

On standard assumptions about PIC, as soon as v is merged, the embedded TP is spelled out, the result of which is that anything inside this TP will not be accessible to further operations.16, 17 vP

(37)

v

VP managed

CP

C

TP T′

PRO T

vP PRO go for walk

Thus, in order to allow Agree between T and PRO in such cases, Landau modifies the Phase Impenetrability Condition. He does not take the route Bošković took (essentially exempting Agree from PIC). Instead, he allows Agree into the complement domain under certain circumstances, namely when the Goal in this

16

17

The only difference between subject and object control in both Exhaustive and Partial Control cases is that the first Agree relationship is between v and the object (rather than T and the subject). As far as PIC is concerned, the same problem arises: the Agree between v and PRO in Partial Control cases also violates it. Or even earlier if we assume the earlier version of PIC, on which the complement of a phase head becomes inaccessible as soon as the next head is merged, which in this case is the matrix V.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

41

[23–57] 23.11.2013 11:49AM

Multiple Spell-Out

41

Agree operation is interpretable (or more accurately, has interpretable features). Landau’s version of PIC is given in (38). (38) In a structure [. . .X. . . [YP . . .Z. . .] ], where YP is the only phase boundary between X and Z, Z is accessible to X. a. only at the head or edge of YP, if Z is uninterpretable; b. anywhere in the YP phase, if Z is interpretable. (Landau 2000: 69)

The possibility of Agree not being subject to the PIC raises a very fundamental question concerning the relationship between Transfer and PIC. Implicit in the discussion so far has been the assumption that the reason the complement of a phase head becomes inaccessible at a certain point in the derivation (i.e. when the next phase head is merged) is the fact that it has been transferred to the interfaces. From this perspective, Agree that violates PIC, which would be Agree after Transfer, should never be possible. But is the conclusion that PIC determines Transfer a logical necessity? Could it be that PIC determines locality but not Transfer? For the proposals we discussed in this section, PIC determines both Transfer and locality. This, however, is only one possibility. Another possibility would be to assume that in certain cases (such as the cases of LDA considered here), Transfer to the interfaces is simply delayed. This would keep the concept of Transfer intact, in that once an object is transferred, it is no longer visible to syntactic processes such as Merge, Move or Agree, but would allow delayed Transfer, the possibility we will discuss in more detail in Chapter 6.18

2.5 Multiple Spell-Out The concept of Multiple Spell-Out was implicit in the discussion of the Phase Impenetrability Condition in the previous section, where we saw that different versions of PIC make different predictions with respect to how long in the derivation the complement of a phase head is accessible. The most straightforward explanation for why this should be the case is that at a given point, this complement is not accessible to the syntax because it is no longer part of the syntax; it is transferred to the two interfaces. The technical term used to describe this moment is Transfer; however, it is often referred to as SpellOut, even though – technically speaking – Spell-Out refers only to transfer to the SM interface. To the best of my knowledge, there is no analogous term to describe transfer to the C-I interface. Interpret-Out is a term that comes to mind, but, admittedly, it does not have the same intuitive appeal as Spell-Out. I will thus follow the field and perpetuate the misconception in referring to Transfer as Spell-Out. This terminological note aside, it is clear that given PIC and the

18

This is essentially what Richards (2012) does by taking Spell-Out to occur when the complement of a phase head contains only interpretable features.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

42

[23–57] 23.11.2013 11:49AM

42

Introducing phases

fact that (most) derivations contain more than one phase, Spell-Out is going to occur more than once per derivation. In this respect, phase-theoretical syntax constitutes a major departure from its predecessors (minimalist and preminimalist alike), where there was only one point at which syntax could ‘communicate’ with phonology and semantics. The difference is illustrated schematically in (39) and (40). (39)

SINGLE SPELL-OUT

C-I

SM

(40)

MULTIPLE SPELL-OUT

C-I

SM

C-I

SM

C-I

SM

Logically speaking, the Transfer to each interface could happen at different times (the Non-Simultaneous Spell-Out of Marušič 2005, for example), as shown below.19 (41)

SM C-I SM C-I SM C-I

A variant of the general non-simultaneous approach would be to allow Multiple Spell-Out to apply only to one of the two interfaces, as illustrated in (42a) and (42b), respectively. (42) a.

SM SM SM

C-I 19

Since there is only one point of Transfer under the so-called inverted T (or Y) model illustrated in (39), non-simultaneous Transfer was not an option.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

43

[23–57] 23.11.2013 11:49AM

Multiple Spell-Out

43

b. C-I C-I C-I

SM

Even though Chomsky leans towards maintaining the LF and PF integrity of phases, there is nothing that would exclude such non-simultaneous scenarios on principled grounds. Since this is nevertheless not something that is standardly assumed, I will abstract away from Non-Simultaneous Spell-Out for now, and come back to it in Chapter 6. The concept of Multiple Spell-Out goes back to Uriagereka (1999) (see, however, Uriagereka 2012 for a fuller exposition). The details of Uriagereka’s Multiple Spell-Out, however, are different from the Multiple Spell-Out that became standard in Phase Theory, in which phase heads determine when SpellOut takes place, and the complement of a phase head is what gets spelled out. For Uriagereka, non-complements are spelled out first and become ‘giant lexical compounds’ (Uriagereka 1999: 256), whose terms become unavailable to further syntactic operations. For example, specifiers are spelled out before complements, as shown schematically in (43). This can explain, for example, why extraction from subjects is more constrained than extraction from complements. If the specifier is spelled out, nothing can move out of it.20 XP

(43)

YP Y

X′ ZP X

WP

With the concept of Multiple Spell-Out in mind, let us look at how the theoretical pieces we have seen so far come together. The example we will look at is given in (44a); its Numeration, given in (44b), contains two phase heads, indicated in bold. This means that that there are two subarrays, each containing one phase head. If DP is a phase – a possibility which I abstract away from until Chapter 4 – the Numeration will be (44c) instead. (44) a. Parrots like nuts. b. N = { {C1, T1}, {Parrots1, v1, like1, D2, nuts1} } c. N = { {C1, T1}, {D1, parrots1}, {v1, like1}, {D1, nuts1} }

20

It also helps reconcile Bare Phrase Structure with Kayne’s (1994) Linear Correspondence Axiom. If the specifier becomes one big ‘lexical item’, it is not an issue that its subparts are not ordered with respect to the elements in the rest of the structure.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

44

[23–57] 23.11.2013 11:49AM

44

Introducing phases

The relevant features are given in (45), and the derivation proceeds along the lines given in (46): (45) CiForce[Decl] TiT[pres], uφ[ ],EPP DuC[ ] parrotsiφ[3pl] vuφ[ ] nutsiφ[3pl] (46) a. Merge D with nuts

DPiϕ[3pl], uC[ D uC[

]

nutsiϕ[3pl]

]

b. Merge like with [DP nuts]

VP like

DPiϕ[3pl], uC[ DuC[ ]

]

nutsiϕ[3pl]

c. Merge VP with v Agree between v and DP21 vP

vuϕ[3pl]

VP

like

DPiϕ[3pl]uC[Acc] DuC[Acc]

nutsiϕ[3pl]

d. Move (Internal Merge) like to v

v v like vuϕ[3pl] like

V DPiϕ[3pl], uC[Acc] DuC[Acc]

21

nutsiϕ[3pl]

The Agree between v and the direct object results in v getting its φ-features valued by the object. This Agree is realized overtly in languages in which verbs agree with both subjects and objects.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

45

[23–57] 23.11.2013 11:49AM

Multiple Spell-Out

45

e. Merge [NP parrots] with D

DPiϕ[3pl], uC[ DuC[

]

]

parrotsiϕ[3pl]

f. Merge [DP parrots] with [vP like nuts]

vP DPuC[

v′

],iϕ[3pl]

VP

D parrots v like vuϕ[3pl] like

DPiϕ[3pl], uC[Acc] DuC[Acc]

nutsiϕ[3pl]

g. Merge T with vP Agree between T and [DP parrots] TP

TEPP,uϕ[3pl]

vP

DPuC[Nom],iϕ[3pl] D

v′

parrots v

VP

like vuϕ[3pl] like

DPiϕ[3pl],uC[Acc] DuC[Acc]

nutsiϕ[3pl]

h. Move (Internal Merge) [DP parrots] to [Spec, TP], checking the EPP feature of T

TP T′

DPuC[Nom],iϕ[3pl]

DuC[Nom] parrotsiϕ[3pl] TEPP,uϕ[3pl] vP DP D

v′

parrots v like vuϕ[3pl] like

VP DPiϕ[3pl],uC[Acc] DuC[Acc]

nutsiϕ[3pl]

{DESKTOP}NEW FOLDER//9781107040847C02.3D

46

[23–57] 23.11.2013 11:49AM

46

Introducing phases

i. Merge C with TP, Transfer VP to SM and C-I

CP C

TP DPuC[Nom],iϕ[3pl]

T′

DuC[Nom] parrotsiϕ[3pl] TEPP,uϕ[3pl]

vP v′

DP D

parrots v

VP DPiϕ[3pl], uC[Acc]

like vuϕ[3pl] like D

nuts

j. Transfer CP to SM and C-I

2.6 Feature Inheritance This section illustrates perhaps the most straightforward (and perhaps expected, given the discussion in the previous sections) implementation of the idea that phase heads are the driving force behind each derivation. One way to implement it is to assume that only phase heads are the locus of uninterpretable features. This might not seem like a big departure from what we have seen so far, but it does have one important consequence when taken at face value. If all uninterpretable features start their derivational lives on phase heads, non-phase heads can either have no uninterpretable features whatsoever, or they can acquire uninterpretable features from phase heads in the course of the derivation. The latter scenario is what Feature Inheritance (FI) refers to. Generally speaking, the idea is that phase heads have uninterpretable features but non-phase heads can inherit them in the course of the derivation. This, in addition to maintaining the intuition that the ability to host uninterpretable features is a prerogative of phase heads, avoids the conclusion that T is a phase head, or that T has no uninterpretable features at all. Either conclusion about T is both empirically and conceptually problematic. T heads, which we normally do not think of as phase heads, must have the uninterpretable φ-features that allow them to undergo Agree with the subjects in their domains. T heads are also assumed to have EPP features. The EPP feature is an odd feature in the taxonomy of features we have so far, which takes features to be either interpretable or uninterpretable; EPP functions as an uninterpretable feature but it has no interpretable counterpart. One possibility is to assume that EPP is the only uninterpretable/formal feature that has no place on the valued/unvalued spectrum. An alternative is

{DESKTOP}NEW FOLDER//9781107040847C02.3D

47

[23–57] 23.11.2013 11:49AM

Feature Inheritance

47

to assume that it is not the same kind of feature as all other features, and is more of a second-order feature in the Adger & Svenonius (2011) sense. If we treat EPP as a second-order feature, it becomes less of a mystery that it lacks an interpretable counterpart.22 Taking T to be a phase head instead would also cause both conceptual and empirical problems. Conceptually, TPs are not complete or independent at the interfaces in the same sense that CPs are. Empirically, if TP were a phase, we would encounter similar problems to the ones we encountered in the previous section with the earlier version of PIC, so-called PIC1. If T were a phase head, however, the same problems would arise even with PIC2 in place. In particular, Agree between T and a VP-internal object (which is what is needed in (47a)) would never be possible: (47) a. Henni höfðu leiðst Þeir. her.DAT had.3PL bored.at they.NOM ‘She had found them boring.’ (Sigurðsson 2002: 692)

b.

[Icelandic]

TP Tuϕ[

]

v

Agree between T and DP blocked

vP VP ...DPiϕ[val], uC[ ]...

If T inherits its uninterpretable φ-features from C, we can reconcile the idea that phase heads are hosts of uninterpretable features with the idea that T undergoes Agree with the subject (and sometimes with the object), thus valuing the unvalued/uninterpretable features of both. A corollary of the claim that phase heads have uninterpretable features is the claim that derivations are evaluated at phase levels. Independent evidence in favor of this claim comes from locality considerations. The idea that locality is evaluated at phase levels eliminates the need for equidistance, which arose from a locality violation in run-of-the-mill wh-questions. A simple formulation of equidistance is given in (48).23 22

23

Adger (2003) treats EPP as a D-feature. This is a bit of an oversimplification given that other elements besides DPs can check it. The original formulation of the equidistance principle (from The Minimalist Program) is given below: (i) If α, β are in the same minimal domain, they are equidistant from γ. (Chomsky 1995: 184)

See Chomsky (1995: 177–80) for a detailed explanation of what counts as a minimal domain.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

48

[23–57] 23.11.2013 11:49AM

48

Introducing phases

(48) Terms of the edge of HP are equidistant from probe P.

(Chomsky 2001: 27)

At the point in the derivation illustrated in (49b), Agree between T and the subject is blocked by the intervening wh-object. The movement of the subject over the intervening wh-object is also blocked by its presence. Equidistance, which made specifiers of the same head equidistant from any higher Goal, was essentially a stipulation to allow such movement (or Agree) to take place. (49) a. What do parrots like?

b.

CP

C

TP TEPP,uϕ[ ]

vP

whatiϕ[3sg], uC[Acc] Agree between T and subject blocked

parrotsuC[

vP ],iϕ[3pl]

v′

vEPP

VP

like

what

If locality is evaluated at the T level (as is unavoidable if T is a phase), and if equidistance is not available, Agree between T and the subject is not possible. This Agree, however, is not a locality violation if locality is evaluated at the C level and, more generally, if operations happen simultaneously at the phase level. What this means is that Agree between T and the subject, movement of the subject to [Spec,TP], and movement of the object from [Spec,vP] to [Spec,CP] happen simultaneously at the phase level.24 Thus there is no sense in which the subject moves across the object, or T undergoes Agree with the subject across the object. The copy of the wh-object does not intervene because only entire chains can intervene, and in (50) only one link of the chain does. This argument from intervention also leads Chomsky (2004) to conclude that

24

The equidistance principle was set against a different theoretical backdrop, with Agreement projections still in place, and before the emergence of concepts like Probe, Goal and Agree. The basic logic behind it has not changed, though. The idea that all operations take place simultaneously at the phase level also removes the countercyclicity in Holmberg’s Generalization; the verb had to move to T before the object could undergo object shift. Chomsky (2001) addresses the countercyclic nature of head movement, which together with other problematic aspects of head movement (such as the fact that the moved head does not c-command its trace and is subject to different locality conditions than other types of movement) led him to suggest that head movement is phonological.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

49

[23–57] 23.11.2013 11:49AM

Feature Inheritance

49

wh-phrases cannot remain in [Spec,vP] and that the landing site of overt object shift cannot be [Spec,vP], either.25 (50)

CP C′

CiForce[Q], EPP TP T′

Tiϕ[3sg], uC[Acc] vP x whatiϕ[3sg], uC[Acc] vP parrotsuC[Nom ],iϕ[3pl]

v′

vEPP

VP like

what

Feature Inheritance thus provides a concrete way of implementing the idea that operations happen simultaneously at the phase level. If uninterpretable φ-features of T start out on C and get inherited by T from C, there is simply no way for T to have these features before C is merged. This is what Chomsky (2008) proposes. Richards (2008) argues that Feature Inheritance is obligatory and deduces its obligatoriness (as stated in (51)) from the interplay of the two independently motivated principles, Value-Transfer Simultaneity (given in (52a)) and the formulation of the Phase Impenetrability Condition (given in (52b)).26 (51) Feature Inheritance uF must spread from edge to nonedge (i.e. from C to T, v* to V, etc.) (Richards 2008: 567) (52) a. Value-Transfer Simultaneity Value and Transfer of uFs must happen together. b. Phase Impenetrability Condition The edge and nonedge (complement) of a phase are transferred separately. (Richards 2008: 566–8)

25

26

Chomsky states explicitly that ‘no clear case of stranding in Spec-v is known’ (Chomsky 2004: 123). This conclusion might be too strong, in light of the data to be discussed in Section 4.2 of Chapter 4, which does seem to suggest that stranding sometimes is possible in the intermediate [Spec,vP] position. Strictly speaking, (52b) is a consequence of PIC not PIC itself.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

50

50

[23–57] 23.11.2013 11:49AM

Introducing phases

First, Richards argues that feature valuation and Transfer have to happen at the same time. The relative timing of these two operations is something that we have not been explicit about so far. Uninterpretable features (being uninterpretable, thus by definition not readable by the interfaces) have to be deleted by the time of Transfer. However, if feature interpretability reduces to feature valuation, with uninterpretable features being simply unvalued (and interpretable features valued), unvalued features that were valued during the derivation via Agree become indistinguishable from features that were valued to begin with. In essence, uninterpretable features become indistinguishable from interpretable features. Thus, once they are valued there is no way to know that they have to be deleted (unless they come with ‘I acquired my value’ versus ‘I was born with my value’ tags, a somewhat speculative solution, to say the least). The problem disappears if feature valuation and feature deletion happen simultaneously.27 This is precisely what Feature Inheritance allows. There are nevertheless some nontrivial questions that Feature Inheritance raises, listed in (53). (53) a. Do all uninterpretable features have to be inherited? b. Can other features (besides φ-features) be inherited? c. Can a non-phase head inherit its uninterpretable features from a phase head below it?

Interpretable features, being inherently associated with lexical elements, do not get inherited. Given the logic above, there is also no reason why interpretable features would need to be inherited; they never get deleted. While there is a growing consensus in minimalist theory that non-phase heads (such as T or V) lack formal features and can only get them from phase heads via the process of Feature Inheritance (see Chomsky 2008, Richards 2004, 2008, 2011, among others), the precise mechanism behind Feature Inheritance remains somewhat elusive. Given Richards’ reasoning above about the timing of feature valuation, Transfer and deletion, the only possible answer to (53a) is yes. However, there is a growing body of evidence showing that under certain circumstances φ-features can remain on C, or wh-features can be inherited by T, which suggests that Feature Inheritance perhaps does not have to be obligatory. Well-known cases of complementizer agreement of the kind found in Germanic or Bantu languages (discussed by Carstens 2005, Carstens & Diercks 2011, Diercks 2011, Haegeman 1992, among many others) has been taken to mean that φ-features can remain on C (or be present on both C and T). In (54a) the

27

This is also a point Epstein & Seely (2002) make. The conclusion they draw is different. In order to ensure that feature valuation and Spell-Out happen simultaneously, they propose Spell-Out happens whenever feature valuation happens.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

51

[23–57] 23.11.2013 11:49AM

Feature Inheritance

51

complementizer agrees with the embedded subject, in (54b) it agrees with the matrix subject, and in (54c) the verb moved to C agrees with the wh-phrase in the specifier of CP. (54) a. Kpeinzen dan-k (ik) morgen goan. [West Flemish] I.think that-I (I) tomorrow go ‘I think that I’ll go tomorrow.’ (Carstens 2005: 222, citing Haegeman 1992) b. Alfredi ka-bol-el-a baba-ndu a-li ba-kha-khil-e. [Lubukusu] 1Alfred 1S-said-AP-FV 2-person 1-that 2S-FUT-conquer ‘Alfred told the people that they will win.’ (Diercks 2011: 1) c. Bikí bi-á-kás-íl-é bábo bíkulu mwámí mu-mwílo? [Kilega] 8what 8CA-A-give-PERF-FV 2that 2woman 1chief 18–3village ‘What did those women give the chief in the village?’ (Carstens 2005: 220)

A somewhat lesser known case of complementizer agreement comes from Slavic languages. The Polish example in (55) shows that the complementizer agrees with the embedded pro subject in Number and Person features.28 Chce¸ z·eby-s ´cie skon´czyli. want.1SG that-2PL finish.PST.PL.MASC ‘I want you to finish.’

(55)

28

The complementizer agreement is different from the so-called floating inflection (discussed by Booij & Rubach 1987, Embick 1995, Franks & Bański 1999, among many others), illustrated in (i). (i) My-s´my we-AUX.1PL

wyjechali

z

Wrocławia

rano.

leave.PART.VIR.PL

from

Wrocław

morning

‘We left Wrocław in the morning.’ (Migdalski 2006: 231)

Unlike floating inflection, complementizer agreement is obligatory. Compare the grammatical unfloated variant of (i) in (ii) with the ungrammatical unfloated variant of (55) in (iii). (ii) Rano morning

wyjechali-s ´my

z

Wrocławia

leave. PART.VIR.PL-AUX.1PL

from

Wrocław

‘We left Wrocław in the morning.’ (Migdalski 2006: 230) (iii) *Chce¸ z·eby skon´czyli-s ´cie. want.1SG

that

finish.PST.PL.MASC

‘I want you to finish.’

Second, not only complementizers can host floating inflection: (iv) Ty

to

widziałes ´

you

it

saw

(v) Tys´

to

widział.

(iii) Ty

tos ´

widział

this

saw

you

‘You saw it.’ (Embick 1995, citing Dogil 1987)

{DESKTOP}NEW FOLDER//9781107040847C02.3D

52

[23–57] 23.11.2013 11:49AM

52

Introducing phases

A slightly different argument in favor of C being able to keep its φ-features comes from the work of Ouali (2008) on Berber. Ouali (2008) discusses three logical possibilities concerning the distribution of uninterpretable φ-features on C and T heads and argues that all three are attested. To use his terminology, C can donate its φ-features to T (which is the standard scenario), keep its features or share its features with T. The three possibilities are illustrated in (56a–c). (56) a. b. c.

DONATE: SHARE: KEEP

C Cuφ Cuφ

Tuφ Tuφ T

The latter two options are the ones that are problematic if Feature Inheritance is obligatory.29 Ouali attributes the so-called anti-agreement effect in Tamazight Berber, illustrated in (57) below, to the ability of C to not pass the uninterpretable φ-features to T (the KEEP option given in (56c)). The anti-agreement effect refers to the fact that in examples involving subject extraction, subject–verb agreement is blocked. The data below provide a concrete illustration. (57a) is the base example showing that verbs typically agree with subjects, (57b) shows that this is impossible if the subject is wh-moved, and (57c) shows that in such cases the non-agreeing participial form surfaces (hence the term antiagreement). (57) a. thamttut thʕla araw. woman 3SF.see.PERF boys ‘The woman saw the boys.’ b. *mani thamttut ag thʕla araw. which woman COMP 3SF.see.PERF boys ‘Which woman saw the boys?’ c. mani thamttut ag ʕlan araw. which woman COMP see.PERF.PART boys ‘Which woman saw the boys?’ (Ouali 2008: 164)

[Tamazight Berber]

Ouali argues that in (57b) T does not inherit any φ-features from C, and, as a result, subject–verb agreement is impossible. Another plausible case of C keeping its uninterpretable φ-features comes from Legate’s (2011) analysis of subject-initial orders in V2 languages. Legate refers to it as under-inheritance, but it is equivalent to Ouali’s KEEP option. What is interesting about subject-initial V2 clauses is that the subject behaves as if it were in an A (rather than A-bar) position, in spite of the fact that it occupies the [Spec,CP] position. The evidence that the [Spec,CP] subject position is an Aposition comes from reconstruction effects. Legate presents the following contrast from Dutch, due to Craenenbroeck & Haegeman (2007), to show that the

29

Ouali also argues that the three options are ordered with respect to each other, but nothing here hinges on this argument.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

53

[23–57] 23.11.2013 11:49AM

Feature Inheritance

53

movement of the subject bleeds Principle C reconstruction. (58a) constitutes a Principle C violation if the pronoun t ‘it’ is coindexed with the R-expression lemmeken ‘lamb’. Such coindexation is possible in (58b), which shows that the moved subject does not have to reconstruct. (58) a. Nou ein-t*i/j den aaigeneir van t lemmekeni zelf muutn now has-it the owner of the lamb himself must doewtuun. kill ‘Now the owner of the lamb has had to kill it (not the lamb) himself.’ b. Den aaigeneir van t lemmekeni ein-ti/j zelf muutn doewtuun. the owner of the lamb has-it himself must kill ‘The owner of the lamb has had to kill it (possibly the lamb) himself.’ (Craenenbroeck & Haegeman 2007: 173)

Furthermore, Legate derives under-inheritance from a form of economy: the sequence of operations in which C first donates its φ-features to T, and the subject raises to [Spec,TP] only to subsequently raise to [Spec,CP] is less economical than the scenario in which T never inherits φ-features from C and the subject raises to [Spec,CP] directly.30 The prediction that feature valuation and feature deletion (therefore Transfer, if deletion is part of Transfer) have to happen simultaneously and, consequently, that feature deletion is obligatory (see the discussion of Richards 2008 above) only arises if feature interpretability reduces to feature valuation. However, as we saw in Chapter 1, this equivalence has been questioned in the literature (see Bošković 2011, Pesetsky & Torrego 2007, among others). If we abandon the equivalence between uninterpretable and unvalued features, and allow uninterpretable features to remain uninterpretable throughout the derivation and thus be marked for deletion with no further stipulations, the need for obligatory Feature Inheritance disappears. The cases of Feature Inheritance discussed so far involve a lower non-phase head inheriting uninterpretable features from a phase head above it. This is a countercyclic process, but perhaps the countercyclicity does not matter so much if derivations are evaluated at phase levels. It does, however, raise the question of whether in addition to this type of Feature Inheritance, which we might call Feature Inheritance from Above (FIFA), the standard cases of a non-phase head inheriting its features from a higher phase head, schematized in (59a–b)), there exists Feature Inheritance from Below (FIFB), by means of which a nonphase head can inherit its features from a lower phase head (matrix V from 30

Interestingly, as pointed out by Richards (2011), PIC2 (the weaker version) is ‘informulable’ with Feature Inheritance in place. If the Merge of C is what triggers the Spell-Out of VP, the Nominative object becomes inaccessible when C is merged. However, T does not inherit its uninterpretable φ-features, which are the features that allow it to undergo Agree with the object, until C is merged.

{DESKTOP}NEW FOLDER//9781107040847C02.3D

54

[23–57] 23.11.2013 11:49AM

54

Introducing phases

embedded C or T from v, as in (60a–b)), yielding the following four possibilities, with arrows indicating directionality of Feature Inheritance. (59)

a.

CP C

TP

FIFA

Tuϕ b.

vP VP

v

Vuϕ

(60)

a.

VP

FIFB

CP

Vuϕ

C

b.

TP Tuϕ

vP v

The question now is whether the two FIFB options in (60) are theoretically possible and empirically justified. There does not seem to be anything that excludes them on theoretical grounds; if anything, they seem conceptually more plausible if we pursue the analogy between Feature Inheritance and movement. The FIFA options in (59a) and (59b) would be the options corresponding to lowering movement operations and the FIFB options would correspond to standard (upwards) movement operations. Nevertheless, let me end with a piece of speculation concerning the potential empirical support for (60a). Perhaps this is what happens in Subject-to-Object Raising (SOR) / Exceptional Case Marking (ECM) configurations, where the matrix V inherits uninterpretable features from the embedded C, in addition to inheriting them from the matrix v. The latter option was suggested by Chomsky (2008). On either account, SOR targets [Spec,VP]; however, FIFB might provide a new take on the old puzzle of why some languages allow SOR and others, such as Polish (as shown in (61a–b)), do not (see Bošković 1997, Dziwirek 2000, Kayne 1981, Lasnik 1998, Rooryck 1997, among others, for a discussion of crosslinguistic variation with respect to SOR).

{DESKTOP}NEW FOLDER//9781107040847C02.3D

55

[23–57] 23.11.2013 11:49AM

Feature Inheritance

(61) a. *Maria uwaz·a Jana byc´ ma˛drym. Maria.NOM considers Jan.ACC be.INF smart.INSTR ‘Maria considers Jan to be smart.’ b. *Maria spodziewa sie¸ Jana byc´ na czas. Maria expects REFL Jan.ACC be.INF on time ‘Maria expects Jan to be on time.’

55

[Polish]

This could only work on an independent assumption that control and ECM clauses are categorically non-distinct and both are CPs (see, for example, Bošković 1997 and Pesetsky 1991). This seems problematic given the defective tense of ECM clauses (see Stowell 1981 and Martin 1996, among others). But maybe this defective nature of CP is what motivates FIFB. This is not what happens in control clauses, whose CPs bear irrealis tense. Perhaps more generally, this is impossible in Slavic languages, whose CPs are more temporally independent than their English counterparts, as evidenced, for example, by the lack of Sequence of Tense (SOT) (see Khomitsevich 2007 for a recent discussion of SOT in Russian, and Lasnik 1998 on the connection between tense and the lack of ECM, also in Russian). Let me conclude this chapter with a concrete illustration of how exactly Feature Inheritance affects Narrow Syntax. In particular, let us look at how the derivation of the example (44) above, repeated below as (62a), changes with Feature Inheritance in place. There are two cases of Feature Inheritance to consider: T inherits its uninterpretable φ-features from C, and V inherits its uninterpretable φ-features from v. What is significant is that a number of processes happen simultaneously at phase levels (the steps (63c) and (63h)). These are: Feature Inheritance, Agree (and feature valuation), movement and Transfer. (62) a. Parrots like nuts. b. N = { {C1, T1}, {parrots1, v1, like1, D2, nuts1} } (63) a. Merge D with nuts

DPiϕ[3pl], uC[ DuC[

]

nutsiϕ[3pl]

]

b. Merge like with [DP the nuts] VP like

DPiϕ[3pl], uC[

DuC[ ]

]

nutsiϕ[3pl]

c. Merge [VP like nuts] with v V inherits uφ-features from v

{DESKTOP}NEW FOLDER//9781107040847C02.3D

56

[23–57] 23.11.2013 11:49AM

56

Introducing phases

Agree between V and DP

vP vuϕ[

VP

]

likeuϕ[3pl]

DPij[3pl], uC[Acc]

DuC[Acc]

nutsij[3pl]

d. Move (Internal Merge) like to v

v v

V

v likeuϕ[3pl]

like

DPiϕ[3pl], uC[Acc]

DuC[Acc]

nutsiϕ[3pl]

e. Merge [NP parrots] with D

DPiϕ[3pl], uC[ DuC[

]

parrotsiϕ[3pl]

]

f. Merge [DP parrots] with [V P like nuts] vP DPuC[

],iϕ[3pl]

D parrots

v′ VP

v

likeuϕ,[3pl] v like

DPiϕ[3pl], uC[Acc] DuC[Acc] nutsiϕ[3pl]

g. Merge T with vP

TP T

vP

DPuC[ D

],iϕ[3pl]

v′

parrots v

like v

VP like

DPiϕ[3pl], uC[Acc] D

nuts

{DESKTOP}NEW FOLDER//9781107040847C02.3D

57

[23–57] 23.11.2013 11:49AM

Feature Inheritance

57

h. Merge C with TP, Transfer VP to PF and LF T inherits uφ-features and the EPP feature from C31 Agree between T and [DP parrots] Move (Internal Merge) [DP parrots] to the specifier of TP DP checks the EPP feature of T

CP Cuϕ[3pl]

TP

DPuC[Nom],iϕ[3pl] T′ D parrots

TEPP,uϕ[3pl] vP DPuC[Nom],iϕ[3pl] v′

D

parrots v

VP DPiϕ[3pl], uC[Acc]

like vuϕ:val like D

nuts

j. Transfer CP to PF and LF

In sum, we have seen the motivation for phases stemming from the Merge over Move principle, as well as reviewed the existing characterizations of phases (e.g. phases ‘defined’ in terms of subarrays versus phases defined in terms of the properties of phase heads). We have also seen how the introduction of phases changes the overall architecture of the grammar – in that it results in Multiple (and potentially Non-Simultaneous) Spell-Out – and how the property of phase heads as the loci of uninterpretable features leads to Feature Inheritance. Throughout this chapter, we have assumed that certain types of vPs and CPs (and perhaps DPs) are phases, without providing much evidence for this assumption, and without applying concrete diagnostics to these categories. This is what we turn to in the next two chapters. In Chapter 3, we discuss various diagnostics that have been suggested in the literature to determine the phasehood of a given category, with an eye towards determining which of these diagnostics are real, and which ones are not. The diagnostics established in this chapter will serve as the basis for the discussion of specific phases in the chapters that follow.

31

I assume that the EPP feature of T is also inherited from C. This is not uncontroversial, but it seems innocuous in the case at hand, where TP is dominated by C and where C does not need any EPP features of its own. More interesting are cases in which TP is not dominated by C but nevertheless has an EPP feature, as in the successive cyclic raising example in (i), and cases in which both C and T have an EPP feature, as in the object wh-question in (ii). (i) Parrots seem to be likely to like nuts. (ii) What do parrots like?

{DESKTOP}NEW FOLDER//9781107040847C03.3D

58

[58–70] 23.11.2013 11:45AM

3 Phasehood diagnostics

3.1 Anatomy of a phasehood diagnostic1 A common characterization of phases, which we alluded to in previous sections (Section 2.3 in particular), is their independence at the interfaces. Thus, to determine if a given XP is a phase, we need to ask the two questions in (1). (1) a. Is XP semantically independent? b. Is XP phonologically independent?

This, however, only restates the question of how to determine if XP is a phase as the question of what it means to be independent at the PF or LF interface, and, more importantly, how this independence can be diagnosed. The bulk of the discussion in the current chapter is devoted to making these questions more concrete and establishing more tangible phasehood diagnostics. Alternatively, we can define phases by extension (by simply listing them) and this is essentially all we have been doing so far, merely alluding to independence at the interfaces as a possible factor. The categories we have been assuming to be phases are vPs, CPs and possibly DPs. This is hardly an optimal way to proceed, as it raises the obvious question of why these three categories, as opposed to any other logically possible subset of syntactic categories, should count as phases. A more desirable way to proceed is to define phases by intension, establishing independent diagnostics. The following alternative, often employed in the literature, comes to mind: if we know that categories X, Y are phases (but Z is not), and we establish that X and Y share properties p1 and p2 (which is what distinguishes them from Z), we can conclude that properties p1 and p2 are phasehood diagnostics, and apply these diagnostics to other categories to determine their phasehood. However, this method also has its drawbacks. For one thing, it relies on having some a priori 1

I borrow the title of this section from the title of chapter 1 of Levin & Hovav’s (1995) book Unaccusativity: At the Syntax–Lexical Semantics Interface, which is ‘The anatomy of a diagnostic: the resultative construction’.

58

{DESKTOP}NEW FOLDER//9781107040847C03.3D

59

[58–70] 23.11.2013 11:45AM

Anatomy of a phasehood diagnostic

59

division of syntactic categories into phases and non-phases. Furthermore, not all common properties of phases will work as diagnostics. For example, it seems to be the case that all phases are maximal projections, but we do not think of being a maximal projection as a phasehood diagnostic, since the common view is that not all maximal projections are phases (see, however, Epstein & Seely 2002, 2006, Lahne 2008, Müller 2004, 2011 for proposals that all phrases are indeed phases).2 A natural alternative, which all the proposals surveyed so far assume, is that only a certain well-defined subset of syntactic categories constitutes the set of phases. Now the question is which subset, and what exactly distinguishes this subset from its complement (the set of non-phases). There are many logical possibilities that come to mind; one is that only phrases headed by core functional heads constitute phases. However, conceptually it is not clear why things should work this way, as opposed to the other way around. Why couldn’t all phrases headed by lexical categories constitute phases instead? Neither approach seems to yield correct empirical results, with either TP being problematically classified as a phase if we take all functional categories to be phase heads, or VPs and NPs being problematically classified as phases (and CP and vP as non-phases) if we take the opposite route. So what is it that CPs and vPs have in common that they do not share with TPs? A trivial answer is that things would go terribly wrong if a different set of categories were phases. A more promising path to pursue is to establish independent phasehood diagnostics, based on what the interfaces need phases to be, and apply these diagnostics to test whether a given syntactic category is a phase or not. Since phasehood is often characterized in terms of interface considerations, the diagnostics tend to be classified into two major groups: LF diagnostics and PF diagnostics, each relating to the two interfaces. Some researchers, such as Matushansky (2005), divide phasehood diagnostics into more types, adding a separate class of syntactic (or morphosyntactic) diagnostics to the two interface-based ones. It would be preferable on conceptual grounds if we were able to diagnose phases by appealing only to interface-based considerations, but ultimately it is an empirical question of whether there are some phasehood diagnostics that are purely syntactic. The answer depends on the somewhat murky issue of where the boundary between syntax and semantics or phonology lies. Furthermore, it is not always easy to tell whether a given diagnostic is a true syntactic diagnostic, as opposed to being a syntactic reflex of a semantic diagnostic. The classification is harder for some potential 2

I do not think anyone would question the idea that all phases have to be maximal projections, since maximal projections have the level of independence and completeness that non-maximal projections do not have. However, the question of whether something would go wrong at the interfaces if phases were not maximal projections, or if phases were not always maximal projections, is ultimately an empirical question. But allowing non-maximal projections to be phases would devoid phases of completeness associated with phasehood.

{DESKTOP}NEW FOLDER//9781107040847C03.3D

60

60

[58–70] 23.11.2013 11:45AM

Phasehood diagnostics

diagnostics than for others. For example, being a domain for nuclear stress assignment would be a fairly straightforward example of a PF diagnostic, whereas being propositional would count as a solid case of an LF diagnostic. But some are harder to classify. For example, is being a target for Quantifier Raising (QR) or reconstruction a semantic or a syntactic diagnostic? This depends on whether we think of QR (or reconstruction) as a semantic or a syntactic process, and we know that there are different views on this matter. Similarly, if ellipsis is a diagnostic, should it count as a PF, LF or a syntactic diagnostic? Surely, ellipsis has an effect on both pronunciation and interpretation, and depending on whom we ask, it is treated as either a PF process (literal deletion) or an LF process (reconstruction of missing material at LF). Difficulties in classification notwithstanding, I will proceed on the assumption that phasehood diagnostics can indeed be subdivided into three types: syntactic, semantic and phonological, with the caveat that sometimes the distinctions between the three types might be less than perfect. I will thus not attach too much importance to this classification and leave open the possibility that the same diagnostic can belong to more than one group, and, for example, be both a syntactic and a semantic diagnostic. It is not even clear whether the issue of classification is a substantive, as opposed to a terminological issue. After all, does it really matter whether a given diagnostic is a PF, LF or a syntactic phasehood diagnostic, as long as it is a reliable phasehood diagnostic? The two more concrete questions we can ask about phases concern the properties of phase heads. If phases are ‘syntactic engines’, to borrow Richards’ (2011) term, the following questions come to mind: (2) a. Does X trigger Spell-Out? b. Is X the source of uninterpretable features?

The answer to (2a) will help us diagnose the PF and LF status of phase; if phase heads determine Spell-Out, the domain that is spelled out is a domain that will feed phonology and semantics. We might therefore expect both phonological and semantic processes to apply to this domain. Sections 3.2 and 3.3 will thus be devoted to PF and LF phasehood diagnostics, respectively. The affirmative answer to (2b), on the other hand, will help us uncover syntactic diagnostics; if only phase heads are inherently endowed with uninterpretable features, then they are going to trigger operations that rely on these features. This will lead us to concrete syntactic diagnostics in Section 3.4.

3.2 PF diagnostics This section focuses on the questions of what it means for a phase to be independent at PF, what it means for a phase head to trigger Spell-Out, and – more importantly – how we know that a phase is PF independent or has

{DESKTOP}NEW FOLDER//9781107040847C03.3D

61

[58–70] 23.11.2013 11:45AM

PF diagnostics

61

triggered Spell-Out. Below is a list of PF phasehood diagnostics that are often discussed (or alluded to) in the literature on phases. The ones in (3a–b), discussed explicitly by Matushansky (2005), speak to PF independence, whereas the ones in (3c–d) speak to the ability of phase heads (and only phase heads) to determine Spell-Out. (3) a. Can XP be isolated? b. Can XP be moved? c. Does XP (or X) determine phonological domains? d. Can the complement of X be elided?

Perhaps the most natural potential PF diagnostic is the one in (3a), involving phonetic isolability. Simply put, according to this diagnostic, if XP can stand alone (and function as a response to a question, for example), XP is a phase. The diagnostic in (3b) says that if a given XP can be moved (by processes like fronting of any kind, extraposition, clefting etc.), it should count as a phase. However, these are standard constituency tests (as any introductory syntax textbook can confirm), so a much larger number of categories ‘pass’ as phases according to this diagnostic. Matushansky (2005) also notes that the ability to be isolated or undergo movement does not seem to be a very reliable phasehood diagnostic (although she nevertheless applies it to DPs).3 The reason why we might be tempted to think of it as such is that taking phonetic independence to be a property of phases has an intuitive appeal. However, both isolability and movement are broader characteristics of phrases, or perhaps of constituents even more generally. Thus, example (4aʹ) shows that DPs can stand alone as sentence fragments. However, the examples in (4b–f) show that many other phrases can do so as well. (4) a. What did Mary study? aʹ. [DP Western philosophy] b. How many tests did Mary take to become a doctor? bʹ. [NumP Two] c. Where did Mary go? cʹ. [PP To Cambridge] d. What does everyone believe? dʹ. [CP That the earth is flat] e. What did Tom try? eʹ. [CP To be a good syntactician] f. What color is this page? fʹ. [AP White]

3

Matushansky also considers extraposition, clefting and though-movement as PF phasehood diagnostics. She starts of by comparing TPs, CPs and vPs, and then applies the same diagnostics to test DPs for phasehood.

{DESKTOP}NEW FOLDER//9781107040847C03.3D

62

62

[58–70] 23.11.2013 11:45AM

Phasehood diagnostics

Similarly, many phrases may function as pivots (bracketed in (5)) in pseudocleft constructions, which is also a sign of isolability.4 (5) a. What Mary read was [DP Chomsky’s Syntactic Structures]. b. What Tom wondered was [CP whether Pluto is a planet]. c. Where Tom went was [PP to the store]. d. *What Tom seemed was [TP to be a hero]. e. *How many books Tom read was [NumP five]. f. *What color was Tom’s new car was [AdjP yellow].

Since different types of verb phrases have been claimed to differ in terms of their phasehood status (something we will discuss in detail in Section 4.2 of Chapter 4), I am going to keep them distinct for the purposes of the (putative) diagnostics considered here. (6) a. What Mary did was [vP read Chomsky’s Syntactic Structures] b. What John did was [vP/VP arrive late]. c. *What John did/was [vP/VP given a book].

UNERGATIVES UNACCUSATIVES PASSIVES

The examples in (7a–g) show that quite a few types of phrases can undergo movement: (7) a. [DP This new movie], we all admired. b. [PP To this store], John went after the meeting. c. [TP To become a top syntactician], Mary is very likely. d. [CP To become a top syntactician], John really wanted. e. [CP That the earth is not flat], nobody doubts. f. *[AdjP Interesting], we read books. g. *[NumP Five], we read books.

All three verb phrase types under consideration here behave alike in this respect: (8) a. [vP Read Chomsky’s new manuscript], all linguists should. b. [vP/VP Arrive at the station], the train should have by now. c. [vP/VP Eaten by sharks], nobody wants to be.

UNERGATIVES UNACCUSATIVES PASSIVES

And the examples below make a similar point with respect to Right Node Raising (RNR):5 (9) a. John read and Bill reviewed [DP a new article on RNR]. b. John knows and Bill believes [CP that syntactic phases exist].

4

5

A reasonable question to ask is what type of derivation pseudoclefts involve; in particular whether they involve movement of the pivot. Derivational details notwithstanding, it is clear that there is a pause before the pivot of the pseudocleft, which is suggestive of some degree of phonetic independence. Similar considerations apply to Right Node Raising. See also Bošković (2002) for a discussion of RNR (of TPs in particular) as an argument against treating isolability as a diagnostic.

{DESKTOP}NEW FOLDER//9781107040847C03.3D

63

[58–70] 23.11.2013 11:45AM

PF diagnostics

63

c. John should and Bill won’t [vP read a new article on RNR]. d. John should but Bill won’t [vP/VP arrive on time].

However, RNR seems to be a very permissive diagnostic, given that RNR can even apply to subwords:6 (10) a. My analysis over and your analysis under-generates. b. Ex and current smokers have high blood pressure. c. Sheila likes pre- and post-colonial architecture. (Barros & Vicente 2010: 9)

This shows that isolability (the ability to stand alone and the ability to undergo movement) is not a very reliable diagnostic. It is also not clear that there are any reasons to expect that movement or the ability to stand alone should count as phasehood diagnostics to begin with. There is nothing in the way syntactic derivations proceed under phase-theoretical assumptions that should make us believe that only phases should be able to undergo syntactic movement, or that only phases should be able to stand alone. For example, we could conclude from the grammaticality of sentence fragments of the kind given in (11b) that AdjP are phases. But we could also conclude that they involve a more complex (phasal) structure with some portion deleted (namely (11c)). Needless to say, this makes the diagnostic vacuous, unless we have a clear sense of what it means for a given string to stand alone. (11) a. What color is the sky today? b. [AdjP Grey]. c. [CP it is grey]

Phase Theory also does not seem to make any predictions about the movability of a given constituent. It only makes predictions about the movement path (namely, the Phase Impenetrability Condition states that movement out of a phase is possible only if it proceeds through the phase edge). Thus, even though both the ability to stand alone and the ability to function as a unit for the purposes of movement seem like reasonable and intuitive ways to understand ‘phonetic independence’, there is nothing in the theory that leads to such a prediction; the conditions imposed by the interfaces do not lead to a prediction that only phases should be able to move or stand alone. However, the ability of phase heads to determine Spell-Out does lead to two concrete diagnostics: one having to do with determining phonological domains and the other one having to do with ellipsis. If phase heads determine Spell-Out, they should also determine domains for indisputably PF-related processes, such as nuclear stress assignment (and, arguably, linear order). We will discuss this in more detail in Chapter 7, where we focus on the syntax–phonology interface and the role phases play at this interface. For now, suffice it to note that given the role phase heads play in determining 6

This is only a counterexample if Right Node Raising involves movement.

{DESKTOP}NEW FOLDER//9781107040847C03.3D

64

[58–70] 23.11.2013 11:45AM

64

Phasehood diagnostics

Spell-Out, we expect them to be involved in determining phonological domains. Crucially though, saying that phases might be relevant in determining phonological domains is not necessarily equivalent to saying that phases are the relevant phonological domains. Furthermore, such phonological domains serve as useful (direct or indirect) phasehood diagnostics only if they can be established independently. Yet another potential PF-related diagnostic comes from the domain of ellipsis. It is a well-documented fact that elided strings have to be licensed in a sense that is somewhat debatable (but not directly relevant to our purposes), and that only certain types of heads can license ellipsis (see Chung, Ladusaw & McCloskey 1995, Hankamer & Sag 1976, Lobeck 1995, Merchant 2001, Sag 1976, Zagona 1988, among many, many others). What is relevant for our purposes is that Phase Theory does seem to make a concrete prediction about ellipsis. In particular, if we take ellipsis to be a PF phenomenon (and thus something that happens at SpellOut or after Spell-Out), it is only natural to link ellipsis to phases and assume that only phase heads can license ellipsis. This is the view on ellipsis that has been taken by Craenenbroeck 2010, Gengel 2007, 2009, Holmberg 2001, Rouveret 2012, among others.7 The following examples provide an illustration; what is elided in each case is the complement of a phase head, the complement of C in sluicing, the complement of v in VP Ellipsis (VPE) and the complement of D in nominal ellipsis. In all of these examples, the phase head is in bold and the missing constituent is indicated by a Δ symbol.8 (12) a. A parrot flew somewhere but I don’t know where [CP C [TP Δ] ]. b. A macaw ate a nut and a cockatoo did [vP v [VP Δ] ], too. c. Parrots like Randy’s biscuts but they prefer [DP Barbara’s D [NP Δ] ]. 7

This is not to say that being a complement of a phase head is sufficient to license ellipsis. Rouveret (2012), for example, proposes the licensing condition in (i) to capture crosslinguistic variation with respect to VPE. (i) Licensing Condition on VPE VPE is available in a given structure if, and only if, v’s uninterpretable [tense] feature is valued at the v-level.

8

He argues that in some languages, the tense feature is valued at the v level, whereas for others it is valued at the T level. This view of ellipsis does raise questions about other ellipsis types, such as gapping, pseudogapping, Right Node Raising or argument ellipsis, which do not readily lend themselves to being treated as involving deletion of a complement of a phase head. For some of them, there are alternative treatments that do not involve ellipsis. For example, Right Node Raising has been reanalyzed in a multidominant fashion (see Citko 2011a, b and the references therein), and gapping has been argued to involve ATB movement (see Johnson 2000, 2009). Logically speaking, it could also be the case that ellipsis is not directly linked to Spell-Out, and, consequently, the grammar has to include some process directly responsible for ellipsis that goes beyond spelling out the complement of a phase head as null. Aelbrecht (2009), for example, argues explicitly against the claim that only phase heads can license ellipsis.

{DESKTOP}NEW FOLDER//9781107040847C03.3D

65

[58–70] 23.11.2013 11:45AM

PF diagnostics

65

This is not the only view of ellipsis that has been proposed within the confines of Phase Theory. Bošković (to appear) argues that either complete phases or complements of phase heads can be elided. To illustrate with a concrete example, he argues, following Şener & Takahashi (2009), that (13b) involves ellipsis of a complete noun phrase, since it allows for both a strict and sloppy reading, the latter being characteristic of ellipsis.9 (13) a. Taroo-wa sannin-no sensei-o Taro-TOP three-GEN teacher-ACC ‘Taro respects three teachers.’ b. Hanako-mo e sonkeisiteiru.

sonkeisiteiru. respects

[Japanese]

Hanako-also respects ‘(Lit.) Hanako respects e, too.’ (Boškovic´ to appear: 22, citing S¸ener & Takahashi 2009: 3)

Possessors, particles and quantifiers can escape ellipsis, which Bošković takes to mean that complements of phase heads can also be elided. Interestingly, he notes that when both a quantifier and a possessor is present, the quantifier can escape ellipsis but the possessor cannot: (14) *B sensei-wa hotondo-no Ziroo-no tikoku-o yurus-anakat-ta. Prof. B-TOP most-GEN Ziro-GEN tardiness-ACC forgive-NEG-PST ‘Prof. B didn’t forgive most of Ziro’s tardiness.’ (Boškovic´ to appear 28, citing Takahashi 2001)

This is related to an independent claim that phases are dynamic in that the highest projection in a given extended projection constitutes a phase. In (14), the highest projection is the QP hosting the quantifier most. This is the projection that constitutes a nominal phase. The possessor occupies a specifier of a lower (non-phasal) head. In Bošković’s theory, (14) is ungrammatical because it involves ellipsis of a complement of a complement of a phase head. We will discuss the possibilities of phases being dynamic in more detail in Chapter 6. For now, let me conclude the discussion of PF diagnostics by listing the questions that survived the scrutiny of this section; positive answers to these questions classify a given constituent as a PF phase. I do so in (15); the answers to (15b) and (15c) establish the answer to (15a). (15) a. Does X trigger Spell-Out? b. Does XP constitute (or determine) a prosodic domain? c. Can the complement of X be elided?

9

See, however, Merchant (2013) for a discussion of some issues with considering sloppy identity a reliable diagnostic of ellipsis.

{DESKTOP}NEW FOLDER//9781107040847C03.3D

66

66

[58–70] 23.11.2013 11:45AM

Phasehood diagnostics

3.3 LF diagnostics A common characterization of phases is that, in addition to being independent at PF, they should also be independent at LF. What this means is that they should be interpretable and complete in some semantically relevant sense. However, as often pointed out, it is not very clear what it means to be complete in a semantic sense. Furthermore, what counts as complete can mean different things depending on the category. One possibility, explored by Matushansky (2005), would be to equate semantic completeness with being saturated; in other words, being of type e or t. From this perspective, only elements of type e (individuals) or type t (truth values) would be phases. Thus only certain types of DPs would be phases: proper names (being of type e), but not quantified DPs (being of type ). On the Generalized Quantifier theory, no DP would ever be a phase, since all DPs are of the unsaturated type , as also noted by Matushansky (2005). Furthermore, even if we take semantic saturation to be a viable diagnostic, it is not going to be a reliable one, given that the semantic type of a given constituent can change if something moves out of it. Chomsky often refers to a slightly different kind of completeness: the kind associated with a complete argument structure. However, it is not clear that an unaccusative verb is any less complete than an unergative one; both require only one argument, yet Chomsky only treats unergative (and transitive) vPs as phases. Treating unergatives as hidden transitives with a hidden object, along the lines of Hale & Keyser (1993), provides a possible solution to this problem. There are other (perhaps more concrete) diagnostics that are arguably also semantic, which I list in (18). One has to do with reconstruction, and the other one with QR. Classifying them as semantic is somewhat controversial, since they both have to do with overt movement out of XP, which is undeniably a syntactic process. However, since they deal with the interpretation of elements moving out of phases, it seems innocuous enough to treat them as semantic (or both syntactic and semantic). (16) a. Can an element moving through the edge of XP be interpreted at the edge of XP? b. Is XP a target for Quantifier Raising?

Another potential LF phasehood diagnostic comes from existential closure and the distinction between nuclear scope and the restrictive clause. It has been proposed that the quantificational partition of the clause into nuclear scope and restrictive clause corresponds to the distinction between a vP and a CP phase10 (see Butler 2005, Biskup 2009a, Carnie & Barss 2006, among others). Since it is

10

Butler also takes a TP to be a phase.

{DESKTOP}NEW FOLDER//9781107040847C03.3D

67

[58–70] 23.11.2013 11:45AM

Syntactic diagnostics

67

not clear that this is a general diagnostic that can be applied to test the phasehood of other categories, I will not rely on it here as a diagnostic, but will come back to it in Section 7.2, which deals with the syntax–semantics interface and the role phases play at that interface. Quantifier Raising might seem odd as a phasehood diagnostic, given the ‘classic’ view that QR targets TPs, non-phases on standard assumptions (see May 1977, 1985, for example). There are, however, many independent factors involved (such as semantic type compatibility, economy or locality considerations), which make it impossible for QR to target many phrases, but the availability of QR in cases where these factors do not intervene does help us establish that movement can target the edge of a given a category, a property associated with phases. Bruening (2001), for example, argues that QR is subject to economy in that it targets the closest constituent of the right semantic type (type t). Thus, vP is always going to be the closest semantically appropriate QR landing site for vPinternal quantifiers. This can explain clause-boundedness of QR; since vP is always closer than a CP, CP is never a potential landing site, and successive cyclic movement from the edge of the embedded vP to the edge of matrix vP would be banned by the Phase Impenetrability Condition. Thus PIC, as shown by Cecchetto (2004), can explain clause-bounded nature of QR. Simply put, if QR could target CPs, there would be no way to explain it. QR should be able to take place successive-cyclically, the way wh-movement takes place.

3.4 Syntactic diagnostics The Phase Impenetrability Condition, which we discussed in detail in Section 2.4 above, and which is repeated in (17) below, gives rise to a number of very tangible diagnostics. Since, for the most part, this class of diagnostics involves syntactic movement (which may or may not have semantic effects), it is only natural to treat them as syntactic. (17) a. [ZP Z . . . [XP X [HP α [ H YP ] ] ] b. The domain of H is not accessible to operations at ZP [emphasis mine, B.C.]; only H and its edge are accessible to such operations. (Chomsky 2001: 14)

The question is whether movement out of XP has to proceed through the edge of XP. I will refer to this property of phases as the edge property.11 This general question can be broken up into more specific questions about the effects of such movement, listed in (18).12 11

12

This is slightly different from perhaps the more typical usage of the term ‘edge property’, which takes it to be the feature that triggers movement to the edge. These are familiar from the vast literature on successive cyclic movement (see Boeckx 2003, 2008a, Felser 2004 and the references therein, for more detailed discussion of successive cyclic movement, and Lahne 2008 in particular for a discussion of these (and other) diagnostics in the context of Phase Theory).

{DESKTOP}NEW FOLDER//9781107040847C03.3D

68

68

[58–70] 23.11.2013 11:45AM

Phasehood diagnostics

(18) a. Can the moved element be interpreted at the edge of XP? b. Can the moved element be pronounced the edge of XP? c. Can the moved element strand anything the edge of XP?

It may well be the case that some of these diagnostics are phonological and some are semantic, or that they cannot be classified as falling neatly into one (or the other) class. As we saw above, deciding whether a given diagnostic is a PF, LF or a syntactic diagnostic might not matter for all practical intents and purposes as long as we assume phase integrity, take the same categories to count as both PF and LF phases, and take Transfer to PF and LF to occur at the same time. This is not uncontroversial; we alluded to the possibility of nonsimultaneous Transfer/Spell-Out in Section 2.5, and we will discuss it in more detail in Chapter 6. There are two other diagnostics that do not seem to be easily classifiable as either PF or LF diagnostics, as they do not capitalize on the idea that phases are independent at the interfaces. Both have to do with uninterpretable features. If phase heads drive syntactic computation, and uninterpretable features are the syntactic engines, uninterpretable features have to be associated with phases. Thus the questions in (19a–b) can also be used to diagnose phases. (19) a. Is XP a domain for feature valuation? b. Is X the source of uninterpretable features?

If phase heads are the only locus of uninterpretable features, and as such, they literally drive syntactic computation, the only way a non-phase head such as T or V can acquire uninterpretable feature is through Feature Inheritance. We have seen above (in Section 2.3) that in many accounts, being the locus of uninterpretable features is in fact the definitional property of phases. Recall the Phase Condition of Gallego (2010), repeated below. (20) Phase Condition Uninterpretable features signal phase boundaries. (Gallego 2010: 151)

One might object to the phasehood diagnostics we have discussed in this chapter on the grounds that they involve certain circularity. We have established a set of diagnostics of the following form: if X and Y share property p, they are phases. However, it could just as well be the case that having property p is a consequence of being a phase rather than a diagnostic of it. I do not think I will be able to resolve this potential issue here; I will proceed on the assumption that property p is a phasehood diagnostic if property p is associated with phasehood. One important syntactic phenomenon that might seem to be conspicuously absent from the discussion of phasehood diagnostics (or phases in general) in this book is island effects. This is not to say that Phase Theory is not concerned with island effects, and it goes without saying Phase Theory would be incomplete at

{DESKTOP}NEW FOLDER//9781107040847C03.3D

69

[58–70] 23.11.2013 11:45AM

Syntactic diagnostics

69

best without an account of islands that is complete with its basic tenets. And islands have indeed received a fair amount of attention in phase-theoretical literature (see, for example, Müller 2011 and Gallego 2010 for specific proposals on how certain islands can be derived). However, since our focus here is on introducing Phase Theory and justifying phases themselves, an adequate discussion of islands from a phase-theoretical perspective, which merits a book-length discussion of its own, would take us too far astray.13 Let me nevertheless comment briefly on the status of islands in Phase Theory. Locality, which is captured by means of the Phase Impenetrability Condition, is an important aspect of Phase Theory but, strictly speaking, there is no direct correlation between islandhood and and phasehood. This makes islands not very useful for the purposes of diagnosing phases. First, it is not simply the category of a given phrase (CP versus TP, or DP versus NP) that determines whether it is an island or not. For example, only certain types of CPs and DPs are islands, namely interrogative CPs and definite DPs. Appealing to a filled edge position (filled [Spec,CP] position and a filled [Spec,DP] position in these two cases) is not going to be helpful, since adding an extra specifier to a phase head is always an option, and a filled specifier does not always lead to island violation (as also pointed out by Boeckx & Grohmann 2007). For example, a filled [Spec,vP] does not make that vP an island. All Phase Theory says is that movement out of phases has to proceed through phase edges; it does not say that movement out of phases (or non-phases, for that matter) is impossible. Furthermore, as we know, the answer to the question of whether a CP or DP is an island depends on many independent factors. DP is an island only for certain movements (in certain languages, as we will see in Chapter 4): movement of a specifier of D, for example, leads to an island violation but movement of (or from) the complement of D only does so sometimes. Furthermore, whether DP or CP is an island depends on what positions it occupies: if it is a complement, movement out of it is possible, but if it is a specifier or an adjunct, it is not. This is Huang’s (1982) Condition on Extraction Domains, which distinguished objects on the one hand, from subjects and adjuncts on the other hand, by appealing to proper government, an obsolete theoretical concept. In phase-theoretical terms, CED effects have been derived from the Phase Impenetrability Condition, combined with certain assumptions about features and activity (see Müller 2011 and Gallego 2010 in particular). To illustrate briefly, Müller proposes that the (optional) EPP feature (edge feature in his terms) can be added to a phase head only if the phase head is still active, where being active means not having had all its features checked. This includes features checked by both Internal and External Merge. Extraction from subjects is banned in his system because the subject is the last element merged in the vP phase. This means that the EPP feature cannot be added after the subject is merged. At that point, the v head is no longer active because it has had all its features checked. Without that EPP feature, movement is going to violate the Phase Impenetrability Condition. Adding 13

I refer the interested reader to Müller (2011) and Boeckx (2012) for book-length discussions of islands specifically.

{DESKTOP}NEW FOLDER//9781107040847C03.3D

70

70

[58–70] 23.11.2013 11:45AM

Phasehood diagnostics

the EPP feature before the subject is merged is possible because the v head is still active; this is what allows movement of the object (or out of the object). But this is not going to help with extraction out of the subject since the subject has not been merged yet. Setting islands aside, let me conclude the current chapter by listing the three types of diagnostics, as they are the ones that will be used as the basis of the discussion in the next two chapters, where we look at specific categories and determine whether they pass as phases according to these diagnostics. In the next chapter (Chapter 4), we will look at three ‘classic’ phases – CPs, vPs and DPs – whose phasehood status is fairly well established in the literature on phases. (21) PF phasehood diagnostics a. Does X trigger Spell-Out? b. Does XP determine phonological domains? c. Can the complement of X be elided? (22) LF phasehood diagnostics a. Can an element moving out of XP be interpreted in the edge position? b. Is XP a target for Quantifier Raising? (23) Syntactic phasehood diagnostics a. Can an element moving out of XP be pronounced (partially or completely) at the edge? b. Can an element moving out of XP strand anything at the edge? c. Is XP a domain for feature valuation? d. Is X the source of uninterpretable features?

We have alluded above to alternative approaches to phases, such as Bošković (to appear), Müller (2004), or Den Dikken (2007), which rely on (or require) different sets of diagnostics, and end up with different categories counting as phases. While these alternatives do avoid problems with imprecise characterizations of phases, they often raise questions of their own. For Bošković, for example, all lexical categories (Vs, Ns, Ps and As) project phases. However, for him it is not necessarily the case that every V, N, P or A head is a phase head.14 Its phasehood is determined relative to the context in which it appears (see also Bobaljik & Wurmbrand 2005) in that the highest projection in the extended domain of each lexical category is a phase. This can yield to a fair amount of crosslinguistic variation if languages differ in terms of how ‘extended’ their extended projections are. For Epstein & Seely (2002, 2006), Lahne (2008) and Müller (2004, 2011), all phrases are phases, which leads to a very different view of successive cyclic movement. For Den Dikken, whose proposal we will discuss in more detail in Chapter 6 (Section 6.1), the definitional characteristic of a phase is the subject–predicate configuration. 14

This is different from proposals that claim that every phrase is a phase (as suggested by Müller 2004, for example).

{DESKTOP}NEW FOLDER//9781107040847C04.3D

71

[71–123] 23.11.2013 11:48AM

4 Classic phases

4.1 CPs as phases This section reviews the evidence pointing towards CPs being phases. Most of it involves well-known (and often discussed) evidence that long-distance whmovement takes place in a successive-cyclic fashion through intermediate [Spec, CP] positions (see, for example, Boeckx 2008a, Felser 2004, Lahne 2008 and the references therein).1 What matters for our purposes is that successive-cyclic movement targeting a certain projection can be taken as evidence in favor of this projection being a phase. We dubbed this property of phases the edge property.2 The questions that can help us diagnose this edge property are given in (1). In the case at hand, the edge refers to the edge of CP, namely the [Spec,CP] position.3 (1) a. Can an element moving out of CP be pronounced at the edge of CP? b. Can an element moving out of CP strand anything at the edge of CP? c. Can an element moving out of CP be interpreted at the edge of CP?

In this section we will also look at the properties of the C head, particularly its featural makeup (whether it has any uninterpretable features and whether it 1

2

3

Many of the phenomena I discuss in this chapter are also discussed by these authors, and are generally well-known from the literature on successive cyclic movement. Lahne (2008), for example, also discusses reconstruction, agreement, quantifier float, partial movement and whcopying. Her conclusions, however, are different from the conclusions reached here, as she takes all phrases to be phases. Felser (2004) focuses on partial movement and wh-copying. Legate (2003) discusses reconstruction, parasitic gaps, QR and nuclear stress assignment (focusing on vPs rather than CPs). However, there is more to establishing whether a CP (or any other category) is a phase than establishing whether it has the edge property, and we have seen many phasehood diagnostics that do not make reference to edges. We will discuss non-edge diagnostics in this section as well. The title of Lahne’s (2008) dissertation ‘Where There Is Fire There Is Smoke’ captures the spirit of the diagnostics considered here; we are looking for the smoke.

71

{DESKTOP}NEW FOLDER//9781107040847C04.3D

72

[71–123] 23.11.2013 11:48AM

72

Classic phases

triggers Spell-Out). Both of these properties are associated with phase heads only. We will defer the discussion of CPs as potential phonological domains till Chapter 7. Perhaps the most straightforward illustration of the edge property comes from the domain of long-distance wh-movement of the kind represented schematically in (2). In such configurations, the wh-phrase moving out of the lowest CP has to pass through (at the very least) each [Spec,CP]. For now, we will not worry about whether it has to pass through any other positions; this is the issue we turn to in the next section. (2) [CP WH [C′ C [TP …

[CP WH [C′ C [TP… [CP WH [C′ C [TP … WH… ] ] ] ] ] ] ] ] ]

Even though the claim that long-distance wh-movement proceeds in such successive-cyclic fashion is hardly controversial, the why-question (i.e. the motivation for movement to the intermediate positions) is somewhat less understood; the Phase Impenetrability Condition offers the first glimpse into this why-question. If both C and v are phase heads, as I hope to convince the reader by the end of Section 4.2, this is the only way a wh-phrase can get from the embedded clause to the matrix [Spec, CP]. Crucially, this is different from the question of what features drive this movement, the question I set aside for the time being. Thus, in a simple case such as the one in (3a) below, represented schematically in (3b), the embedded TP is spelled out when the matrix v is merged. Thus, if the wh-pronoun does not move to [Spec,CP] in the embedded clause, it will not be accessible either to the matrix v or C (via PIC), as shown in (3b). Movement to the specifier of the embedded vP is not indicated in this diagram, but the same logic applies: the complement of v is spelled out when C is merged, so the only way for the wh-phrase to avoid Spell-Out is to move to [Spec,vP]. (3) a. What does Jan think Piotr read? b. vP

v′ v V

TP Spell-Out

CP C

TP ... WH...

{DESKTOP}NEW FOLDER//9781107040847C04.3D

73

[71–123] 23.11.2013 11:48AM

CPs as phases

73

There are a number of ways to tell that wh-movement does indeed target the edge of CP. Perhaps the simplest one is to look at what happens if the edge is filled. And, indeed, unless a language has some other strategy (such as the availability of multiple specifiers), the result is ungrammatical and constitutes a familiar example of a wh-island violation: (4) a. *What did you think [CP when Jan devoured what when]? b. *This is the book which Jan wondered [CP who wrote which]. c. *How did you wonder [CP who fixed the sink how]?

Interestingly, languages can vary in whether they obey the Wh-Island Condition, as noted by Rizzi (1982) for Italian and by Rudin (1988) for Bulgarian, for example. The examples in (5a–b) show that in these languages movement from wh-islands does not result in ungrammaticality. mi domando che storie abbiano (5) a. tuo fratello, a cui your brother, to whom myself ask.1SG which stories have.3PL

raccontato. [Italian] told

‘your brother, to whom I wonder which stories they told’ (Rizzi 1982: 50)

ˇ oveka, kojto b. C the man who.REL

ne znaeš NEG

kakvo kazvat

know.2S what

ˇce

e

kupil

[Bulgarian]

say.3PL that has bought

‘The man who you don’t know what they say that he bought.’

(Rudin 1988: 452)

The common explanation for Bulgarian is that its CPs allow multiple specifiers, as evidenced by the well-known the fact that Bulgarian (along with other Slavic languages) is a multiple wh-fronting language.4 From a phase-theoretical perspective, what such languages teach us is that a phase edge can contain multiple specifiers (at least in those languages in which all wh-phrases move to [Spec,CP], such as Bulgarian and Romanian).5 This explanation, however, is not going to work for Italian, which is not a multiple wh-fronting language, and, to make things even more intriguing, is not even a language that allows English-style multiple wh-questions with one of the wh-phrases in situ, noted by Calabrese (1984) and discussed more recently by Stoyanova (2008).6 The classic account of the grammaticality of wh-island violations in Italian is that of Rizzi (1982); it 4

5

6

The literature on multiple wh-fronting is vast, starting with Wachowicz’s (1974) observation that there are indeed languages that can front all their wh-phrases overtly, and Rudin’s (1988) seminal work on the distinction between two types of multiple wh-fronting languages, which, in most general terms, differ with respect to whether they treat all the fronted wh-pronouns alike or privilege the highest one. See also Boeckx & Grohmann (2003), Bošković (2002b) and Richards (2001), among others, for more recent proposals and refinements of the early empirical generalizations. We have already seen that v, another candidate for a phase head, also allows multiple specifiers. There is a stage in the derivation in which the subject is in the inner specifier and the object in the outer one. Stoyanova also includes Somali, Irish and Berber in her discussion of languages with no multiple wh-questions, and attributes the lack of multiple wh-questions to the existence of a unique focus position in such languages.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

74

74

[71–123] 23.11.2013 11:48AM

Classic phases

involves parameterization of bounding nodes, and, consequently, of the Subjacency Condition itself. If in Italian, a CP (rather than a TP) counts as a bounding node that is relevant for subjacency, the movement of the wh-phrase a cui in (5a) above would cross only one bounding node. An explanation along these lines, however, is somewhat anachronistic, and – perhaps more gravely – incompatible with the claim that crosslinguistic variation is limited to features of individual lexical items, sometimes referred to in the literature as the Borer– Chomsky Conjecture.7 Leaving wh-islands and subjacency aside, let us turn to other wellestablished arguments in favor of movement proceeding through the specifier of CP, and thus, indirectly, in favor of CPs being phases. A fair amount comes from reconstruction effects; situations in which the moved element is interpreted in a lower position than the position in which it is pronounced (see Barss 1986, 2001, Fox 1999, 2000, Lebeaux 2009 and the references therein).8 On a syntactic approach to reconstruction, in which reconstruction amounts to the interpretation of a lower copy, the only way a wh-phrase can be interpreted in an intermediate [Spec,CP] position is if it passed through this position, leaving a copy behind. Consider in this light example (6): the presence of multiple wh-copies gives rise to multiple potential interpretations, involving different antecedents. If the copy in the specifier of CP1 is the one that is interpreted, the most local potential binder for the anaphor is going to be Piotr. If it is the highest copy instead (the one in the specifier of CP2), the most local potential binder is going to be Jan. And if it is the lowest one, the most local potential binder is going to be Adam.

7

8

On a methodological note, it is worth noting that what looks like the same phenomenon can have very different explanations from language to language. The phenomenon of reconstruction is quite complex; there are many factors that play a role, such as the differences in reconstruction possibilities between A and A-bar movement, arguments and predicates, and questions and relative clauses, to name just a few. The very assumption that reconstruction is a syntactic mechanism is far from being uncontroversial. There is also a possibility that different binding principles might exhibit reconstruction behavior. For example, Lebeaux (2009) argues that Binding Principles A, B and C apply at different points in the derivation (contra the standard minimalist view, on which they all apply at LF only), and that the negative principles (Principles B and C) have to apply throughout the derivation while the positive principle (Principle A) applies at LF. As he himself admits, examples of the following sort are a problem, as before movement they violate Principle C. He resorts to a kind of ‘late insertion’ mechanism to avoid such problems. (i) John seems to himself to like cheese.

Since my goal here is not to argue for a specific analysis of binding, I will abstract away from these controversies. All that matters for our diagnostic purposes here is that reconstruction be a syntactic mechanism, indicating the presence of a lower copy.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

75

[71–123] 23.11.2013 11:48AM

CPs as phases

75

(6) Jan4 asked [CP2 [which picture of himself2,3,4]1 Piotr3 thought [CP1 which picture of himself1 that Adam2 liked [which picture of himself2,3,4]1.

Fox (1999, 2000) looks at slightly more complex cases to establish that in successive cyclic movement contexts, wh-pronouns can be interpreted in intermediate [Spec,CP] positions. Building on the work of Lebeaux (1994) and Heycock (1995), he looks at the interaction of scope reconstruction with Binding Principle C, and concludes that reconstruction feeds Principle C. In cases where reconstruction would lead to a Principle C violation, reconstruction is blocked. Fox examines successive cyclic movement cases in which the whphrase undergoing movement contains both an R-expression and a bound pronoun.9 The pronoun, in order to receive a bound variable interpretation, has to be c-commanded by the quantified expression that binds it. The R-expression, on the other hand, cannot be c-commanded by the pronoun coindexed with it. In the configuration given in (7a) there exists a position (indicated by √) in which the wh-phrase can be c-commanded by the quantified DP but not by the pronoun. This is not an option in (7b), as reconstruction to either of the two available positions will result in a Principle C violation.10 (7) a. [which. . . pronoun1. . . r-expression2] . . ..QP1 √ pronoun2 . . . * . . . b. [which . . . pronoun1 . . . r-expression2] . . . pronoun2 . . .*. . . QP1 . . .*. . . (Fox 1999: 173)

The actual examples are given in (8a–b). In (8a), the wh-phrase can be interpreted in the intermediate [Spec,CP] position (indicated by tʹ), in which the pronoun he1 will be bound by every student, but the R-expression Ms. Brown will not be bound by she2. In (8b), on the other hand, this is not an option; reconstruction even to the intermediate position (tʹ) will yield a Principle C effect. This is only possible if movement takes place through the edge of CP, which in turn is only possible on current assumptions if CP is a phase. (8) a. [Which (of the) paper(s) that he1 gave to Ms. Brown2] did every student1 hope tʹ that she2 will read t? b. *[Which (of the) paper(s) that he1 gave to Ms. Brown2] did she2 hope tʹ that every student1 will revise t? (Fox 1999: 173, citing Lebeaux 1990)

This suggests the derivation in (9) (first pass) for long-distance movement. The questions that we will come back to are whether there are any other intermediate landing sites, and what exactly drives the movement to the intermediate [Spec,CP] positions. For now, let us assume the answer to the latter question is some uninterpretable feature with the EPP property.

9

10

Fox’s goal, however, is to establish that Principle C applies at LF, rather than to argue for the phasehood of CPs. ‘The star here indicates a position to which reconstruction will result in a Principle C violation.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

76

[71–123] 23.11.2013 11:48AM

76

Classic phases

(9) a. Who does she think John saw? b.

b.

CP

whoiwh[wh]

C′

CiForce[Q],uwh[wh],EPP TP DPuC[Nom] T′ sheiϕ[3sg]T uϕ[3sg],EPP vP she

v′ v

VP

think vuϕ[3sg] think

CP who

C′

CiForce[Decl],EPP

TP

DPuC[Nom] D

T′

Johniϕ[3sg]Tuϕ[3sg],EPP vP John

v′ v

saw vuϕ[3sg] saw

VP whoiwh[wh]

Another well-known range of arguments in favor of successive cyclic movement taking place through the specifier of CP comes from kinds of questions in which the wh-phrase with a matrix scope can either be pronounced in the embedded [Spec,CP] or leave something behind in this position. There are a number of such cases to consider, illustrated below in pseudo-English. (10a) is allowed only in some varieties of English, (10b) in child’s English, and (10c) and (10d) not at all. In all of them, the scope of the wh-phrase who or which movie is in the matrix [Spec,CP]; however, the embedded [Spec, CP] contains some portion of it (in the form of the stranded quantifier or the nominal), the wh-phrase itself (in case of the partial wh-movement construction) or its copy. (10) a. b. c. d.

wh-quantifier stranding [CP who did John think [CP who all Bill saw who all] [CP who did John think [CP who Bill saw who] wh-copying [CP what did John think [CP who Bill saw who] partial wh-movement [CP which movie did John think [CP which movie Bill saw which movie] split wh-movement

Now, let us turn to some actual data, starting with the phenomenon dubbed whquantifier float by McCloskey (2000) in the varieties of English he refers to as West Ulster English. What distinguishes West Ulster English dialects from standard English is that they allow the quantifier all to be associated with a whpronoun (as shown in (11a)), and – more interestingly for our purposes – allow the quantifier to be separated from the wh-pronoun, as shown in (11b). In this respect, these varieties differ from standard English, which only allows quantifier float under A-movement, thus providing a nice window into the inner workings of successive cyclic wh-movement.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

77

[71–123] 23.11.2013 11:48AM

CPs as phases

77

(11) a. What all did you get for Christmas? [West Ulster English] b. What did you get all for Christmas? (McCloskey 2000: 58)

Crucially, in long-distance questions, the quantifier can appear in the intermediate [Spec,CP] positions (as in (12a)), in addition to being stranded in its base position (as in (12b)), or pied-piped all the way to the matrix [Spec, CP] (as in (12c)). (12) a. What did he say all (that) he wanted? b. What did he say (that) he wanted all? c. What all did he say (that) he wanted t? (McCloskey 2000: 61)

McCloskey adopts Sportiche’s (1988) analysis of quantifier float, in which the position of the floated quantifier can only mean one of two things: that the element undergoing movement has started in the position of a floated/stranded quantifier or that it has moved through this position.11 McCloskey further assumes that the quantifier occupies the D position, and that the wh-pronoun moves to the specifier of DP first before moving out of the DP, thus stranding the quantifier. Such movement suggests that DPs might also be phases, a possibility I discuss in detail in Section 4.3 of this chapter. We might wonder, however, if the movement in (13) violates antilocality (See Grohmann 2003, Abels 2003), which prevents movement of a complement of a given phrase to the specifier of the same phrase; the rationale is that such movement violates economy, not by virtue of being too long (one could not think of a shorter movement, in fact) but by virtue of being too short, hence unmotivated. I do not think this is a major blow to the analysis presented here, given the evidence that DP structure is much more elaborate than the picture in (13a) might suggest. I will not belabor this point here, as DP phasehood (and consequently, DP internal structure) is the main focus of Section 4.3. The addition of a single projection between D and N, shown in (13b), will make the movement of NP to the specifier of DP legitimate from the standpoint of antilocality. Now the wh-pronoun what is not moving from the complement of one head to the specifier of the same head.12

11

12

The stranding analysis is by no means the only analysis of quantifier float. Alternatives include treating the floated quantifier as an adjoined adverbial of sorts. I refer the interested reader to Bobaljik (2003) and the references therein for further discussion. The claim that quantifiers can be stranded in their base positions is also not uncontroversial. Bošković (2004a), for example, argues that stranding in thematic positions is impossible. Abels’ version of antilocality is different from Grohmann’s. Grohmann (2003, 2011) divides clause structure into what he dubs prolific domains (theta domain, case/agreement domain, operator domain) and argues that movement within the same domain (as opposed to movement across domains) is illegitimate.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

78

[71–123] 23.11.2013 11:48AM

78

Classic phases

(13) a.

DP D′

whati D all

b.

ti

DP D′

whati D all

XP X

ti

We might ask here whether quantifiers are the only elements that can be stranded in intermediate landing sites. Two other phenomena come to mind: preposition stranding and what we might call nominal stranding. Felser (2004), following Du Plessis (1977), shows that Afrikaans, for example, allows P-stranding in intermediate positions. The availability of such intermediate P-stranding would have to be parameterized, as not all languages that in principle allow P-stranding allow it in intermediate positions. English, for example, does not, as noted by Postal (1972) and illustrated in (14a–b).13 How to achieve such parameterization (in a way that attributes it to some lexical property of some lexical item rather than simply stating it as a general parameter), however, is not a trivial matter. If stranded prepositions undergo reanalysis (as suggested, for example, by Hornstein & Weinberg 1981), one way to explain why preposition stranding in intermediate [Spec,CP] positions is not allowed would be to assume that the requisite reanalysis is blocked across a CP boundary and to parameterize the reanalysis itself. (14) a. *Who/Whom do you believe to Mary thinks Joan talked? b. *Who/Whom do you believe Mary thinks to Joan talked? (Postal 1972: 213)

Both quantifier stranding and preposition stranding are amenable to a nonstranding analysis, in which the entire wh-phrase moves but part of it is pronounced in [Spec,CP] and part in the pre-movement site, as illustrated schematically in (15a) for preposition stranding and (15b) for quantifier stranding. This does not affect the point made here, which is that movement targets [Spec,CP].

13

Postal himself uses the ungrammaticality of such examples to argue against successive cyclic movement, which is not the point I am making here.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

79

[71–123] 23.11.2013 11:48AM

CPs as phases

79

(15) a. [CP To who [Cʹ C [TP . . . [CP to who [C’ C [TP . . . to who] ] ] ] ] ]? b. [CP All what [Cʹ C [TP . . . [CP all what [Cʹ C [TP . . . all what] ] ] ] ] ]?

The mechanism illustrated here, referred to in the literature as distributed deletion or scattered deletion, has also been employed to handle various types of discontinuous noun phrases (see Fanselow & Ćavar 2002 and the references therein), illustrated in (16) and (17) for Polish with the two parts of the split/discontinuous DP given in italics.14 Distributed deletion is particularly appealing for cases of what would otherwise have to involve non-constituent movement, such as the ones given in (17), in which what is pronounced in the fronted position is the preposition followed by the wh-determiner (to the exclusion of the complement of D). (16) a. Kolorowe Maria zobaczyła papugi. colorful Maria saw parrots ‘Maria saw colorful parrots.’ b. Papugi Maria zobaczyła kolorowe. parrots Maria saw colorful ‘Maria saw colorful parrots.’ c. Ile Maria naliczyła papug? how.many Maria counted parrots ‘How many parrots did Maria count?’

[Polish]

(17) a. O australijskich Maria czytała papugach. about Australian Maria read parrots ‘Maria read about Austrialian parrots.’ b. O papugach Maria czytała australijskich. about parrots Maria read Australian c. O jakich Maria czytała papugach? about what.kind Maria read parrots ‘What kind of parrots did Maria read about?’

Perhaps the most extreme example of a wh-pronoun ‘leaving something behind’ in intermediate movement sites comes from a construction referred to in the literature as wh-copying (see Felser 2004, Lahne 2008 and the references therein), in which a complete copy of the wh-phrase is pronounced in each [Spec, CP] on the wh-pronoun movement path. Representative examples from Felser’s work are given in (18a–d). (18) a. Wen glaubst Du, wen sie getroffen hat? Who think you who she met has ‘Who do you think she has met?’ (Felser 2004: 544) 14

[German]

The idea that copies can be pronounced in a discontinuous manner is related to the idea that under some (fairly restricted) circumstances, pronunciation of a lower copy is forced in languages with overt movement of the relevant kind. This is the case when the pronunciation of the higher copy would lead to a PF violation (see Franks 1998, Bošković 2002b and the references therein).

{DESKTOP}NEW FOLDER//9781107040847C04.3D

80

80

[71–123] 23.11.2013 11:48AM

Classic phases

b. Waarvoor dink julle waarvoor werk ons? [Afrikaans] Wherefore think you wherefore work we ‘What do you think we are working for?’ (Du Plessis 1977: 725; Felser 2004: 544) c. Kas o Demìri mislenola kas i Arìfa dikhla? [Romani] Whom Demir think whom Arifa saw ‘Who does Demir think Arifa saw?’ (McDaniel 1989: 569 note 5; Felser 2004: 544) d. Who do you think who is in the box?

[Child English] (Thornton 1990: 204)

The Copy Theory of Movement provided a simple explanation of this construction: wh-copying could be thought of as copying with no subsequent deletion. While there are many non-trivial questions that wh-copying raises (especially given that movement is no longer thought to be a simple copying operation), what matters most for our purposes is the evidence for movement through [Spec,CP] that it provides. First, how to motivate movement through all the intermediate positions? This is the question Felser dubs the triggering question. This is the question that arises for all cases of successive-cyclic movement (not just wh-copying). Second, what allows the embedded CP to converge in spite of the fact that it seems to contain copies with unvalued features? After all, only one copy (the highest one) determines the scope of the wh-question and values the uninterpretable wh-feature (or Q feature) on C. This is what Felser calls a convergence question. And third, what allows multiple copies to be pronounced. This is the linearization question. If copies count as identical elements with respect to linearization, the result is going to violate the Linear Correspondence Axiom (LCA) of Kayne (1994), which prohibits structures in which the same element both c-commands and is c-commanded by itself.15 Any account of wh-copying, phase-theoretical or not, must answer these questions. The convergence question and the linearization question might appear to be non-distinguishable, but they are actually quite distinct. The convergence question focuses on why the Spell-Out of a wh-phrase with some of its uninterpretable features unchecked/unvalued does not violate Full Interpretation. Note that the wh-phrase does not get its uQ feature valued/ checked till it undergoes Agree with the matrix C (which is the C that bears the corresponding iQ feature and marks the scope of the wh-question). Felser solves the convergence problem in the following way. First, she assumes, following a long line of research going back at least to Katz & Postal (1964) on the morphosyntax of wh-pronouns, that they can be decomposed into the 15

Even though the specifics of the LCA in its original formulation are incompatible with Bare Phrase Structure theory, the idea behind it (the idea that linear order is not a syntactic primitive) is very much minimalist in spirit (see, however, Chomsky 1995 and Kayne 2010 for different ways to make the LCA compatible with BPS).

{DESKTOP}NEW FOLDER//9781107040847C04.3D

81

[71–123] 23.11.2013 11:48AM

CPs as phases

81

quantificational part (the wh-part) and the indefinite part. Second, she posits that the two parts can be pronounced in a discontinuous manner, akin to what saw above in discontinuous DPs (see the Polish examples in (16) and (17)). We also alluded to it being a possible analysis for preposition stranding and whquantifier stranding. In a copying construction, in a way that is similar to split wh-questions, the indefinite part is pronounced in the embedded [Spec,CP] and the quantificational part is pronounced in the specifier of the matrix [Spec,CP]. It just happens to be the case that the two parts sound the same, so to speak, in the languages under consideration here, namely in those that allow wh-copying. A derivation incorporating this insight, including only the relevant features for the sake of simplicity, is given in (19).16 (19)

CP whoiWH,iINDEF

C′ C

TP you

T′ T

Spell-Out whoiWH

vP

you

v′ v

VP

think v think CP whoiWH,iINDEF C′ C

TP John

T′

T

vP

Spell-Out whoiINDEF

John

v′ v

VP

saw v saw

whoiWH,iINDEF

Partial wh-movement (also referred to in the literature as scope marking) is similar to wh-copying in that there is a wh-element in both the matrix and embedded [Spec,CP]. However, the two wh-elements are distinct; the scope of the wh-phrase is marked by the so-called scope marker (typically a default whpronoun), and the contentful wh-phrase is in the embedded [Spec,CP]. The

16

This also leads Felser to argue for a convergence-based view of phases. In (29), the uninterpretable wh-feature of the wh-pronoun is not going to be checked until the wh-pronoun reaches the matrix [Spec,CP]. This suggests that the embedded CP cannot be spelled out before this feature is valued.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

82

82

[71–123] 23.11.2013 11:48AM

Classic phases

literature on scope marking is vast; see, among others, Bruening (2004), Dayal (1993), Felser (2004), Horvath (1997), Lahne (2008), Lutz, Müller & von Stechow (2000), Manetta (2010), McDaniel (1989), Stepanov (2000) and the references therein. The examples below, in which both the real wh-phrase and the scope marker are given in bold, provide an illustration. (20) a. Was glaubt Hans [mit wem Jacob jetzt spricht]? [German] what believe Hans with whom Jacob now talks ‘Who do you think Maria has spoken with?’ (cf. McDaniel 1989: 569) b. Jak mys´lisz [co Maria zrobi]? [Polish] how think.2SG what Maria do.3SG ‘What do you think Maria will do?’ c. Kak vy dumaete, [kogo ljubit Ivan]? [Russian] how you think.2PL whom love.3SG John ‘Who do you think John loves?’ (Stepanov 2000: 1) d. Mit gondolsz, [hogy kit látott János]? [Hungarian] who.ACC saw.3SG John.NOM what.ACC think.2SG that ‘Who do you think that John saw?’ (Horvath 1997: 510)

We might be tempted to analyze scope marking in a way that is similar to partial wh-movement, with the wh-pronoun being spelled-out in a distributed fashion across the main and the embedded [Spec, CP]. However, the two wh-elements in a scope-marking construction do not lend themselves to a partition along the indefinite/quantificational divide of the kind we saw in (19). As Felser also points out, assimilating the two constructions along these lines would leave a number of well-documented differences between them unaccounted for. For example, whcopying is restricted to simple wh-pronouns such as who or what but impossible with complex wh-phrases such as which book. No such restriction is present in scope marking.17, 18 There is also no sense in which the scope marker, which is the wh-word what in some languages, and interestingly, how in Slavic languages, is a complete copy (or a partial one) of the ‘real’ wh-pronoun. It is thus more typical to think of the scope marker as an independent element (perhaps an expletive of sorts, as suggested by McDaniel 1989, for example). The scope marker can either be inserted directly in the matrix [Spec,CP], as shown schematically in (21a), or start out in the embedded [Spec,CP] and move to the matrix [Spec,CP], as shown in (21b). In both cases, the scope marker is capitalized. The

17 18

Interestingly, Felser reports that some speakers accept wh-copying with prepositional phrases. Felser discusses other differences as well, such as the ability to conjoin two embedded clauses and their compatibility with verbs like seem. What matters for our purposes is the fact that both partial wh-movement and wh-copying provide evidence for movement through the embedded [Spec,CP].

{DESKTOP}NEW FOLDER//9781107040847C04.3D

83

[71–123] 23.11.2013 11:48AM

CPs as phases

83

latter option is favored by Mahajan (1996) for Hindi and by Stepanov (2000) for Polish and Russian. (21)

a.

CP C′

WHATiwh[wh]

CiForce[Q],uwh[wh],EPP

CP who iwh

C′

CiForce[Decl], uwh[wh]EPP

TP

who iwh

b.

CP C′ CiForce[Q],uwh[wh],EPP

DP WHATiwh whoiwh

CP C′ CiForce[Decl], uwh[wh]EPP

TP

whoiwh

The existence of so-called agreeing complementizers also provides quite a convincing argument for successive-cyclic movement through the edge of CP, and thus indirectly for CPs being phases. The well-known case of Irish, due to McCloskey’s seminal work on the topic, provides a nice illustration. Irish has different types of complementizers, which some researchers analyze as verbal particles (a point of contention that goes beyond the scope of this chapter). The one relevant for our purposes is the one glossed as aL, which appears in whextraction contexts. More importantly, it also appears in every intermediate [Spec,CP] in long-distance wh-extraction contexts:19

19

In addition to relative clauses and wh-questions, which are perhaps the two best-known cases, we see the same complementizer in equative clauses, comparative clauses and clefts (see McCloskey 2001 for examples and a detailed analysis).

{DESKTOP}NEW FOLDER//9781107040847C04.3D

84

84

[71–123] 23.11.2013 11:48AM

Classic phases

(22) a. an ghirseach a ghoid na siogai. [Irish] the girl aL stole the fairies ‘the girl that the fairies stole away’ b. rud a gheall tui a dheanfa. thing aL promised you aL do.COND-S2 ‘something that you promised that you would do’ (McCloskey 2001:67–8)

Complementizer agreement is also found in other languages (Germanic and Bantu, most notably), and we saw some examples of it in Chapter 3, in the context of the discussion of Feature Inheritance. The relevant examples are repeated in (23) and (24) below. The pattern generalizes widely across the Bantu language group; Carstens reports on a study of over seventy languages that show similar agreement. In (23), the verb is in C and it agrees in noun class with the whpronoun in the specifier of CP.20 (23)

Bikí bi-á-kás-íl-é bábo bíkulu mwámí 8what 8CA-A-give-PERF-FV 2that 2woman 1chief ‘What did those women give the chief in the village?’

mu-mwílo? [Kilega] 18–3village (Carstens 2005: 220)

In long-distance questions, both the matrix and the embedded verb agree with the wh-pronoun in class. This also shows that the wh-phrase has passed through the embedded [Spec, CP]. (24)

Bikí bi-á-ténd-ílé bána bi-á-gúl-ílé nina-bó? 8what 8CA-A-say-PERF 2child 8CA-A-buy-PERF mother-their ‘What did the children say their mother had bought? (Carstens 2005: 247)

The issue of how to account for such complementizer agreement phenomena is somewhat contentious in the literature. For some researchers, such as Kinyalolo (1991), it is a reflex of spec–head agreement. McCloskey (2002) analyzes the Irish complementizer aL as a realization of C with an operator feature and an EPP feature. Carstens (2005), however, analyzes agreeing complementizers in terms of Agree between a C head endowed with uninterpretable φ-features and a wh-element in its scope. Either way, complementizer agreement provides evidence that C is a Probe having uninterpretable features, thus a phase head. The two types of uninterpretable features that we considered in this section were uninterpretable wh-features and uninterpretable φ-features, capturing whmovement and complementizer agreement phenomena, respectively. It is certainly true that non-phase heads such as T can have uninterpretable features (uninterpretable φ-features being an obvious example), but, as we saw in Chapter 2 (Section 2.6), non-phase heads can only have uninterpretable features

20

Crosslinguistically, complementizer agreement is not limited to wh-extraction contexts. The point that matters for our purposes is that it is possible in those contexts.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

85

[71–123] 23.11.2013 11:48AM

CPs as phases

85

if they inherit them from phase heads above. Such Feature Inheritance is not a very likely option for C, especially in matrix contexts, where there is no head (phase or non-phase) dominating it. And in embedded contexts, where C is typically dominated by a non-phase head such as a V or an N, the presence of uninterpretable features on C is highly indicative of C being a phase head. The phenomenon of complementizer agreement (illustrated above with Bantu, West Germanic and, in a slightly different form, Slavic data) also suggests that the uninterpretable φ-features do not have to be obligatorily inherited by T, and that languages can vary in whether these features stay on C (yielding complementizer agreement) or migrate down to T (yielding subject–verb agreement). These are two of the three options proposed by Ouali (2008): SHARE and KEEP, given in (25a–b). If C shares its features with T, the result is both complementizer agreement and subject verb agreement. If C keeps its features, the result is just complementizer agreement. (25) a. Cuφ b. Cuφ

Tuφ SHARE T KEEP

In all the cases of complementizer agreement under consideration here, the same set of φ-features is taken to be responsible for subject–verb agreement and for complementizer agreement. This claim has recently been challenged by Haegeman & Van Koppen (2011), who argue for dissociating the two, based on the fact that in (26), for example, the complementizer agrees with the first member of the conjunct, but the verb agrees with the entire conjunction phrase. (26)

Ich dink de-s [toow en Marie] kump. [Limburgian] I think that-2SG you.SG and Marie come.PL ‘I think that you and Marie will come.’ (Haegeman & Van Koppen 2011: 3)

Leaving complementizer agreement aside, let us turn to the last piece of evidence in favor of CPs being phases. Given that complements of phase heads undergo Spell-Out, the fact that the complement of C can undergo ellipsis is also evidence in favor of CPS being phases. This is what happens in sluicing constructions. On the most straightforward account of this construction (see Merchant 2001 and the references therein for arguments in favor of such an account), sluicing involves movement of the wh-phrase moves to [Spec,CP], followed by deletion of the TP, as schematized in (27). (27) I know Maria likes someone but I don’t Know [CP whoi [C′ C [TP Maria likes ti ] ] ]

Let us briefly consider two of his arguments. One involves a crosslinguistic correlation between the ability to strand prepositions in sluicing constructions and the ability to strand prepositions in run-of-the-mill wh-questions. English, for example, allows preposition stranding in both cases, as shown in (28a–b).

{DESKTOP}NEW FOLDER//9781107040847C04.3D

86

86

[71–123] 23.11.2013 11:48AM

Classic phases

(28) a. I know that Maria talked to someone but I don’t know [CP whoi [TP Maria talked to ti]. b. Whoi did Maria talk to ti?

On the other hand, Polish (patterning together with other Slavic languages in this respect) disallows preposition stranding in both sluicing and wh-constructions, as shown in (29a–b):21 (29) a. *Wiem, z·e Maria rozmawiała z kims´, [Polish] know.1SG that Maria talk.PST.3SG with someone ale nie wiem [CP kimi [TP Maria rozmawiała z ti] ]. but not know.1SG whom Maria talked with ‘I know that Maria talked to someone but I don’t know whom?’ b. *Kimi Maria rozmawiała z ti? with Maria talk.PST.3SG with ‘Who did Maria talk to?’

This led Merchant to the following generalization: (30) A language L will allow preposition stranding under sluicing iff L allows preposition stranding under wh- movement. (Merchant 2001: 92)

Case provides further evidence in favor of a deletion account. The case of the fronted wh-pronoun is the case it is assigned within the elided clause. In (31a) it is Instrumental not Nominative. This also provides an argument against an analysis of sluicing that derives it from a cleft (illustrated in (31b)), which would incorrectly predict Nominative case instead: (31) a. Maria zainteresowała sie˛ czyms´, ale nie wiem czym/*co. Maria got.interested REFL something.INSTR but not know.1SG what.INSTR/*NOM ‘Maria got interested in something but I don’t know what.’ b. Maria zainteresowała sie˛ czyms´, ale nie wiem co Maria got.interested REFL something.INSTR but not know.1SG what.NOM to jest czym Maria sie¸ zainteresowała. it is what.INSTR Maria REFL get.interested.PST.1SG. ‘Maria got interested in something but I don’t know what (it is that Maria got interested in).’

This concludes our discussion of CPs as phases. The table in (32) summarizes the behavior of CPs (and C heads) with respect to the phasehood diagnostics established in Chapter 3.

21

See, however, Szczegielniak (2008) and Stjepanovic (2008) for some interesting counterexamples to this correlation, from Polish and Serbo-Croatian, and Sato (2011) for similar evidence from Indonesian.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

87

[71–123] 23.11.2013 11:48AM

CPs as phases

87

(32) CPs as phases Diagnostic question

Answer Evidence

Does C trigger Spell-Out?

Yes

Can the complement of C be elided? Can an element moving out of CP be interpreted in [Spec,CP]? Can an element moving out of CP be pronounced (partially or completely) in [Spec,CP]?

Yes Yes

VP-internal elements inaccessible sluicing reconstruction

Yes

preposition stranding (subject to crosslinguistic variation) quantifier stranding wh-copying scope marking

Is CP a domain for feature valuation?

Yes

Is C the source of uninterpretable features?

Yes

finite CPs as domains for Nominative case valuation uφ-features uwh-features

Let me conclude this section with a brief comparison of CPs with TPs with respect to the same diagnostics. So far, we have assumed TPs not to be phases, but let us see if the behavior of TPs with respect to the diagnostics actually confirms this assumption. If TPs were phases (and T heads were phase heads), we would expect a number of things from them. First, we would expect T heads to trigger Transfer/ Spell-Out. This would put limits on Agree (since on standard assumptions, no Agree is possible into spelled out constituents). For example, if both T and v were phase heads in (33a), VP would contain an object whose case feature could not be valued. (33) a. Piotrowi podobaja˛ sie¸ te Peter.DAT please REFL ‘Peter likes these flowers.’

b.

kwiaty. flowers.NOM

[Polish]

TP T′ Tuϕ [ ], EEP

vP

DP

v′ VP

v V

DPuC[

]

{DESKTOP}NEW FOLDER//9781107040847C04.3D

88

88

[71–123] 23.11.2013 11:48AM

Classic phases

Second, we would expect ellipsis to be able to target vP. While Merchant does analyze pseudogapping as vP ellipsis, this could be due to the fact that T might have inherited whatever features make it a licensor for vP ellipsis from C. Thus we are only going to look at cases in which T is not dominated by (a finite) C, and it could not have inherited any features from it. Raising infinitives (both subject to subject raising and exceptional case marking) are a case in point. Unlike control infinitives, they are assumed to be TPs, not CPs. Evidence to support this distinction comes from tense considerations. Stowell (1982) argues that raising infinitives are nonfinite and non-tensed (i.e. involve a T[-finite, -tensed]), whereas control infinitives are non-finite but tensed (i.e. involve a T[-finite, +tensed] ). The tense of an embedded control infinitive clause is interpreted as unrealized or future with respect to the tense of the matrix clause. By contrast, the tense of an embedded raising infinitive clause is interpreted as simultaneous with respect to the tense of the matrix clause. This is shown by the contrast in (34). In (34a), the time of parrots flying in the rainforest is typically understood to be future with respect to the time of their desire to do so. In (34b), on the other hand, the two events (my expectation and the event of parrots flying) are understood to be simultaneous. (34) a. Parrots want to fly in the rainforest. b. I expect parrots to fly in the rainforest.

If being tensed is a property of a C-T complex (rather than of T alone), this contrast receives a natural explanation: there is a C head in (34a) but not in (34b). ECM infinitives (and raising infinitives more generally) are thus good diagnostic cases, as there is independent evidence that they do not involve a CP layer (so Feature Inheritance is not a confounding factor). If T were a phase head, we would expect it to be able to license ellipsis by itself. The fact that raising and ECM infinitives do not license VP ellipsis, as shown in (35a–b), suggests that T is not a phase head.22 (35) a. *I expect cockatoos to be funny but I don’t consider blue-headed macaws to Δ. b. *Cockatoos are able to fly but kakapos don’t appear to Δ.

The next diagnostic concerns successive-cyclic movement. If TP were a phase, we would expect wh-movement to target the edge of TP (in addition to targeting the edges of other phases), and the derivation of the wh-question in (36a) to proceed as in (36b) not (36c). (36) a. What do parrots eat? b. [CP WHi [Cʹ C [TP ti [TP parrotsj [vP ti [ tj [vʹ v [VP eat ti]]]]]]]] c. [CP WHi [Cʹ C [TP ti [TP parrotsj [vP ti [ tj [vʹ v [VP eat ti]]]]]]]]

Investigation of such cases, however, is not very illuminating, as the intermediate copy in [Spec,TP] is so close to the final copy. For example, interpreting the copy in 22

See Lobeck (1995) for a classic account of the factors which determine which heads do (and which ones do not) license ellipsis.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

89

[71–123] 23.11.2013 11:48AM

CPs as phases

89

the [Spec,TP] position is indistinguishable from interpreting the copy in [Spec,CP]. Wh-movement in successive-cyclic raising contexts, however, is more informative. Here, the question is whether movement proceeds as in (37b) or (37c). (37) a. Which parrots appear to be likely to nest soon? b. [CP WHi [TP ti [TP…appear…[TP ti [TP…to be likely… [TP ti [TP … ti] ] ] ] ] ] ] c. [CP WHi [TP…appear … [TP ti to be likely [TP … ti…] ] ] ]

Abels (2003) argues against wh-movement proceeding as in (37a), uniformly targeting all the TP edges (hence the term he uses, uniform paths). His crucial contrast is given in (38). If wh-movement took place through the edge of TP, we would expect both (38a) and (38b) to be grammatical. (38) a. Which pictures of himselfi did it seem to Johni that Mary liked? b. *Which pictures of himselfi did Mary seem to Johni to like? (Abels 2003: 30)

The reason (38a) is grammatical is that which pictures of himself can be interpreted in the intermediate position (bolded in (39)), in which John is the closest binder for the anaphor himself contained within the wh-trace/copy: (39) [CP Which pictures of himselfi [C0 did [TP it seem to Johni [CP ti that [TP Mary liked ti ]]]]]?

However, if wh-movement targeted intermediate [Spec, TP] positions, we would expect (38b) to be equally grammatical, with the anaphor himself also bound by John in the intermediate (bolded) position: (40) *[CP Which pictures of himselfi [C0 did [TP Mary seem to Johni [TP ti [TP to like ti ]]]]]?

This indicates that TP cannot be a phase, since wh-movement cannot proceed through its edge. A related question concerns A-movement. The question is whether A-movement takes place through intermediate embedded [Spec,TP] positions (as in (41a)), or movement in one step (as in (41b)).23 Such examples are often brought to bear on the issue of the status of EPP in the grammar. If T were a phase head (and EPP were the uninterpretable feature by assumption only associated with phase heads), we would expect any movement out of TP to have to proceed through the edge of TP. (41) a. The parrots appear to be likely to nest soon.

23

b.

[TP The parrotsi appear [TP ti to be likely [TP ti to nest soon] ] ]

c.

[TP The parrotsi appear [TP to be likely [TP ti to nest soon] ] ]

PIC does not discriminate between A-bar and A-movement. The fact that passive movement does not target the [Spec,vP] as an intermediate landing site does show that passive vPs are not phases, or at least the same kinds of phases as active vP. This is the point I discuss in the next section.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

90

90

[71–123] 23.11.2013 11:48AM

Classic phases

Epstein & Seely (2006) and Grohmann, Drury & Castillo (2000) argue explicitly against A-movement proceeding in a successive-cyclic fashion. Many of their arguments are quite theoretical in nature, involving the general status of EPP in the grammar, or the nature of chains formed by this movement. Empirical considerations of the familiar kind, such as binding of quantifier stranding, however, do seem to suggest that A-movement takes place through the intermediate specifiers of TP.24 Consider the following example: (42) a. Johni seems to Mary to appear to himselfi to be ill. b. *Mary seems to Johni to appear to himselfi to be ill. (Grohmann, Drury & Castillo 2000: 160)

If John and Mary moved directly from their thematic positions, as shown in (43a–b) below, there would be no way to account for this contrast. In both cases, they would be moving from the positions in which they do not bind the anaphor. (43) a. Johnj seems to Maryi [TP to appear to himselfi [TP to be tj ill] ] b. *Maryj seems to Johni [TP to appear to himselfi [TP to be tj ill] ]

Successive-cyclic derivation, on the other hand, can capture this contrast; in (43b), schematized in (44b), there is a copy of the matrix subject Mary in the intermediate subject position, which causes a Principle A violation: (44) a. Johnj seems to Maryi [TP tj to appear to himselfi [TP tj to be tj ill] ] b. *Maryj seems to Johni [TP tj to appear to himselfi [TP tj to be tj ill] ]

Quantifier stranding, also discussed by Grohmann, Drury & Castillo (2000), shows that movement proceeds through the embedded specifier of TP, which is one of the positions in which the quantifier can be stranded: (45) a. The parrots appear all to be ready to nest soon. b. The parrots appear to all be ready to nest soon. c. The parrots all appear to be ready to nest soon.

Bošković (2002a) also presents evidence in favor of intermediate landing sites in raising configurations. In addition to the quantifier float and Condition A data of the kind discussed above, he discusses variable binding (following Lebeaux 1991 and Nunes 1995). In (46a), reconstruction to the position necessary for variable binding results in a Principle C effect. In (46b), on the other hand, there exists a position (the intermediate specifier of TP position) in which the pronoun his can be bound by every man without the R-expression his mother being bound by her. (46) a. *[Hisi mother’sj bread]k seems to herj [TP tk to be known by every mani [TP tk to be tk the best there is.]]

24

Grohmann, Drurg & Castillo (2000) attribute this observation to David Pesetsky, who in turn credits Danny Fox. They themselves do not take it to indicate successive cyclic movement.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

91

[71–123] 23.11.2013 11:48AM

vPs as phases

91

b. [Hisi mother’sj bread]k seems to every mani [TP tk to be known by herj [TP tk to be tk the best there is.]] (cf. Boškovic´ 2002: 180)

The table in (47) summarizes the behavior of TPs with respect to phasehood diagnostics. (47) TPs as phases Diagnostic Question

Answer Evidence

Does T trigger Spell-Out?

No

Can the complement of T be elided?

No

Can an element moving out of TP be interpreted in [Spec,TP]? Can an element moving out of TP be pronounced (partially or completely) in [Spec,TP]? Is TP a domain for feature valuation?

No

Is T the source of uninterpretable features?

Yes

No No

Agree between T and Nominative object would be impossible lack of VP ellipsis with ECM/ raising verbs no reconstruction in successive raising contexts quantifier stranding in [Spec,TP] no Nominative case in nonfinite contexts no φ-feature agreement (unless T is dominated by a finite C)

4.2 vPs as phases In this section, we examine the status of vPs with respect to the phasehood diagnostics we established in Chapter 2 and applied to CPs in the previous section. The diagnostics, relativized to vPs, are repeated in (48). (48) a. Does v trigger Spell-Out? b. Can the complement of v be elided? c. Can an element moving out of vP be interpreted at the edge of vP? d. Is vP a target for Quantifier Raising? e. Can an element moving out of vP be pronounced (partially or completely) at the edge of vP? f. Can an element moving out of vP strand anything at the edge of vP? g. Is vP a domain for feature valuation? h. Is v the source of uninterpretable features?

We will examine different types of vPs, distinguishing those selecting an external argument (transitive and unergative ones) from those not selecting one (passive and unaccusative ones). The question of whether passive and unaccusative vPs are phases relies on the not uncontroversial assumption that there is a vP layer in such clauses to begin with. We know that passive and unaccusative verbs differ from transitive and unergative ones in that they (i) do not value Accusative case,

{DESKTOP}NEW FOLDER//9781107040847C04.3D

92

[71–123] 23.11.2013 11:48AM

92

Classic phases

and (ii) do not select an external argument. This is Burzio’s Generalization, which becomes easy to account for if there are no vPs in passives and unacccusatives. However, irrespective of whether passives and unaccusatives involve the structures in (49b) and (50b) (with a vP) or the structures in (49c) and (49c) (with no vP), the question of whether they constitute phases is still valid. It just becomes a question about passive and unaccusative VPs rather than vPs. (49) a. Tom arrived.

b.

TP T′ Tuϕ [

],iT[pst], EPP

vP VP

v

c.

V arrived

TP

DPiϕ [3sg,masc],uC[

]

Tom

T′ Tuϕ [

],iT[pst], EPPVP

V arrived

DPiϕ [3sg,masc],uC[

]

Tom

(50) a. Tom was chosen.

b.

TP T′ Tuϕ [ was

], iT[pst], EPP

vP VP

v V chosen

DPiϕ [3sg,masc], uC[ Tom

c.

TP T′ Tuϕ [ was

], iT[pst], EPP

V chosen

VP DPiϕ [3sg,masc], uC[ Tom

]

]

{DESKTOP}NEW FOLDER//9781107040847C04.3D

93

[71–123] 23.11.2013 11:48AM

vPs as phases

93

Let us first look at the properties of v heads in order to determine whether v is the locus of uninterpretable features and whether v determines Spell-Out. The answer to the first question seems fairly straightforward, at least for transitive (and unergative) vPs. The idea that in transitive clauses v is responsible for structural Accusative case and that the Agree relationship between v and the object in its domain, shown in (51a–b), values Accusative case is standard, and I will not dwell on it here. (51)

a.

vP vuϕ [

VP

]

V read

DPiϕ [3sg],uC[

]

a book

b.

vP VP

vuϕ [3sg] V read

DPiϕ [3sg], uC[Acc] a book

We also know that the complement of v can be elided; this is VP Ellipsis (VPE), perhaps one of the best-studied ellipsis types. As we saw in Chapter 3, if phase heads determine points of Transfer, VPE can be thought of as a null SpellOut of the complement of v. (52) a. Mary talked to Bill and Tom did, too.

b.

…

CP C

TP Tom

T′

Tuϕ [3sg], EP vP did

v

VP

SPELL-OUT DOMAIN

talk to Bill

Since v in this case is empty, we might wonder why ellipsis couldn’t target vP instead. Merchant (2008) shows that VPE is indeed deletion of a VP (and not a vP),

{DESKTOP}NEW FOLDER//9781107040847C04.3D

94

[71–123] 23.11.2013 11:48AM

94

Classic phases

based on the fact that it allows voice mismatches in which the antecedent clause is active and the elided clause is passive or vice versa.25 (53) a. This problem [VP1 was to have been looked into], but obviously nobody did [VP2 look into this problem]. PASSIVE VP1 /ACTIVE VP2 b. The janitor must [VP1 remove the trash] whenever it is apparent that it should [VP2 be removed]. ACTIVE VP1 /PASSIVE VP2 (cf. Merchant 2008: 169)

Merchant accounts for the availability of such mismatched examples by assuming that v (which includes voice information) is not included in the elided portion, which is what allows the v head in the antecedent clause to be passive and the one in the elliptical clause to be active, as shown in (54). (54) &P &′ CP

but

CP C

T

vP

vP

vivoi[pass] VP look_into

nobody DP

this problem

v′

vivoi[act]

VP

look into this problem

The edge diagnostics, which helped us determine whether movement out of a phase has to proceed through the phase edge, were perhaps the most fruitful diagnostics in the discussion of CPs as phases, so let us apply them to vPs. If vPs are phases, successive cyclic movement would have to proceed as in (55b), rather than proceeding as in (55a), resulting in two ‘extra’ copies, given in bold. (55) a. [CP WH [Cʹ C [TP . . . [CP WH [Cʹ C [TP . . .WH . . . ] ] ] ] ] ] b. [CP WH [Cʹ C [TP [vP WH [vʹ v . . . [CP WH [Cʹ C [TP [vP WH [vʹ v . . .WH . . . ]]]]]]]]]] 25

If pseudogapping, as Merchant (2008) argues (based on the fact that unlike VPE it does not allow voice mismatches), is vP (rather than VP) ellipsis, the generalization that only complements of phase heads allow ellipsis breaks down. Boškovič argues that in pseudogapping cases, it is not a v but a higher (verbal) projection that is a phase head, which makes vP ellipsis the ellipsis of the entire phase head. Boškovič’s relativized approach to phases is something we will discuss in more detail in Chapter 6.

95

{DESKTOP}NEW FOLDER//9781107040847C04.3D

[71–123] 23.11.2013 11:48AM

vPs as phases

95

A related question is whether [Spec,vP] is ever the final (as opposed to the intermediate) landing site for wh-movement, as shown in (56a) for a monoclausal wh-question and in (56b) for wh-movement that spans across clausal boundaries. (56) a. [CP [Cʹ C [TP[vP WH [vʹ v . . . [VP . . .WH . . . ]]]]] ] b. [CP [Cʹ C [TP[vP WH [v 0 v . . . [CP WH [C0 C [TP [vP WH [v 0 v . . .WH . . . ]]]]]]]]]]

This is one way to think about so-called short wh-movement, overt whmovement to a clause-medial position. The examples in (57a–e) provide an illustration: this is how Jayaseelan (2001) analyzes Malayalam, Aldridge (2010) archaic Chinese, Manetta (2010) Hindi–Urdu, and Citko (2010) a subset of wh-questions in Polish.26, 27 It is also a plausible way to think of Hungarian wh-questions. (57) a. nin-ne aarə aTiccu? [Malayalam] you-ACC who beat.PST ‘Who beat you?’ (Jayaseelan) b. Wu shei qi? Qi tian hu? [Archaic Chinese] I who deceive deceive Heaven Q ‘Who do I deceive? Do I deceive Heaven?’ (cf. Aldridge 2010: 2, from Analects 9) c. Hamid-ne kya: par.ha:? [Hindi–Urdu] Hamid-ERG what read ‘What did Hamid read?’(Manetta 2010: 4) d. Co Ewa komu dała? what.ACC Ewa who.DAT gave ‘What did Ewa give to whom?’(Citko 2010: 46) e. János kit mutatott be Marinak?

[Hungarian]

John who.ACC introduced PRT Mary.to ‘Who did John introduce to Mary?’ (Cable 2008: 3)

26

The idea that Polish wh-movement can target a clause medial position goes back to Dornisch (1998), who also points out that this medial position has to be an A-bar position, based on the fact that it licenses parasitic gaps: (i) Za

co

by

Piotr

kogoi

wyrzucił

nie

Piotr

who

throw.out

not

wysłuchawszy pgi

przedtem?

listen.PART before ‘For what would Peter throw out whom without having listened to?’ (Dornisch 1998: 160) for

27

what

COND

However, many researchers working on Hungarian take this position to be a more specific focusdesignated projection (see, for example, Kiss 1987). Cable (2008), however, argues against this position.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

96

96

[71–123] 23.11.2013 11:48AM

Classic phases

In successive-cyclic movement contexts, there are a number of ways we could diagnose the presence of these extra copies. While I am not aware of any cases in which all the copies are realized (as in (58a)), there are cases in which the copies surface in [Spec,vP] (as in (58b)).28 (58) a. [CP WH [Cʹ C [TP [vP WH [vʹ v . . . [CP WH [C 0 C [TP [vP WH [v 0 v . . .WH . . . ]]]]]] b. [CP WH [Cʹ C [TP[vP WH [vʹ v . . . [CP WH [C 0 C [TP [vP WH [v 0 v . . .WH . . . ]]]]]]

We saw above that Manetta (2010) analyzes wh-movement in Hindi–Urdu as targeting the [Spec,vP] position. In this respect, she contrasts Kashmiri (a language in which C is the phase head ‘responsible’ for wh-movement) with Hindi–Urdu (a language in which v head is the phase head responsible for whmovement). However, since Hindi–Urdu is an SOV language, (57c) above in itself is not evidence for wh-movement targeting [Spec,vP] position, as it is also compatible with the wh-pronoun never leaving its initial position. The example in (59), however, is incompatible with the in-situ analysis, as the wh-phrase originates in the embedded clause.29 (59) Sita-ne kis-ko soca: ki Ravi:-ne dekha:? Sita-ERG who-ACC thought that Ravi-ERG saw ‘Who did Sita think that Ravi saw?’ (Manetta 2010: 1)

Hindi–Urdu also has a wh-scope-marking construction, in which the scope marker is at the edge of the matrix vP and the ‘real’ wh-phrase remains inside the embedded clause, as shown in (60a–b)). (60b) also shows that such expletive wh-elements appear in every clause separating the clause containing the real wh-phrase and the clause in which the wh-phrase gets its scope.30 (60) a. Sita-ne kya: soca: ki Ravi:-ne kis-ko dekha:? [Hindi–Urdu] Sita-ERG EXPL thought that Ravi-ERG who-ACC saw ‘Who did Sita think that Ravi saw?’ (Manetta 2010: 1) b. Ra:m-ne kya: soca: ki Ravi:-ne kya: kaha: ki kon sa a:dmi: a:ya:? Ram-ERG EXPL thought that Ravi-ERG EXPL said that which man came ‘Which man did Ram think that Ravi said came?’ (Manetta 2010: 24)

28

Example (59) could also involve the derivation in (i), in which there are no deleted copies in the specifiers of CP, just Agree between C heads and the wh-phrases in [Spec,vP]: (i) [CP [C’ C [TP[vP WH [v’ v . . . [CP [C’ C [TP [vP WH [v’ v . . .WH . . . ]]]]]]

29

30

Manetta also gives independent evidence to argue against the claim that Hindi–Urdu is a wh-in-situ language, coming from adverb placement, for example. The preverbal position that the wh-word occupies is also a ‘generalized’ focus position, something which is also the case in other languages with this type of short wh-movement. Manetta describes the scope marker kya: as a minimal wh-word.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

97

[71–123] 23.11.2013 11:48AM

vPs as phases

97

In principle, the difference between languages like English or Kashmiri, in which wh-phrases surface at the edge of the CP phase, and languages like Hindi–Urdu, in which they surface at the edge of the vP phase, could mean one of the following two things: either wh-movement proceeds through the edge of CP in some languages but through the edge of vP in others, or movement uniformly proceeds through the edges of both CP and vP, but languages ‘choose’ which copies to pronounce. The first possibility is reminiscent of Rackowski & Richards (2005), who argue, based primarily on data from Tagalog, that long-distance movement proceeds directly from the embedded [Spec,vP] to the matrix [Spec,vP]. In Tagalog, as in other Austronesian languages, we see morphological reflexes of this movement in the form of verbal agreement: the extracted element agrees with the verb. In (61) what gets extracted is the indirect object, and we see the dative agreement marker on the verbal complex.31 (61)

Sino ang b-in-igy-an ng lalaki ng buluklak? [Tagalog] who ANG ASP-give-DAT CS man CS flower ‘Who did the man give the flower to?’ (cf. Rackowski & Richards 2005: 587)

The differences between languages in which wh-movement targets the edge of CP and the languages in which it targets the edge of vP could mean that C is a phase head in some languages but v is in others, and, more generally, that phasehood could be subject to parametric variation. We will discuss the latter possibility in more detail in Chapter 6.32 This, however, is neither the route that Manetta takes to explain the difference between Kashmiri and Hindi–Urdu, nor the one Rackowski and Richards take to explain the difference between Tagalog and English. The morphological evidence from Tagalog in (61) parallels in spirit the morphological evidence we used in the previous section to establish that C is a phase head. There we looked at complementizer agreement phenomena; here we are looking at what might be called v-agreement phenomena. This vagreement is indeed what we find in Austronesian languages (see, among others, Chung 1994, 1998, Lahne 2008 on Chamorro, Sato 2012 on Malay/ Bahasa Indonesia, Rackowski & Richards 2005 and Aldridge 2005 on Tagalog). We saw above that the verb agrees with the moved wh-phrase.

31

32

The Tagalog-specific glosses used in these examples are: CS – case ANG – subject, topic, pivot marking Rackowski and Richards actually generalize their claim to English-type languages, and argue that even in such languages, wh-movement proceeds from the embedded [Spec,vP] to the matrix [Spec, vP]. This leaves all the evidence in favor of wh-movement proceeding through the edge of [Spec, CP] in need of an alternative explanation. While not impossible (and Rackowski and Richards do offer some insightful suggestions), it does raise some non-trivial questions.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

98

98

[71–123] 23.11.2013 11:48AM

Classic phases

Conversely, arguments that do not agree with the verb cannot be extracted. This is the well-known extraction restriction of Austronesian syntax. Aldridge attributes it to the Phase Impenetrability Condition: if movement out of a phase proceeds through the phase edge, only elements at the edge can undergo further movements. Let us look at a couple of other examples. Chung (1994) analyzes wh-agreement in Chamorro as ‘agreement with a trace’. The wh-agreement morphology is sensitive to the case of the moved element, which, among other things, leads her to conclude that wh-agreement is not a reflex of C agreeing with a wh-phrase in its specifier. The contrast between (62a) and (62b) provides an illustration. When the wh-trace is Nominative, the wh-agreement is realized as the infix -um-; when it is objective (or oblique), it is realized as -in-. (62) a. Humällum si Maria [na ha-pänak si Juan i pätgun]. AGR.assume Maria COMP AGR-spank Juan the child ‘Maria assumes that Juan spanked the child.’

[Chamorro]

b. Hayi hinalomña si Maria [ t pumänak t i pätgun]? who? WH.assume Maria WH.spank the child ‘Who does Maria assume spanked the child?’ (Chung 1994: 1)

A slightly different, yet related, illustration comes from Indonesian, in which the morphological marking involves the deletion of the active voice prefix men-. Sato (2012) analyzes this as a reflex of the wh-phrase moving through the edge of vP, where it checks the uninterpretable D-feature of v. This is what blocks the insertion of the active voice prefix and forces the insertion of a null one instead. What is crucial for our purposes is that the deletion of the menprefix marks the path site. (63) a. Siapai yang Bill (*mem)-beritahu ibu-nya [CP yang ti *(men)-cintai Fatimah]? who that Bill AV-tell mother-his that AV-love Fatimah ‘Who does Bill tell his mother that loves Fatimah?’ b. Apai yang Ali (*mem)-beri ti kepada Fatimah? to Fatimah what that Ali AV-give ‘What did Ali give to Fatimah?’ (Sato 2012: 34, citing Cole & Hermon 1998: 231–2)

Yet another way to establish the edge property of vPs is to look at stranding possibilities. The question here is whether quantifiers, prepositions or nominals can ever be stranded in [Spec, vP] positions, the same way they can in [Spec,CP] positions. (64) a. [CP WH [Cʹ C [TP . . . [vP WH all [vʹ v . . . [VP . . .WH . . . ] ] ] ] ] ] b. [CP WH [Cʹ C [TP. . . [vP P WH [vʹ v . . . [VP . . .WH . . . ] ] ] ] ] ] c. [CP which [Cʹ C [TP. . . [vP which NP [vʹ v . . . [VP . . .WH . . . ] ] ] ] ] ]

McCloskey shows that West Ulster English, a language that allows quantifier stranding in intermediate [Spec,CP] positions (as we saw in the previous section), disallows analogous stranding in [Spec,vP] positions.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

99

[71–123] 23.11.2013 11:48AM

vPs as phases

99

(65) a. *Whati did he tell all him (that) he wanted ti? [West Ulster English] b. *What did he tell all his friends/Mickey (that) he wanted t? (McCloskey 2001:63)

Preposition stranding yields similar results. It should not come as a surprise that prepositions in English cannot be stranded in [Spec,vP] positions, given that English does not allow P-stranding in [Spec,CP] positions either: (66) Whoj did John [vP [PP to tj]i talk ti ] ]?

We saw, however, in Section 4.1 above that Afrikaans does allow P-stranding in the intermediate [Spec,CP]. However, as noted by Rackowski & Richards (2005), it does not allow stranding at the edge of vP: (67) *Wat dink julle [vP oor dink die bure stry ons]? [Afrikaans] what think you about think the neighbors argue we ‘What do you think the neighbors think we argue about?’ (Rackowski & Richards 2005: 593)

Some evidence in favor of nominal-stranding in the specifier of v comes from languages that allow Left Branch Extraction (LBE). Wiland (2010) shows that in Polish, one such language, the stranded nominal marks the movement path. Thus, the fact that it can be stranded in [Spec,vP] (either in simple or longdistance questions) shows that movement does indeed proceed through the specifier of vP.33 (68) a. [CP Jaki [TP Paweł [vP samochód [VP what Paweł.NOM car.ACC ‘What car did Paweł buy for his wife?’

kupił bought

swojej his.DAT

z·onie t]]]]? wife.DAT

b. %[CP Jaki [TP Maria [vP samochód mys´lała, [CP t z·e Paweł [vP t kupił what Maria.NOM car.ACC thought that Paweł.NOM bought swojej z·onie t]]]]]? his.DAT wife.DAT ‘What car did Maria think Paweł bought his wife?’ (cf. Wiland 2010: 335–6)

So far we have seen mixed results regarding the pronunciation of the whpronoun (or its ‘remnant’) at the edge of vP. On the one hand, we have seen that it can be the site of the wh-copy in Hindi–Urdu, that it can be the landing site of wh-movement in languages like Hungarian, and that the nominal can be stranded in this position in languages with Left Branch Extraction like Polish. On the other hand, we have seen that the wh-related quantifier all cannot be stranded in this position and that prepositions cannot be stranded in this position. I do not take these to be a fatal blow; both phenomena are known to be sensitive to prosodic factors, which could influence the grammaticality of

33

The % symbol indicates speaker variation. I find this example perfectly acceptable.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

100

100

[71–123] 23.11.2013 11:48AM

Classic phases

(65) or (66), as claimed explicitly by McCloskey (1998) with respect to whquantifier stranding in West Ulster English. Furthermore, if P-stranding requires reanalysis, it is also conceivable that such reanalysis is only possible if the verb is in a certain relationship with the stranded preposition, and the specifier of vP is simply not such a position. Let us now turn to the evidence for vPs being phases on the interpretative side. Here the question is whether the wh-phrase can be interpreted at the edge of vP even if it is not its ultimate resting place. The evidence that bears on this issue comes from various types of reconstruction effects (see Agüero-Bautista 2001, Fox 1999, 2000, Lahne 2008, Legate 2003 among others). The reconstruction argument is very similar in spirit to the argument made in Section 4.1 to establish reconstruction into the intermediate [Spec,CP] position. The relevant configuration is similar to the configurations we considered in the previous section (where we established that wh-movement targets [Spec,CP] in a successive-cyclic manner). This time, however, the reconstruction site we are interested in is the edge of vP, not CP. The examples in (69) provide evidence that wh-movement does indeed proceed through this edge. In all of them, the fronted wh-phrase contains a bound pronoun and an Rexpression. They are all grammatical because there exists a reconstruction site, marked with a checkmark, below the quantifier and above the pronoun. This is the edge of the vP.34 (69) a. [Which of the books that he1 asked Ms. Brown2 for] did every student1 _√_ get from her2 * ? b. [Which (of the) paper(s) that he1 wrote for Ms. Brown2] did every student1 __ get her2 * to grade? (Fox 1999: 175)

If we reverse the position of the quantifier and the pronoun, as in (70) below, the results become ungrammatical because there is no reconstruction site such that Principle C is obeyed and the variable is bound; interpreting the wh-phrase in the specifier of the matrix vP results in a Principle C violation and a lack of variable binding. (70) a. *[Which of the books that he1 asked Ms. Brown2 for] did she2 * give every student1 * ? b. *[Which (of the) paper(s) that he1 wrote for Ms. Brown2] did she2 get * get every student1 * to revise? (Fox 1999: 174)

A similar argument, due to Agüero-Bautista (2001), comes from the availability of the so-called Pair List (PL) readings in wh-questions with quantifiers. The 34

Den Dikken (2006a) raises the possibility that the reconstruction site could also be the edge of TP (thus providing an argument in favor of TPs being phases). It is not clear whether the moved whphrase could be tucking in below the subject. He also points out that the argument relies on the assumption that reconstruction is the only way to get the bound variable reading. This does seem to be the simplest assumption.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

101

[71–123] 23.11.2013 11:48AM

vPs as phases

101

crucial assumption that this argument relies upon is that in questions containing a universal quantifier, the wh-phrase has to be interpreted in a position in which it is c-commanded by the quantifier. Thus, the distribution of PL readings in such questions will help us determine possible reconstruction sites. The baseline contrast is given in (71a–b). In (71a) the PL interpretation is possible because the wh-pronoun can reconstruct to a position below the quantified subject. In (84b) there is no such position; hence the PL reading is not available. The only available reading is Single Answer (SA) reading. (71) a. Who do you think everyone saw __ at the rally? SA, PL b. Who __ thinks everyone saw you at the rally? SA, *PL (Agüero-Bautista 2001: 142)

Augero-Bautista shows that the availability of PL readings in the Fox-style examples points towards [Spec,vP] as a reconstruction site, and is thus consistent with the conclusions that Fox drew based on Principle C and variable binding data. Consider in this light the contrast in (72). In (72a), the only possible interpretation is a SA interpretation. This implies lack of reconstruction, as reconstruction to any site would result in a Principle C violation. In (72b), on the other hand, reconstruction to a position below the quantifier is a possibility (and, consequently, a PL reading becomes available), since a potential Principle C violation is not a factor. (72) a. [Which (of the) books that John asked Ms. Brown2 for ] did she2 give every student * ? SA, *PL b. [Which (of the) books that John asked her2 for] did Ms. Brown2 ___ give every student * ? SA, PL (Agüero-Bautista 2001: 145)

Yet another well-known argument comes from the domain of parasitic gaps. It relies on a particular account of parasitic gaps, namely one that involves complex predicate formation, defended in most detail by Nissenbaum (2000), who in turn builds on Larson’s (1988) insights. Such an analysis, Nissenbaum shows, is only possible if the wh-movement licensing the parasitic gap moves through the edge of vP. (73) below illustrates the configuration necessary for parasitic gap formation, with irrelevant details omitted.35

35

I refer the interested reader to Nissenbaum’s work (and the references therein) on how such an account captures the properties typically associated with parasitic gaps, listed in (i) (and to the contributions in Culicover & Postal (2000) for a discussion of some counterexamples to these generalizations). Nissenbaum himself notes, for example, that covert movement can license parasitic gaps in multiple wh-questions if the overtly moved wh-phrase also licenses a parasitic gap. Null operators might seem suspect and there are quite a few of alternative accounts that do not involve null operators (or complex predicates, for that matter). See Hornstein & Nunes (2002) for a sideward movement account, and Kasai (2007) for a multidominant account.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

102

[71–123] 23.11.2013 11:48AM

102

Classic phases

(73) CP which paper C′ did

TP T′

Johni T

vP

which paper

vP1

vP2 John file which paper

PP OPi without PRO reading ti

(cf. Nissenbaum 2000, Larson 1988)

The interpretation Nissenbaum assigns to the relevant nodes is given in (74a–c). (74) a. vP2: λe. λx. there is an event e of John filing x. b. PP: λe. λ.x. there is no subevent of e such that it is the reading of x by PRO. c. vP1: λeλx. there is an event e of John filing x and there is no subevent e’ of e such that e’ is the reading of x by PRO.

The null operator movement inside the adjunct PP forms a predicate, and movement of the wh-phrase to the edge of vP2 forms a predicate out of this vP. Thus, only if vP2 contains a trace of the moved wh-phrase, the vP and PP can combine yielding a complex (conjoined) predicate via predicate modification. This predicate can subsequently apply to the wh-pronoun. If vP didn’t contain a trace of the wh-phrase, it could not combine with the adjunct containing the parasitic gap, because it would not be of the right semantic type. Antecedent-Contained Deletion (ACD) provides yet another argument in favor of movement targeting the edge of vP. This time, the movement in question is Quantifier Raising (see also Bruening 2001 for arguments that vP can be the target for QR). The issue is how to avoid the so-called infinite regress in cases of VP ellipsis in which the antecedent VP contains the missing VP (hence the term Antecedent-Contained Deletion). Consider a simple example in (75a). It contains a missing VP (indicated by Δ). The only VP that could help us recover the content

(i) The antecedent of a P-gap must be in A-bar position. (ii) A P-gap is licensed only at S-structure. (iii) The antecedent of a P-gap must be an NP. (iv) The true gap cannot c-command the P-gap (v) The P-gap is in a chain with the antecedent of the true gap. (Culicover 2001: 65)

{DESKTOP}NEW FOLDER//9781107040847C04.3D

103

[71–123] 23.11.2013 11:48AM

vPs as phases

103

of the missing one is the one containing the missing VP, namely saw every movie Tom did. This leads to infinite regress.36 (75) a. Bill saw every movie Tom did [VP Δ ] b. Bill saw every movie Tom did [see every movie Tom did [VP Δ ]] c. Bill saw every movie Tom did [see every movie Tom did [see every movie Tom did [VP Δ ]] d. Bill saw every movie Tom did [see every movie Tom did [see every movie Tom did [see every movie Tom did [VP Δ ]]]]

QR solves the infinite regress in the following way.37 First the quantified object undergoes QR, as in (76b). In this particular example, it does not matter whether QR targets vP or TP. Bruening (2001) argues in favor of the former option based on economy considerations: adjunction to vP involves shorter movement than adjunction to TP, and is thus more economical. Second, the missing VP gets copied, as in (76c). (76) a. Bill saw every movie Tom did [VP Δ ] b. [TP Bill [vP [DP every movie Tom did [VP Δ]]i [vP v [VP saw ti] ] ] ] c. [TP Bill [vP [DP every movie Tom did [VP saw ti]]i [vP v [VP saw ti] ] ] ]

QR COPY VP

Legate (2003) ingeniously solves the question of why QR could not always target the matrix TP (as is perhaps more standardly assumed) by making the quantified element a Negative Polarity Item (NPI), or by making the subject quantified and testing if the subject can have wide scope with respect to the

36

Copying just see every movie, as in (i) below, is not an option, as this string is not a constituent; irrespective of the specifics of the structure (complementation versus adjunction) and derivation (raising versus deletion) for relative clauses, the string saw every movie Tom did is a constituent, but saw every movie is not. (i) Bill saw every movie Tom did [VP see every movie] (ii)

VP V saw

DP D every

NP NP

37

CP

movie Tom did [VP…] See, among others, Larson & May (1990), Kennedy (1997) for QR-based accounts of ACD, and Baltin (1987), Hornstein (1994) for non-QR based alternatives, which rely on movement to a case position or extraposition instead.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

104

104

[71–123] 23.11.2013 11:48AM

Classic phases

QR-ed object. The first case is illustrated by (77), modeled upon Legate’s examples. In (77a), the only way to resolve infinite regress without moving the NPI out of the scope of negation is to adjoin any movies that Tom did to vP, as shown in (77b), and then copy the missing VP, as shown in (77c). (77) a. Bill didn’t [VP1 see [DP any movies that Tom did [VP2 Δ ]]] b. Bill didn’t [vP [DP any movies that Tom did [VP2 Δ]]i [vP [VP1 see ti]]] c. Bill didn’t [vP [DP any movies Tom did [VP1 see ti]]i [vP [VP1 see ti]]]

QR COPY

VP

Likewise, in (78a), the way to ensure that some linguist maintains wide scope over every paper Bill did is to adjoin the latter to vP, as shown in (78b). (78) a. Some linguist [VP1 read [DP every paper Bill did [VP2 Δ ]]] b. Some linguist [vP [DP every paper Bill did [VP2 Δ ]]i [vP [VP1 read ti]]]

QR

c. Some linguist [vP [DP every paper Bill did [VP1 read ti]] [vP [VP1 read ti]]] COPY

VP

All the examples and arguments discussed in this section involved vPs containing transitive and unergative verbs. For Chomsky, such vPs are complete in terms of their argument structure, whereas their passive and unaccusative counterparts are defective, which leads him to conclude that they are not phases. This conclusion, however, has been challenged by Legate (2003), who applies by now familiar diagnostics to argue that vPs headed by passive and unaccusative verbs are phases. And, interestingly, the conclusion she reaches is that with respect to these diagnostics, unaccusative and passive verb phrases behave the same way as transitive and unergative ones do. This, in principle, could mean one of two things: the diagnostics are ill-conceived (and are perhaps tests for something else); or that passive and unaccusative verb phrases are phases as well, contrary to Chomsky’s view. As I alluded to at the very beginning of this section, the issue is independent of the issue of whether passive and unaccusative verb phrases have a vP or not. The presence of v in these structures seems pretty innocuous, but perhaps this is the reason to be suspicious of its existence; since it does not value Case and does not select an external argument, it might be dispensable in the same way that Agreement projections became dispensable in the mid nineties (see Chomsky 1995, for example). However, there might be independent reasons in favor of having v projections in every verbal structure. This is the view taken by Marantz (2007), who argues explicitly that phase heads are responsible for providing categorial status to categoryless roots. On this view, associated with the Distributed Morphology view of the lexicon, such category-neutral roots acquire their verbal (or nominal) status only by virtue of merging with an appropriate light functional head (a verbal v or a nominal n, for example). This leaves us with the

{DESKTOP}NEW FOLDER//9781107040847C04.3D

105

[71–123] 23.11.2013 11:48AM

vPs as phases

105

structures in (79b) and (80b) as the only possible structures. What is at issue here is not so much the VP versus vP status of passive and unaccusative verbal projections, but the phase versus non-phase status of such verbal projections, be they VPs or vPs. The most significant (and syntactically most visible) consequence of treating VP/vP as a phase is that passive (or unaccusative) movement will proceed through the phase edge, as shown below. This could also be taken as an argument in favor of the presence of the v head, as movement of the complement of V to the specifier of V would violate antilocality, which, as we saw above, prohibits movements that are too short. However, movement from a complement of v to the specifier of vP would not constitute a violation of antilocality.

(79) a. Tom arrived.

b.

TP T′ TiT[pst] uϕ[3sg],EPP vP vP VP

v V arrived

DPiϕ[3sg],uC[Nom] Tom

c.

TP T′ TiT[pst] uϕ [3sg],EPP VP V′ V arrived

DPiϕ [3sg],uC[Nom] Tom

{DESKTOP}NEW FOLDER//9781107040847C04.3D

106

[71–123] 23.11.2013 11:48AM

106

Classic phases

(80) a. Tom was chosen. b.

TP T′ TiT[pst] uϕ[3sg], EPP vP was vP VP

v

V chosen

DPiϕ[3sg], uC[Nom] Tom

c.

TP T′ TiT[pst] uϕ [3sg], EPPVP V′ V chosen

DPiϕ [3sg], uC[Nom] Tom

Most of Legate’s arguments come from phenomena that diagnose the presence of this extra movement step. The logic of the argument is familiar by now; if the whpronoun moved to the edge position, we should find some evidence of this movement – either on the phonological side (in the form of a stranded quantifier, a copy or an expletive wh-phrase), or on the semantic side (in the form of scope reconstruction, for example). In (70) above, we saw evidence for movement through [Spec,vP] coming from the interaction of Principle C reconstruction and variable binding. Those examples all involved transitive or ditransitive verbs like get, read or ask. We also saw that in all those cases, there was evidence for reconstruction to the vP edge position, which was a position that allowed for a bound variable interpretation without inducing a Principle C effect. Parallel passive examples are given in (81a–b); they also show that reconstruction to the edge of vP has to be possible.38 (81) a. [CP [At which of the parties that hei invited Maryj to]k was [TP every mani [vP ti introduced to herj * ] ] ]? b. [CP [At which of the parties that hei invited Maryj to] was [TP shej [vP * introduced to every mani * ] ] ]? (Legate 2003: 507) 38

I refer the reader to Legate’s paper for parallel data involving unaccusative verbs. And to Den Dikken’s (2006a) response for a discussion of some issues stemming from Legate’s proposal.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

107

[71–123] 23.11.2013 11:48AM

vPs as phases

107

We also saw above that the edge of vP can be a landing site for Quantifier Raising. The examples involved Antecedent-Contained Deletion (ACD) with an NPI to ensure that QR does not target the matrix TP. If it did, the NPI would no longer be licensed. The grammaticality of such examples showed that there has to be a QR site below negation, which makes the edge of vP a plausible candidate. Legate shows that passive vPs behave the same way. Example (82a), modeled on Legate’s example, provides evidence. It shows that QR has to target vP; if it targeted a TP instead, the negative polarity item would be outside the scope of negation. (82) a. The movie wasn’t [VP1 seen by [DP anyone that wanted to [VP2 Δ ]]] b. The movie wasn’t [ [ anyone that wanted to [ vP DP VP2 Δ ]]i [vP [VP1 see ti]]]

QR

c. The movie wasn’t [vP [DP anyone that wanted to [VP2 see ti]]i [vP [VP1 seen by ti]]] COPY VP

Legate also brings forth nuclear nuclear stress as an argument in favor of passive and unaccusative vPs being phases. I defer the discussion of nuclear stress (and the question of how phases may play a role in the distribution of nuclear stress) till Chapter 7. Before moving on to DPs, let me summarize the behavior of various types of verb phrases with respect to various types of phasehood diagnostics. (83) vPs as phases

Diagnostic Is vP a domain for feature valuation? Is v the locus of uninterpretable features? Does movement out of vP proceed through the edge? Is the complement of v a Spell-Out domain?

Transitive Unergative Passive Unaccusative vPs vPs vPs vPs Yes

Yes

No

No

Yes

Yes

No

No

Yes

Yes

Yes

Yes

Yes

Yes

No

No

The evidence that transitive and unergative vPs are phases is quite strong; they behave as phases with respect to the diagnostics established in Chapter 3. The evidence in favor of passive and unaccusative vPs being phases is somewhat less unequivocal, however. They do not trigger Spell-Out but they do require movement to proceed through the phase edge. This is the reason why they are sometimes referred to as weak phases, and thus distinguished from strong (transitive and unergative) vP phases.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

108

[71–123] 23.11.2013 11:48AM

108

Classic phases

4.3

DPs as phases

In this section, we turn to the phase status of DPs. There are two – in principle independent, questions to ask. One is whether DPs themselves are phases, and the other one is whether DPs contain other phases. The latter question arises especially in the context of the many well-documented similarities between noun phrases (be they DPs or NPs) and clauses (be they TPs or CPs). If CPs contain other phases (vPs, whose phasehood we established in the previous section), it may be the case that DPs contain phases as well. The well-documented similarities between DPs and CPs involve argument structure, agreement and case morphology, and extraction possibilities. See Bernstein (2001a) for a succinct summary (and Abney 1987, Chomsky 1970, Hiraiwa 2005, Szabolcsi 1983, among many, many others, for the original observations). In many languages, possessors and subjects bear the same case. A classic illustration comes from Szabolcsi’s (1983) work on Hungarian; she shows that in Hungarian DP-internal possessors and clausal subjects bear the same (Nominative) case: (84) a. (a) Mari-ø the Mary-NOM ‘Mary’s guest’

vendég-e-ø guest-POSS-3SG

b. Mari-ø alud-t-ø Mary-NOM sleep-PAST-3SG ‘Mary slept.’ (Szabolcsi 1983: 89–90)

The distinction between A and A-bar positions, familiar from a clausal domain, is operative inside DPs as well. Valois (1991), Szabolcsi (1983) and Tellier (1988), among others, show that the specifier of DP is an A-bar position, providing an escape hatch for movement, and Bernstein (2001b) shows that there is a DP-internal focus position.39 We will look at some of their evidence shortly. For now, suffice it to note that given the many similarities between noun phrases and clauses, it is not unreasonable to expect DPs to be phases, paralleling CPs in this respect.40 The answer to the question of whether DPs contain phases, and if so, what these DP-internal phases might be, depends largely on what the internal structure of DPs is. So before delving into DP phasehood, let us digress into DP-internal syntax. There are a number of DP-internal projections that have been proposed over the years, mainly to accommodate all the DP-internal elements, and to capture the relative ordering of these elements as well as the crosslinguistic

39

40

On a related note, there is also interesting evidence that CP and DP-internal movements can be equally impaired in agrammatism (see, for example, Rausch, Butchert & De Bleser 2005 for a case study involving a German agrammatic aphasic). Chomsky initially did not include DPs in the list of phases, but suggests this as a possibility in later writings.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

109

[71–123] 23.11.2013 11:48AM

DPs as phases

109

variation with respect to this ordering (see Alexiadou, Haegeman & Starrow 2007, Bernstein 1993, 2001a, 2008, Cinque 2002, Ritter 1992, 1993, Valois 1991, Zamparelli 1996, among many, many others). Svenonius (2007b), based on a crosslinguistic study of the ordering of articles, demonstratives, plural markers, classifiers, numerals and adjectives, arrives at the following hierarchy: (85) Dem > Art > Num > unit > Pl/sort > Adj > n > N

Below are some of the more common nominal functional projections posited to accommodate these elements; others include Kase Phrase (Bittner & Hale 1996), Quantifier Phrase, Person Phrase, and Demonstrative Phrase.41 (86) a. Number Phrase (NumP) (Ritter 1992) b. Gender/Classifier Phrase (GenP/ClassP)42 (Picallo 1991) c. Possessive Phrases(PossP) (Valois 1991) d. nP

I will henceforth assume the following (somewhat streamlined) DP structure:43 DP

(87)

D′ D

PossP Poss′ Poss

NumP Num′ Num

nP n′ n

41

42

43

NP

The list gets bigger if we assume that nouns of different semantic types project different structures. Zamparelli (1996) adds Kind Phrases, Predicative Phrases and Strong DPs to the inventory of nominal projections. See, however, Ritter (1993) for an argument that there is no syntactic Gender projection, and gender can be either part of the Number head or the Noun head itself. This hierarchy abstracts away from the relative position of different types of adjectives with respect to each other and with respect to other DP-internal elements. If adjectives occupy specifiers of functional heads, the structures will get exponentially more complex (see Scott 2002, Cinque 2010, Laenzlinger 2005, Bernstein 1994 for some representative accounts). An example hierarchy, based on Scott (2002), is given in (i). For him, this adjectival hierarchy reflects the hierarchy of heads licensing them, making the internal structure of DP quite elaborate. (i) Ordinal > Cardinal > Subject Comment > Evidential > Size > Length >Height > Speed > Depth > Width > Temperature > Wetness > Age > Shape > Color > Nationality/Origin > Material

{DESKTOP}NEW FOLDER//9781107040847C04.3D

110

110

[71–123] 23.11.2013 11:48AM

Classic phases

With these brief remarks on DP-internal structure as background, we can apply to DPs the diagnostics we established during our discussion of CPs and vPs in the previous sections. They are given in (88a–i).44 (88) a. Is DP a domain for feature valuation? b. Is D a source of uninterpretable features? c. Does movement out of DP proceed through the edge of DP? d. Can an element moving out of DP be interpreted at the edge of DP? e. Can an element moving out of DP be pronounced at the edge of DP? f. Can an element moving out of DP strand anything at the edge of DP? g. Is DP a target for Quantifier Raising? h. Does D trigger Spell-Out? i. Can the complement of D be elided?

Some of the diagnostics are subsumed by others; the way we can tell whether movement out of DP proceeds through the edge of DP is by looking at reconstruction (is there a copy in this position?), ordering (can a copy be pronounced in this position?), stranding (can something be stranded at this position?). Similarly, the way we can tell whether D triggers Spell-Out is by looking at DP-internal ellipsis and stress assignment. The discussion of DPs as phases in this section is to a large extent informed by Matushansky (2005), who discusses the behavior of DPs with respect to many of the diagnostics listed in (89), and concludes that since DPs behave as phases with respect to some diagnostics (PF diagnostics) but not with respect to others (LF diagnostics), ‘the behavior of DPs does not bode well for phases’ (Matushansky 2005: 179). Two of the diagnostics in (88) involve uninterpretable features: only phase heads can be inherent hosts of uninterpretable features (although non-phase heads can inherit them), and only phases can be domains for feature valuation. I take it to be uncontroversial that DPs have case features, and at least for structural case, that these case features are uninterpretable.45 However, the uninterpretable case features on DPs are different from the uninterpretable φ-features on C and v heads. An uninterpretable Case feature is not a feature that makes DP (or its head) a Probe, which is what we saw with uninterpretable φ-features on v or C heads.46 Instead, an uninterpretable case feature of a DP makes it an active Goal. This brings us to the next question, the question of whether DPs are (or contain) domains for feature valuation. There is evidence that DPs can contain a 44

45

46

The list of questions is ordered slightly differently to reflect the order of presentation in this section. This is independent of whether a case feature is analyzed as a sui generis feature (uC[ ] feature, as we have assumed throughout) or an uninterpretable counterpart of some other interpretable feature (uT[ ] for Nominative case and uAsp[ ] for Accusative case, the possibility we alluded to in Section 1.3). Strictly speaking, a phrase (as opposed to a head) cannot be a Probe.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

111

[71–123] 23.11.2013 11:48AM

DPs as phases

111

separate case domain, which suggests that either a D (or some DP-internal head) has to be the source of this case. Genitive case is a prime example. We can see it even in a very morphologically impoverished language like English, if we take the preposition of to be a marker of genitive case (see Chomsky 1986, for example). A more straightforward illustration comes from Slavic languages, in which – descriptively speaking – nouns assign Genitive case (so-called adnominal genitive), as illustrated below with Polish data. In (89a) the adnominal genitive is a possessor, but this genitive is not limited to possessors, as shown in (89b–d). (89) a. ksiaz·ka studenta book.NOM student.GEN ‘a student’s book’ b. artykuł profesora article.NOM professor.GEN ‘an article (written) by a professor’ c. kon´ wysokiej klasy horse.NOM high.GEN class.GEN ‘a high-class horse’ d. obje¸cie władzy assumption.NOM power.GEN ‘assumption of power’ e. kawałek chleba piece.NOM bread.GEN ‘a piece of bread’

[Polish] POSSESSOR

AGENT

ATTRIBUTE

THEME

PARTITIVE

(Swan 2002: 330–2)

Another illustration comes from the domain of numerals. Slavic numerals higher than five are also case assigners; the lower ones (one, two, three, four) agree in case with their complements (as shown in (90a)), and only get Genitive case if the entire DP gets Genitive case, as shown in (90b) with respect to the so-called Genitive of negation. The higher numerals, on the other hand, assign Genitive case (the so-called Genitive of quantification) to their complements, as shown in (91a). Not surprisingly, this Genitive case stays Genitive under negation.47

47

This pattern is only found in structural case contexts. The situation changes in lexical case contexts: (i) Jan

zainteresował

sie¸

dwoma

je¸zykami

słowian´skimi.

Jan

got.interested

REFL

two.INSTR

languages.INSTR

Slavic.INSTR

‘Jan got interested in two Slavic languages.’ (ii) Jan

zainteresował

sie¸

szes´cioma

je¸zykami

słowian´skimi.

Jan

got.interested

REFL

six.INSTR

languages.INSTR

Slavic.INSTR

‘Jan got interested in six Slavic languages.’

{DESKTOP}NEW FOLDER//9781107040847C04.3D

112

112

[71–123] 23.11.2013 11:48AM

Classic phases

(90) a. Przeczytałam jedna˛/dwie/trzy/cztery ksia˛z·ki. read.1SG one/two/three/four.ACC books.ACC ‘I read one/two/three/four books.’ b. Nie przeczytałam jednej/dwóch/trzech/czterech not read one/two/three/four.GEN (91) a. Przeczytałam pie¸c´/szes´c´/siedem ksia˛z·ek. read five/six/seven.ACC books.GEN ‘I read five/six/seven books.’ b. Nie przeczytałam pie¸ciu/szes´ciu/siedmiu not read five/six/seven ‘I didn’t read five/six/seven books.’

ksia˛z·ek. books.GEN

ksia˛z·ek. books.GEN

The distribution of adnominal genitive case thus suggests that there has to be a distinct case domain within DP, and is compatible with some DP-internal head (like the Num head) being the source of this genitive case. A different argument, also pointing towards DPs being self-contained domains, comes from binding.48 Well-known examples of the kind given in (92) show that only certain types of DPs (namely the ones that contain possessors) count as domains for binding. This raises the possibility that maybe only certain types of DPs are phases (or strong phases), the possibility we considered in the last section regarding different types of vPs. (92) a. *Johni likes [Mary’s descriptions of himselfi]. b. Johnj likes [Mary’si descriptions of herselfi]. c. Johni likes these descriptions of himselfi.

The contrast in (93a–b), due to Chomsky (1986), is also compatible with this view if DPs can contain a PRO possessor/subject, which is coindexed with the matrix subject in (93a) but not in (93b). (93) a. The childreni heard [PROi stories about each otheri] b. The childreni heard [PROj stories about themi]

48

This argument builds on Quinoli’s (2008) proposal that binding is determined cyclically at a phase level. His focus, however, is on CPs and vP as binding domains, rather than DPs. While a full review of his arguments goes beyond the scope of this chapter, he does show, for example, that Huang’s (1993) predicate fronting cases, such as the one given in (i), receive a natural explanation if binding is determined derivationally and vPs, being phases, constitute binding domains. (i) [CP [vP t1 criticize himself1/*2] John2 never thought [tvp that [Bill1 would tvP] ] ]

Quinoli also uses the ambiguity of examples like (ii) to argue against TPs being binding domains (and phases). If TP were a binding domain, anaphoric interpretation would be fixed at the embedded TP level. (ii) Johnj wonders which pictures of himselfi/j Billi saw. (Quinoli 2008: 313)

{DESKTOP}NEW FOLDER//9781107040847C04.3D

113

[71–123] 23.11.2013 11:48AM

DPs as phases

113

The next set of diagnostics refers to what we called above the edge property. Here, the question is whether movement from a DP has to proceed through the edge of DP. We know that movement out of a phase has to proceed through the edge of a phase. In cases of other phases, we looked at reconstruction, Quantifier Raising, various types of stranding, copying and agreement phenomena to determine movement paths. For independent reasons, however, not all of them will be applicable to DPs. For example, since DP-internal elements often agree in φ-features and case (due to whatever mechanism underlies case concord), it is not a surprise that they will continue to agree if a subpart of DP gets extracted. This is shown in (94a–c) for Polish. (94a) shows that DP-internal elements agree, and (94b–c) show that movement preserves this agreement. Thus, agreement on extracted elements is not going to tell us whether movement took place through the specifier of DP or not. (94) a. Jan przeczytał ciekawa˛ ksia˛z·ke ¸. Jan read interesting.3SG.FEM.ACC book.3SG.FEM.ACC ‘Jan read an interesting book.’ b. Ciekawa ˛i Jan przeczytał [DP ti ksia ˛z·ke¸]. interesting.3SG.FEM.ACC Jan read book.3SG.FEM.ACC c. Która˛i Jan przeczytał [DP ti ksia ˛z·ke¸]? which.3SG.FEM.ACC Jan read book3SG.FEM.ACC ‘Which book did Jan read?’

[Polish]

The fact that movement from a DP is blocked if the specifier of DP is filled might be taken as evidence that movement does indeed have to proceed through the edge of DP. (95) *Whoi did Jan like Maria’s picture of ti?

However, the ungrammaticality of examples of this kind, as pointed out by Matushansky (2005), is part of a larger phenomenon, more likely linked to specificity than definiteness. First, a D head filled by a definite article also blocks movement, and so does one filled by an indefinite article (as long as the noun phrase in question is interpreted as specific). Second, not all definite noun phrases block extraction.49, 50 (96) a. *Whoi did Jan like the picture of ti? b. *Whoi did Jan like a specific/particular/certain picture of ti? c. Whoi did Jan take the best picture of ti?

49 50

These examples are modeled on Matushansky’s examples. Valois (1991) makes a similar point based on French data, showing that demonstratives in French (which he takes to occupy [Spec,DP] positions) block not only movement but also negative polarity item licensing and parasitic gap licensing. However, similar caveats apply.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

114

[71–123] 23.11.2013 11:48AM

114

Classic phases

More generally, as discussed by Davies & Dubinsky (2003), for example, extraction from English DPs is a fairly complex phenomenon, sensitive to factors like the semantics of the selecting verb, argument structure and incorporation possibilities, rather than just a filled versus empty specifier position. Furthermore, it is not clear what aspect of Phase Theory (or the Phase Impenetrability Condition) leads to the expectation that movement from a phase is impossible if the specifier of that phase head is filled. We might think that extraction from CPs proceeds in this fashion; this is the standard way to account for wh-island effects. However, a C head occupied by an interrogative complementizer also results in ungrammatical extraction, so a filled specifier cannot be the only culprit. We also know that there are languages that do not show wh-island effects, which (at least for some of them) is attributed to the availability of multiple [Spec,CP] positions. Thus it is not unreasonable to expect the same parametric option to be available for DPs. Likewise, a filled specifier does not block movement out of the vP. If it did, nothing would ever be able to move out of transitive vPs, since the specifier is filled by the external argument. This also shows that adding an extra (non-thematic) specifier must be an option. These considerations do not support the conclusion that a filled phase edge necessarily blocks extraction out of that phase. There is, however, a fair amount of crosslinguistic evidence pointing towards movement out of DPs proceeding through DP edges. Szabolcsi’s work on Hungarian DPs provides perhaps the best-known argument in favor of the specifier of DP providing an escape hatch for movement out of DPs. First, Szabolcsi shows that Hungarian has two types of possessors, differing in case and position relative to the determiner. One follows the determiner and is marked with Nominative case, as shown in (97a). The other one precedes the determiner and is marked with Dative case, as shown in (97b). (97) a. a the

Mari-ø Mary-NOM

b. Mari-nak Mary-DAT

a the

vendég-e-ø guest-POSS-3SG vendég-e-ø guest-POSS-3SG (Szabolcsi 1983: 89–91)

Furthermore, Szabolcsi shows that the specifier of DP is a landing site for DPinternal operators and that this DP-internal movement turns the entire DP into an operator: (98) a. *(a) ki-ø vendég-e-ø the who-NOM guest-POSS-3SG ‘whose guest’ b. ki-nek a vendég-e-ø who-DAT the guest-POSS-3SG ‘whose guest’ (Szabolcsi 1983: 91)

{DESKTOP}NEW FOLDER//9781107040847C04.3D

115

[71–123] 23.11.2013 11:48AM

DPs as phases

115

Thus, only wh-phrases that are dative-marked possessors can undergo further movement, as shown by the following contrast with respect to topicalization. (99) a. Péter-neki, csak Peter-DAT only

Mari Mari

látta [DP ti saw

a the

kalap-ját] hat-POSS.3SG.ACC

b. *Peter-øi csak Mari látta [DP ti a kalap-ját] Peter-NOM only Mari saw the hat ‘As for Peter, only Mari saw his hat.’ (Gavruseva 2000:750, citing Szabolcsi 1994: 205)

The availability of a DP-related escape hatch position could then be what accounts for the contrast between languages that allow possessor extraction and those that do not. This is the line of thought taken by Gavruseva (2000), who contrasts Hungarian, Tzotzil and Chamorro (languages that allow possessor extraction) with Germanic languages, which overwhelmingly do not allow it. More specifically, she links the availability of possessor extraction to strong (possessor) agreement.51 In this respect, her account is the opposite (as she herself points out) of that of Uriagereka (1988), Corver (1990), Bošković (2005, 2009), which links the availability of Left Branch Extraction (of which possessor extraction is perhaps the best-known type) to the lack of a D projection. For these researchers, the evidence for the lack of D projection, in turn, comes from the lack of overt articles. Gavruseva, on the other hand, posits a DP shell-like structure akin to the one given in (100), and takes possessor extraction to proceed in two steps: movement to the specifier of the lower DP is linked to case and agreement, and movement to the specifier of the higher DP is linked to a strong operator or Q features. This higher specifier position is the escape hatch position necessary for possessor extraction.52 In her analysis, the higher [Spec,DP] position is an A-bar position corresponding to [Spec,CP] in the clausal domain, whereas the lower one is an A position, corresponding to [Spec, TP]. In our terms, the higher D head could be thought of as a phase head endowed with both operator features and case/agreement features. I will call this head a light d head, to make it analogous to a v head dominating a VP, but nothing hinges on this

51

52

Gavruseva also discusses Aissen’s (1996) analysis of Tzotzil, in which regular possessives appear postnominally, but wh-possessors can appear pre-nominally. Aissen analyzes it as movement of a wh-possessor to [Spec,DP]. The question of whether movement out of DPs proceeds through the specifier of DP is different from the question of whether the edge of DP can be the final landing site. Hungarian dative possessors suggest that it can indeed be a landing site. English DP inversion structures of the kind given in (i) have also been analyzed in this way (see Valois 1991, for example). (i) How big a nut can a parrot eat?

{DESKTOP}NEW FOLDER//9781107040847C04.3D

116

[71–123] 23.11.2013 11:48AM

116

Classic phases

terminological choice. If d is a phase head the lower head (D head in (100)) could inherit its uninterpretable φ-features. (100)

dP d′ duOp[

], uϕ [ ]

DP D′ D

NP

The licensing of parasitic gaps was an important edge diagnostic in the discussion of vPs as phases. For DPs, the two questions to consider are: (i) does the empty operator target the [Spec,DP] position, and (ii) can the wh-movement licensing the parasitic gap proceed through [Spec,DP]? With respect to the first question, Emonds (2001) argues that the empty operator inside the gerund adjunct clause moves to [Spec,DP], based on the fact that the status of gerund-internal parasitic gap examples worsens if the specifier of DP is filled:53, 54 (101) a. Which paintings could she scrutinize without (*the owner/*the owner’s) bringing to the gallery? b. These are the dishes you should leave out instead of (*John/*John’s) putting away. (Emonds 2001: 93)

However, since the behavior of the empty operator in these parasitic gap examples is similar in this respect to the behavior of wh-phrases undergoing overt movement (as evidenced by the contrast in (102a–b)), it might also be the case that the issue is not a filled versus empty [Spec,DP] position but other factors (specificity, definiteness etc.):55 (102) a. What did John enjoy showing off at the party? b. What did John enjoy Mary’s showing off at the party? (Emonds 2001: 104)

In our discussion of parasitic gaps in vPs, we focused on the wh-movement licensing the parasitic gap, not the empty operator movement inside the constituent containing the parasitic gap. And we saw that movement to the edge of vP, repeated in (103a) was necessary in order for the parasitic gap to be possible (otherwise the

53 54

Only the variants without ‘s are from Emonds 2001. Engdahl, however, considers examples parallel to (101b) to be grammatical: (i) Which boy did Mary’s talking to pg bother t most? (Engdahl 2001: 69)

55

Valois (1991), following Tellier (1988), also analyzes nominal parasitic gaps as involving empty operator movement to the specifier of DP.

{DESKTOP}NEW FOLDER//9781107040847C04.3D

117

[71–123] 23.11.2013 11:48AM

DPs as phases

117

result is a semantic type mismatch when the adjunct combines with the vP). An analog in the DP domain would involve wh-movement from a DP internal position to the edge of DP licensing a DP internal parasitic gap, as shown in (103b). (103)

a.

CP WHj vP WHj

vP Adjunct

vP … tj …

b.

OPi…ti…

CP WHj DP DP

WHj DP

Adjunct

… tj…

OPi…ti…

Common examples in which the parasitic gap is DP-internal, such as (104a), are not going to be relevant, as they are licensed by another DP-external gap. However, (104b) is the right configuration; it involves ‘real’ extraction from DP, licensing a parasitic gap inside the adjunct relative clause. (104) a. Mary’s the kind of woman OPi that [DP people who meet pg] usually end up inviting ti into their homes. b. *Whoi did Mary [DP tell a story about ti [CP that really impressed pg] ]? (Matushansky 2005: 168)

While I agree with Matushansky’s judgment regarding (104b), the reasons for its degraded status might be non-syntactic in nature. (105) below, involving an analogous configuration, improves markedly: (105) Whoi did Mary take [DP pictures of ti [CP that weren’t that flattering to pg ] ]?

The next edge diagnostic concerns Quantifier Raising. Here the question is whether the edge of DP can be a landing site for QR. To determine that, let us look at cases of complex DPs containing two quantifiers to see if the lower quantifier can take scope over the higher one without scoping outside the DP. Discussion of such cases abounds in the literature (see Charlow 2009, Larson 1985, May 1977,

{DESKTOP}NEW FOLDER//9781107040847C04.3D

118

[71–123] 23.11.2013 11:48AM

118

Classic phases

Sauerland 2005, among many others). Consider the example in (106). The most felicitous reading for it is the inverse scope reading, in which every team scopes over one member. (106) [DP One member of every team] won an MVP (Most Valuable Player) award.

A more complex example is given in (106). Its most natural reading is one in which the relative scope of the three quantifiers is the mirror image of their linear ordering. This reading can naturally be paraphrased as ‘There is a large California city in which every freeway has some exits that are badly constructed.’ (107) [Some exits from every freeway in a large California city] are badly constructed. (Larson 1985: 1)

The issue is whether inverse scope in such cases is the result of the lower quantifier adjoining to DP or moving out of DP and adjoining to TP, as shown in (108a–b) respectively. (108) a.

TP DP DPi

T′

DP

T

vP

every state two delegates from ti

b.

TP DPi

TP

every state DP two delegates from ti T

T′ vP

The availability of the bound variable reading in (109) suggests the latter possibility; however, given Kayne’s definition of c-command, both adjuncts and specifiers (which are indistinguishable for him) can c-command out of DPs containing them. (109) Someone from every city despises it. (Larson 1985: 1, citing May 1977)

However, if inverse scope were the result of movement of the lower quantifier to a TP-adjoined position (as opposed to a DP-adjoined position), we would expect the scope of the lower quantifier to be independent of the scope of the higher quantifier in cases involving yet another (DP-external this time) quantifier. Larson (1985) also discusses such cases, concluding that the scope of the lower quantifier is ‘dependent’ on the higher one. To illustrate with a concrete example, in (110), every city can have scope over two politicians, but only if someone has

{DESKTOP}NEW FOLDER//9781107040847C04.3D

119

[71–123] 23.11.2013 11:48AM

DPs as phases

119

scope over two policitians. This is unexpected if every city could move out of DP and adjoin to TP directly. (110) Two politicians spy on someone from every city. every city> someone > 2 politicians (Larson 1985: 5)

The literature on inverse scope centers around the issue of whether DP is a scope island or not, with examples of the kind given in (110) suggesting that it is.56 However, this issue is independent of the issue at stake here, which is the issue of whether the edge of DP is a landing site for QR. And the fact that every city can scope over someone does suggest that QR can target the edge of DP. This brings us to the question of whether QR can ever target a DP-internal position. If so, it might suggest the presence of another DP-internal phase. The following example, from Matushansky, shows that there has to be such a site; the quantifier a/any foreign country has to undergo QR in order to be interpreted. But at the same time, being a Negative Polarity Item, it has to remain within the scope of the negative quantifier no. (111) No student from a/any foreign country was admitted. (Matushansky 2005: 168)

Thus the edge diagnostics do show that DPs are phases and possibly contain other phases. We also saw evidence that DPs are case and binding domains, which is also compatible with DPs being phases. What is left is the semantic status of DPs (are they semantically complete in the same way CPs and vPs are?) and their phonological status (do D heads determine Spell-Out?). Let us discuss these questions in turn. For CPs and vPs, the criterion of semantic completeness had to do with their propositional status and full argument structure. The obvious parallelism between CPs like Romans destroyed the city and DPs like Roman destruction of the city suggests that DPs are complete in the requisite sense. Matushansky, however, interprets semantic completeness slightly differently, focusing of semantic types instead. The question of whether DPs are semantically complete, as she points out, is complicated in view of the fact that DPs come in different semantic guises, with referential DPs being of type e, quantified DPs of type , and predicative DPs of type . If we take completeness to be saturation, only referential DPs are complete, and on the Generalized Quantifier Theory (which treats all DPs as generalized quantifiers), no DPs would ever be semantically complete.

56

Sauerland (2005) argues that DP is not a scope island (see also Marušič 2009, Charlow 2009 for relevant discussion), based on the fact that examples of the kind given in (i) allow the interpretation in which the most deeply embedded quantifier these two countries can scope over want and someone. (i) Mary wants to marry someone from these two countries. (Sauerland 2005: 306)

{DESKTOP}NEW FOLDER//9781107040847C04.3D

120

[71–123] 23.11.2013 11:48AM

120

Classic phases

The next set of diagnostics involves Spell-Out. If D is a phase head, we expect its complement to be able to undergo ellipsis. The result is a pretty ubiquitous ellipsis type, dubbed Nʹ anaphora (or Nʹ deletion) by Jackendoff (1977), illustrated in (112a–b). (112) a. John wanted to read the dossiers of famous linguists, and/but he succeeded in reading Morris’s/yours/these/one/none/some/many/three __. (Jackendoff 1977: 115) b. The students attended the play but most/some/all/each/two __ went home disappointed. (Lobeck 1995: 42)

Interestingly, however, there are a number of restrictions on NP ellipsis (see Lobeck 1995 and the references therein). Only some determiner-like elements can license NP ellipsis, namely the quantifiers all, each, some, both, many, few, numerals, possessives and plural demonstratives. By contrast, the quantifier every, definite and indefinite articles and singular demonstratives do not license ellipsis. For Lobeck, the crucial factor involved in the licensing of ellipsis (not just NP ellipsis but also VP ellipsis and sluicing, the two other types she considers) is what she dubs strong agreement. Strong agreement refers to morphological strength, which Lobeck takes to be defining ‘agreement in a productive number of cases’ (Lobeck 1995: 51). Lobeck also shows that DPinternal heads such as Number heads can license ellipsis. The evidence comes from examples containing more than one DP internal element, such as the ones in (113a–c). (113) a. My sister’s two boys are wild, but John’s two __ are really quite well-behaved. b. Many students enrolled in the class, but the few __ who dropped it later said it was too difficult. c. Although all twelve __ did well in the class, only two students got As. (Lobeck 1995: 71–2)

If we take the higher element to occupy either the specifier of DP or its head, and the lower one to occupy either the specifier of NumP or its head (as in (114), adapted from Lobeck 1995: 87), these case will involve ellipsis of the complement of the Num head, thus pointing towards Num being a phase head as well. (114) DP Possessor D all both these ’s

D′ NumP Num′ Num six/few

NP ø

{DESKTOP}NEW FOLDER//9781107040847C04.3D

121

[71–123] 23.11.2013 11:48AM

DPs as phases

121

Lobeck’s work in many respects builds on Jackendoff (1977), who distinguishes two types of NP specifiers, differing in the level of attachment (N΄΄ versus N΄΄΄ in his terms). The structure in (115) below ‘translates’ his insights into the more current structure (see Jackendoff 1977: 105 for the original one). (115)

DP Fred D the both those which

D′ NumP Num′ Num many few several

NP dwarfs

A potential problem for the idea that D heads trigger Spell-Out involves case. It is well known that in a language like Polish, which marks case morphologically, DP internal elements usually (but not always, see the discussion of adnominal genitive or genitive of quantification above) agree in case. This is the familiar phenomenon of concord, which involves not only case, but gender, number and person features as well (cf. Danon 2011, Carstens 2000 and the references therein for a detailed discussion of concord from a more crosslinguistic perspective). ta˛ ciekawa˛ ksia˛z·ke¸. [Polish] read.1SG this.3SG.FEM.ACC interesting.3SG.FEM.ACC book.3SG.FEM.ACC ‘I am reading this interesting book.’ b. Nie przeczytam tej nudnej ksia˛z·ki. not read.PERF.1SG this.3SG.FEM.GEN boring.3SG.FEM.GEN book.3SG.FEM.GEN ‘I will not read this boring book.

(116) a. Czytam

The simple question, raised by Matushansky (2005), is how case spreads onto all the DP-internal elements if D is a phase head. In more technical terms, the question is what the uninterpretable case feature (uC feature) is a feature of. The logical possibilities to consider are: the uC feature could be a feature of D (percolating to DP), N, or both, as illustrated in (117a–c) for an uC feature of a direct object.57

57

The latter two options are compatible with the idea that only phase heads can have uninterpretable features, as N can inherit its features from D. See the discussion of Feature Inheritance in Chapter 2 (Section 2.6).

{DESKTOP}NEW FOLDER//9781107040847C04.3D

122

[71–123] 23.11.2013 11:48AM

122

Classic phases

(117)

a.

vP v′ vuϕ [

VP

]

V

DP DuC [

b.

NP

]

N

vP v′ vuϕ [

VP

]

DP

V D

c.

NP NuC [

vP

]

v′ vuϕ [

VP

]

V

DP DuC [

]

NP NuC [

]

Each of these options raises questions. If we assume the structure in (117a), the question is how case ‘spreads’ downwards to N. If NP (being a complement of the phase head D) is spelled out, the question about (117b) is how case ‘spreads’ upwards to D. And, finally, if we assume the structure in (117c), the question is how v can value uninterpretable case feature on two elements simultaneously. If it values uC feature of D first, then this D head with a valued case feature might act as an intervener, and block Agree between v and N. We could appeal to Hiraiwa’s (2001) Multiple Agree mechanism, which allows a single Probe to undergo Agree with multiple Goals. However, in typical cases of Multiple Agree, the two Goals are in the same Spell-Out domain. If D is a phase head, its complement NP is spelled out as soon as the next phase head (v in the case at hand) gets merged. This would also result in the uC feature on N being unvalued. Matushansky (implicitly) assumes the option in (117a), pointing out that it is not clear how case ‘gets’ onto the N head, if NP gets spelled out when the v head is merged. But it does seem reasonable to assume that case ‘spreads’ DP

{DESKTOP}NEW FOLDER//9781107040847C04.3D

123

[71–123] 23.11.2013 11:48AM

DPs as phases

123

internally via an agreement or concord mechanism (see Baker 2008b, Carstens 2000, among others, for explicit proposals), perhaps taking place in the postsyntactic component. This concludes our discussion of DPs as phases and the table below summarizes the behavior of DPs with respect to the phasehood diagnostics established in Chapter 3. (118) DPs as Phases Diagnostic

Evidence

Is DP a domain for feature valuation? Yes adnominal Genitive Is D the locus of uninterpretable features? Yes genitive case Does movement out of DP proceed through Yes parasitic gaps constraints on movement out the edge? of DPs

Does D determine Spell-Out? Is DP a binding domain? Can D serve as a Probe? Can QR target the edge of DP?

Yes Yes Yes Yes

Nʹ ellipsis picture noun phrases genitive case inverse scope

The next chapter examines the idea of extending the idea of phasehood to three other categories, all of which have been argued to be phases. The three are: Predicatation Phrases (PrPs), Prepositional Phrases (PPs) and Applicative Phrases (ApplPs). The discussion in this chapter will be somewhat more speculative for two reasons. First, the very existence of two of the three categories is not uncontroversial. A PrP structure is not the only possible structure for small clauses (see Citko 2011c for an overview of possibilities), and an Applicativetype structure is certainly not the only structure that has been proposed for double object constructions. While the very existence of PPs is fairly settled, there are quite a few of unsettled matters regarding the types, structure, and lexical vs. functional status of prepositions.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

124

[124–160] 23.11.2013 11:35AM

5 Other ph(r)ases

5.1 PrPs as phases In this section, we turn to the idea that there is a relationship between predication and phasehood. There are two possible ways to look at this relationship. One is to first expand the inventory of functional projections to include a Predication Phrase (PrP), as proposed by Bowers (1993), and, next, to take the head of PrP to be a phase head, in addition to the other heads whose phasehood we established in the previous chapter: C, v and D heads. A very different view, due to Den Dikken (2006), is to essentially abandon the standard characterization of phases, and, instead, to take the subject–predicate relationship to be a definitional characteristic of phases. For Den Dikken, whenever there is a predication relationship, there is a phase.1, 2 I focus here on the former, more specific view, which is that Predication Phrases are phases, as argued for explicitly by Bowers (2002), Harves (2002), Matushansky (2000) and Tanaka & Yokogoshi (2010), among others. For Bowers, all predication relationships are mediated by a Pr head; simple transitive clauses also include a PrP in addition to the usual functional projections, some phasal and some not. I take a more modest view and assume that a Pr head is only present when it needs to be: when there is no other functional head to relate the subject and the non-verbal predicate (see also Svenonius 1994 for a similar view). Small clauses are a case in point. In both matrix and embedded contexts, Pr heads the small clause.

1

2

Den Dikken differs from Bowers in that he assumes that predication relationships can be mediated by a number of functional categories (which he collectively refers to as relators), rather than a predication-specific functional category. See also Carnie & Barss (2006) for a slightly different view. For them, phases have to contain some predicative element, a single argument, and a temporal operator that situates the predicate and argument in time and space.

124

{DESKTOP}NEW FOLDER//9781107040847C05.3D

125

[124–160] 23.11.2013 11:35AM

PrPs as phases

125

(1) a. Icarus is a parrot.

b.

TP T′ T is

PrP DP

Pr′

Icarus

Pr

DP a parrot

(2) a. I consider parrots smart.

b.

TP T′ T

vP v

VP consider

PrP

DP parrots Pr

Pr′ AdjP smart

The Predication Phrases in both cases are certainly semantically complete since they contain a predicate with all its arguments. This is the behavior we expect of phases. Tanaka & Yokogoshi (2010) also bring forth general architectural considerations as an argument in favor of Predicatation Phrases being phases. For them, treating PrP as a phase ‘fits in well with the configuration of phases envisaged by Chomsky (2001): F-XP, where a functional category F selects a substantive category XP, e.g. C-TP and v-VP’ (Tanaka & Yokogoshi 2010: 26). However, it is not clear what exactly they mean by ‘substantive category’. A TP does not seem more substantive than a CP, both being functional. We have also seen some issues for the characterization of phases as semantically complete; what counts as complete may vary from category to category. Therefore, let us look at more concrete diagnostics, listed in (3), to determine the phasehood of PrPs. These are the same diagnostics we used in previous chapters to determine the phasehood of other categories.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

126

[124–160] 23.11.2013 11:35AM

126

Other ph(r)ases

(3) a. Is PrP a domain for feature valuation? b. Is Pr the source of uninterpretable features? c. Does Pr trigger Spell-Out? d. Does movement out of PrP proceed through the edge of PrP?

And to tell whether movement out of PrP proceeds through the edge of PrP, we will ask the more specific questions in (4). (4) a. Can an element moving out of PrP be interpreted in the edge position? b. Can QR target PrP? c. Is stranding at the edge of PrP possible?

Many languages have what can be referred to as a predicative case, i.e. case assigned to predicate nominals that cannot be attributed to a default case mechanism or a case agreement mechanism (see Matushansky 2008 and the references therein for a typological overview of case patterns on predicates). To illustrate with one concrete example, predicate nominals are marked with Instrumental case in Slavic languages like Polish or Russian, as shown in (5a) and (5b), respectively.3 jest studentka˛/*studentka. Maria.NOM is student.INSTR/*NOM ‘Maria is a student.’ ego durakom/*duraka. b. Ja scˇitaju I consider him.ACC fool.INSTR/*ACC ‘I consider him a fool.’ (Bailyn & Citko 1999: 19)

(5) a. Maria

[Polish]

[Russian]

Bailyn & Rubin (1991) propose that the Pr head heading the small clause is responsible for Instrumental case assignment. In phase-theoretical terms, their claim can be restated as a claim about the Pr head valuing the uninterpretable case feature on the predicate as Instrumental. This is only possible if Pr is a phase head. It can value case, which means that it has to have uninterpretable features which make it an active Probe, and since it is not a complement of a phase head (in (1b) PrP is a complement of T and in (2b) of V), it could not have inherited its uninterpretable features from a phase head above it. There is also no (other) Instrumental nominal in either sentence with which the predicate could agree, and there are no reasons to consider Instrumental case to be default, even if we grant the existence of default case in general. Interestingly, in Russian (and to a much lesser degree in Polish), predicate nominals can also be marked with Nominative case, as shown in (6a–b). This alternation (Instrumental versus Nominative) is a very well-documented case alternation in Slavic linguistics (see Bailyn & Citko 1999, Citko 2008b, Harves 2002, Matushansky 2000, Pereltsvaig 2007, Richardson 2007, among many

3

This is not the only option. I turn to the other option, predicates marked with Nominative case, shortly.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

127

[124–160] 23.11.2013 11:35AM

PrPs as phases

127

others), and, roughly speaking, it correlates with a difference in interpretation along the individual versus stage-level dimension. byl ucˇitel’. [Russian] brother.NOM was teacher.NOM ‘(My) brother was a teacher (by nature).’ b. Brat byl ucˇitelem’. brother.NOM was teacher.INSTR ‘(My) brother was a teacher (for a certain period of time).’ (Bailyn & Citko 1999: 30)

(6) a. Brat

Interestingly, if the head of the PrP is filled by an overt predicator (such as the particle za ‘as’ in the examples below), the predicate nominal is no longer Instrumental. Instead, it agrees with the DP it is predicated of: in (7a–b) the adjective is predicated of the direct object and both are Accusative. (7) a. Uwaz·am

Marie¸ za I.consider Maria.ACC as b. Ja prinimaju ego I take him.ACC ‘I consider him as a fool.’

ma˛dra˛ studentke¸/*ma˛dra˛ studentka˛. clever.ACC student.ACC/*INSTR za duraka/*durakom. as fool.ACC/*INSTR (Bailyn & Citko 1999: 22)

[Polish] [Russian]

Likewise, in Polish copular sentences, the Instrumental case ‘disappears’ in the presence of the demonstrative-like element to, often characterized as a pronominal copula. When it is present, either in addition to or instead of the verbal copula ‘być’ be, both the subject and the predicate are Nominative (see Citko 2006, 2008b for a more detailed discussion of this pattern): (8) Maria to ( jest) Maria.NOM DEM is ‘Maria is a student.’

studentka. student.NOM

In phase-theoretical terms, the fact that Pr head is involved in Instrumental case assignment means that it contains uninterpretable φ-features that make it a Probe capable of valuing case on its complement, as shown in (9a–b): (9) a.

TP T′ Tuϕ:[3sg],EPP

PrP

DPuC:[Nom],iϕ:[3sg] Pr′ Maria

Pruϕ:[3sg]

DPuC:[Instr],iϕ:[3sg] a student

{DESKTOP}NEW FOLDER//9781107040847C05.3D

128

[124–160] 23.11.2013 11:35AM

128

Other ph(r)ases

b.

TP T′ Tuϕ:[1sg],EPP

vP

DPuC:[Nom],iϕ:[1sg] v′ I

VP

vuϕ:[3sg] consider

PrP

DPuC:[Acc],iϕ:[3sg] Maria

Pr′

Pruϕ:[3sg]

DPuC:[Instr],iϕ:[3sg] a student

The fact that the Pr head can lose its ability to value Instrumental case suggests that it can lose its uninterpretable φ-features, and, consequently, stop being a phase head.4 The fact that neither the pronominal copula to nor the prepositional predicator za ‘as’ agrees with the subject or the predicate is consistent with Pr lacking φ-features. Instead, predicate nominals agree in case with the subject of predication, and end up either with Nominative case (by Multiple Agree with T) or with Accusative case (by Multiple Agree with v): (10) a.

TP T′ T uϕ:[3sg],EPP

PrP

DP uC:[Nom],iϕ:[3sg] Pr′ John

Pr

DPuC:[Nom],iϕ:[3sg] a student

4

The insight that a filled Pr head loses the ability to assign Instrumental case is due to Bailyn & Citko (1999), who capture it under a slightly different (non-phase-theoretical) set of assumptions. The insight is reminiscent of the classic account of passive morphology as absorbing Accusative case (see Baker, Johnson & Roberts 1989).

{DESKTOP}NEW FOLDER//9781107040847C05.3D

129

[124–160] 23.11.2013 11:35AM

PrPs as phases

b.

129

vP vuϕ:[3sg],EPP VP consider

PrP

DPuC:[Acc],iϕ:[3sg] Pr′ John

Pr

DPuC:[Acc],iϕ:[3sg] a student

The ability of Pr to value Instrumental case (or the lack thereof) has been shown to correlate with the possibility of extracting the complement of a Pr head. The contrast in (11a–b), discussed in Citko (2006, 2008b), shows that copular clauses with Nominative and Instrumental predicates behave differently with respect to movement: Instrumental predicates can undergo wh-movement, but Nominative ones cannot. Russian behaves similarly in this respect, as discussed by Matushansky (2000) and Harves (2002), among many others.5 (11) a. Kimi chcesz, z·eby Maria była ti? who.INSTR want.2SG COMP Maria was ‘Who do you want Maria to be?’ b. *Ktoi chcesz, z·eby Maria (to) była ti? who.NOM want.2SG COMP Maria DEM was ‘Who do you want Maria to be?’

[Polish]

This contrast has also been accounted for by taking the Pr head selecting an Instrumental predicate (but not one selecting a Nominative predicate) to be a phase head.6 Matushansky (2000) and Harves (2002) analyze Russian in a similar way; however, for them the relevant difference between the two types of Pr heads is the difference between defective and non-defective phase heads, rather than phase versus non-phase heads. What both types of accounts share is the insight that the extraction asymmetry between Nominative and Instrumental predicates is due to the presence of an EPP feature on the phasal

5

6

The Polish complementizer żeby is a subjunctive complementizer, selecting past tense complements (see Tomaszewicz 2012 for a recent discussion of its many interesting properties). The reason I use subjunctive clauses here is that for some speakers, extraction out of indicative clauses is degraded. The idea that there are two variants of the same head, one phasal and one not, is not new. This is a common way to think of v heads, for example. In Section 5.3 of this chapter, we will look at two types of applicative heads, which also vary in terms of phasehood.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

130

[124–160] 23.11.2013 11:35AM

130

Other ph(r)ases

Pr head, which allows the Instrumental wh-predicate to move to the outer edge of PrP, as shown in (12).7 (12) Structure of (11a) CP C′ C

TP T′ Tuϕ:[3sg],EPP

PrP PrP DPuC:[Nom],iϕ:[3sg] Pr′ Maria

Pruϕ:[3sg],EPP

DPuC:[Instr],iϕ:[3sg] who.INSTR

If Pr lacked the EPP feature, the wh-predicate would not be able to move past the subject, as is the case with Nominative predicates. This movement would be blocked either by the Phase Impenetrability Condition if Pr remains a phase head, or by the presence of the intervening subject either in [Spec,PrP] or [Spec,TP] if Pr is no longer a phase head. The two options are given in (13a) and (13b), respectively. In (13a), the subject intervenes and C cannot undergo Agree with the wh-phrase.8 In (13b), the Merge of C (the next phase head up) triggers the Spell-Out of the complement of Pr (the nominal predicate). Since it is impossible to prevent a phase head from having an EPP feature in a nonspeculative way, I will not pursue this option any further.

7

8

Note that the movement of the Instrumental DP from the complement of Pr head to its (outer) specifier is a violation of antilocality. This is not a problem if Instrumental DPs involve a richer structure, in which what moves is not a complement of the phase head but a complement of a complement of phase head). This configuration is very similar to the configuration involving object wh-movement out of simple transitive clauses. It is important to note that the reason movement is not allowed in (13a) is not PrP not being a phase, as movement out of non-phases is certainly possible. The reason is the intervention of the subject. One might wonder why the subject, not being a wh-phrase itself, blocks wh-movement. This I take be similar to the cases we saw in Section 2.6, in which the whobject in the outer [Spec,vP] blocked Agree between T and the subject.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

131

[124–160] 23.11.2013 11:35AM

PrPs as phases

(13)

a

131

CP C′ C

TP T′ Tuϕ:[3sg],EPP

PrP PrP DPuC:[Nom],iϕ:[3sg] Pr′ Maria

Pr

DPuC:[Nom],iϕ:[3sg] who.NOM

b. CP C′ C

TP T′ Tuϕ :[3sg],EPP

PrP

DPuC:[Nom],iϕ:[3sg] Pr′ Maria

Pruϕ :[3sg]

DPuC:[Nom],iϕ :[3sg] who

The claim that what allows wh-extraction of Instrumental predicates is Pr being a phase head seems in conflict with Tanaka & Yokogoshi’s (2010) proposal that it is the phase status of Pr that prevents (certain types of) inversion in English small clauses. They argue that English small clauses historically went from being headed by a lexical category (the predicate) to being headed by a functional category (the Pr head). They further argue that with the emergence of Pr as a phase head, the inversion of the kind illustrated in (14) became impossible.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

132

[124–160] 23.11.2013 11:35AM

132

Other ph(r)ases

(14) *I consider smart Maria.

Tanaka & Yokogoshi attribute the ungrammaticality of (14) to the Phase Impenetrability Condition, arguing that the predicate cannot move because it is within the Spell-Out domain of Pr. However, the argument only works on the stronger version of the Phase Impenetrability Condtion (PIC1), on which the complement of Pr is spelled out as soon as the next head up is merged, as shown in (15a). Under the assumptions of PIC2, the later and the more empirically adequate of the two, the complement of Pr is not spelled out until v (the next phase head up) is merged. Thus, movement of the predicate should in principle be possible, as shown in (15b). The configurations schematized here are very similar to the configuration involving Agree between T and a quirky Nominative object, which was excluded by PIC1 but allowed by PIC2. (15) a.

VP V

PrP Pr′

Maria Pr

AdjP smart

b.

vP v

VP V′ V

PrP Maria

Pr′ Pr

AdjP smart

Furthermore, we do not want to exclude all cases of predicate inversion, which a PIC2-based account would do. The contrast in grammaticality between the so-called predication copular clause in (16a) and the specificational one in (16b) is often attributed to inversion around the copula (e.g. Moro 1997 and much later work):

{DESKTOP}NEW FOLDER//9781107040847C05.3D

133

[124–160] 23.11.2013 11:35AM

PrPs as phases

133

(16) a. Maria is my best friend. b. My best friend is Maria.

On Tanaka & Yokogoshi’s assumptions, this should not be allowed. The movement in (17a–b) is similar to the movement in (15a–b). However, given PIC2, the complement of Pr does not get spelled out until C is merged, which allows either the subject or the predicate to move to [Spec,TP], as shown in (17a–b).9 (17) a. CP C

TP T′ T

PrP DP

Pr′

Maria Pr

DP my best friend

b.

CP C

TP T′ T

PrP DP Maria Pr

Pr′ DP my best friend

The difference in extraction possibilities between Instrumental and Nominative predicates in Slavic is different in spirit from the evidence we used in Chapter 4 to argue that movement out of CPs, vPs and DPs has to proceed through their respective edges. There, the issue was not whether extraction was possible (we saw that it was), but whether movement had to proceed through the edge of the presumed phase. One of the ways we could tell was by looking at

9

The timing becomes problematic with Feature Inheritance in place.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

134

[124–160] 23.11.2013 11:35AM

134

Other ph(r)ases

whether movement could strand a quantifier or a nominal (in cases of split constituents) at the edge of the phase. Let us now apply the same logic to PrPs. The grammaticality of the examples in (18a–b) shows that the subject of the Predicate Phrase can move stranding all, but it does not necessarily show that it has to move through the edge of PrP. (18) a. I consider the students all smart. b. The cat regards the kids as all hopeless cases. (cf. Starke 1995: 242)

We have seen that in Polish – a language that allows Left Branch Extraction – the position of the nominal can mark the movement path. The grammaticality of (19a), then, is also consistent with the nominal being stranded in its underlying position (shown in (19b)) or in the matrix object position (shown in (19c)), so it does not tell us much about movement through the edge of PrP.10 (19) a. Którego which.ACC

uwaz·asz

studenta

za

najlepszego

składniowca?

consider.2SG

student.ACC

as

best.ACC

syntactician.ACC

[Polish]

‘Which student do you consider to be the best syntactician?’

b. [CP which student [TP you

[vP which student

[vP consider [VP which student [PrP

which student [Prʹ as [DP the best syntactician]]]]]]]]

c. [CP which student [TP you

[vP which student

[vP consider [VP which student [PrP

which student [Prʹ as [DP best syntactician]]]]]]]]

Extraction from the predicate position will thus be more telling. If movement proceeds as schematized in (20a–b), we should expect that the nominal can be strandable in the bolded outer [Spec,PrP] position. (20) a. [CP WH [TP [vP WH v [VP [PrP WH [PrP SUBJECT [Pr’ Pr [ … WH…]]]]]]]] b. [CP WH [TP [PrP WH [PrP SUBJECT [Prʹ Pr [ … WH …]]]]]]

Since the subject moves from the specifier of PrP in both cases (to the matrix [Spec,VP] position in (20a) and to the matrix [Spec,TP] in (20b)), the ‘surface’ configurations we end up with are as shown in (21a–b), which again makes it difficult to distinguish some of these cases from cases in which the moving wh-predicate strands something in its base position. (21) a. [CP WH [TP [vP WH v [VP SUBJECTi [PrP WH [PrP ti [Prʹ Pr [ … WH… ]]]]]]]] b. [CP WH [TP SUBJECTi [PrP WH [PrP ti [Prʹ Pr [ … WH …]]]]]]

The examples below provide an illustration. (22a) is a baseline example. The contrast between (22b) and (22c) shows that the predicate can only be extracted if it pied-pipes the prepositional predicator za ‘as’, which I am treating as the

10

I do not have information on the status of the following examples in Irish English. However, if they are acceptable, a similar caveat applies to them; the nominal could be stranded in situ or in the matrix object position. (i) Who do you consider all smart?

{DESKTOP}NEW FOLDER//9781107040847C05.3D

135

[124–160] 23.11.2013 11:35AM

PrPs as phases

135

head of PrP.11 (22d) and (22e) show that the nominal can be stranded. (22d) is compatible with the nominal being stranded in its underlying position. However, in (22e) the matrix subject remains inside PrP (as evidenced by the position of the adverbial), which makes the outer specifier of PrP the most likely culprit for the position of the stranded nominal. (22) a. Uwaz·am Marie¸ za najlepsza˛ przyjaciółke¸. consider.1SG Maria.ACC as best.ACC friend.ACC ‘I consider Jan to be (my) best friend.’ b. Za jaka˛i przyjaciółke¸ uwaz·asz Marie¸ ti? as what.ACC friend.ACC consider.2SG Maria.ACC c. *Jaka˛i przyjaciółke¸ uwaz·asz Marie¸ za ti? what.ACC friend.ACC consider.2SG Maria.ACC as ‘What kind of friend do you consider Maria to be?’ d. Za jaka˛i uwaz·asz Marie¸j [PrP tj przyjaciółke¸ ]? as what.ACC consider.2SG Maria.ACC friend.ACC ‘What kind of friend do you consider Maria to be?’ uwaz·asz po tym [PrP [ti przyjaciółke¸]k [PrP Marie¸ tk ] ]? e. Za jaka˛i as what.ACC consider.2SG after this friend.ACC ‘What kind of friend do you consider after this Maria to be?’

Maria.ACC

Similar considerations apply to copular sentences. (23a) is the baseline example. (23b) shows that wh-movement from the predicate nominal is possible. (23c) shows the noun sąsiada ‘neighbor’ can be stranded by wh-movement. However, in this case it is stranded in its underlying position (the complement of Pr head), not at the phase edge. (23d), however, does show that it can also be stranded at the edge of PrP. (23) a. Maria jest przyjaciółka˛ naszego sa˛siada. Maria is friend.INSTR our.GEN neighbor.GEN ‘Maria is our neighbor’s friend.’ b. Czyjego sa ˛siadai Mariaj jest [PrP tj [Prʹ Pr [DP przyjaciółka˛ ti]]]? whose.GEN neighbor.GEN Maria is friend.INSTR c. ?Czyjegoi Mariaj jest [PrP tj [Prʹ Pr [DP przyjaciółka˛ [DP ti sa˛siada]]]]? whose.GEN Maria is friend.INSTR neighbor.GEN d. Czyjegoi Mariaj jest [PrP [ti sa˛siada]k [PrP tj [Pr’Pr[DP przyjaciółka˛ tk ]]]]? whose.GEN Maria is neighbor.GEN friend.INSTR ‘Whose neighbor is Maria a friend of?’

11

This does not seem to have anything to do with za ‘as’ being prepositional (and Polish disallowing P-stranding), as the English as resists stranding as well, as pointed out by Emonds (1985), for example: (i) ??Who do you consider John as?

{DESKTOP}NEW FOLDER//9781107040847C05.3D

136

[124–160] 23.11.2013 11:35AM

136

Other ph(r)ases

In cases of other phases (vPs and DPs in particular), we also looked at Quantifier Raising to determine whether movement proceeds through the phase edge. Let us now apply this diagnostic to PrPs. Ambiguity of examples like (24) below is not directly relevant; the surface scope reading is in principle compatible with the quantified predicate nominal remaining in its surface position, and the inverse scope reading is compatible with the lower of the two quantifiers adjoining to TP. (24) Everyone is someone.

EVERYONE

>

SOMEONE, SOMEONE

>

EVERYONE

However, the first option disappears if quantified elements need to QR for independent semantic reasons even if this movement does not establish a new scopal relationship. Then someone will have to move in both cases, and it will have to move to the closest position of type t, as is commonly assumed in the literature on QR (see Bruening 2001, for example). In copular constructions such as (24), the closest category of the right semantic type to which someone could adjoin is PrP.12 Negative Polarity Licensing also suggests that PrP is a potential QR site. In (25), for example, anyone has to undergo QR for semantic reasons, and it has to remain within the scope of negation to be licensed as a Negative Polarity Item. Thus the only possible QR site is the PrP adjoined position. (25) a. John isn’t anyone. b. [TP Johnj [Tʹ isn’t [PrP anyonei [PrP tj [Prʹ Pr ti ]]]]]

Antecedent-Contained Deletion points towards the same conclusion. (26) a. John isn’t [anyone Bill is Δ] b. [TP Johnj [Tʹ isn’t [PrP [anyonei Bill is Δ]i [PrP tj [Prʹ Pr ti ]]]]]

The last diagnostic we will consider involves the ability of Pr heads to determine Spell-Out. One way to diagnose this is by looking at whether the complement of Pr can be deleted. Here, however, the evidence is somewhat inconclusive. The grammaticality of (27a) does not necessarily involve deletion of the complement of the Pr head, and the ungrammaticality of (27b) seems to be related to a more general ban on as stranding illustrated in (27c): (27) a. John is my best friend but Bill isn’t Δ. b. *Mary considers Bill as her best friend and Ann considers Mike as Δ. c. *Whoi does Mary consider Bill as ti?

This concludes our discussion of PrPs as potential phases. The table in (28) provides a brief summary of the main findings in this section. We saw that the evidence for the phasehood of PrPs is less unequivocal than the evidence for the 12

These cases are different from the cases of frozen scope in examples of the kind given in (i), discussed by Hornstein (1995), Den Dikken (2007), among others, which show that the smallclause-internal elements cannot scope out of the small clause. (i) Someone considers every parrot endangered. SOMEONE

SOMEONE

>

EVERY PARROT,

*EVERY

PARROT

>

{DESKTOP}NEW FOLDER//9781107040847C05.3D

137

[124–160] 23.11.2013 11:35AM

PPs as phases

137

phasehood of other categories, and that different types of PrPs behave differently with respect to some of the diagnostics. Nevertheless, I hope to have established in this section that at least some languages have PrPs that do behave like phases with respect to most phasehood diagnostics. (28) PrPs as phases Diagnostic Question

Answer Evidence

Is Pr the source of uninterpret- Yes able features? Does movement out of PrP Yes have to proceed through the edge of PrP? Can QR target PrP?

Yes

Can the complement of Pr be deleted? Is PrP semantically complete (propositional)?

No

5.2

Yes

Instrumental case on Slavic predicates contrast with respect to extraction between Instrumental and Nominative predicates possibly stranding in [Spec,PrP] scope, Antecedent-Contained Deletion no stranding of overt predicators PrP consists of a subject and a predicate

PPs as phases

The category we focus on in this section, with an eye towards determining its phase status, is a Prepositional Phrase. What makes determining the phasehood of PPs somewhat more elusive than determining the phasehood of CPs or vPs is the fact that there are quite a few background issues that have to be settled before we can even figure out whether PPs are phases. We first need to establish: (i) are prepositions functional or lexical?, (ii) what is the internal structure of PPs?, (iii) do different types of PPs have different structure?13, 14 Thus a fair amount of the discussion in this section will revolve around these background questions. This is not to say that these questions do not matter for other phasal categories; the answers to them, however, seem relatively more settled and less controversial. 13

14

Perhaps an even more fundamental question is what prepositions are to begin with, as the line distinguishing them from case particles or verbal particles is not always clear. It is not even clear if all languages have prepositions, which, if we establish that prepositions are phases and that some languages lack prepositions, would bear directly on one of the more general issues concerning phases under consideration here, namely the issue of whether there is any crosslinguistic variation with respect to phasehood. Typically, however, the issue is whether a given category – such as ApplP or DP– counts as a phase in a given language, which is different from the issue of whether a given language might lack a potentially phasal category. See Cinque’s ‘Mapping Spatial PPs: An Introduction’ in Cinque & Rizzi (2010) for a very illuminating review of these (as well as a host of other) issues.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

138

138

[124–160] 23.11.2013 11:35AM

Other ph(r)ases

The logical possibilities concerning the phasehood of PPs, all of which have been defended in the literature, are given in (29). (29) a. Prepositional phrases are phases (Boškovic´ 2004b, Drummond, Hornstein & Lasnik 2010, Kayne 1999, 2004) b. Prepositional phrases are not phases (Den Dikken 2010) c. Prepositional phrases in some languages are phases (Abels 2003)

In order to determine the phasehood of PPs, we will look at the familiar diagnostics: (30) a. Is PP a domain for feature valuation? b. Is P a source of uninterpretable features? c. Does P trigger Spell-Out? d. Does movement out of PP proceed through the edge of PP?

And, as in the previous cases, we will look at reconstruction, stranding and QR to see whether movement proceeds through the edge of PP: (31) a. Can an element moving out of PP be interpreted in the edge position? b. Can QR target PP? c. Is stranding at the edge of PP possible?

First, however, let us digress a bit into the background issues alluded to above: the functional versus lexical status of prepositions, as well as the structural differences between different types of prepositions. These are directly relevant for the following reason: if only functional categories are capable of being phases, whether or not PPs are phases might depend on whether prepositions are functional or lexical.15 And if different types of PPs involve different structures, we might expect only certain types of PPs to be phases. The assumption that only functional categories are capable of being phases has been implicit throughout the discussion in this book. Being a functional category, however, would not make a very useful phasehood diagnostic, as the implication goes only one way: while all the phase heads we have encountered so far are functional, not all functional heads are phase heads. Examples of uncontroversially functional heads that are not considered to be phasal include (but are not limited to): Tense heads, Aspect heads or various functional heads within the (extended) nominal domain. Prepositions have always straddled the lexical/functional divide, which reflects the fact that different types of prepositions might behave differently with respect to some of the criteria distinguishing lexical from functional categories. For some researchers, prepositions are essentially lexical (but may be dominated by a functional layer) (e.g. Den Dikken 2010, Jackendoff 1977); for 15

The view that some or all prepositions are lexical (thus non-phasal) is compatible with the idea that they are dominated by some P-related phasal functional head (little p perhaps, by analogy to a little v dominating VP), as is sometimes suggested.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

139

[124–160] 23.11.2013 11:35AM

PPs as phases

139

others, they are functional (e.g. Baker 2003, Emonds 1985, Grimshaw 1991, Kayne 2004); and yet for others some prepositions are lexical and others functional (e.g. Abraham 2010, Van Riemsdijk 1998, 1990, Yadroff 1999)16 Abney (1987) characterizes functional elements as belonging to closed lexical classes, lacking descriptive content and being phonologically and morphologically dependent. Even though there are more prepositions than articles or conjunctions, the number of prepositions in English, to choose a familiar example, is much smaller than the number of nouns or verbs (less than a hundred, according to most estimates).17 Prepositions also pattern with other functional categories with respect to the remaining criteria: they tend to be phonologically and morphologically dependent, they do not undergo derivational processes, they often cliticize, and they tend to lack ‘descriptive content’.18 Baker (2003) also notes that they cannot be incorporated, which he attributes to the Proper Head Movement generalization, a generalization that makes it impossible to move from a functional head to a lexical one (see Baker 2003: 306). A slightly different host of arguments in favor of treating prepositions as functional and/or phasal comes from the parallelism between prepositions and other (uncontroversially) functional heads, most notably C heads. This is what Emonds (1985), Kayne (1999) and Abels (2003), among others, do. Complementizers often develop from prepositions, with the English preposition for or the Romance de/di serving as good examples.19

16

17

18

19

Strictly speaking, Van Riemsdijk treats prepositions as belonging to the so-called semi-lexical category. Baker lists English as having around fifty prepositions, and points out that in some languages, the class of prepositions is really small: Mohawk has four simple adpositions, Edo three and Chichewa three or five (depending on the criteria used). However, small does not mean closed. Psycholinguistic evidence also seems to points towards treating prepositions as functional. Froud (2001), as reported by Cinque (2010), shows that they pattern with functional categories in aphasia studies. Frederici (1982), as reported by Muysken (2008), shows that with respect to some types of speech errors, prepositions pattern with lexical classes, but with respect to others they pattern with other (functional) categories. Emonds (1985) actually treats complementizers as prepositions. Kayne’s (1999) main arguments in favor of treating prepositions and complementizers alike comes from the fact that both can act as Probes and attract phrases to their specifiers. More specifically, he argues that prepositional complementizers (such as di in (i)) do not form a constituent with their clausal complements. Instead, the main verb merges with the infinitival clause and the prepositional complementizer di is merged later. It attracts the infinitive clause to its specifier. Next, it moves to a higher head, which subsequently attracts the VP to its specifier: (i) Gianni John

ha

tentato

di

cantare.

has

tried

di

sing.INF

‘John has tried to sing.’ (Kayne 1999: 51)

[Italian]

{DESKTOP}NEW FOLDER//9781107040847C05.3D

140

[124–160] 23.11.2013 11:35AM

140

Other ph(r)ases

Interestingly, with respect to some of the criteria for distinguishing lexical from functional categories, (at least some) prepositions behave like lexical categories. For example, borrowed prepositions do exist, which casts doubt on the claim that prepositions constitute a closed class. Yadroff (1999: 61) gives the following examples: vis-a-vis, circa, via, per, qua. Van Riemsdijk (1998) also argues that prepositions do not constitute a purely ‘closed’ lexical class, as new prepositions can be added to the lexicon. He gives upcreek from here as an example of a prepositional phrase coinage (Van Riemsdijk 1998: 17), and takes the fact that prepositions can participate in derivation, as evidenced by the forms like offish, inner, to also argue against their functional nature. Abney (1987) notes that prepositions can appear in compounds but other functional categories cannot. Given the mixed nature of the evidence in favor of prepositions being lexical or functional, with some diagnostics pointing in one direction and others in the other, an obvious possibility to consider is that prepositions are a mixed category, with some prepositions being lexical and others functional. This is suggested, for example, by Van Riemsdijk (1990, 1998), Yadroff (1999), Yadroff & Franks (2001), Cinque (2010) and Abraham (2010).20, 21 Yadroff (1999) provides a very systematic discussion of the differences between the two types, based on the data from Russian. Functional prepositions tend to be unstressed, monosyllabic and polysemous (due to the more abstract meaning). Lexical prepositions, on the other hand, are more complex: they tend to be stressed, polysyllabic (often polymorphemic) and have a more concrete (hence fixed) meaning. Cinque (2010) also points out that in some languages only functional prepositions assign case. The table in (32) provides examples of the two types from Polish; I refer the reader to Yadroff (1999: 65) for the Russian counterparts. (32) Functional versus lexical prepositions in Polish Functional prepositions bez za z na od

20

21

‘without’ ‘behind’ with, ‘from, out of’ ‘on’ ‘from’

Lexical prepositions mie¸dzy zamiast powyz·ej poza podczas

‘between/among’ ‘instead of’ ‘above’ ‘except’ ‘during’

To be accurate, Yadroff argues that there is ‘no such thing as a preposition in Universal Grammar’ (Yadroff 1999: 26). For him, functional prepositions occupy a functional slot within the extended projection of N/D (Yadroff 1999: 162). Yadroff & Franks (2001) take a similar view, arguing that prepositions are an epiphenomenon: functional prepositions arise through fission of a functional projection dominating NP, and lexical prepositions are bleached lexical categories. Abraham (2010) argues that the two types are merged in different positions: lexical prepositions are merged high and serve as Probes, and grammatical ones are merged low.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

141

[124–160] 23.11.2013 11:35AM

PPs as phases

141

Lexical prepositions can be stressed and are often morphologically complex (for example, podczas ‘during’ morphologically consists of pod ‘under’ and czas ‘time’). The upshot of the discussion so far is that at least some prepositions are functional. On the assumption that only functional heads are phase heads, we might expect only these prepositions to be phase heads. However, the existence of two types of prepositions is also consistent with all PPs being phases, if lexical prepositions are not phase heads but are dominated by functional heads that are. Before turning to this possibility, let me introduce another distinction: the distinction between locative and directional prepositions, as this distinction is often correlated with a difference in structure. Many prepositions are ambiguous and can be either locative or directional, depending on the context. This is shown in (33a–b). (33) a. The statue sat on/under/behind the table. b. John put the statue on/under/behind the table.

LOCATIVE DIRECTIONAL

P P

As is well documented in the literature, many languages disambiguate the two uses by means of case. The contrast in polish between the a and b examples in (34)–(36) shows that the same preposition can assign two different cases, depending on whether it is used in a locative or a directional sense. (34) a. Jan jest na wakacjach. Jan is on holidays.LOC ‘Jan is on vacation.’ b. Jan pojechał na wakacje. Jan went on holidays.ACC ‘Jan went on vacation.’ (35) a. Kot jest pod stołem. cat is under table.LOC ‘The cat is under the table.’ b. Kot uciekł pod stół. cat ran under table.ACC ‘The cat ran under the table.’ (36) a. Kon´ jest za płotem. horse is behind fence.LOC ‘The horse is behind the fence.’ b. Kon´ pobiegł za płot. horse ran behind fence.ACC ‘The horse ran behind the fence.’

[Polish]

One way to think about it is too assume that directional prepositions are structurally more complex than locative ones, and that perhaps directional prepositions are literally built from locative ones. Structural implementations

{DESKTOP}NEW FOLDER//9781107040847C05.3D

142

142

[124–160] 23.11.2013 11:35AM

Other ph(r)ases

of this general idea range from very simple ones as in (37a) to cartographically rich ones as in (37b).22 (37) a. [PP Pdir[PP Ploc DP]] b. [CP(Place) (er) [C(Place) [DegP(Place) (two meters) [Deg(Place) [PlaceP (er) [Place [PP (P) [AgrP pronouns Agr [PP (P) DP]]]]]]]]] (cf. Koopman 2000: 61)

With this as background, we can turn to the phasehood diagnostics. As we apply them, we will distinguish lexical from functional prepositions and directional from locative ones, to see if these differences correlate with differences in phasehood. We already saw that both locative and directional prepositions can value case (Accusative versus Locative in Polish); so, according to this diagnostic, both behave like phase heads. The ability to value case also does not seem to distinguish between lexical and functional prepositions. (38a) is an example with the functional preposition za ‘behind’, and (38b) is an example with the lexical preposition poza ‘except’. The two assign the same (Locative) case: (38) a. Jan wyszedł za Jan.NOM left behind ‘Jan left after Maria.’

22

Maria˛. Maria.LOC

Many proposals build on this basic insight, proposing a finer-grained structure in which there are a number of projections dominating PP, the specifiers of which provide landing sites for PP-internal movement of pronouns and adjunction sites for different types of measure phrases. Koopman’s structure in (37b) is one example. Other representative examples, which also distinguish structurally locative from directional prepositions, are given in (i–ii). (i) [PPStat (at) [DPplace [AxPartP under [PP P [NPplace the table [PLACE]]]]]] (Cinque 2010: 5) (ii) [DirP (DegP) Dir [LocP (DegP) Loc [DP N/CP]]] (cf. Biskup 2009b: 22)

Den Dikken (2010) takes Koopman’s structure as a starting point but modifies it substantially. For him, both locative and directional prepositions are lexical categories that project an extended functional layer, similarly to what other lexical categories do. Thus, verbs, nouns and prepositions all have a C-like projection, a Deixis-like projection and an Aspect-like projection. The precise nature of this projection varies depending on the nature of the head. (iii) a. [CP C[FORCE] [DxP Dx[TENSE] [AspP Asp[EVENT] [VP V …]]]] b. [CP C[DEF] [DxP Dx[PERSON] [AspP Asp[NUM] [NP N …]]]] c. [CP C[SPACE] [DxP Dx[SPACE] [AspP Asp[SPACE] [PP P …]]]] (Den Dikken 2010: 100)

Radkevich (2010) also posits a more exploded structure for PPs, in which the core P-meaning (Path, Location) can be dominated by three levels of functional projections, subject to crosslinguistic variation. While there is no variation with respect to the availability of the core structure and all languages have the locative/directional complex, Radkevich shows that languages can differ with respect to how much functional structure they allow in PP. The diagnostics Radkevich uses to ‘diagnose’ the level of functional structure involve the compatibility of PPs with measure phrases, the availability of quantifier float, and the behavior of PPs in a given language with respect to binding.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

143

[124–160] 23.11.2013 11:35AM

PPs as phases

b. Wszyscy poza Maria all except Maria.LOC ‘Everyone except Maria left.’

143

wyszli. left

So according to the ability to value cases, all prepositions ‘pass’ as phase heads. Prepositions can also show agreement. The following examples from Scottish Gaelic, due to Adger & Ramchand (2003), show that prepositions agree in definiteness with their complements: (39) a. ri tidsear with.INDEF teacher ‘with a teacher’ b. ris an tidsear with.DEF the teacher ‘with the teacher’ (Adger & Ramchand 2003: 348)

[Scottish Gaelic]

The next host of diagnostics concerns extraction from PPs. There are two questions to consider here. One is whether movement out of PPs is possible at all (especially movement of the complement of P, which gives rise to preposition stranding), and the other one is whether movement proceeds through the edge of PP (or some other functional head within the extended P projection). If the answer to both is yes, the P head is a phase head. Preposition stranding is considered to be typologically a rare phenomenon. Van Riemsdijk (1978) is quite explicit about not allowing any crosslinguistic variation with respect to extraction out of PPs, taking all PPs to be islands, and allowing extraction out of PPs only under exceptional circumstances: (40) It is impossible to relate X, Y in the structure … Xi … [Pʹʹʹ … [Pʹ … Y …] ] …Xj unless (a) Y = r-pronoun (b) Y = modifying clause (c) Y = motional postposition (Van Riemsdijk 1978: 159)

For Van Riemsdijk, movement out of PP is possible if that PP has a special escape hatch position, which he dubs the R-position in Dutch (the position to which so-called R-pronouns move) and the COMP position in English. R-pronouns (which are lexically marked +R because they contain the /r/ phoneme) are locative forms, such as daar ‘there’ er ‘there’, ergens ‘somewhere’, waar ‘where’ (see Van Riemsdijk 1978: 40).23 The R-rule inverts the R-pronoun with the preposition, which Van Riemsdijk analyzes as movement of the pronoun to [Spec,PP]. 23

Van Riemsdijk also points out that English has R-forms, such as therewith, whereby, hereafter. These, however, are more restricted and typically frozen.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

144

[124–160] 23.11.2013 11:35AM

144

Other ph(r)ases

(41) a. Ik had eri niet op ti gerekend. I had there not on counted ‘I had not counted on it.’ (cf. Van Riemsdijk 1978: 192)

[Dutch]

b. PP P′

eri P

ti

The same R-rule is operative with wh-pronouns, and accounts for the contrast in grammaticality between (42a) and (42b). In (42a), the R-rule applies first and P-stranding is allowed, whereas in (42b), the R-rule does not apply and P-stranding is disallowed. (42) a. Waari heb je [PP ti op ti] gerekend? where have you on counted ‘What did you count on?’ b. *Wiei heb je [PP op ti ] gerekend? who have you on counted ‘Who did you count on?’ (Van Riemsdijk 1978: 135–7)

[Dutch]

From the perspective of Van Riemsdijk’s theory, movement of R-pronouns in Dutch and, more generally, the ability to strand prepositions is more of an exception than the rule.24, 25 Van Riemsdijk’s evidence that movement proceeds through the edge of PP comes from the fact that movement is blocked if the edge is filled: (43) *Waari heeft zij er vaak over ti gesproken? where has she there often about spoken ‘What has she often spoken about there?’ (Van Riemsdijk 1978: 211)

This is very similar to some of the evidence we discussed in Section 4.3 showing that movement out of DPs is blocked if the specifier of DP is filled. There, however, we saw that the issue was more likely a semantic definiteness effect than a syntactic constraint on movement. In cases of PPs, however, definiteness is not a factor. 24

Preposition stranding, even in a language like English, which allows it in principle, is not without exceptions. As also pointed out by Van Riemsdijk, prepositions like during, despite or notwithstanding resist it. Others require it: (i) a. *As what is this house famous? b. What is this house famous as? (Emonds 1985: 276)

25

Strictly speaking, Van Riemsdijk allows two distinct mechanisms to permit preposition stranding: reanalysis for pseudopassives and movement through the escape hatch for wh-movement.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

145

[124–160] 23.11.2013 11:35AM

PPs as phases

145

Even though movement proceeds through this intermediate position, a wh-phrase cannot remain there, as shown in (44a). Interestingly, it can do so under sluicing, as shown in (44b), in a process referred to as swiping (see Craenenbroeck 2010 and the references therein).26 (44) a. Who read [PP whati [Pʹ about ti ] ]? b. John talked about something but I don’t know [PP whati [Pʹ about ti ] ].

The ability to strand prepositions (or the lack thereof) does not seem to distinguish between lexical and functional prepositions or locative and directional ones. In English all four types in principle allow it: (45) a. Whati were you thinking of ti? b. Whati were you hiding under ti? c. Whati was the statue resting against ti? d. Whoi were you running towards ti?

FUNCTIONAL LEXICAL LOCATIVE DIRECTIONAL

P P P P

And in a language like Polish, which disallows P-stranding, all four types disallow it, as shown in (46a–d). The pied-piped counterparts of (46a–d) are all perfectly well formed. (46) a. *Kimi rozmawiałas´ z ti? who.LOC talk.PST.2SG with ‘Who did you talk to?’ b. *Kogoi posłałas´ Jana zamiast ti? who.GEN send.PST.2SG Jan.ACC instead.of ‘Who did you send Jan instead of?’ c. *Jakich wakacjachi Jan był na ti? what.LOC vacation.LOC Jan.NOM was on ‘What kind of vacation was Jan on?’ d. *Jakie wakacjei Jan pojechał na ti? what.ACC vacation.ACC Jan.NOM went on ‘What vacation did Jan go on?’

FUNCTIONAL

P

LEXICAL

P

LOCATIVE

P

DIRECTIONAL

P

The availability of preposition stranding is only one part of a more general issue of whether extraction from PPs is allowed, and, if so, whether it has to proceed through the specifier of PP. Abels’ (2003) Stranding Generalization disallows on principled (economy-based) grounds movement of an entire complement of a phase head to the specifier of the same phase head. Such movement would have to target the edge (i.e. the specifier of the phase

26

The fact that swiping is only allowed in sluicing constructions led Craenenbroeck (2010) to a different analysis of (44b), in which the entire PP moves to [Spec,CP] first, and, next, the whpronoun moves to the specifier of a higher CP in a CP shell-like structure: (i) [CP whati [C’ C [CP [PP about tj]i [C’ C [TP … ti …] ] ] ]

{DESKTOP}NEW FOLDER//9781107040847C05.3D

146

[124–160] 23.11.2013 11:35AM

146

Other ph(r)ases

head) first, which – Abels argues – is too short: it does not establish any new relationship between H and XP, since the two are already in a very local relationship. (47)

HP H′ XP

H EPP

For Abels, movement that does not stop at the edge of HP is also not an option, as it is a violation of the Phase Impenetrability Condition. Whether the specifier of HP is filled or not should not matter, as an extra optional specifier is in principle allowed.27 (48)

ZP

Z′ Z

HP (YP)

H′ H

XP

The ban on movement in (48) does not take into account the distinction between the two versions of PIC, which, interestingly, make different predictions in this case. In particular, PIC1 disallows any movement of XP (or out of XP), as XP is spelled out as soon as HP is complete (namely when the next head Y is merged): (49)

ZP Z′ Z

HP (YP)

H′ H

XP Spell-Out domain

27

This depends of whether the filled specifier is an A or A-bar position. A specifier of vP occupied by a subject does not block wh-movement out of vP, but a specifier of CP occupied by a wh-phrase does block wh-movement of another wh-phrase out of this CP.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

147

[124–160] 23.11.2013 11:35AM

PPs as phases

147

But PIC2 does not spell out XP until the next phase head is merged. Thus, it allows XP in (50a) to move to a higher specifier position without stopping at the edge of HP, as long as this higher specifier position is not a specifier of a phase head. Spell-Out does not occur until later, as shown in (50b). (50) a.

YP Y′ Y

H (YP)

H′ H

b.

XP

ZP Z

YP Y′

XPi Y

HP (YP)

H′ H

ti Spell-Out domain

This predicts the existence of languages that might allow preposition stranding with A movement but not with A-bar movement. These are languages in which the following contrast would hold:28 (51) a. J O H N W A S b. *WHOi D I D

T A L K E D A B O U T ti. W E T A L K A B O U T ti?

However, Abels is very explicit about it being not an attested pattern. (52) All languages that allow P-stranding under A-movement, i.e. pseudopassivization, also allow P-stranding under A’-movement. (Abels 2003: 230)

28

I use small caps here to indicate that these are hypothetical/pseudo-English examples.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

148

[124–160] 23.11.2013 11:35AM

148

Other ph(r)ases

The grammatical (51a) involves precisely the configuration in (53), in which movement targets a specifier position which is lower than the (phase) head that determines Spell-Out:29 (53)

CP C

TP

T′

Johni T

vP VP

v V

PP P

ti Spell-Out domain

Another prediction of the Stranding Generalization (and the account of preposition stranding that relies on it), examined in detail by Abels, concerns extraction from – as opposed to of – the complement of the P head. This is allowed: (54), for example, involves movement of the PP über welches Thema ‘about which topic’ from the complement of the PP headed by malnach ‘after’. (54) Über welches Themai hast du mich noch [PP malnach ti ] einem Buch gefragt? about which topic have you me again after a book asked ‘Which topic did you ask me about a book on again?’ (Abels 2003: 211)

Furthermore, if some prepositions are more complex than others (for example, if directional prepositions are more complex than locative ones), we might expect to find cases in which locative prepositions can be stranded but directional ones cannot. This does not seem to be the case either in P-stranding or non-Pstranding languages. The next diagnostic concerns binding: can PPs be domains for binding? Interestingly, as often noted, different types of prepositions do behave differently in this respect (see Büring 2005, Chomsky 1986, Radkevich 2010, Reinhart & Reuland 1993, among many others). For example, PPs headed by prepositions like near or apart from can constitute a distinct binding domain, but PPs headed by prepositions like about or with do not.30, 31 29 30 31

The passive v-V complex, irrespective of whether it is a phase or not, does not determine Spell-Out. The examples are from Reinhart & Reuland (1993); I have added the indices and bolding. Radkevich (2010) notes a fair amount of crosslinguistic variation in this respect and shows that languages that allow coindexed pronouns inside PPs have more functional structure than languages that do not. PPs in Polish, Czech and Russian are structurally ‘impoverished’, and

{DESKTOP}NEW FOLDER//9781107040847C05.3D

149

[124–160] 23.11.2013 11:35AM

PPs as phases

149

(55) a. Maxi saw a gun near himselfi/himi. b. Luciei counted five tourists in the room apart from herselfi/heri. (56) a. Maxi speaks with himselfi/*himi. b. Lucie’si joke about herselfi/*heri (Reinhart & Reuland 1993: 661)

A related diagnostic concerns stranding and the question of whether the quantifier all or a nominal in a split constituent can be stranded at the edge of PP. West Ulster English, which we know allows quantifier stranding under whmovement (see McCloskey 2000 and the discussion of CPs as phases in Chapter 4), is a good language to look at. The evidence is mixed at best. McCloskey shows that it is possible to strand all in the specifier of the prepositional complementizer for: (57) Who did you arrange all for your mother to meet at the party? (McCloskey 2000: 70)

However, while stranding is fine in PPs in principle (see (58)), it is impossible in the [Spec,PP] position (see (59)). (58) a. ?Who did you talk to all (at the party)? b. ?Who was he laughing at all? (59) a. *Who did you talk all to? b. *What were you laughing all at? (McCloskey 2000: 65?)

Bošković (2004b) links the availability of PP-internal quantifier stranding, which is slightly different from the quantifier stranding given in (58) above, as the nominal remains inside the PP (or its extended projection), to the availability of object shift. English, which disallows both, differs in this respect from Icelandic, which allows both. Bošković also argues that both are subject to the same restrictions: both quantifier stranding and object shift, for example, are excluded in periphrastic constructions and with indefinite objects. (60) a. *John the booksi/themi read ti. b. *John talked the books/them about ti.

lack the functional layer that would allow them to count as a separate binding domain. SerboCroatian and Slovak, on the other hand, behave like English; they allow the coreferential pronoun inside the PP, which suggests more structure, and potentially a clausal boundary. (i) Mariai Maria

połoz·yła

bron´

koło

siebiei/*nieji.

put

gun

near

herself/her

[Polish]

‘Maria put a gun near herself.’ (ii) Maria Maria

položila

zbranˇ

blizko

sebai/?neji.

put

gun

near

herself/her

‘Maria put a gun near herself.’ (Radkevich 2010: 126)

[Slovak]

{DESKTOP}NEW FOLDER//9781107040847C05.3D

150

150

[124–160] 23.11.2013 11:35AM

Other ph(r)ases

(61) a. Halldór lasj bækurnari ekki tj ti. Halldor read the.books not ‘Halldór didn’t read the books.’ b. ?Ég talaði (i gaer) við stúdentanai

[Icelandic]

alla ti. I talked yesterday with the-students all ‘Lit. I talked yesterday with all the students.’ (cf. Boškovic´ 2004b: 102)

The simplest derivation that would yield (61b) is the one in (62), in which the NP simply moves to [Spec,PP], and the quantifier is stranded in D.32 (62) a. [PP with [DP all students]] b. [PP studentsi [Pʹ with [DP all ti ]]] c. [pP with [PP studentsi [Pʹ t [DP all ti]]]]

However, a more complex derivation is also possible, in which stranding does indeed happen in [Spec,PP].33 (63) a. [PP with [DP all the students]] b. [PP all the studentsi [Pʹ with ti ]] c. [FP the studentsj [PP [all tj]i [Pʹ with ti ]]] d. [pP withk [FP the studentsj [PP [all tj]i [Pʹ tk ti ]]]]

The last two diagnostics we will consider involve Quantifier Raising and ellipsis. If QR can target PP, we have extra evidence that movement can proceed through the edge of PP. However, PPs are not of the right semantic type to serve as the target for QR. If the complement of P can undergo ellipsis, we have evidence that P can determine Spell-Out. Here, the evidence also suggests that P heads cannot license ellipsis, but it is reducible to independent factors. Thus, the ungrammaticality of (64a) in Polish is due to the lack of preposition stranding. And the ungrammaticality of (64b) in English, a language with preposition stranding, could be due to the lack of argument ellipsis in English. (64) a. *Jan zadzwonił do Marii a Piotr poszedł do Δ. Jan phoned to Maria and Piotr went to ‘Jan called Maria and Piotr went to see her.’ b. *John talked about a holiday in Paris and Bill dreamt about Δ.

[Polish]

As we can see, neither QR nor ellipsis provides an argument in favor of PPs being phases.

32

33

The derivation in which ‘students’ moves to [Spec,DP] stranding all in D is excluded as a violation of antilocality. This is more in line with Bošković’s actual derivation, given in (i). For Bošković, however, there are extra steps involved, as he disallows stranding in thematic positions: (i) [PP withj [OP studentsi tj [AgrpP [all ti] tj [PP tj ti(cf. ]]]] Boškovic´ 2004b: 108)

{DESKTOP}NEW FOLDER//9781107040847C05.3D

151

[124–160] 23.11.2013 11:35AM

ApplPs as phases

151

This concludes our discussion of PPs as phases. The behavior of PPs with respect to phasehood diagnostics is mixed, as summarized in the following table. (65) PPs as phases Diagnostic Question

Answer Evidence

Is P the source of uninterpret- Yes able features? Is PP a domain for feature Yes valuation? Does movement out of PP have to proceed through the edge of PP?

Yes

Can QR target PP? No Can the complement of P be No deleted? Can PP be a binding domain? Yes

structural ‘prepositional’ cases Locative versus Accusative case on complements of directional versus spatial prepositions blocked by filled edge position quantifier stranding in [Spec,PP] (subject to crosslinguistic variation), ungrammaticality of preposition stranding (subject to crosslinguistic variation) PP is not the right semantic type ungrammatical in languages that lack argument ellipsis with certain prepositions

5.3 ApplPs as phases In this section, we turn to the idea, due to McGinnis (2001), that certain types of double object constructions are phases. This idea relies on the assumption that in a double object construction, the relationship between the direct and the indirect object is mediated by a functional head called an applicative head (see Pylkkänen 2008 and the references therein for arguments in favor of such an analysis):34

34

There are many alternative structures for double object constructions, and many issues surrounding double object constructions, which I will abstract away from here. For example, the relationship between the two objects in a double object construction is in principle independent from the issue of what the relationship is between the so-called double object and dative constructions (expressed by means of the preposition to in English). More specifically, the issue is whether the two sentences in (i) and (ii) are related transformationally or not. (i) Jan gave Piotr a cookie. (ii) Jan gave a cookie to Piotr.

I refer the interested reader to Green (1974) and Oehrle (1976) for some early diagnostics and observations about the differences between the two. To illustrate briefly, double objects are infelicitous with locations (as shown in (iia–b)) and imply a successful transfer of possession

{DESKTOP}NEW FOLDER//9781107040847C05.3D

152

[124–160] 23.11.2013 11:35AM

152

Other ph(r)ases

(66)

ApplP IO

Appl′ Appl

DO

This straightforwardly captures the fact that the indirect object c-commands the direct object, as can be shown by applying the familiar Barss & Lasnik (1986) c-command diagnostics, which I will refrain from doing here. The structure in (66) also captures the fact that in some languages, the applicative head is realized overtly, as shown by the following examples from Bukusu (a northeastern Bantu language).35 (67) a. n-a-keend-el-a omu-xasi. 1SS-tense-walk-APP-FV CL1-woman ‘I walked for the woman.’ b. n-a-ar-ir-a e-nyuungu luu-saala 1SS-tense-break-APP-FV CL9-pot CL11-stick ‘I broke the pot with the stick.’ c. wanjala e-er-er-a en-goxo e-yaywa Wanjala 3sS.tense-kill-APP-FV CL9-chicken CL9-axe ‘Wanjala killed the chicken with the axe.’ (Peterson 2007: 7)

[Bukusu]

Pylkkänen distinguishes two types of applicative structures, which she refers to as high and low applicatives.36 This distinction is empirically motivated by the fact that applicative constructions can behave differently from language to language, and, furthermore, that different types of applicatives can behave differ-

(either literal or metaphorical). Thus (iiib) implies that the students successfully mastered Polish; no such implication is present in a DP PP frame. (ii) a. Maria sent a book to Warsaw. b. #Maria sent Warsaw a book. (iii) a. Maria taught Polish to the students. b. Maria taught the students Polish.

35

36

In languages that allow both orders of the two objects, the issue of some contention is which of the objects is basic and which one derived (see Miyagawa & Tsujioka 2004 for a discussion of Japanese, and Bailyn 2009 and Dyakonova 2007 for a discussion of Russian). See also the brief discussion of dative shift in Chapter 6. The applicative head can convey a large number of meanings: benefactive, malefactive, instrumental, locative and comitative meanings, among others. Other types of applicative markers, discussed by Peterson (2007), include prioritive ones (which indicate that the action was completed before a given object) or relinquitive ones, which indicate that the action was accomplished without the (applied) object. See also Marantz (1993) for the idea that different types of applicative heads might occupy different levels, even in a language like English.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

153

[124–160] 23.11.2013 11:35AM

ApplPs as phases

153

ently even within a single language. The differences involve transitivity restrictions, compatibility of an applicative structure with static verbs and the ability for a depictive modifier to modify the indirect object: (68) High versus low applicatives

possible with intransitive verbs possible with stative verbs possible for depictive modifiers to modify indirect object

High applicatives

Low applicatives

√ √ √

* * *

According to these diagnostics, English lacks high applicatives. First, it disallows applicative objects with intransitive verbs, such as the unergative run or the unaccusative die. (69) a. *Maria ran Jan. b. *Maria died Jan.

They are not implausible semantically as applicatives; (70a) could be a benefactive and (70b) a malefactive: (70) a. Maria ran for the benefit of Jan. b. Maria died to the detriment of Jan.

Second, English disallows applicative objects (indirect objects) with stative verbs like hold. This is illustrated by the ungrammaticality of (71a), whose intended interpretation is the one in (71b). (71) a. *I held Maria the door. b. I held the door for Maria.

And, third, English disallows the indirect object to be modified by depictives: (72) *I gave Mariai the book curiousi

In a language with high applicatives, all three applicative sentence types are possible. Pylkkänen applies these diagnostics to a variety of languages; below I illustrate with data from Luganda. (73) a. Mukasa ya-tambu-le-dde Katonga. Mukasa 3SG.PAST-walk-appl-PAST Katonga ‘Mukasa walked for Katonga.’ b. Katonga ya-kwaant-i-dde Mukasa ensawo. Katonga 3SG.PAST-hold-APPL-PAST Mukasa bag ‘Katonga held the bag for Mukasa.’ (Pylkkänen 2008: 20)

[Luganda]

{DESKTOP}NEW FOLDER//9781107040847C05.3D

154

[124–160] 23.11.2013 11:35AM

154

Other ph(r)ases

c. Nd-o-shum-el-a Katonga o neta. 1SG-PAST-work-APPL-FV Katonga 3SG tired ‘I worked for Katonga while he was tired.’ (Pylkkänen 2008: 32)

Pylkännen derives the differences between the two types of applicatives from either structural or interpretative considerations. First, she establishes a different structure for high and low applicatives. In a high applicative structure, given in (74) below, the direct object is the complement of the verb, and the indirect object is the specifier of a high applicative head (which itself takes a VP as a complement). A high applicative head establishes a relationship between the indirect object and the event described by the VP containing the direct object: (74)

ApplP IO

Appl′ Appl

VP V

DO

In a low applicative structure, the asymmetric relationship between the indirect and direct object remains the same in that the indirect object c-commands the direct one; however, this relationship is more direct. The Applicative Phrase, with the indirect object as its specifier and the direct object as its complement, is a complement of the verb: VP

(75) V

ApplP IO

Appl′ Appl

DO

Pylkännen argues that the differences illustrated above follow from this difference in structure. Given that the low applicative head takes a direct object as its complement and the indirect object as its specifier, the ungrammaticality of applicative structures in which the direct object is missing becomes a matter of lexical selection; in the English examples in (69a–b) above the selectional requirements of the applicative head are not satisfied. Since in a high applicative structure, the applicative head selects the VP, it does not matter if this VP is transitive or not. The next two differences between high and low applicatives, namely the fact that only high applicatives are compatible with static verbs and allow the indirect object to be modified by a depictive phrase, follow from the semantics of the two types of applicative heads.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

155

[124–160] 23.11.2013 11:35AM

ApplPs as phases

155

So far, nothing we have said about the two types of applicatives has any bearing on their phasal status, as the diagnostics distinguishing the two types are independent of the diagnostics that distinguish phase heads from non-phase heads. However, McGinnis (2001) argues that high applicatives differ from low applicatives in that the high applicative head is a phase head but a low one is not. As such, it can have the EPP feature, which allows movement of the direct object over the indirect one. What this captures nicely is the distinction between languages in which only the higher of the two objects (the indirect object) can passivize, and languages in which either the direct or the indirect objects can passivize. The examples in (76a–b) show that Bukusu is of the latter type: in benefative applicatives, either object can undergo passive movement:37 (76) a. omu-xasi a-kul-il-w-a sii-tabu nee-wanjala. [Bukusu] CL1-woman 3SS-buy-APP-PASS-FV CL7-book by-Wanjala ‘The woman was bought the book by Wanjala.’ b. sii-tabu sy-a-kul-il-w-a omu-xasi (?nee-wanjala). CL17-book CL7S-tense-BUY-APP-PASS-FV CL1-woman by-Wanjala ‘The book was bought for the woman (?by Wanjala).’ (Peterson 2007: 8)

From a purely locality-centric perspective, this pattern is surprising; if the two objects stand in an asymmetric relationship with respect to each other, only the higher one should be able to passivize. McGinnis attributes the ability of either object to passivize to the fact that the direct object moves to the (outer) specifier of the ApplP head first. From this position, it is going to be closer than the indirect object to T for the purposes of passivization. Passivization of the direct object instead (illustrated in (77b)) would constitute a standard locality violation, irrespective of how locality is implemented (Attract Closest, Minimal Link Condition, Shortest Link).38

37

38

The language McGinnis focuses on is Kinyarwanda, another Bantu language in which either object can in principle passivize. In this language (as in many Bantu languages), different types of applicatives behave differently with respect to passivization. For example, in locative applicatives, only the locative argument can passivize. McGinnis speculates that perhaps more generally the difference in phasal status of the two types of applicatives is due to the fact that a sister of a VP can head a phase only if it has an argument in its specifier. This correctly distinguishes high applicatives from the low ones, but it does not seem to be a general fact about phase heads; for example, it is not the case that C heads only constitute phases if the specifier of CP is filled. Perhaps this is due to the fact that the specifier of CP is not a thematic position, which would mean that McGinnis’s generalization could be restated as a statement that the sister of VP counts as a phase head only if it selects an external argument. If, however, passive and unaccusative vPs are phase heads, as argued by Legate (2003) (whose arguments we reviewed in Section 4.2), this might not be the right characterization, either.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

156

[124–160] 23.11.2013 11:35AM

156

Other ph(r)ases

(77) a. High applicative TP T′ Tuϕ:[

], EPP

ApplP ApplP

DOi IO

Appl′ VP

ApplEPP V

ti

b. Low applicative TP DOi

T′

T VP V

ApplP Appl′

IO Appl

ti

Movement of the direct object to the outer specifier of the ApplP (over the indirect object), which is analogous to what we have seen above in a high applicative structure, can be ruled out in two – in principle – independent ways. We could appeal to the idea that the low applicative head, being a non-phase head, does not have the EPP feature, which is the feature driving the movement of the direct object to the outer specifier.39 Alternatively, we could appeal to antilocality, which is what Jeong (2006) does. We have seen above what allows passivization of direct objects in high applicatives. We have yet to see what allows passivization of indirect objects; if 39

Feature Inheritance is not an option either, since a low applicative phrase is dominated by a nonphase head (the V head).

{DESKTOP}NEW FOLDER//9781107040847C05.3D

157

[124–160] 23.11.2013 11:35AM

ApplPs as phases

157

the direct object were always to move across the indirect object, passivization of the direct object should be the only possibility. We could either assume that the direct object does not move to [Spec, ApplP], which implies that the EPP feature on the high applicative head is optional (not an unreasonable assumption), or that it does move, but that it does so in a tucking-in fashion, below the indirect object, the latter option being the one favored by McGinnis.40 These two possibilities are schematized in (78a–b). (78) a.

TP IO

T′ T ApplP Appl′

ti Appl

VP V

b.

DO

TP T′

IOi T

ApplP ti

ApplP DOj

Appl′ ApplEPP V

VP tj

So far, the discussion in this section has established that high and low applicatives differ with respect to extraction, and that passivization of the lower object in a high applicative proceeds through the edge of the ApplP. The extraction diagnostic thus distinguishes between low and high applicatives,

40

Equidistance is a logical possibility, too. However, since equidistance became ‘obsolete’ with Feature Inheritance in place, I will not elaborate on it here.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

158

[124–160] 23.11.2013 11:35AM

158

Other ph(r)ases

but it is not clear that other diagnostics do. For example, the Appl head is implicated in case valuation in both high and low applicatives.41 To complete the discussion of high Applicative Phrases as phases, let us apply the remaining phasehood diagnostics to high applicatives. Polish is a language which, according to Pylkkanen’s diagnostics, has high applicatives (see Citko 2010, 2011b for a more detailed discussion). Dyakonova (2007) makes a similar point about Russian, and the Polish examples below are based on her Russian ones. (79) a. Be¸de¸ wam s´piewac´ i tan´czyc´. [Polish] be.FUT you.DAT sing.INF and dance.INF ‘I will sing and dance for you.’ b. Siedzisz mi na sukni. sit.2SG me.DAT on dress ‘You’re sitting on my dress.’ c. Maria odebrała nam klucze juz· troche˛ pijanym. Maria took.away us.DAT keys already bit tipsy.INSTR ‘Maria took away the keys from us (while we were) already a bit tipsy.’

Direct objects can be passivized, which implies movement through the outer specifier of ApplP. The ungrammaticality of (80b) is unexpected in a high applicative structure, but receives an independent explanation based on case: indirect object is dative and dative objects do not passivize, something I discuss in detail in Citko (2011b).42 (80) a. Ksia˛z·kai została wysłana Marii ti (przez book.NOM became sent Maria.DAT by ‘A book was sent to Maria’ b. *Mariai została wysłana ti ksia˛z·ke¸. Maria.NOM became sent book.ACC ‘Maria was sent a book.’

Jana). Jan.ACC

Either object can be wh-moved: (81) a. Komui Jan wysłał ti ksia˛z·ke¸? who.DAT Jan.NOM sent book.ACC ‘Who did Jan send a book to?’ b. Coi Jan wysłał ti? Marii Maria.DAT what.ACC Jan.NOM sent ‘What did Jan send Maria?’

41

42

This is something we have not discussed in a lot of detail here. But it is quite standard to assume that the Applicative head values the case feature of the indirect object (often as dative). Even though the dative indirect object cannot move, it could still act as an intervener (of a defective kind), so similar considerations apply.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

159

[124–160] 23.11.2013 11:35AM

ApplPs as phases

159

When both move, the ordering is free, which is not a surprising fact for a language with no superiority effects: (82) a. Coi komuj Jan wysłał tj ti? what.ACC who.DAT Jan.NOM sent ‘What did Jan send to whom?’ b. Komuj coi Jan wysłał tj ti? who.DAT what.ACC Jan.NOM sent ‘What did Jan send to whom?’

I assume the underlying order is one in which the indirect object precedes the direct one, even though this is an issue of some controversy in the literature on Slavic double object constructions (see, for example, Bailyn 2009 and Dyakonova 2007 for opposing views on this matter). This implies that the whmovement of the direct object is always going to proceed through the edge of ApplP, and that the structure of (83a) is that of (83b). (83) a. Coi Jan wysłał Marii ti? what.ACC Jan.NOM sent Maria.DAT ‘What did Jan send to Maria?’ b. [CP whati [TP Jan [vP sent [ApplP ti [ApplP Maria [VP ti ]]]]]]

The edge of ApplP may also be the shifted position for ‘scrambled’ orders in which the direct object precedes the indirect one: (84) a. Jan wysłał ksia˛z·ke¸i Marii ti. Jan sent book.ACC Maria.DAT ‘Jan sent a book to Maria.’ b. [TP Jan [vP sent [ApplP a booki [ApplP Maria [VP ti]]]]]

What is relevant for our purposes is the fact that a complex wh-phrase moving through the edge of ApplP can strand a nominal in that position: (85) a. Ilei Jan wysłał ksia˛z·ek Marii ti? how.many.ACC Jan.NOM sent books.GEN Maria.DAT ‘How many books has Jan sent Maria?’ b. [CP how.many [TP Jan [vP sent [ApplP how.many books [ApplP Maria [VP how.many books]]]]]]

Ellipsis of the complement of the high applicative head is also possible:43 43

The fact that the English example in (i) does not readily allow the interpretation in (ii) is thus consistent with English Applicative Phrases not being phases. (i) John sent Maria letters and Peter Eve. (ii) John sent Maria letters and Peter sent Eve letters.

{DESKTOP}NEW FOLDER//9781107040847C05.3D

160

[124–160] 23.11.2013 11:35AM

160

Other ph(r)ases

(86) a. Jan wysłał Marii listy, a Piotr Ewie. Jan.NOM sent Maria.DAT letters.ACC and Piotr.NOM Ewa.DAT ‘Jan sent Maria letters and Peter (sent letters to) Eve.’ b. [TP Piotr.NOM [vP [ApplP Ewa.DAT [Applʹ Appl [VP sent letters ]]]]]

The behavior of (high) Applicative Phrases with respect to phasehood diagnostics is summarized in (87). (87) High ApplP as phases Diagnostic Question

Answer Evidence

Is Appl the source of uninterpretable features? Does movement out of ApplP have to proceed through the edge of ApplP?

Yes

Does Appl determine Spell-Out?

Yes

Yes

Dative case on indirect objects Passivization of direct objects, discontinuous nominals VP ellipsis in high applicatives

More generally, we have seen in this chapter how extending phasehood to three new categories (Predication Phrases, Prepositional Phrases and certain types of Applicative Phrases) fares. The results are somewhat more mixed. In the next chapter, we turn to variation in phasehood, which, broadly speaking, refers to the idea that the status of a given category as a phase might vary not only from language to language but also within a single language. There are a couple of possibilities to consider. One is that a given head can acquire (or lose) its phasehood status in the course of the derivation. This will be the topic of Section 6.1, where I focus on proposals like those of Gallego or Den Dikken, for whom head movement (of a phase head to a non-phase head dominating it) can result in what would otherwise be a non-phase head becoming a phase head. The second type of variation concerns crosslinguistic variation. Such variation could arise trivially if a language lacks a given category (which is considered universally to be a phase). We have seen some instances of this scenario already. For example, applicatives ‘count’ as phases only in languages that have high applicatives. There are other cases to consider, as we will see in Section 6.2 of next chapter. And the last type of variation, which is referred to in the literature as nonsimultaneous phasehood, will be the topic of Section 6.3.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

161

[161–184] 23.11.2013 11:05AM

6 Variation in phasehood

6.1

Dynamic phases

This chapter focuses on the idea that the phase status of a given category can change throughout the derivation. This is very different from the view that has been implicit throughout this book, which is that phasehood, once fixed, remains constant. The ideas we will examine in this chapter allow categories to become phases or stop being phases. Such a dynamic approach to phasehood has been articulated most explicitly by Den Dikken (2007) and Gallego (2010), and in what follows, I focus on their proposals. What they share is the basic idea that a non-phase head can become a phase head under certain well-defined circumstances. Where they differ is in the details of implementation, as well as in their fundamental assumptions about the nature of phases. For both Den Dikken and Gallego, the process responsible for the variability of the phase status of a given head is head movement. Both thus rely on head movement being a syntactic movement, as opposed to a post-syntactic PF movement. Den Dikken, whose formulation is given in (1), refers to this process as Phase Extension, and Gallego refers to it as Phase Sliding. (1) Phase Extension Syntactic movement of the head H of a phase α to the head X of the node β dominating α extends the phase up from α to β; α loses its phasehood in the process, and any constituent on the edge of α ends up in the domain of the derived phase β as a result of Phase Extension. (Den Dikken 2007: 1)

According to (1), HP is a phase in (2a). However, after it moves to X in (2b), it stops being a phase, and XP becomes a phase instead. What changes is not just the status of XP and HP; the edge of the phase and the Spell-Out domain also get adjusted accordingly. In (2b), HP will eventually become the Spell-Out domain, but not until the phase is complete and the next head is merged (on PIC1) or the next phase head up is merged (on PIC2).

161

{DESKTOP}/9781107040847C06.3D

162 [161–184] 23.11.2013 12:21PM

162

Variation in phasehood

(2)

a.

XP PHASE X

HP ZP

H′ YP

H

SPELL-OUT DOMAIN

EDGE

PHASE

b.

XP WP

X′ X

HP

EDGE

SPELL-OUT DOMAIN ZP

H′ H

YP

Logically speaking, it is also possible for the lower head not to lose its phase status. On this scenario, we end up with two phase heads (X and H), two edges (circled below), and two Spell-Out domains: (3)

XP X′

WP

HP

X

SPELL-OUT DOMAIN 2 ZP

H′ H

YP

SPELL-OUT DOMAIN 1

There are two arguments against adopting the scenario in (3), in which a phase head moves up to another phase head and keeps its phase status. First, it is incompatible with the assumption that every phase head has to dominate a non-phase head, as argued by Richards (2008), whose reasoning we reviewed in the section on Feature Inheritance in Chapter 2 (Section 2.6). Second, it will not allow Agree between X and YP, which – as we will see shortly – was Den Dikken’s main empirical motivation for Phase Extension. Phase Extension is not the only innovation in Den Dikken’s system. As we saw earlier, his definition of a phase also departs crucially from the standard definition. For Den Dikken, it is neither propositionality, nor complete argument structure, nor being the locus of uninterpretable features that defines a phase. For him, what defines a phase is a subject–predicate configuration. Such

{DESKTOP}NEW FOLDER//9781107040847C06.3D

163

[161–184] 23.11.2013 11:05AM

Dynamic phases

163

configurations (and only such configurations) constitute inherent phases. Others (i.e. non-subject–predicate configurations) can become phases only as a result of Phase Extension.1 Den Dikken (2007: 27) considers the ‘classic’ criteria for phasehood, such as semantic and phonological independence, being the loci of uninterpretable features and EPP features, contrasting his dynamic approach with Chomsky’s ‘static’ approach. Den Dikken’s approach is both stronger and weaker than Chomsky’s. On the one hand, there are some categories which count as phases given Chomsky’s diagnostics (or, perhaps to put it in more accurate terms, given the field’s understanding of Chomsky’s diagnostics), but which do not always count as phases on Den Dikken’s diagnostics. A CP is a good example: for Den Dikken it can only become a phase by extension.2 On the other hand, some of the structures that Den Dikken analyzes as predication structures (and thus phases) are not typically considered to be phasal. Given the conclusions of Section 5.3, English double object structures are a case in point. Since the concept of Phase Extension is in principle independent of redefining phases as predication structures, I will consider Phase Extension on its own merits. Den Dikken illustrates how Phase Extension works (and why it is necessary) with three well-documented types of inversion structures, illustrated in (4)–(6): predicate inversion, locative inversion and dative shift. In all of them, the subject and the predicate are given in italics, with the b examples being the inverted variants. (4) a. Spix’s Macaw is the world’s rarest parrot b. The world’s rarest parrot is Spix’s macaw. (5) a. All the parrots flew up the river. b. Up the river flew all the parrots. (6) a. Leslie gave each parrot a piece of fruit b. Leslie gave a piece of fruit to each parrot

1

As we saw in Chapter 2, such an understanding of phases relies on an independent understanding of predication; something which is the focus of his earlier work (Den Dikken 2006b). The list in (i–vii) provides examples of the kinds of structures that Den Dikken (2006b) considers to be predication structures. While many of them can be pretty straightforwardly analyzed as such, it is not completely clear whether we would want to treat all of them as phases. (i) I have [SUBJECT a parrot] for [PREDICATE a friend] (ii) [SUBJECT Parrots], [PREDICATE I really admire] (iii) I gave [SUBJECT a nut] to [PREDICATE a parrot] (iv) I put [SUBJECT the parrot] on [PREDICATE its perch] (v) [SUBJECT Parrots screech] [PREDICATE loudly]. (vi) [PREDICATE the monster] of [SUBJECT a parrot] (vii) [SUBJECT Mary]’s [PREDICATE parrot]

2

If CP is not a phase, wh-movement through (or to) [Spec,CP] can only take place if there is v-T-C movement; something that happens universally at LF, Den Dikken argues.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

164

[161–184] 23.11.2013 11:05AM

164

Variation in phasehood

I will focus here on copular constructions, since we have devoted a fair amount of attention to them in the previous chapter. The PrP structure presented there is different from Den Dikken’s structure, but it is consistent with his assumptions that the relationship between the subject and the predicate is mediated by a functional projection. On Den Dikken’s assumptions, all the b examples in (4)– (6) involve inversion of the predicate around the subject, as illustrated in (7a–b): (7) a. Spix’s Macawi is [PrP ti [Prʹ Pr the world’s rarest parrot] ]. b. The world’s rarest parroti is [PrP Spix’s Macaw [Prʹ Pr ti] ].

A similar inversion mechanism applies to locative inversion and dative shift cases, the only difference being that the functional head mediating the relationship between the subject and the locative PP in (5a–b) or the two objects in (6a–b) is different.3 Den Dikken argues that in all of them, the relationship between the subject and the predicate is mediated by a functional category he refers to as a relator, which projects its own phrase, the Relator Phrase (RP). In the cases at hand, the RP is dominated by an FP (where FP stands for any phrase headed by a functional head F). In copular clauses, RP and FP stand for PrP and TP, respectively: (8)

FP(TP) F/T

RP(PrP)

SUBJECT

R′(Pr′)

R(Pr)

PREDICATE

Den Dikken’s main theoretical argument in favor of Phase Extension involves the following issue: if RP (TP) in (9) is a phase, how is it possible for the functional head F(T) to establish an Agree relationship with the predicate, and, consequently, for the predicate to move to [Spec,TP]? (9)

FP/TP F′(T′) F/T

RP(PrP)

SUBJECT

R′(Pr′)

R(Pr)

3

PREDICATE

We might wonder whether dative shift (which in Den Dikken’s terms is inversion around the preposition to) is compatible with the approach to double object verbs advocated in Section 5.3 above, which relied on the presence of an applicative head. While this is not the approach Den Dikken takes, nothing in principle would prevent the inversion from taking place around the applicative head.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

165

[161–184] 23.11.2013 11:05AM

Dynamic phases

165

If, however, R raises to F, as shown below, then RP stops being a phase and F (with R adjoined to it) becomes one. Now, on Den Dikken’s assumptions, F can undergo Agree with the predicate; the complement of the derived phase head (i.e. RP) is not spelled out until FP is merged with a higher head. (10)

FP(TP) F/ T

RP(PrP) R′(Pr′)

SUBJECT

R(Pr)

PREDICATE

Note, however, that predicate inversion in (9) is only a problem if the complement of the phase head R, the predicate itself, is spelled out as soon as the phase is complete, i.e. when F is merged. This is what the earlier version of PIC, referred to as PIC1 throughout this book, mandated. However, if F is not a phase head (as is the case if F is T), the domain of the lower phase head R is not spelled out till the next phase head, C in this case, is merged. Thus, under the assumptions of PIC2, Agree between F and the complement of R in (9) (as well as movement of the complement of R to the specifier of TP) should be possible, unless F is independently shown to be a phase head. For Den Dikken, however, Phase Extension is what allows movement of the predicate to the edge of the derived phase.4 FP/TP

(11) PREDICATEi

F/ T

F′/ T′ RP(PrP) SUBJECT

R′(Pr′)

R(Pr)

ti

Now the (original) subject of PrP becomes part of the Spell-Out domain when a higher head is merged:5

4

5

Den Dikken relies on equidistance in order to explain why the subject in the specifier of RP does not intervene. In order to rule out the derivation in which the subject moves to the edge of FP first, Den Dikken appeals to a ban against adjunction to empty categories.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

166

[161–184] 23.11.2013 11:05AM

166

Variation in phasehood

(12)

HP(CP) H(C)

FP/ TP

PREDICATEi

F/ T

F′/ T′ RP(PrP) R′(Pr′)

SUBJECT

R(Pr)

ti

Phase Extension can not only explain what allows the predicate to move to [Spec,FP], but also why the postverbal subject (the original subject in [Spec, RP]) cannot move. The immobility of postverbal subjects in inverted copular constructions, illustrated in (13a–b), is well documented in the literature (see Den Dikken 2006, Moro 1997, 2000, among others). On the Phase Extension account, the ungrammaticality of the b examples is a consequence of the structure and derivation in (11), where the original subject is ‘buried’ inside the Spell-Out domain of the derived phase head T.6 (13) a. Which parroti do you think is ti the world’s rarest parrot? b. *Which parroti do you think the world’s rarest parrot is ti? (14) a. Which parrotsi do you think ti flew up the river? b. *Which parrotsi do you think up the river flew ti?

Let me conclude this section with a brief comparison of Den Dikken’s Phase Extension with Gallego’s Phase Sliding. Phase Sliding is similar in spirit, but it relies on a different (more standard) definition of a phase, and it applies to a different empirical domain. Gallego shares Den Dikken’s misgivings about interface-based criteria for distinguishing phases from non-phases, suggesting a need for a definition of phases that does not rely on the interface diagnostics.

6

Den Dikken extends this reasoning to DPs, and takes the ban on extraction out of specific noun phrases of the kind illustrated in (i) to be a consequence of a DP becoming a phase by extension due to the raising of the relator head (the head mediating the relationship between picture of who and maria) to D, as shown schematically in (iia–c). After Phase Extension, the wh-phrase who is ‘buried’ inside the complement of the derived phase headed by the complex head D with an R head adjoined to it. Since DP is an argument, adjunction to it is blocked by assumption. (i) Whoi did Jan see Maria’s picture of ti? (ii) a. [RP picture of who R Maria] b. [DP R+D [RP picture of R Maria] ] c. [DP Mariai’s R+D [RP picture of ti] ]

{DESKTOP}NEW FOLDER//9781107040847C06.3D

167

[161–184] 23.11.2013 11:05AM

Dynamic phases

167

He treats the interface ‘behavior’ of phases as a consequence of phasehood rather than a diagnostic of phasehood. However, he also points out problems with Den Dikken’s predication-based definition of phases, proposing instead the familiar by now definition based on the presence of uninterpretable features: (15) Phase Condition Uninterpretable features (uFF) signal phase boundaries. (Gallego 2010: 51)

Gallego thus does not redefine phases; he still takes core phases to be CPs and vPs, and he shares with Den Dikken the idea that head movement can extend phases. His empirical focus is on v to T movement, and the contrast between Romance languages (Spanish and Catalan primarily) and English in this respect. He frames his discussion in more general terms, suggesting that Phase Sliding could be what distinguishes Null Subject Languages (NSLs) from non-Null Subject Languages. I illustrate his proposal below with data from Polish, also a NSL. As is well known, NSLs differ from non-NSLs in that they do not show that-trace effects: (16) a. Ktoi Jan chce, z·eby ti przeczytał ten who Jan wants COMP read this ‘Who does Jan want to read this poem?’ b. *Whoi do you think that ti read this poem?

wiersz? poem

[Polish]

The classic account of this contrast is to attribute the grammaticality of (16a) to the fact that the subject moves from a postverbal position so there is no that-trace configuration whatsoever. This is what Gallego does as well, relying on Pesetsky & Torrego’s (2001) treatment of the Nominative case feature as an uninterpretable Tense feature on the DP (uT[ ]), and allowing case to be valued in situ in NSLs.7 This also allows him to account for the obligatory presence of the complementizer in embedded clauses in NSLs:8 (17) a. John knows (that) Mary wrote this poem. b. Jan wie *(z·e) napisała Maria ten Jan knows that wrote Maria this ‘Jan knows that Maria wrote this poem.’

wiersz. poem

[Polish]

For Pesetsky & Torrego (as well as for Gallego), these two facts are related. C has an uninterpretable T feature, which can be ‘checked’ either by the complementizer that or by the movement of the subject, which also has the uT feature (namely, its Nominative case feature). With Phase Sliding, the subject (being inside vP) is spelled out as soon as C is merged, as shown in (18).9 The subject thus cannot move to [Spec,CP] to check the uT feature of C, which is what happens in English, according to Pesetsky & Torrego. What forces the presence 7

8 9

This suggests that T in NSLs either has no EPP feature, or that its EPP feature is checked by verb movement to T, as proposed, for example, by Alexiadou & Anagnostopoulou (2001). A postverbal position is not the neutral position for subjects in Polish. Gallego thus assumes the weak PIC (PIC2), in contrast to Den Dikken.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

168

[161–184] 23.11.2013 11:05AM

168

Variation in phasehood

of the complementizer is the need for the uT feature of C to be checked. In languages like English, the subject (which moves to [Spec,TP]) remains accessible and can go on and move to [Spec,CP] to check C’s uT feature. Note, however, that it would also remain accessible in [Spec,vP] with no Phase Sliding in place. CP

(18) CuT [

]

TP

TiT [val]

vP

SUBJECTUT[NoM]

v

v′ DP

This concludes our discussion of dynamic phases. While the proposals we looked at in this section are quite programmatic in nature, the very idea that phasehood is dynamic and might be affected by independent factors (such as the presence or absence of head movement) is quite appealing. It also opens up the possibility that phasehood may be subject to crosslinguistic variation, which we will explore in the next section.

6.2 Crosslinguistic variation Crosslinguistic variation with respect to phasehood has already been raised as a possibility in this book. In Section 5.2 above, for example, we saw proposals, most notably Abels (2003), that allow PPs to be phases in some languages (i.e. languages that disallow preposition stranding) but not in others (i.e. languages that allow preposition stranding). Chomsky’s own writings on phases do not seem to leave much room for crosslinguistic variation. Certainly, if the conceptual motivation for phases comes from the fact that they reduce computational complexity, we would expect phases to be universal. However, conceptual considerations lead us only to the expectation that the existence of phases is universal, not that their identity is universal. One question we need to ask before entertaining the possibility of crosslinguistic variation with respect to phasehood any further is whether the diagnostics we developed in Chapter 3 (both interface-based diagnostics and syntactic diagnostics) allow for crosslinguistic variation, and if so, how such crosslinguistic variation might come about. We saw that one of the core criteria for phasehood was independence at the interfaces. This, in turn, was diagnosed by things like propositional status or full argument structure (on the LF side), or the ability to determine Spell-Out (on the PF side). If propositional status or full argument structure is an LF diagnostic of phasehood, then there does not seem to be a lot of room for crosslinguistic variation. It is not clear what it would mean for a given category (i.e. CP or vP) to

{DESKTOP}NEW FOLDER//9781107040847C06.3D

169

[161–184] 23.11.2013 11:05AM

Crosslinguistic variation

169

be propositional in some languages but not in others, or for a transitive verb phrase to be complete in terms of its argument structure in some languages but not in others. LF diagnostics thus point towards a more static and crosslinguistically stable approach to phasehood. PF diagnostics, however, seem somewhat more flexible: if isolability or movability is a PF diagnostic, we might expect to find some variation with respect to PF phasehood. However, we saw in Chapter 3 that isolability is not a reliable phasehood diagnostic.10 Setting the diagnostics aside, there are (at least) three different ways, listed in (19a–c), in which crosslinguistic variation in phasehood can arise. I will consider them in turn in the rest of this section. (19) a. A language lacks a category which is a phase head in other languages. VARIATION AS LEXICAL GAP

b. Independent factors create an illusion of crosslinguistic variation with VARIATION AS EPIPHENOMENON respect to phasehood. c. A given category may be a phase in some languages but not in others. VARIATION AS PARAMETRIC SETTING

The first possibility is for a given language to simply lack a category that is considered to be a phase universally. The second one might be a misnomer, since, technically, there is no variation on this view. Independent factors can affect phases in ways that ultimately lead to different categories behaving like phases in different languages.11 The third option is to simply assume that variation in phasehood does occur, and posit a parameter to capture this variation accordingly. A straightforward illustration of the first option (and not unexpected, given the discussion in Section 5.3) comes from the domain of double object constructions. We saw that only certain types of applicative heads are phase heads (i.e. so-called high applicative heads), which can explain why passivization of either the higher (i.e. the indirect) object or the lower (i.e. the

10

11

We also saw in Chapter 3 that phase heads determine Spell-Out. Crosslinguistic variation with respect to what heads determine Spell-Out (and thus are phase heads) would thus have consequences for determining case valuation domains and ellipsis types. And both are places in the grammar where we do find some crosslinguistic variation. Logically speaking, however, crosslinguistic variation with respect to ellipsis could mean one of two things: either that the inventory of phase heads (which are the heads that license ellipsis of their complements) can vary from language to language, or that ellipsis is not a reliable phasehood diagnostic either. Processes like Phase Extension or Phase Sliding can endow new heads with phasehood characteristics, without changing the definitional characteristics of phases. As a result of Phase Extension, TP can be a phase in one language but not in another. For Den Dikken, similar considerations apply to CPs: a CP can only be a phase if a phase head moves to it. On this view, there is no (or very little) crosslinguistic variation with respect to what categories count as inherent phases. Gallego (2010), in fact, is quite explicit about not making the claim that there is crosslinguistic variation with respect to phasehood.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

170

[161–184] 23.11.2013 11:05AM

170

Variation in phasehood

direct) object is possible in high applicative constructions.12 We also saw that the existence of high applicatives in a language can be independently established by the behavior of its applicatives with respect to three diagnostics, due to Pylkkänen (2008): compatibility with intransitive verbs, compatibility with stative predicates, and the ability for the indirect object to be modified by a depictive phrase. What allowed passivization of the lower object was the EPP feature on the high applicative head, as proposed by McGinnis (2001), something that we took to be possible only if the high Applicative head is a phase head.13 A common place to look for variation with respect to phasehood is in the domain of noun phrases (or determiner phrases), given the long line of research on the crosslinguistic validity of the DP Hypothesis. The logic of the argument is very simple: if DPs are phases (as we saw in Section 5.2), and if some languages lack DPs (as is often argued for languages with no overt articles), 12

13

We also saw in Section 5.1 above that Slavic languages like Polish or Russian have two types of Predication Phrases, one of which is a phase head and one is not. This makes it quite plausible that some languages might lack the Pr head that is a phase head and only allow the one that is not. English might seem like a plausible candidate for such a language: predicate nominals do not receive a designated predicative case. However, unlike Slavic PrPs with Nominative predicates, they do allow extraction. In principle, there could be another way for a high Applicative head to acquire its EPP feature, without necessarily being a phase head. Given that high applicatives are dominated by a vP, the high Applicative head could also acquire its uninterpretable features via Feature Inheritance, by analogy to T inheriting its uninterpretable features from C. However, strictly speaking, this would not be a case of Feature Inheritance, but rather of feature spreading (since v has to retain its uninterpretable features), and it would also have to be possible with passive vPs. The two options are also hard to distinguish empirically, since in both cases we end up with uninterpretable features on both v and Appl heads: (i)

vP vuϕ [

], (EPP)

ApplP Appl′

IO Appluϕ [

], (EPP)

V (ii)

VP DO

V AND

APPL

AS PHASE HEADS

vP vuϕ [

], (EPP)

ApplP

IO Appluϕ [

Appl′ ], (EPP)

V

VP DO

V AS PHASE HEAD SHARING UΦ-FEATURES WITH

APPL

HEAD

{DESKTOP}NEW FOLDER//9781107040847C06.3D

171

[161–184] 23.11.2013 11:05AM

Crosslinguistic variation

171

such languages are predicted to lack one phase when compared to languages with overt articles. This is the case that I will focus on in what follows. Bošković (2005, 2009, to appear) revives the idea that languages can vary with respect to how much nominal structure they project: languages with overt articles project their nominals – Traditional Noun Phrases (TNPs), in Bošković’s terms – up to DP, whereas languages with no articles projects their TNPs only up to NP. However, the claim that the lack of overt articles implies the lack of a DP projection is far from uncontroversial; see Citko (2010), Pereltsvaig (2007), Progovac (1998), Rappaport (2001), Rutkowski (2002) and Rutkowski & Progovac (2005), among many others, for a defense of the universality of the DP Hypothesis.14 14

A complete discussion of the arguments in favor of (and against) the Universal DP Hypothesis would take us too far astray here and is somewhat tangential to the issue of crosslinguistic variation in phasehood. Let me nevertheless mention a couple of potential issues for the Parametrized DP Hypothesis. While languages at the core of the debate (Slavic languages like Polish, SerboCroatian or Russian) lack overt articles, they do have other elements that are typically analyzed as D heads: demonstratives, possessive pronouns and quantifiers, as shown in (ia–c). (i) a. z with

ta ˛

papuga˛

this.3SG.FEM.INSTR

parrot.3SG.FEM.INSTR

[Polish]

‘with this parrot’ b. z

moja˛

papuga˛

with

my.3SG.FEM.INSTR

parrot.3SG.FEM.INSTR

‘with my parrot’ c. z kaz·da ˛ papuga˛ with

every

parrot.3SG.FEM.INSTR

‘with every parrot’

Bošković analyzes such elements essentially as adjectives, based on the fact that they show the same agreement with the nouns they modify as adjectives do. He also points out that they can cooccur with each other and that some of them can appear in predicative positions: (ii) ta this (iii) Ta this

moja

papuga

my

parrot

[Polish]

papuga

jest

moja.

parrot

is

my

This, however, is not a convincing argument unless we have independent evidence that DP-internal concord is limited only to adjectives. A different argument against treating TNPs as Noun Phrases (as opposed to Determiner Phrases) comes from contrasts of the following kind, which can be attributed to the fact that pronouns occupy D positions (see Progovac 1998, Rutkowski 2002, for example). The contrast between (iii) and (iv) shows that pronouns have to follow nominal modifiers like sam ‘alone’. This follows nicely from an analysis in which pronouns either occupy a D position to begin with, or raise to it. (iii) Ona she

sama

przyszła

na

zebraine.

alone

came

to

meeting

‘She alone came to the meeting.’

[Polish]

{DESKTOP}NEW FOLDER//9781107040847C06.3D

172

[161–184] 23.11.2013 11:05AM

172

Variation in phasehood

Following Uriagereka (1988) and Corver (1990, 1992), Bošković links the grammaticality of Left Branch Extraction, well documented in Slavic literature and illustrated in (20) with data from Polish, to the lack of a DP projection.15 (20) a. Ilei papugi jedza˛ ti orzechów? how.many parrots eat nuts.GEN ‘How many nuts do parrots eat?’ b. Jakiei papugi najbardziej lubia˛ ti what.kind.of parrots best like ‘What kind of nuts do parrots like best?’ c. S´wiez·ei papugi lubia˛ ti orzechy. fresh parrots like nuts ‘Parrots like fresh nuts.’

orzechy? nuts

It might be tempting to attribute the grammaticality of LBE to DP not being a phase. However, it is not generally the case that movement out of phases is impossible. What is generally the case (and what we have seen throughout this book) is that movement out of a phase has to proceed through the phase edge, whereas movement out of a non-phase is not subject to such a restriction. It would also have the unwanted consequence of imposing similar restrictions on movement of complements from DPs. This is not to say that parameterizing the phasehood of DP is not a logical possibility. The crucial point here is that it is not a possibility that is going to yield the right empirical results with respect to LBE, a point made very explicitly by Bošković.16 (iv) *?Sama

ona

przyszła

na

zebranie.

alone

she

came

to

meeting

‘She alone came to the meeting.’ 15

16

The fact that there are languages that allow violations of the Left Branch Condition was noted by Ross (1967). Bošković also points out that such an explanation will not explain why LBE violations sometimes emerge, for example, when extraction takes place from the left branch of the complement of the noun. (i) *Cˇijei whose

je

on

vidio [NP [Nʹ

prijatelja [NP ti

majke]]]?

is

he

seen

friend

mother

[Serbo-Croatian]

‘Whose mother did he see a friend of?’ (Boškovic´ 2005: 8)

While the equivalent of (i) in Polish is also degraded, it improves markedly if the Genitive precedes the head noun: (ii) ?*Czyjeji

Jan

zobaczył

przyjaciela ti

matki?

whose

Jan

saw

friend

mother

‘Whose mother did Jan see a friend of?’ (iii) Czyjeji

Jan

zobaczył ti

matki?

przyjaciela

whose

Jan

saw

mother

friend

‘Whose mother did Jan see a friend of?’

[Polish]

{DESKTOP}/9781107040847C06.3D

173 [161–184] 23.11.2013 12:21PM

Crosslinguistic variation

173

Bošković considers a number of other phase-based explanations of the contrast between English-type and Serbo-Croatian-type languages briefly, but ends up rejecting them all. If DP is a phase, one way to exclude Left Branch Extraction of adjectival modifiers (which is the type of LBE that Bošković focuses on) is to assume that D cannot have an ‘extra’ escape hatch. He rejects this assumption on the grounds that it would rule out any movement from DP, not just movement from the left branch. He further modifies it by allowing D to have an extra specifier position, but deems movement from an adjoined NP position (the position for adjectives) to be too short (via a form of antilocality of the kind proposed by Abels 2003). This would rule out movement of adjoined APs, but it is not clear how it would rule out movement of other left branch elements. Demonstratives and determiners are a case in point; they are already at the edge of DP (so antilocality would not be able to explain why they cannot move out of DP). Since our focus here is not on the specifics of LBE, or even on the issue of whether Slavic noun phrases are noun phrases or determiner phrases, I will not dwell on the specifics of Bošković’s account. In short, he settles on the parameter concerning the position of adjectives to distinguish languages that allow LBE from the ones that do not. In English-type languages (languages with no LBE), adjectives are heads taking NP as complements (see (21a)), whereas in Serbo-Croatian-type languages (languages that allow LBE), they are adjuncts (see (21b)). Given this structural distinction, in languages like English, LBE would have to involve non-constituent movement.17 (21) a.

DP AdjP

D Adj

NP

LANGUAGES WITH NO LBE

b. NP AdjP

NP

LANGUAGES WITH LBE

To try and settle the question of whether DPs are phases in languages with no articles, let us apply to them the same diagnostics we applied to English DPs in Section 4.3. As usual, I will use Polish to illustrate. The examples below show that with respect to these diagnostics, Polish Traditional Noun Phrases (be they DPs or NPs) behave like phases. First, they contain heads that are capable of valuing so-called adnominal Genitive case: (22) a. kawa Jana coffee.NOM Jan.GEN ‘Jan’s coffee’ 17

[Polish]

LBE, being a fairly straightforward case of A-bar movement, could not be easily reanalyzed as head movement, which could in principle be possible out of the structure in (21a).

{DESKTOP}/9781107040847C06.3D

174 [161–184] 23.11.2013 12:21PM

174

Variation in phasehood

b. kawa dobrego coffee.NOM good.GEN ‘good quality coffee’

gatunku quality.GEN

Second, Polish TNPs are domains for the valuation of uninterpretable features. DP-internal concord provides evidence: (23) a. Jan zobaczył pie¸kna˛

zielona˛ papuge¸. [Polish] Jan saw beautiful.3SG.FEM.ACC green.3SG.FEM.ACC parrot.3SG.FEM.ACC ‘Jan saw a beautiful green parrot.’

b. Pie¸kna zielona papuga green.3SG.FEM.NOM parrot.3SG.FEM.NOM beautiful.3SG.FEM.NOM przeleciała Janowi nad over flew Jan.DAT ‘A beautiful green parrot flew over Jan’s head.’

głowa˛. head

Third, demonstratives and numerals determine Spell-Out, as evidenced by the following nominal ellipsis types: (24) a. Ja wezme¸ tego Δ. I take this.3SG.MASC.ACC ‘I will take this one.’ b. Ja wezme¸ tych I take these.3PL.MASC.ACC ‘I will take these two.’

[Polish]

dwóch Δ two.3PL.MASC.ACC

And fourth, DP-internal inverse scope is also possible, which shows that the edge of DP can be a target for QR. The inverse scope reading is the most natural reading for the following example. (25) Dwóch senatorów z kaz·dego miasta be¸dzie na tym zebraniu. two senators from every city will.be at this meeting ‘Two senators from every city will be at the meeting.’

Logically speaking, there are two conclusions we can draw from the behavior of Polish TNPs. One is that Polish noun phrases are DPs and behave accordingly (thus supporting the Universal DP Hypothesis), and the other one is that in some languages NPs can be phases as well (thus supporting the Parameterized DP Hypothesis). Bošković (to appear) argues for the latter, relativized approach to phasehood, in which the highest projection is a phase (see also Bobaljik & Wurmbrand’s 2005 work on relativized domains). If in a given language TNPs are NPs, then NP is going to be a phase; if TNPs are DPs, then DP is going to be a phase; and if TNPs are KPs, then KP is going to count as a phase, as shown schematically in (26a–c). What is impossible is the situation illustrated in (25d), in which TNPs are KPs but a DP counts as a phase.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

175

[161–184] 23.11.2013 11:05AM

Crosslinguistic variation

(26) a.

VP V

b. NP

VP V

c. DP

D

175

VP

d. * VP

V NP

KP K

V DP

D

KP K

NP

DP D

NP

phases

Bošković’s proposal is interesting in that it only indirectly attributes crosslinguistic variation in phasehood to lexical variation (e.g. the presence or absence of a D projection in a language). It does, however, show that a different inventory of categories might yield differences in phasehood. This is what distinguishes his proposal from others considered in this section: the number of phases remains the same, only their identity changes. This brings us to the third option for variation in phasehood, which simply attributes it to variation in parameter setting. This is what Abels (2003) proposes in order to explain crosslinguistic variation with respect to preposition stranding. For Abels, such variation is simply due to the different settings of the following parameter:18 (27) Parameter 1: [+/−] P° is a phase head. (Abels 2003: 233)

A more general question that a proposal like Abels’ raises is whether such a parameter is a plausible parameter on theoretical grounds. If the phasehood of P is subject to crosslinguistic variation, we would expect the phasehood of other phase heads to be subject to similar parametric variation. More generally then, the question is whether the parameter in (28), which I will call the Phasehood Parameter, is a plausible parameter. X in (28) ranges over the categories that we have seen to be phase heads. (28) [+/−] X° is a phase head.

(where X° = C, v, D, P, Appl, Pr)

Of course, the answer to this question hinges on many other, more general questions concerning parameters, such as what exactly parameters are, what is the locus of parametric variation in the grammar, and how to best state parameters. These are not questions to which I can do justice here; for very 18

Capitalizing on the same intuition, Bošković offers a slightly different take on the difference between P-stranding and non P-stranding languages. In particular, he attributes this difference to the possibility that PPs in languages that allow P-stranding project more structure than PPs in languages that disallow P-stranding.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

176

176

[161–184] 23.11.2013 11:05AM

Variation in phasehood

relevant and illuminating discussion, I refer the reader to Biberauer et al. (2010), Biberauer (2008) and the references therein. The common minimalist assumption concerning parameters is that they ‘reside’ in the lexicon as features of individual functional heads. This assumption is sometimes referred to as the Borer–Chomsky conjecture, which attributes crosslinguistic variation to the differences in featural make-up of functional heads.19 Being a phase head, if it is a feature at all, is not a feature of the same kind as other syntactic features we know: wh-features, case features, EPP features or φfeatures. So it seems that a parameter like (27) is not a parameter that can be stated in a way that is compatible with this view of parametric variation. However, there are views of phase heads that are more amenable to phasehood being subject to parametric variation. If, however, all there is to being a phase head is the ability to bear uninterpretable features (as proposed explicitly by Gallego 2010, for example), the parameter in (29a) could easily be restated as (29b).20, 21 (29) a. [+/−] X° is a phase head. (where X° = C, v, D, P, Appl) b. [+/−] X° is the locus of uninterpretable features. (where X° = C, v, D, P, Appl)

However, allowing such a parameter into the grammar raises a number of questions. For example, are all phases subject to similar crosslinguistic variation? More specifically, are there any languages in which C is not a phase head or v is not a phase head? This is ultimately an empirical question, but I take it to be a telling fact that, to the best of my knowledge, we have not stumbled upon such languages. Given the properties associated with the phase head C, it is hard to imagine a language in which C is not a phase head. This would have to be a language in which CPs are not propositional, a language that lacks sluicing, a language in which the C-T complex does not value Nominative Case, or a language in which successive cyclic movement does not proceed through [Spec,CP]. We have only scratched the surface of the issues concerning crosslinguistic variation in phasehood. There are many others to consider, such as whether variation in phasehood is something that might change over time. And if so, what would it mean for a given head to lose or acquire phasal status?

19

20

21

The name ‘Borer–Chomsky conjecture’ refers to Borer’s seminal work on parametric variation (Borer 1984) and Chomsky’s endorsement of it, and the term itself is attributed to Baker (2008a). To be fair, Baker himself has a different take on the nature of parameters, and argues explicitly in favor of so-called macro-parameters, which cannot easily (if at all) be stated as features of individual functional heads. As we saw in the earlier chapters, for Gallego uninterpretable formal features are ‘the key formal criterion’ of phasehood. Feature Inheritance thus muddies the picture somewhat, as (29b) can only refer to heads that can inherently count as bearers of uninterpretable features (as opposed to heads that get to inherit them).

{DESKTOP}NEW FOLDER//9781107040847C06.3D

177

[161–184] 23.11.2013 11:05AM

Non-simultaneous phases

177

6.3 Non-simultaneous phases In this section, we turn to the possibility that there might be a mismatch between what counts as a phase with respect to phonological considerations and what counts as a phase with respect to semantic considerations. This would mean that a given constituent could be a PF-only phase, an LF-only phase, both a PF and LF phase or neither a PF or LF phase. I alluded to these possibilities in Chapter 2, where we looked at Multiple Spell-Out, as well as in Chapter 3, where we looked at phasehood diagnostics, distinguishing between PF and LF diagnostics. We also saw that the same constituent can pass as a phase with respect to one set of diagnostics but not with respect to others. This was, for example, Matushansky’s (2005) conclusion about the phase-theoretical status of DPs: she argued that they are PF phases but not LF phases. All these possibilities are consistent with the general architecture of the grammar in Phase Theory. In fact, the concept of Multiple SpellOut, which is at the core of Phase Theory, only raises the question of whether Transfer to each interface could happen at different points in the derivation. And, as we will see in this section, there are reasons to believe that it can. And, if fact, quite a few researchers have either argued for such Non-Simultaneous Spell-Out explicitly (see Cecchetto 2004, Felser 2004, Marušič 2005, 2009, for example) or assumed it implicitly. In what follows, I will use the term Transfer rather than Spell-Out, and to avoid using the term Spell-Out to describe transfer to the semantic interface. The standard scenario we have been assuming throughout is for Transfer to PF and LF to happen simultaneously, as shown in (30) for a simple clause with two phase heads (C and v).22 (30) Simultaneous Transfer CP C

TP DP

PF2, LF2

T′ T

vP v′

ti

VP

v V

22

PF1 LF1 DP

For the sake of clarity, I only mark the first two cases of Transfer in the diagrams that follow. Technically speaking, the edge of the CP phase gets transferred separately. PF1 stands for the first Transfer to PF, and LF1 for the first Transfer to LF.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

178

[161–184] 23.11.2013 11:05AM

178

Variation in phasehood

There are different ways to think of what a Non-Simultaneous Multiple Transfer might be. One is to allow Multiple Transfer to apply only to one of the two interfaces (with Single Transfer to the other interface), as illustrated in (31a) and (31b), respectively. The cases we will look at in the rest of this section will involve these two options. (31) a. Multiple Transfer to PF, Single Transfer to LF

CP C

TP DP

T′ T

PF2

LF1

VP

PF1

vP v′

ti

v V

DP

b. Multiple Transfer to LF, Single Transfer to PF

CP C

TP DP

LF2

T′

PF1

vP

T

v′

ti

VP

v V

LF1 DP

One could also imagine scenarios in which there are Multiple Transfers to both PF and LF, but one is delayed with respect to the other. This might be the case if the CP in (31a) or (31b) were embedded and Transfer to both interfaces proceeded simultaneously once the embedded CP is complete.23

23

There is also another possibility, which I will not consider here: for Transfer to be totally ‘asynchronic’.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

179

[161–184] 23.11.2013 11:05AM

Non-simultaneous phases

179

Now that we have seen that Non-Simultaneous Multiple Transfer is conceptually plausible, we can turn to the question of how a grammar with it compares to a grammar without it. More specifically, we turn to the empirical motivation for Non-Simultaneous Multiple Transfer. What Non-Simultaneous Transfer is particularly well designed to handle is a situation in which a given element is phonologically visible, so to speak, in one position but semantically visible in another one. Marušič argues that Non-Simultaneous Transfer is what happens in cases of reconstruction (in which a moved element is interpreted lower than it is pronounced) and covert movement (in which a seemingly unmoved element is interpreted higher than it is pronounced), illustrated in (32a) and (32b), respectively. In (32a), we are interested in the non-specific reading of the indefinite a Kenyan, which can be paraphrased as ‘It is likely that some Kenyan (or other) will win the marathon.’ And in (32b), we are interested in the inverse scope reading, in which the object every problem has wide scope over the subject some student and we are not necessarily talking about one smart student. Both are ambiguous and also allow interpretations in which the scope of the relevant elements parallels their relative order.24 (32) a. A Kenyan is likely to win the marathon. b. Some student solved every problem.

LIKELY EVERY PROBLEM

>

>

A KENYAN

SOME STUDENT

Marušič is careful about distinguishing cases like (32a), in which the entire constituent reconstructs (total reconstruction cases), from cases in which only a part of it does (partial reconstruction cases), and focusing on the former. The latter is what happens with wh-movement, in which the operator part is interpreted in its scopal position and the remainder reconstructs. The early Transfer to LF (illustrated in (31b) above) gives rise to a situation in which a moving element is interpreted ‘earlier’ than it is pronounced. This could be a general way to think about all cases of semantically vacuous overt movement, of which (32a) above was one example. Another plausible example comes from cases of scrambling which are (arguably) semantically vacuous (see Bošković & Takahashi 1998, Saito 2003, among many others).25 Consider the following contrast. (33a) is the unscrambled variant, (33b) shows that the wh-pronoun can be scrambled out of the embedded CP; the scrambled variant is nevertheless interpreted as an indirect question. Saito takes it to mean that it has to undergo total reconstruction; if there were a trace (or copy)

24

25

Marušič puts both of these cases in the more general context of attested mismatches between form and meaning. He points towards idioms as another common example of a mismatch between phonological and semantic units: they function as one unit semantically but multiple units phonologically. See, however, Bailyn (2001), Miyagawa (1997), for example, for a different take on scrambling.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

180

[161–184] 23.11.2013 11:05AM

180

Variation in phasehood

left in the matrix clause position, the result would be a violation of the Proper Binding Condition, the condition that requires traces to be bound.26 (33) a. John-ga

[CP Mary-ga dono hono-o yonda ka] siritagatteiru. [Japanese] which book-ACC read Q want.to.know John-NOM Mary-NOM ‘John wants to know which book Mary read.’

b. ?Dono hon-oi John-ga [CP Mary-ga ti yonda ka] siritagatteiru. what-ACC John-NOM Mary-NOM read Q want.to.know ‘John wants to know which book Mary read.’ (cf. Saito 2003: 484)

In a grammar with Multiple Non-Simultaneous Transfer, such cases of scrambling can be analyzed as involving Transfer to LF before Transfer to PF. In (33b) the embedded vP phase containing the wh-pronoun nani-o ‘what’ is transferred to LF at a lower phase level.27 However, its phonological features will still be accessible, since Transfer to PF has not occurred yet. This means that the pronoun is going to be interpreted low no matter where it ends up surfacing. Reconstruction under A-movement of the kind illustrated in (32a) above is different though. Sauerland & Elbourne (2002) analyze such movement as PF movement, which, strictly speaking, would not involve reconstruction of any type. We might be tempted to analyze such cases in a way parallel to vacuous scrambling as also involving Transfer to LF before Transfer to PF. This is the derivation Marušič argues explicitly for, a derivation in which the indefinite subject (a Kenyan in (32a)) is transferred to LF before it is transferred to PF. However, given that we are dealing with a raising construction, the only two (non-DP) phases are the vP associated with the embedded verb phrase win the marathon and the CP associated with the matrix clause, both circled in (34) below. In (34), all three copies of the indefinite subject a Kenyan are in the same domain (the complement of the matrix C) and get transferred at the same stage. This does not change even if verb phrases headed by verbs like seem or be likely (or passive and unaccusative verb phrases in general) are phases; they are thought to be weak phases; weak in that they do not trigger Transfer to the interfaces.28

26

27

28

This is not to say that all cases of scrambling are semantically vacuous in this way. In the literature on Japanese scrambling, it is standard to distinguish long-distance and short-distance scrambling, and assume radical reconstruction (or some process akin to it) only for the long-distance cases. The question is whether this necessarily means that the embedded phase has to be an LF phase but not a PF phase, and whether there is any independent evidence for treating it as such. This is not a problem for Marušič, who treats non-finite TPs as LF phases but not PF phases. I do not think this conclusion is generally warranted for raising TPs, and is incompatible with the assumptions about phases defended here. Raising TPs are not complete in the requisite sense; they are not capable of valuing Nominative case, for example.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

181

[161–184] 23.11.2013 11:05AM

Non-simultaneous phases

(34)

181

CP C

TP T′

DPuC[Nom]

a Kenyaniϕ [3sg] Tuϕ [3sg],EPP vP AP

v likely

TP

DP

T′

a Kenyaniϕ [3sg] Tuϕ [3sg],EPP vP

v′

DP a Kenyaniϕ [3sg]

v

VP win the marathon

The intuition is clear; either the copy below or above the predicate be likely gets interpreted. Depending on which one it is, we either get the reconstructed or the non-reconstructed interpretation. However, this seems to be more a case of Transfer targeting different features differently (PF features versus LF features) than Transfer happening at different points in the derivation. Thus, the interpretation of the lower copy does not seem to have much to do with Transfer to the two interfaces happening non-simultaneously, given the assumption that phase heads determine the points of Transfer and the complement of a lower phase head is transferred to the interfaces when a higher phase head is merged.29 Another illustration of Non-Simultaneous Transfer, which we alluded to above, comes from the domain of Quantifier Raising. The idea here is that QR (and perhaps covert movement more generally) involves Transfer to PF before Transfer to LF. In (32b), repeated below as (35a), there are two Transfer domains, with copies of the quantified object every problem present in both. Given that every problem is pronounced low but interpreted high, it is reasonable to assume

29

The issue of reconstruction under A-movement (or the lack thereof) is quite a thorny issue. It is only allowed with certain elements (indefinites contrast in this respect with universal quantifiers, for example). That is why the mechanism that is used to capture it is sometimes different from the mechanism used to capture A-bar reconstructions. It has been given many different accounts: quantifier lowering by May (1977, 1985); covert insertion of an expletive there by Boeckx (2001), PF movement by Sauerland & Elbourne (2002) (see Iatridou & Sichel 2011 for a recent proposal teasing the different complicating factors apart).

{DESKTOP}NEW FOLDER//9781107040847C06.3D

182

[161–184] 23.11.2013 11:05AM

182

Variation in phasehood

that it the lowest copy is transferred to PF during the first PF cycle. In the diagram that follows I indicate quantification type with subscripts on the relevant DPs.30 (35) a. Some student solved every problem.

b.

CP C

TP DP∀

TP

DP∃

PF2

T′ T

vP LF1 DP∀

vP

DP∃

v′ PF1

VP

v V

DP∀

The concept of Non-Simultaneous Transfer we have focused on so far in this section is conceptually related to the concept of Delayed Spell-Out, due to Bachrach & Katzir (2009). For consistency’s sake, I will refer to it here as Delayed Transfer. Delayed Transfer, as the name implies, allows Transfer to skip a phase. In cases of Non-Simultaneous Transfer we have seen above, Transfer to one interface is delayed, but, in principle, delay could affect both interfaces. This is what Bachrach & Katzir (2007, 2009) argue happens in a construction known as Right Node Raising (RNR), an example of which is given in (36). (36) Leslie recommended and Terry watched a new documentary on Australian parrots.

Bachrach & Katzir assume a multidominant structure for RNR, in which the RNR element is literally shared between the two conjuncts, as shown in (37a).31 Furthermore, they argue that the italicized string a new documentary on Australian parrots is not spelled out when (and where) we would expect it to be spelled out, namely inside the embedded vP at the time when that embedded vP is spelled out. Instead, it is spelled out during the next Spell-Out cycle, above the coordination level.

30

31

Marušič treats these cases slightly differently, deriving the effects of QR from DPs being PF phases but not LF phases. See Citko (2011a,b and the references therein) for a defense of such structure for RNR, and for a more general discussion of multidominance in the grammar.

{DESKTOP}NEW FOLDER//9781107040847C06.3D

183

[161–184] 23.11.2013 11:05AM

Non-simultaneous phases

(37)

a.

183

CP C

&P &′ and

TP Terryi

TP Lesliei

T′

vP

T

T′ T

vP

v′

ti

ti

v′

VP

v

VP

v

Transfer 1 recommended

watched DP a new documentary on Australian parrots

b.

CP &P

C

&′ and

TP T′

Terryi T

Transfer 2 TP T′

Lesliei T

vP ti

v′ v

vP ti

VP

recommended

v′ v

VP watched

DP

a new documentary on Australian parrots

For Bachrach & Katzir, Transfer can only affect completely dominated constituents, which is what allows the shared element to escape early (or standard, perhaps more accurately) Transfer; in (37a–b) it is not completely dominated by either VP.32 Such a multidominant structure for RNR (coupled with Delayed Transfer) can explain why the shared element (often referred to as the pivot in the literature on RNR) has to be right-peripheral; otherwise the result would be a linearization contradiction, with the pivot both preceding the verb watched and following it, in violation of the antisymmetry requirement on linear order.33 It also captures (and reconciles) two seemingly irreconcilable properties

32 33

The first node that completely dominates the shared DP is the &P node. Since neither linearization of multiply dominated structures or constraints on RNR are our primary focus here, I will refrain from digressing into the details and refer the reader to the literature on the

{DESKTOP}NEW FOLDER//9781107040847C06.3D

184

184

[161–184] 23.11.2013 11:05AM

Variation in phasehood

of RNR: the fact that it is impervious to islands (as noted by Wexler & Culicover 1980 and discussed in much later literature on RNR) and the fact that the pivot can scope over material inside both conjuncts.34 These two properties come together, somewhat paradoxically, in (38). The lack of an island violation implies a non-movement analysis but the wide scope interpretation for the pivot implies a movement analysis.35 (38) John knows [someone who speaks __ ], and Bill knows [someone who wants to learn __], every Germanic language. (∃ > ∀ , ∀ > ∃ ) (Sabbagh 2007: 367)

On Bachrach & Katzir’s proposal, the pivot is transferred only when it becomes fully integrated into the structure (where full integration amounts to complete dominance), which is when the two conjuncts are combined to form a coordinate phrase. Only then will the pivot be fully dominated (by the &P node). It means that it will only incur island violations from this point on (hence the grammaticality of (37)) and that it will be interpreted in this high position (hence wide scope for the pivot). Even though the concept of Delayed Transfer illustrated here is different from the concept of Non-Simultaneous Transfer, the two are alike in that they constitute departures from the standard Multiple Transfer architecture. I take this flexibility to be a welcome consequence of Phase Theory, rather than a potential issue for it, given the facts it can capture. In the next chapter, we turn to the relationship between syntax and phonology on the one hand, and syntax and semantics on the other hand, and we will look at this relationship from the perspective of Phase Theory. We will examine the evidence in favor of phases being (or determining) both phonological and semantic domains, focusing on the following questions: (i) how do phases affect linearization, (ii) how do phases affect prosodic structure?, and (iii) how do phases affect interpretation?

34

subject (see, for example, Citko (to appear) for a handbook-style overview of these (and related) issues surrounding RNR and the references therein for the details of the specific accounts of RNR). The contrast below between the a examples, involving run-of-the-mill wh-movement, and the b examples, involving RNR, provides an illustration: (i) a. *Whati did Leslie wonder [WH-ISLAND when Terry saw ti]? b. Leslie asked [WH-ISLAND when John directed ti] and Terry asked [WH-ISLAND when Bill saw ti ] a new documentary on Australian parrotsi. (ii) a. *What does Leslie know [COMPLEX DH-ISLAND someone who directed ti]? b. Leslie knows [COMPLEX DP-ISLAND someone who directed ti] and Terry knows [COMPLEX DP-ISLAND someone who saw ti] a new documentary on Australian parrots.

35

Covert movement is not an option here since it would have to be covert Across-the-Board (ATB) movement of the pivot, and covert ATB movement is generally disallowed (see Bošković & Franks 2000 and Citko 2000, 2005, among others, for data and analysis).

{DESKTOP}NEW FOLDER//9781107040847C07.3D

185

[185–204] 23.11.2013 10:57AM

7 Phases and the interfaces

7.1 Phases and PF The focus of this section is on the relationship between phases (and/or Spell-Out domains) and PF. Given that phase heads determine Transfer (in that domains of phase heads get transferred to the two interfaces, sometimes non-simultaneously, as we saw in the previous section), it is not unreasonable to assume that phases will play a role in PF processes. In fact, many researchers point out that the theory of the syntax–phonology mapping in which phases or Spell-Out domains constitute phonological domains is the null hypothesis (see, among others, Adger 2007, Dobashi 2003, Ishihara 2003, 2007, Kahnemuyipour 2003, 2004, 2005, Kratzer & Selkirk 2007, Pak 2008, Samuels 2011, Scheer 2012). For the two ‘core’ phases CPs and vPs, the relevant phonological units on the null hypothesis would be TPs and VPs. The very idea that there is a principled mapping between syntactic domains and phonological domains is by no means novel (or original) to Phase Theory, and I cannot do justice here to the vast literature on the subject of the syntax– phonology interface; I refer the interested reader to Elordieta (2008) and the references therein for a thorough overview of this interface instead. Perhaps the most relevant PF fact about language is linear order. If linearization is not part of Narrow Syntax (as made explicit by Kayne’s 1994 Linear Correspondence Axiom), and if linear order is established at Spell-Out (as seems reasonable given what Spell-Out is), and if Spell-Out happens phase by phase, linearization will happen phase by phase, too. This is what the title of Fox & Pesetsky’s (2005) ‘Cyclic Linearization’ paper, the gist of which I give below, refers to. The claim that linear order is established at Spell-Out, in a cyclic fashion, as expected in a Multiple Spell-Out architecture, is only one part of their proposal. The other part is the proposal, which they refer to as Order Preservation, that the ordering established within one Spell-Out domain cannot be changed in subsequent Spell-Out domains. Future Spell-Out domains can add information but not delete or change previously established information (see Fox & Pesetsky 2005: 6). Fox & Pesetsky consider the following scenario to

185

{DESKTOP}NEW FOLDER//9781107040847C07.3D

186

[185–204] 23.11.2013 10:57AM

186

Phases and the interfaces

illustrate how Order Preservation works. In (1) below, there is a single Spell-Out domain (D1) consisting of three elements, X, Yand Z. The ordering statements for this domain D1 are given in (1b). (1) a. X Y Z

b. Ordering statements within D1: X >Y, Y > Z

If another element α is merged outside D1, it will be linearized in the next domain (D2). (2) Merge α D2

α D1 X Y Z

Given the Order Preservation Principle, a number of things could happen next; crucially, however, the relative order of elements established within D1 cannot change.1 This means that X, being at the edge of D1, can freely move out of D1. After movement, it will still precede all the elements within D1. (3) a. Move X

D2 Xi

α D1

ti Y

Z

b. Ordering within D1: X>Y, Y>Z c. Ordering within D2: X > α, α > D1

1

See, however, Richards (2007) for a proposal that phase-internal movements are order preserving but transphasal movements are not.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

187

[185–204] 23.11.2013 10:57AM

Phases and PF

187

The ordering created at the next Spell-Out domain (D2) does not change the ordering previously established, it merely adds new orderings to the ones established at D1. This is how Fox & Pesetsky derive the fact that movement from a given domain has to proceed through the edge of that domain. If, on the other hand, movement from D1 takes place from a non-left-edge position, as in (4a), the new linearization statements are going to contradict the previously established ones. Here α both precedes Y (by virtue of preceding D1, which includes Y) and follows Y (due to Y having moved out of D1). (4) a. Move X

D2 Y1

α D1

X ti

Z

b. Ordering within D1: X>Y, Y>Z c. Ordering within D2: Y > α, α > D1 → α >Y

However, if it is essentially PF that rules it out, as Fox & Pesetsky point out, we expect to find ways to fix it. And this is indeed what happens. If both X and Y move but their relative order is preserved, the new linearization statements are not going to contradict the previously established ones.2 (5) a. Move X and Y

D2 Xj Yi

α D1

tj ti

Z

b. Ordering within D1: X>Y, Y>Z c. Ordering within D2: X > Y, Y > α, α > D1

This is how Fox & Pesetsky derive Holmberg’s Generalization, illustrated by the contrast in (6). This contrast shows that the object can undergo object shift only if the verb has shifted even further and the order of the two is preserved:

2

Ellipsis is another way to rescue a structure in which a non-final element moves out of a domain D.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

188

[185–204] 23.11.2013 12:19PM

188

Phases and the interfaces

(6) a. [CP

Jag kysstej hennei inte [VP tj ti]. [Swedish] I kissed her not jag hennei inte [VP kysste ti]. b. *[CP. . . att ... that I her not Kissed (cf. Fox & Pesetsky 2005: 17)

In the grammatical (6a), this is what happens. The ordering established at the CP level does not contradict the ordering established at the VP level: (7) a. Ordering within VP: KISSED > HER b. Ordering within CP: I > KISSED, KISSED > HER, HER > NOT, N O T >

VP

However, in the ungrammatical (7b), the ordering established at the CP level does contradict the ordering established at the VP level: henne precedes kysste at the CP level but follows it at the VP level. (8) a. Ordering within VP: K I S S E D > H E R b. Ordering within CP: THAT > I, I > H E R ,

HER

> NOT,

NOT

> VP →H E R >

KISSED

3

Fox & Pesetsky’s Cyclic Linearization proposal establishes a direct link between Spell-Out domains and linearization domains. It is not clear, however, what in their system prevents the object from moving to the edge of VP first, in which case it could continue to move to the next Spell-Out domain with the verb remaining within VP without violating the Order Preservation Principle. The relationship between Fox & Pesetsky’s Spell-Out domains and phases (or complements of phase heads) is also not quite transparent. Given that vP is a phase, the Spell-Out domain should be a VP, which is consistent with their proposal. However, on standard assumptions about phases and successive cyclic movement, movement takes place through the edge of a phase (vP) not through the edge of the Spell-Out domain (VP). In the rest of this chapter, I will be concerned with the relationship between phases (or Spell-Out domains) and phonological domains, and the issue of whether phases correspond to (or help determine) phonological domains. If so, what are these phonologically relevant domains, and what phonological processes apply to them? The commonly assumed prosodic units are listed in (9) (see Nespor & Vogel 1995: 11). The question is which of them correspond to Spell-Out domains. Since phases correspond to larger chunks of structure, the most likely culprits are higher-level units such as intonational phrases. (9) a. phonological utterance b. intonational phrase c. clitic group d. phonological word d. foot e. syllable 3

This follows from transitivity. Given that henne precedes inte and inte precedes VP (and kysste is included within VP), henne will end up having to precede kysste.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

189

[185–204] 23.11.2013 12:19PM

Phases and PF

189

We will focus below on sentential stress (also referred to as nuclear stress), which is a good example of a phonological process that applies to larger units and which has been shown to be sensitive to syntactic domains. The more specific questions about nuclear stress we will ask are: (10) a. Do phases play a role in determining domains for nuclear stress assignment? b. If so, does movement across phasal domains affect nuclear stress? c. How can we account for the crosslinguistic variation (or the lack thereof) with respect to nuclear stress?

To ground the discussion empirically, let us start with some observations about the nature of nuclear stress and a brief summary of previous (not necessarily phase-based) accounts. Earlier accounts, such as Chomsky & Halle’s (1968), or Halle & Vergnaud’s (1987), often refer to directionality of stress assignement. The rule governing the distribution of sentential stress (the so-called Nuclear Stress Rule (NSR)) operates cyclically and takes the element bearing the rightmost word-level stress in a given domain to project its stress to the next level. Applied to a concrete example, the NSR yields the following: (11) ( (* ) (* ) Parrots

* ) ( * ) (* ) ( * ) like NUTS

Cinque (1993) points out both conceptual and empirical problems with Halle & Vergnaud’s formulation and suggests the revision given in (12). The crucial innovation in his proposal lies in the fact that it makes no reference whatsoever to the directionality of stress assignment (right versus left). (12) a. Interpret boundaries of syntactic constituents as metrical boundaries. b. Locate the heads of line N constituents on line N + 1. c. Each rule applies to a maximal string containing no internal boundaries. d. An asterisk on line N must correspond to an asterisk on line N – 1. (Cinque 1993: 244)

In most general terms, nuclear stress is sensitive to the level of embedding rather than directionality. This explains, for example, why there is not as much crosslinguistic variation with respect to sentential stress as there is with respect to word stress. It also explains why there is not as much crosslinguistic variation with respect to sentential stress as there is with respect to word order (as noted by Cinque 1993, Kahnemuyipour 2004 and the references therein). To illustrate, in simple transitive sentences, the direct object is the

{DESKTOP}NEW FOLDER//9781107040847C07.3D

190

[185–204] 23.11.2013 12:19PM

190

Phases and the interfaces

element that bears nuclear stress irrespective of the word order type involved (SVO, SOV or VSO):4 (13) a. John read a BOOK. b. Ali KETAAB ye Ali a book ‘Ali bought a book.’

SVO [Persian]

xarid. bought

c. Chuala Seonag CALUM. heard Seonag Calum ‘Seonag heard Calum’ (Kahnemuyipour 2004: 12)

SOV [Scottish Gaelic]

Other factors, however, do play a role. For example, as often noted, unaccusative and unergative verbs behave differently: with unergative verbs, stress can fall either on the verb or on the subject (see Kahnemuyipour 2004, Selkirk 1995, Zubizaretta & Vergnaud 2005 and the references therein). With unaccusative verbs, on the other hand, only the subject can bear nuclear stress: (14) a. The PARROT is screeching. b. The parrot is SCREECHING. (15) a. The PARROT disappeared. b. *The parrot DISAPPEARED.

Not surprisingly, passives pattern with unaccusatives in this respect. (16) My BIKE was stolen. (Legate 2003: 513)

NSR is a cyclic rule: stress is assigned cyclically (domain by domain) and the highest stress within a given domain projects to the next level. Furthermore, the contrast between the a and be examples in (17–19), due to Bresnan (1971), who attributes the first pair to Newman (1946), suggests that the stress assigned within one domain can sometimes survive as the highest stress if the stressed element moves out of that domain. All the a examples involve noun complement clauses (thus no movement from the embedded CP), whereas all b examples involve relative clauses (which on the Head Promotion Account means that the head moves from within the relative CP). Interestingly, the relative clause head keeps its stress.5

4

5

This is not to say that there is no crosslinguistic variation whatsoever. Zubizarreta (1998), for example, in order to account for the differences between Romance and Germanic languages, posits the existence of two ‘types’ of nuclear stress rules, one sensitive to selection (which she dubs selectional nuclear stress rule) and the other one sensitive to c-command. They might interact differently in different languages (or one of them might be absent altogether in a given language), which is what gives rise to crosslinguistic variation. Legate reinterprets Bresnan’s findings in phase-theoretical ways and takes this to mean that movement across phases does not affect nuclear stress. Kahnemuyipour takes issue with this assertation based on the fact that the moved wh-phrase does not keep its stress.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

191

[185–204] 23.11.2013 12:24PM

Phases and PF

191

(17) a. George has plans to LEAVE. b. George has PLANS to leave. (18) a. Helen left directions for George to FOLLOW. b. Helen left DIRECTIONS for George to follow. (19) a. Mary liked the proposal that George LEAVE. b. Mary liked the PROPOSAL that George left. (Bresnan 1971: 258–9)

With this as background, let us turn to the question of how Nuclear Stress Rule might be recast in phase-theoretical terms. A natural starting point is the hypothesis, stated in various forms in (20a–d), that the Spell-Out domain of the phase is the domain for prosodic rules like the Nuclear Stress Rule. (20) a. The spellout domain of a phase is the prosodic domain for phrase stress. (Kratzer & Selkirk 2007: 104) b. On the construction of each phase, the complement of the phasal head is Spelled Out, so it is this category that undergoes phonological rules. (Adger 2007: 247) c. Phonological rules apply directly to the material that is spelled out at eachphase. (Pak 2008: 10) d. Spell-Out domains are mapped onto prosody as MaPs. (Ishihara 2007: 144)

However, even with these proposals about the mapping between Spell-Out domains and prosodic domains, there is still the question of how exactly NSR applies within these domains. There are two logical possibilities. One is that the lowest (e.g. the most deeply embedded) element within the Spell-Out domain is prosodically the most prominent one, and the other one is that it is the highest element within a given domain that is prosodically the most prominent one. Interestingly, both of them have been defended in the literature, not for different languages. This is not necessarily a contradiction. However, depending on which route we take, we will require a different set of background assumptions. Adger (2007) provides a very straightforward implementation of the Cinquestyle Nuclear Stress Rule in a phase-based system.6 Let us look at concrete examples, parallel to Adger’s original (parrot-free) examples. (21) The parrot ate a nut.

6

Adger’s implementation of it, however, does avoid some of the problems Cinque faces. To illustrate, complex noun phrases of the following sort, as pointed out by Adger, are problematic for Cinque. (i) [the [man [from [Philadelphia]]]]’s hat. (Cinque 1993: 268, crediting Richard Kayne)

Since specifiers are at the edge, they escape Spell-Out. Consequently, the noun hat can be the most prominent element in spite of the fact that it is not the most deeply embedded constituent in this noun phrase.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

192

[185–204] 23.11.2013 12:24PM

192

Phases and the interfaces

The steps in the derivation that are relevant for determining nuclear stress (namely Spell-Out domains) are given below. (22a) shows that the complement of the lowest phase head (D) is spelled out when the next phase head (v in this case) is merged. (22)

vP VP

v V ate

DP D a

NP nut

At this stage, the lexical stress on the noun nut projects to the next level, as shown in (23). Similar considerations apply to the subject the parrot. If Spell-Out domain is the domain for stress assignment, and NSR amounts to projecting the highest stress within the Spell-Out domain. This is actually a departure from Adger’s proposal; for him NSR applies to entire phases, and phase edges are not included in the calculation of stress due to extrametricality. Here, NSR applies to Spell-Out domains, which eliminates the need for extrametricality. (23) Spell-Out NP, apply NSR to NP * * nut]

[NP

* * [NP parrot]

The next Spell-Out occurs when C (the next phase head up) is merged. At this point the complement of VP, being the complement of v, is spelled out. (24) a.

CP C

TP DPi

T′

the parrot T

vP v′

ti

VP

v V ate

DP D a

NP nut

{DESKTOP}NEW FOLDER//9781107040847C07.3D

193

[185–204] 23.11.2013 10:57AM

Phases and PF

193

b. Spell-Out VP, Project highest stress to the next level

[VP

* * * a nut]

* ate

In the last Spell Out-cycle, the TP spelled out, and the stress on the noun nut, still being the highest one, is projected to the next level. This correctly derives the placement of nuclear stress on the noun nut in The parrot ate the nut. (25)

a.

CP C

TP DPi

T′

the parrot T

vP ti

v′ VP

v V ate

DP D a

NP nut

b.

[TP

* * The parrot

* ate

a

* * * * nut]

This algorithm can also account for the contrast between unergatives and unaccusatives. We have seen above that unaccusatives differ from unergatives in that they only allow stress on the subject (recall the contrast between (15) and (16) above).7 On the assumption that unaccusative vPs are not strong phases in that they do not trigger Spell-Out, there are two Spell-Out domains, the complement of D and the complement of C.

7

Crucially, Adger assumes that the unaccusative v does not trigger Spell-Out. If it did, the verb would be the most stressed element in this domain.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

194

[185–204] 23.11.2013 10:57AM

194

Phases and the interfaces

(26)

CP C

TP DPi

T′

the parrot T

vP v′

VP

V disappeared

ti

In the first domain the stress on parrot, being the only stress (the determiner the, being functional, is not stressed), is the one that projects to the next level, as shown in (27a). In the second domain, the stress on parrot is still the one that projects to the next level, being the highest stress in this Spell-Out domain, as shown in (27b)).8 (27) a. Spell Out NP, Project stress on parrot * * [NP parrot]

b. Spell Out TP, Project stress on parrot * * * [TP the parrot

* dissappeared]

Examples involving unergatives are different, since we are dealing with an ‘extra’ Spell-Out domain, the complement of v. This means that stress on both the subject and the predicate can project up to the next level (since they are in distinct domains). As a result, they end up equally stressed at the next level. This does not mean that they end up equally stressed at the sentential level. Instead, the indeterminacy means that either can project to the sentential level, which is how Adger derives the fact that either the subject or the predicate can be stressed.

8

Legate’s view is different. She assumes there is a phase boundary here, and assumes that in those cases it preserves the stress it was assigned at an earlier cycle. The reason parrot in (i) is stressed is thus similar to the reason proposal is stressed in (ii). (i) The PARROT disappeared. (ii) I liked the PROPOSAL that George left.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

195

[185–204] 23.11.2013 10:57AM

Phases and PF

(28) a.

195

CP C

TP T′

DPi D the

NP T parrot

vP ti

v′ v

VP screeched

b.

* * * [TP the parrot

* * screeched]

* * [TP the parrot

* * * screeched]

c.

Adger’s phase-based theory of nuclear stress is quite different from Kahnemuyipour’s (2003, 2004, 2005) one, which also crucially relies on phases, but which takes the elements at the edge of the Spell-Out domain to receive highest stress. Kahnemuyipour’s Sentential Stress Rule is given in (29). (29) Sentential Stress Rule Sentential stress is assigned at the phase to the highest element (i.e. the phonological border) of the spelled out constituent or the SPELLEE. (Kahnemuyipour 2004: 88)

At first glance, this seems to make wrong predictions for a language like English: within the Spell-Out domain of a transitive v, a direct object is not the highest element. However, it becomes one if the verb moves out of VP (or the object moves to some higher position within this Spell-Out domain). In verb-final languages (of which Kahnemuyipour 2004 focuses on Persian and German), stress falls on the direct object in simple transitive clauses, just as in English. Kahnemuyipour accounts for it as follows; he assumes the existence of a low aspectual projection, sandwiched between vP and VP (see MacDonald 2006, Travis 2010 and the references therein for a justification of such a projection). He furthermore assumes that the direct object moves to the specifier of this projection, even in a language with no object shift like English. This means that the Spell-Out domain (SPELLEE in his terms) of the phase head v is this low aspectual projection. Thus, at the point in the derivation illustrated in (30)

{DESKTOP}NEW FOLDER//9781107040847C07.3D

196

[185–204] 23.11.2013 10:57AM

196

Phases and the interfaces

(namely when C is merged), the complement of the lower phase head, in this case an AspP is spelled out: (30)

CP TP

C

Subjectj

T′

T

vP tj

v′ AspP

v

OBJECTi

Asp′ Spell-Out Domain VP

Asp V

ti

Even though Kahnemuyipour (2004) and Adger (2007) agree that phases are relevant to the determination of nuclear stress, we saw that they differ in the details of implementation. For Adger, it is the element with the highest level of stress that gets its stress ‘augmented’ by one stress mark. For Kahnemuyipour, it is structurally the highest element that does. They also differ in how they handle the difference between sentences with passive/unaccusative verbs and unergative ones. We have seen above how this difference can be derived in Adger’s system. For Kahnemuyipour (2004), this difference is tied to the assumption that unaccusative vPs are not (strong) phases in that they do not trigger Spell-Out. The only phase is the CP phase, and, consequently, the subject being the highest element in the Spell-Out domain of C is the element that bears nuclear stress. Kahnemuyipour’s system gives rise to two straightforward predictions. First, if the highest element moves out of what would otherwise become its Spell-Out domain, it should lose nuclear stress. Specific objects in Persian are a case in point; the contrast between (31a) and (31b) shows that specific objects differ from their non-specific counterparts in that they do not receive nuclear stress. (31) a. Ali QAZAA Ali food ‘Ali ate food.’ b. Ali Ali

qazaa-ro food-ACC

xord. ate

[Persian]

XORD.

ate (cf. Kahnemuyipour 2004: 96)

Kahnemuyipour provides independent evidence that specific objects move to a higher position. In (32a), the object is the highest element in the Spell-Out domain of v, whereas in (32b) the verb is.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

197

[185–204] 23.11.2013 10:57AM

Phases and PF

197

(32) TP

a.

Alij

TP

b.

Alij

T′ T

T′ T

vP tj

qazaa-roi

v′ v

vP

QAZAA

v′

tj

AspP Asp′

Asp

VP xord

v′ v

AspP Asp′

tj ti

Asp

VP XORD

ti

Second, if some (extra) element is merged higher than the non-specific direct object (which would otherwise be the highest element inside the Spell-Out domain, that element should bear nuclear stress. For example, if an adverb is merged inside the Spell-Out domain, it will bear nuclear stress. So far, we have focused in this section mostly on the Nuclear Stress Rule, and mostly on how it works in English. However, phases and Spell-Out domains have been implicated in a number of other phonological phenomena in a number of other languages. Let me thus conclude this section with a couple of examples. Pak’s (2008) analysis of High Tone Anticipation (HTA) in Luganda makes crucial use of phases. HTA, illustrated in (33a–b) below, is a kind a spreading in which high tone spreads leftward across a given domain, which Pak argues is the Spell-Out domain of a phase head. The verb and the material following it are treated as a single domain, whereas preverbal elements belong to a separate domain. Pak argues that preverbal elements are in the specifier of C, thus by hypothesis outside the Spell-Out domain of C. (33) a. omulenzi a-gul-ir-a Mukasa kááwà. 1.boy SBJ1-buy-APPL-IND 1.Mukasa 1A.coffee ‘The boy is buying Mukasa some coffee.’ b. (òmùlènzì) (à-gúl-ír-á Mukásá kááwà. 1.boy SBJ1-buy-APPL-IND 1.Mukasa 1A.coffee ‘The boy is buying Mukasa some coffee.’ (Pak 2008: 15)

[Luganda]

McGinnis (2001) , building on Seidl (2000), shows that in Bantu languages, high and low applicatives have different prosodic structures. In high applicatives, the two objects belong to the same phonological domain, whereas in low applicatives, they belong to two different domains. The boundaries of phonological domains can be ‘diagnosed’ by processes like Penultimate Vowel Lengthening (PVL) or Vowel Length Shift (VLS); both apply at the right edge of a phonological domain. What is interesting is that what counts as the right edge is different in low and high applicatives. In (34a), which is a low applicative, the

{DESKTOP}NEW FOLDER//9781107040847C07.3D

198

[185–204] 23.11.2013 10:57AM

198

Phases and the interfaces

two objects are in the same Spell-Out domain, whereas in (34b), which is a high applicative, the direct object and the indirect object are in two different domains. This follows if the Spell-Out domain of the high applicative phase head is a VP containing the direct object but not the indirect object, which is part of the next domain. (34) a. [Tu-kά-βi-tum-ir-ά omúkali valinά:nde.] we-PST-T-send-APPL-FV woman Valinande ‘We have just sent Valinande to the woman.’ b. [Ni-mw-ɑndik-il-ile nu:ru:] [xɑti.] SP-OP-write-APPL-FV Nuru letter ‘I wrote Nuru a letter.’ (McGinnis 2001: 23)

[Kinande]

[Chi-Mwhi:ni:]

Ishihara (2003, 2007) shows that in Japanese, Spell-Out domains are mapped onto Major Phrases and that scope in Japanese cannot extend beyond Major Phrases. The empirical support comes from the contrast, due to Miyagawa (2003), illustrated in (35a–b), showing that a universal quantifier unambiguously scopes over negation in SOV word order, but can get either wide or narrow scope with respect to negation in an OVS word order type. (35) a. zen’in-ga sono tesuto-o all-NOM that test-ACC ‘All did not take that test.’

uke-nakat-ta. take-NEG-PST

(SOV) *NOT>>ALL,

ALL>>NOT

b. sono tesuto-oi zen’in-ga ti uke-nakat-ta. (OSV) that test-ACC all-NOM take-NEG-PST ‘That test, all didn’t take.’ NOT >> ALL, (ALL>>NOT) (Ishihara 2007: 139, citing Miyagawa 2003: 183–4)

In (37a), the subject and negation are not in the same Major Phrase, as schematized in (36): (36) [TP ALLi

[vP ti THAT TEST NOT TOOK] ]

MaP

(SOV)

MaP

The scrambled example (36b), on the other hand, is structurally ambiguous, and depending on the structure assigned to it, the subject and negation are in the same Major Phrase, as in (37a), which yields the reading in which negation has scope over the subject, or the subject and negation are in distinct domains, as shown in (37b), which yields a reading in which the subject has scope over negation. (37) a. [TP THAT TESTi MaP

[vP

NOT TOOK] ]

ALL ti

MaP

(OSV)

{DESKTOP}NEW FOLDER//9781107040847C07.3D

199

[185–204] 23.11.2013 10:57AM

Phases and LF

b. [ THAT TEST CP i MaP

[TP ALLj [vP tj ti NOT TOOK] ] ] MaP

199

(OSV)

MaP

This concludes our discussion of the role phases play in determining phonological domains. The question we focused on is whether phonological cycles correspond to syntactic cycles, and more importantly, given the focus of this book, whether phases or complements of phase heads (being the Spell-Out domains) either are the relevant phonological units or map onto the phonological relevant units.We saw that phases and Spell-Out domains (as determined by phase heads) help us understand PF phenomena such as linear order preservation effects or nuclear stress assignment. We also saw that phases play a role in determining phonological processes such as Vowel Lengthening or High Tone Anticipation in Bantu, and that Major Phrases as determined by phases play a role in constraining scope in Japanese. In the next (final) section of this book, we ask similar questions about the mapping between syntax and semantics, and the role phases might play in that mapping.

7.2 Phases and LF This section focuses on the question of whether phases (or complements of phase heads) correspond to semantically relevant units, and if so, what these semantically relevant units might be. We have seen throughout this book that semantic completeness is a commonly assumed LF phasehood diagnostic. Semantic completeness in this context refers to phases being propositional and/or complete in some other relevant sense, for example, complete with respect to their argument structure. However, we have also seen that there are issues with both of these as diagnostics; issues that have led researchers to propose alternative definitions of phases giving rise to possibly different categories counting as phases. For example, Den Dikken (2007) defines phases as subject–predicate structures, Bošković (to appear) takes each phase to be the top node in an extended projection of each lexical category, and Epstein & Seely (2002, 2006), Müller (2004, 2011) take all phrases to be phases. While these approaches avoid the problems associated with limiting phasehood to propositional elements, they do end up with a very different view of what phases are. For example, for Den Dikken CPs are phases only under certain circumstances, and for Bošković AdjPs can be phases. Given the ontology of semantic types where the types are t (truth value), e (entity) (and perhaps s (intension or world)) and various combinations thereof, the types for the categories that we have assumed to be phases are quite diverse. Furthermore, the lack of clear distinctions between the semantic types of the phrases listed in (38a), which are taken to be phases, and the ones in (48b), which are not taken to be phases, shows that semantic types are not a reliable diagnostic distinguishing phases from non-phases.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

200

[185–204] 23.11.2013 10:57AM

200

Phases and the interfaces

(38) a. Phases and semantic types Syntactic category

Semantic type

Determiner Phrase (DP)

e (proper names) (quantified noun phrases) t (questions) t (declaratives) (relative clauses) t

verb Phrase (vP) Complementizer Phrase (CP)

Predication Phrase (PrP) Prepositional Phrase (PP) b. Non-phases and semantic types Syntactic category

Semantic type

Noun Phrase (NP) Verb Phrase (VP) Tense Phrase (TP) Adjective Phrase (AdjP) Adverb Phrase (AdvP)

t

Thus, a more fruitful line of inquiry into the role of phases in the syntax– semantics interface might come from the question of what semantically relevant domains phases might correspond to. For reasons that are similar to the reasons we discussed in the last section (which dealt with phases and PF), given how Phase Theory and Multiple Spell-Out work, it may well be the case that it is not phases per se that are the relevant domains, but, rather, complements of phase heads. Interestingly, there is an important (and independently motivated) semantic distinction that has been argued to correlate with the boundary between vP and CP phases. It is the distinction between the nuclear scope (the domain of existential closure) and the restrictive clause in a tripartite quantificational structure, as proposed explicitly by Biskup (2009a) (see, however, Butler 2005, Carnie & Barss 2006 for different implementations of the same general idea).9 This is Diesing’s (1992) Mapping Hypothesis, stated in (39) in phase-theoretical terms, and illustrated in (40) in arboreal terms.

9

Carnie and Barss’s approach to phases is slightly different, though, from the approach taken here. They argue for a relativized approach to phases, in which each phase has a predicative element, a single argument and a temporal operator.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

201

[185–204] 23.11.2013 10:57AM

Phases and LF

201

(39) The Mapping Hypothesis10 a. Material from vP phase maps into nuclear scope. b. Material from CP phase maps into a restrictive clause. (40) CP C

Restrictive Clause TP

DPSUBJ T

T′ vP Nuclear Scope

DPSUBJ v′ VP

v V

DPOBJ

What is important is that the two subject positions (the thematic vP-internal position and the EPP-related [Spec,TP] position) end up in different domains. Diesing shows how it can explain ambiguity of bare plural subjects in sentences of the following sort: (41) a. Parrots are loud. b. Parrots are intelligent.

With a stage-level predicate (such as be loud, available, present), the sentence is ambiguous between a generic reading, on which being loud is a general property associated with parrothood, and an existential reading, on which parrots may be loud at a specific time (for example, before they roost for the night). However, with an individual-level predicate (such as be intelligent, green, have yellow beaks), the only available reading is the generic one. Diesing accounts for it by proposing that generic bare plural subjects are interpreted in [Spec,TP], whereas existential ones are interpreted in [Spec,vP], as shown in (42a–b).11

10

Diesing’s original formulation, given in (i), does not make reference to phases (but to VPs and IPs instead). This, of course, is not surprising, since her formulation predates Phase Theory by a decade. (i) Material VP is mapped into the nuclear scope. (ii) Material from IP is mapped into a restrictive clause. (Diesing 1992: 10)

11

To be consistent with the analysis of small clauses assumed in this book (see the discussion in Section 5.1), I use PrPs instead of vPs here. This does not impact upon the Mapping Hypothesis in any way: the higher (CP) phase is the restrictive clause and the lower (PrP) phase is the nuclear scope. In fact, this ambiguity could be seen as another argument in favor of treating PrP as phases.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

202

[185–204] 23.11.2013 10:57AM

202

Phases and the interfaces

(42) a. Existential reading CP Restrictive Clause C

TP

parrots

T′ T

PrP parrots

Nuclear Scope

Pr′ Pr

AdjP loud

b. Generic reading CP C

TP

Restrictive Clause T′

parrots T

PrP parrots Pr′ Pr

Nuclear Scope

AdjP intelligent

For Diesing, the position of the subject on the generic interpretation is due to Quantifier Lowering. Given the somewhat unclear current status of Quantifier Lowering, I take it to be due to the option of interpreting the lower versus the higher copy. There is still the question of what disallows the lower interpretation with individual-level predicates. For Diesing, this is due to the fact that these are in fact control clauses and there is simply no subject in the PrP- (or vP-)internal position. Similar considerations apply to objects. Diesing shows that indefinite objects are in principle ambiguous between specific (or definite interpretations) and existential interpretations. Biskup (2009a), following Diesing’s insights, derives this ambiguity from the fact that they can be interpreted either inside the domain of existential closure (within the vP phase) or outside it (within the CP phase). In languages like German (i.e. languages with overt scrambling), the two interpretations can be disambiguated by scrambling. In (43a), the object remains within the vP (the domain of existential closure) and is interpreted existentially. In (43b), on the other hand, the object has moved out of the domain of existential closure (as evidenced by its position relative to the manner adverb immer ‘always’), which makes the existential interpretation no longer available.

{DESKTOP}NEW FOLDER//9781107040847C07.3D

203

[185–204] 23.11.2013 10:57AM

Phases and LF

(43) a. daß Otto immer Bücher über Wombats that Otto always books about wombats ‘that Otto always reads books about wombats’

203

liest reads.

[German]

b. daß Otto Bücher über Wombats immer liest that Otto books about wombats always reads ‘that Otto always reads books about wombats’ (Diesing 1992: 107–8)

Biskup (2009a) goes a step further, and, in addition to making the explicit proposal that Diesing’s Clause Partitioning corresponds to the split between the CP and the vP phase, also argues that the boundary between these two phases corresponds to the information-structure distinction between topic (old information) and focus (new information), or, to use his terms, background information and informational focus. Thus, vP-internal elements are interpreted as new information (informational focus), whereas vP-external elements (which for subjects implies movement to [Spec,TP], and for objects scrambling above vP) as old information (background, topic). This is illustrated by the contrast in (44a– b). In (44a), the subject Maria, being vP-internal, is interpreted as new information. This sentence would be an appropriate response to the question ‘Who bought a new car?’ However, in (44b), being vP-external, it is interpreted as background information. This would be a felicitous response to the question ‘What did Maria buy?’ (44) a. Nowy samochód kupiła Maria. new car bought Maria ‘Maria bought this new car.’ b. Maria kupiła nowy samochód. Maria bought new car ‘Maria bought a new car.’

[Polish]

The structures of the two are given in (45a–b), with insignificant details omitted and/or simplified. (45) a. Structure of (44a) CP C

Old Information (Topic)

TP

a new car

T′

bought

vP

Maria

New Information (Focus)

v′ v

VP bought a new car

{DESKTOP}NEW FOLDER//9781107040847C07.3D

204

[185–204] 23.11.2013 10:57AM

204

Phases and the interfaces

b. Structure of (44b) CP C Maria

TP

Old Information (Topic) T′

bought

vP v′

Maria v

New Information (Focus) VP

bought a new car

Considerations of this sort show that the boundary between vP and CP phases is relevant not only to syntax, semantics and phonology, but also to information structure.

{DESKTOP}NEW FOLDER//9781107040847C08.3D

205

[205–206] 23.11.2013 10:53AM

8 Summary

In lieu of a conclusion, let me briefly summarize each chapter, highlighting its core ideas. The first chapter served mostly as background, providing the rationale for writing a book about phases, as well as an introduction to those aspects of the minimalist program that are relevant to the understanding of phases. Chapter 2 ‘Introducing phases’ motivated the need for phases, provided the necessary historical background on phases, and reviewed the existing characterizations of phases, contrasting the views of phases ‘defined’ in terms of subarrays with the views of phases defined in terms of the properties of phase heads. The property that played an important role throughout the book was the ability of phase heads (and only phase heads) to be the locus of uninterpretable features. This chapter also showed how this property led to Feature Inheritance, the idea that non-phase heads can only acquire uninterpretable features by inheriting them from phase heads, and raised some questions concerning the obligatory nature of Feature Inheritance. This chapter also showed how phases change the overall architecture of the grammar in that they allow transfer to the interfaces to happen more than once during the derivation, perhaps even non-simultaneously, as was further explored in Chapter 6. Chapter 3 turned to the diagnostics that have been suggested in the existing literature to determine the phasehood of a given category, with an eye towards determining which of these diagnostics are real, and which ones are only apparent. While the diagnostics discussed in this chapter were organized into three groups, distinguishing interface diagnostics (PF and LF diagnostics) from syntactic diagnostics (as seems common practice in the literature on phases), I refrained from attaching too much import to a given diagnostic being a PF, LF or a syntactic diagnostic, as it is often difficult to tell, and the process underlying a given diagnostic might have both semantic and phonological effects. The ones that end up playing an important role in the discussion in subsequent chapters were: the ability of phase heads to determine Spell-Out (as determined by ellipsis types and the effects of the Phase 205

{DESKTOP}NEW FOLDER//9781107040847C08.3D

206

206

[205–206] 23.11.2013 10:53AM

Summary

Impenetrability Condition), the ability of phase heads to serve as Probes (be the loci of uninterpretable features), and the requirement that movement out of phases proceed through phase edges. This chapter also alluded to alternative approaches to phases, which might rely on (or require) different sets of diagnostics, and end up with different categories counting as phases. The next chapter (Chapter 4) turned to three ‘core’ categories that are typically assumed to be phases: CPs, vPs and DPs. It applied the diagnostics established in the previous chapter to these three categories, and concluded that, indeed, with respect to these diagnostics, these three categories do behave as phases. It showed that C, v and D heads trigger Spell-Out, and can act as Probes (and thus value uninterpretable features), and that movement out of them proceeds through the edge (as it can trigger agreement, reconstruct to the edge position or be blocked when the edge is filled). This chapter also examined the issue of whether passive and unaccusative vPs are phases. Here the evidence was somewhat less unequivocal. For example, we saw that such vPs do not trigger Spell-Out, but they do require movement to proceed through the edge. Chapter 5 extended the idea of phasehood to three other categories whose status as phases is somewhat more elusive: Predication Phrases (PrPs), Prepositional Phrases (PPs) and Applicative Phrases (ApplPs) and applied the diagnostics to them. It concluded that only certain types of PrPs and ApplP are phases, and that the behavior of PPs with respect to phasehood diagnostics is mixed. Chapter 6 dealt with variation in phasehood, construed quite broadly, to include both interlanguage and intralanguage variation. It examined the question of whether phasehood can be affected by independent factors (such as head movement), and the possibility that phasehood itself might be subject to crosslinguistic variation. It also examined the possibility that there might be variation in phasehood with respect to PF and LF diagnostics. And the last chapter (Chapter 7) examined the role phases play at the interfaces. It showed how phases can help us understand PF-related phenomena such as linear order preservation effects or nuclear stress assignment, as well as the mapping between phases and semantic and information structure domains. To conclude ever so briefly, I do not aspire to have covered (or even touched upon) all the aspects of Phase Theory in this book. There are many questions that I raised that remain unanswered. I do, however, hope to have shown where phases came from, what they can do, and where they are going.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

207 [207–222] 23.11.2013 10:45AM

References

Abels, Klaus. 2003. ‘Successive Cyclicity, Antilocality and Adposition Stranding.’ Doctoral dissertation, University of Connecticut. Abney, Steven P. 1987. ‘The English Noun Phrase in its Sentential Aspect.’ Doctoral dissertation, MIT. Abraham, Werner. 2010. ‘Misleading Homonymies, Economical PPs in Microvariation, and P as a Probe.’ In Mapping Spatial PPs, ed. G. Cinque & L. Rizzi, 261–94. New York: Oxford University Press. Adger, David. 2003. Core Syntax: A Minimalist Approach. Oxford University Press. 2007. ‘Stress and Phasal Syntax.’ Linguistic Analysis 33: 238–66. Adger, David & Gillian Ramchand. 2003. ‘Predication and Equation.’ Linguistic Inquiry 34: 325–59. Adger, David & Peter Svenonius. 2011. ‘Features in Minimalist Syntax.’ In The Oxford Handbook of Minimalist Syntax, ed. C. Boeckx, 27–51. Oxford University Press. Aelbrecht, Lobke. 2009. ‘You Have the Right to Remain Silent: The Syntactic Licensing of Ellipsis.’ Doctoral dissertation, Catholic University of Brussels. Agüero-Bautista, Calixto. 2001. ‘Cyclicity and the Scope of Wh-Phrases.’ Doctoral disseration, MIT. Aissen, Judith. 1996. ‘Pied-Piping, Abstract Agreement, and Functional Projections in Tzotzil.’ Natural Language and Linguistic Theory 14: 447–91. Aldridge, Edith. 2005. ‘Phase Theory Account of Absolutive Extraction in Tagalog.’ In Perspectives on Phases, ed. M. McGinnis & N. Richards, 1–28. Cambridge, MA: MIT Working Papers in Linguistics. 2010. ‘Clause-internal Wh-movement in Archaic Chinese.’ Journal of East Asian Linguistics 19: 1–36. Alexiadou, Artemis & Elena Anagnostopoulou. 2001. ‘The Subject-in-Situ Generalization and the Role of Case in Driving Computations.’ Linguistic Inquiry 32: 193–231. Alexiadou, Artemis, Liliane M. V. Haegeman & Melita Stavrou. 2007. Noun Phrase in the Generative Perspective. Berlin and New York: Mouton De Gruyter. Bachrach, Asaf & Roni Katzir. 2007. ‘Spelling out QR.’ Proceedings of Sinn und Bedeutung 11, ed. E. Puig-Waldmüller, 63–75. Barcelona: Universitat Pompeu Fabra. 2009. ‘Right Node Raising and Delayed Spell-Out.’ In Interphases: Phase-Theoretic Investigations of Linguistic Interfaces, ed. K. K. Grohmann, 283–316. Oxford University Press.

207

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

208

208 [207–222] 23.11.2013 10:45AM

References

Bailyn, John Frederick. 2001. ‘On Scrambling: A Reply to Bošković and Takahashi.’ Linguistic Inquiry 32: 635–58. 2009. ‘What’s inside VP? New Evidence on VP Internal Structure in Russian.’ Paper presented at FASL 18, Cornell University. Bailyn, John Frederick & Barbara Citko. 1999. ‘Case and Agreement in Slavic Predicates.’ In Formal Approaches to Slavic Linguistics 7: The Seattle Meeting 1998, ed. K. Dziwirek, H. Coates & C. Vakareliyska, 17–39. Ann Arbor: Michigan Slavic Publications. Bailyn, John Frederick & Edward Rubin. 1991. ‘The Unification of Instrumental Case Assignment in Russian.’ Cornell Working Papers in Linguistics 9: 99–126. Baker, Mark C. 2003. Lexical Categories: Verbs, Nouns, and Adjectives. Cambridge University Press. 2008a. ‘The Macroparameter in a Microparametric World.’ In The Limits of Syntactic Variation, ed. T. Biberauer, 351–74. Cambridge University Press. 2008b. The Syntax of Agreement and Concord. Cambridge University Press. Baker, Mark, Kyle Johnson & Ian Roberts. 1989. ‘Passive Arguments Raised.’ Linguistic Inquiry 20: 219–51. Baltin, Mark R. 1987. ‘Do Antecedent-Contained Deletions Exist?’ Linguistic Inquiry 18: 579–95. Barros, Mathew & Luis Vicente. 2010. ‘Right Node Raising Requires Both Multidominance and Ellipsis.’ Handout from a talk given at the 34th Penn Linguistics Colloquium. Barrs, Andrew 1986. ‘Chains and Anaphoric Dependence.’ Doctoral dissertation, MIT. 2001. ‘Syntactic Reconstruction Effects.’ In The Handbook of Syntactic Theory, ed. M. Baltin & C. Collins, 670–96. Malden, MA: Blackwell. Barss, Andrew & Howard Lasnik. 1986. ‘A Note on Anaphora and Double Objects.’ Linguistic Inquiry 17: 347–54. Belletti, Adriana. 1988. ‘The Case of Unaccusatives.’ Linguistic Inquiry 19: 1–34. Belletti, Adriana & Luigi Rizzi. 1988. ‘Psych-Verbs and Θ-Theory.’ Natural Language and Linguistic Theory 6: 291–352. Bernstein, Judy. 1993. ‘Topics in the Syntax of Nominal Structure across Romance.’ Doctoral dissertation, CUNY Graduate Center. 2001a. ‘The DP Hypothesis: Identifying Clausal Properties in the Nominal Domain.’ In The Handbook of Contemporary Syntactic Theory, ed. M. Baltin & C. Collins, 536– 61. Malden, MA: Blackwell. 2001b. ‘Focusing the “Right” Way in Romance Determiner Phrases.’ Probus 13: 1–29. 2008. ‘Reformulating the Determiner Phrase Analysis.’ Language and Linguistics Compass 2: 1246–70. Bhatt, Rajesh. 2005. ‘Long Distance Agreement in Hindi–Urdu.’ Natural Language and Linguistic Theory 23: 757–807. Biberauer, Theresa (ed.) 2008. The Limits of Syntactic Variation. Amsterdam and Philadelphia: John Benjamins. Biberauer, Theresa, Anders Holmberg, Ian Roberts & Michelle Sheehan. 2010. Parametric Variation: Null Subjects in Minimalist Theory. Cambridge University Press. Biskup, Petr. 2009a. ‘The Phase Model and Adverbials.’ Doctoral dissertation, University of Leipzig. Biskup, Petr. 2009b. ‘The Syntactic Structure of PPs.’ In Czech in Formal Grammar, ed. M. Dočekal & M. Ziková, 9–29. Munich: Lincom.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

209 [207–222] 23.11.2013 10:45AM

References

209

Bittner, Maria & Ken Hale. 1996. ‘The Structural Determination of Case and Agreement.’ Linguistic Inquiry 27: 1–67. Bobaljik, Jonathan David. 2003. ‘Floating Quantifiers: Handle With Care.’ In The Second Glot International State-of-the-Article Book, ed. L. Cheng & R. Sybesma, 107–48. Berlin and New York: Mouton de Gruyter. Bobaljik, Jonathan David & Dianne Jonas. 1996. ‘Subject Positions and the Roles of TP.’ Linguistic Inquiry 27: 195–236. Bobaljik, Jonathan David & Susi Wurmbrand. 2005. ‘The Domain of Agreement.’ Natural Language and Linguistic Theory 23: 809–65. Boeckx, Cedric. 2001. ‘Scope Reconstruction and A-Movement.’ Natural Language and Linguistic Theory 19: 503–48. 2006. Linguistic Minimalism: Origins, Concepts, Methods, and Aims. Oxford University Press. 2008a. Understanding Minimalist Syntax: Lessons from Locality in Long-Distance Dependencies. Malden, MA: Blackwell. 2008b. Bare Syntax. Oxford University Press. 2012. Syntactic Islands. Cambridge University Press. Boeckx, Cedric & Kleanthes K. Grohmann 2003. Multiple Wh-Fronting. Amsterdam and Philadelphia: John Benjamins. 2007. ‘Remark: Putting Phases in Perspective.’ Syntax: 10: 204–22. Booij, Geert & Jerzy Rubach. 1987. ‘Postcyclic versus Postlexical Rules in Lexical Phonology.’ Linguistic Inquiry 18: 1–44. Borer, Hagit. 1984. Parametric Syntax: Case Studies in Semitic and Romance Languages. Dordrecht: Foris. Bošković, Željko. 1997. The Syntax of Nonfinite Complementation: An Economy Approach. Cambridge, MA: MIT Press. 2002a. ‘A-Movement and the EPP.’ Syntax 5: 167–218. 2002b. ‘On Multiple Wh-Fronting.’ Linguistic Inquiry 33: 351–83. 2003. ‘Agree, Phases and Intervention Effects.’ Linguistic Analysis 33: 54–96. 2004a. ‘Be Careful Where You Float Your Quantifiers.’ Natural Language and Linguistic Theory 22: 681–742. 2004b. ‘Object Shift and the Clause/PP Parallelism Hypothesis.’ In WCCFL 23: Proceedings of the 23rd West Coast Conference on Formal Linguistics, ed. B. Schmeiser, V. Chand, A. Kelleher, & A. Rodriguez, 101–14. Somerville, MA: Cascadilla Press. 2005. ‘On the Locality of Left Branch Extraction and the Structure of NP.’ Studia Linguistica 59: 1–45. 2007. ‘On the Locality and Motivation of Move and Agree: an Even More Minimal Theory.’ Linguistic Inquiry 38: 589–644. 2009. ‘More on the No-DP Analysis of Article-Less Languages.’ Studia Linguistica 63: 187–203. 2011. ‘On Unvalued Interpretable Features.’ In NELS 39: Proceedings of the 39th Annual Meeting of the North East Linguistic Society, ed. S. Lima, K. Mullin & B. Smith, 109–20. Amherst, MA: GLSA. (To appear). ‘Now I’m a Phase, Now I’m Not a Phase: On the Variability of Phases with Extraction and Ellipsis.’ Linguistic Inquiry. Bošković, Željko & Steven Franks. 2000. ‘Across-the-Board Movement and LF.’ Syntax 3: 107–28.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

210

210 [207–222] 23.11.2013 10:45AM

References

Bošković, Željko & Daiko Takahashi. 1998. ‘Scrambling and Last Resort.’ Linguistic Inquiry 29: 347–66. Bowers, John. 1993. ‘The Syntax of Predication.’ Linguistic Inquiry 24: 591–656. 2002. ‘Transitivity.’ Linguistic Inquiry 33: 183–224. Bresnan, Joan W. 1971. ‘Sentence Stress and Syntactic Transformations.’ Language 47: 257–81. Bruening, Benjamin. 2001. ‘QR Obeys Superiority: Frozen Scope and ACD.’ Linguistic Inquiry 32: 233–73. 2004. ‘Two Types of Wh-Scope Marking in Passamaquoddy.’ Natural Language and Linguistic Theory 22: 229–305. Büring, Daniel. 2005. Binding Theory. Cambridge University Press. Butler, Jonny. 2005. ‘The Phase Structure of Tense.’ In Perspectives on Phases ed. M. McGinnis & N. Richards, 69–85. Cambridge, MA: MIT Working Papers in Linguistics. Cable, Seth. 2008. ‘Wh-fronting in Hungarian Is Not Focus Fronting.’ Manuscript. Available at Lingbuzz (lingbuzz/000674). Calabrese, Andrea 1984. ‘Multiple Questions and Focus in Italian.’ In Sentential Complementation, ed. W. De Geest & T. Putseys, 67–74. Dordrecht: Foris. Carnie, Andrew & Andrew Barss. 2006. ‘Phases and Nominal Interpretation.’ Research In Language 4: 127–32. Carstens, Vicki. 2000. ‘Concord in Minimalist Theory.’ Linguistic Inquiry 31: 319–55. 2003. ‘Rethinking Complementizer Agreement: Agree with a Case-Checked Goal.’ Linguistic Inquiry 34: 393–412. 2005. ‘Agree and EPP In Bantu.’ Natural Language and Linguistic Theory 23: 219–79. Carstens, Vicki & Michael Diercks. 2011. ‘Agreeing How? Implications for Theories of Agreement And Locality.’ Manuscript. Available at Lingbuzz (lingbuzz/001389). Castillo, Juan Carlos, John E. Drury & Kleanthes K. Grohmann. 2009. ‘Merge over Move and the Extended Projection Principle: MOM and the EPP Revisited.’ Iberia: An International Journal of Theoretical Linguistics 1: 53–114. Cecchetto, Carlo. 2004. ‘Explaining the Locality Conditions of QR: Consequences for the Theory Of Phases.’ Natural Language Semantics 12: 345–97. Cecchetto, Carlo & Caterina Donati. 2010. ‘On Labeling: Principle C and Head Movement.’ Syntax 13: 241–78. Charlow, Simon. 2009. ‘Can DP Be a Scope Island?’ Talk given at the ESSLLI09 Student Session. Cheng, Lisa Lai-Shen. 1991. ‘On The Typology of Wh-Questions.’ Doctoral dissertation, MIT. Chomsky, Noam. 1970. ‘Remarks on Nominalization.’ In Readings In English Transformational Grammar, by R. A. Jacobs & P. S. Rosenbaum, 184–221. Waltham, MA: Ginn & Co. 1986. Knowledge of Language: Its Nature, Origin, and Use. New York: Praeger. 1994. Bare Phrase Structure. Working Occasional Papers in Linguistics 5. Cambridge, MA: MIT. 1995. The Minimalist Program. Cambridge, MA: MIT Press. 2000. ‘Minimalist Inquiries: The Framework.’ In Step By Step. Essays on Minimalist Syntax in Honor of Howard Lasnik, ed. R. E. A. Martin, 89–155. Cambridge, MA: MIT Press.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

211 [207–222] 23.11.2013 10:45AM

References

211

2001. ‘Derivation by Phase.’ In Ken Hale: A Life in Language, ed. M. Kenstowicz, 1–52. Cambridge, MA: MIT Press. 2004. ‘Beyond Explanatory Adequacy.’ In Structures and Beyond: The Cartography of Syntactic Structures, vol. 3, ed. A. Belletti, 104–31. Oxford University Press. 2005. ‘Three Factors in Language Design.’ Linguistic Inquiry 36: 1–22. 2007. ‘Approaching UG from Below.’ In Interfaces + Recursion = Language?, ed. U. Sauerland & H. -M. Gärtner, 1–30. Berlin: Mouton De Gruyter. 2008. ‘On Phases.’ In Foundational Issues in Linguistic Theory, ed. R. Freidin, C. P. Otero & M. L. Zubizarreta, 133–66. Cambridge, MA: MIT Press. 2013 ‘The Problem of Projection.’ Lingua 130: 33–49. Chomsky, Noam & Morris Halle. 1968. The Sound Pattern of English. New York: Harper & Row. Chung, S., W. A. Ladusaw & J. McCloskey. 1995. ‘Sluicing and Logical Form.’ Natural Language Semantics 3: 239–82. Chung, Sandra. 1994. ‘Wh-Agreement and “Referentiality” in Chamorro.’ Linguistic Inquiry 25: 1–44. 1998. The Design of Agreement: Evidence from Chamorro. University of Chicago Press. Cinque, Guglielmo. 1993. ‘A Null Theory of Phrase and Compound Stress.’ Linguistic Inquiry 24: 239–97. 2002. Functional Structure in DP and IP. Oxford University Press. 2010. The Syntax of Adjectives: A Comparative Study. Cambridge, MA: MIT Press. Cinque, Guglielmo & Luigi Rizzi. 2010. Mapping Spatial PPs. Oxford University Press. Citko, Barbara. 2000. ‘Parallel Merge and the Syntax of Free Felatives.’ Doctoral dissertation, Stony Brook University. 2005. ‘On the Nature of Merge: External Merge, External Merge, and Parallel Merge.’ Linguistic Inquiry 36: 475–97. 2006. ‘Copula Sentences Reconsidered.’ In Formal Approaches to Slavic Linguistics 14: The Princeton Meeting 2005, ed. J. E. Lavine, S. Franks, M. TassevaKurktchieva & H. Filip, 83–98. Ann Arbor: Michigan Slavic Publications. 2008a. ‘Missing Labels.’ Lingua 118: 907–44. 2008b. ‘Small Clauses Reconsidered: Not So Small and Not All Alike.’ Lingua 118: 261–95. 2010. ‘On the (A)Symmetric Nature of Movement: A View from Polish Wh- and Passive Movement.’ In Formal Approaches to Slavic Linguistics 18: The Second Cornell Meeting 2009, ed. C. Browne et al., 38–57. Ann Arbor: Michigan Slavic Publications. 2011a. ‘Multidominance.’ In The Oxford Handbook of Linguistic Minimalism, ed. C. Boeckx, 119–42. Oxford University Press. 2011b. Symmetry in Syntax: Merge, Move and Labels. Cambridge University Press. 2011c. ‘Small Clauses.’ Language and Linguistics Compass 5: 748–63. Cole, Peter & Gabriella Hermon. 1998. ‘The Typology of WH Movement: WH Questions in Malay.’ Syntax 1: 221–58. Collins, Chris. 2002. ‘Eliminating Labels.’ In Derivation and Explanation in the Minimalist Program, ed. S. D. Epstein & T. D. Seely. Malden, MA: Blackwell. Corver, Norbert. 1990. ‘The Syntax of Left Branch Extraction.’ Doctoral dissertation, Tilburg University.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

212

212 [207–222] 23.11.2013 10:45AM

References

1992. ‘Left Branch Extraction.’ In NELS 22: Proceedings of the 22nd Annual Meeting of the North East Linguistics Society, ed. K. Broderick, 67–84. Amherst, MA: GLSA. Craenenbroeck, Jeroen van. 2010. The Syntax of Ellipsis: Evidence from Dutch Dialects. Oxford University Press. Craenenbroeck, Jeroen van & Liliane M. V. Haegeman. 2007. ‘The Derivation of SubjectInitial V2.’ Linguistic Inquiry 38: 167–78. Culicover, Peter W. & Paul Martin Postal (eds.). 2001. Parasitic Gaps. Cambridge, MA: MIT Press. Danon, Gabi. 2011. ‘Agreement and DP-Internal Feature Distribution.’ Syntax 14: 97–317. Davies, William D. & Stanley Dubinsky. 2003. ‘On Extraction from NPs.’ Natural Language and Linguistic Theory 21: 1–37. Dayal, Veneeta. 1996. Locality in Wh-Quantification: Questions and Relative Clauses in Hindi. Dordrecht: Kluwer. 1993. ‘Scope Marking as Indirect Wh-Dependency.’ Natural Language Semantics 2: 137–70. Den Dikken, Marcel. 2006a. ‘A Reappraisal of vP Being Phasal – A Reply to Legate.’ Unpublished manuscript, CUNY Graduate Center. 2006b. Relators and Linkers: The Syntax of Predication, Predicate Inversion, and Copulas. Cambridge, MA: MIT Press. 2007. ‘Phase Extension: Contours of a Theory of the Role of Head Movement in Phrasal Extraction.’ Theoretical Linguistics 33: 1–41. 2010. ‘On the Functional Structure of Locative and Directional PPs.’ In Mapping Spatial PPs: The Cartography of Syntactic Structures, ed. G. Cinque & L. Rizzi, 74–126. Oxford University Press. Diercks, Michael. 2011. ‘Against Feature Inheritance: Phase Heads Are Not Defined by Unvalued Φ-features.’ Manuscript. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Dobashi, Yoshihiko. 2003. ‘Phonological Phrasing and Syntactic Derivation.’ Doctoral dissertation, Cornell University. Dogil, Grzegorz. 1987. ‘Lexical Phonology and Floating Inflection in Polish.’ In Phonologica 1984, ed. W. U. Dressler, 39–47. Cambridge University Press. Donati, Caterina & Carlo Cecchetto. 2011. ‘Relabeling Heads: A Unified Account for Relativization Structures.’ Linguistic Inquiry 42: 519–60. Dornisch, Ewa. 1998. ‘Multiple Wh-Questions in Polish: The Interactions between WhPhrases and Clitics.’ Doctoral dissertation, Cornell University. Drummond, Alex, Norbert Hornstein & Howard Lasnik. 2010. ‘A Puzzle about PStranding and a Possible Solution.’ Linguistic Inquiry 41: 689–92. Dyakonova, Marina. 2007. ‘Russian Double Object Construction.’ ACLC Working Papers 21: 1–28. Dziwirek, Katarzyna. 2000. ‘Why Polish Doesn’t Like Infinitives.’ Journal of Slavic Linguistics 8: 57–82. Elordieta, Gorka. 2008. ‘An Overview of Theories of the Syntax-Phonology Interface.’ International Journal of Basque Linguistics and Philology (ASJU) 42(1): 209–86. Embick, David. 1995. ‘Mobile inflections in Polish.’ In NELS 25: Proceedings of the 25th Annual Meeting of the North East Linguistics Society, ed. J. N. Beckman, 127–42. Amherst, MA: GLSA. Emonds, Joseph. 1985. A Unified Theory of Syntactic Categories. Dordrecht: Foris.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

213 [207–222] 23.11.2013 10:45AM

References

213

2001. ‘The Lower Operator Position with Parasitic Gaps.’ In Features and Interfaces in Romance, ed. J. Herschensohn, E. Mallén & K. Zagona, 85–106. Amsterdam and Philadelphia: John Benjamins. Enç, Mürvet. 1991. ‘The Semantics of Specificity.’ Linguistic Inquiry 22: 1–25. Engdahl, Elisabet. 2001. ‘Parasitic Gaps.’ In Parasitic Gaps, ed. Peter W. Culicover & Paul M. Postal, 69–97. Cambridge, MA: MIT Press. Epstein, Samuel. 2007. ‘On I(nternalist)-Functional Explanation in Minimalism.’ Linguistic Analysis 33: 20–53. Epstein, Samuel D. & T. Daniel Seely. 2002. ‘Rule Applications as Cycles in a Level-Free Syntax.’ In Derivation and explanation in the Minimalist Program, ed. S. D. Epstein & T. D. Seely, 65–89. Oxford: Blackwell. 2006. Transformations and Derivations. Cambridge University Press. Fanselow, Gisbert & Damir Čavar. 2002. ‘Distributed Deletion.’ In Theoretical Approaches to Universals, ed. A. Alexiadou, 65–107. Amsterdam and Philadelphia: John Benjamins. Felser, Claudia. 2004. ‘Wh-Copying, Phases, and Successive Cyclicity.’ Lingua 114: 543–74. Fox, Danny. 1999. ‘Reconstruction, Binding Theory, and the Interpretation of Chains.’ Linguistic Inquiry 30: 157–96. 2000. Economy and Semantic Interpretation Cambridge, MA: MIT Press. Fox, Danny & David Pesetsky. 2005. ‘Cyclic Linearization of Syntactic Structure.’ Theoretical Linguistics 31: 1–46. Frampton, John & Sam Gutmann. 2000. ‘Agreement is Feature Sharing.’ Unpublished manuscript, Northeastern University. 2002. ‘Crash-Proof Syntax.’ In Derivation and Explanation in the Minimalist Program, ed. S. D. Epstein & T. Daniel Seely, 90–105. Malden, MA: Blackwell. Franks, Steven. 1998. ‘Clitics in Slavic.’ Paper Presented at the Comparative Slavic Morphosyntax Workshop. Spencer, IN. Franks, Steven & Piotr Bański. 1999. ‘Approaches to “Schizophrenic” Polish Person Agreement.’ In Formal Approaches to Slavic Linguistics: The Seattle Meeting, ed. K. Dziwirek, H. Coats & C. M. Vakareliyska, 123–44. Ann Arbor: Michigan Slavic Publications. Frascarelli, Mara (ed.). 2006. Phases of Interpretation. Berlin: Mouton de Gruyter. Frederici, Angela. 1982. ‘Syntactic and Semantic Processes in Aphasic Deficits: the Availability of Prepositions.’ Brain And Language 15: 269–58. Froud, Karen. 2001. ‘Prepositions and the Lexical/Functional Divide: Aphasic Evidence.’ Lingua 111: 1–28. Gallego, Ángel J. 2010. Phase Theory. Amsterdam and Philadelphia: John Benjamins. 2012. Phases: Developing the Framework. Berlin and Boston: De Gruyter Mouton. Gavruseva, E. 2000. ‘On the Syntax of Possessor Extraction.’ Lingua 110: 743–72. Gengel, Kirsten. 2007. ‘Focus and Ellipsis a Generative Analysis of Pseudogapping and Other Elliptical Structures.’ Doctoral dissertation, Stuttgart University. 2009. ‘Phases and Ellipsis.’ Linguistic Analysis 35: 21–42. Green, Georgia M. 1974. Semantics and Syntactic Regularity. Bloomington, IN: Indiana University Press. Grimshaw, Jane. 1991. ‘Extended Projection.’ Manuscript, Brandeis University. Grohmann, Kleanthes K. 2003. Prolific Domains: On the Antilocality of Movement Dependencies. Amsterdam and Philadelphia: John Benjamins.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

214

214 [207–222] 23.11.2013 10:45AM

References

(ed.) 2009a. Explorations of Phase Theory: Features And Arguments. Berlin: Walter de Gruyter. 2009b (ed.). Explorations of Phase Theory: Interpretation at the Interfaces. Berlin: Walter de Gruyter. 2009c (ed). Interphases: Phase-Theoretic Investigations of Linguistic Interfaces. Oxford University Press. 2011. ‘Anti-Locality: Too-Close Relations in Grammar.’ In The Oxford Handbook of Linguistic Minimalism, ed. C. Boeckx, 260–90. Oxford University Press. Grohmann, Kleanthes K., John Drury & Juan Carlos Castillo. 2000. ‘No More EPP.’ In WCCFL 19: Proceedings of the 19th West Coast Conference on Formal Linguistics, ed. R. Billerey & B. D. Lillehaugen, 153–66. Somerville, MA: Cascadilla Press. Haegeman, Liliane. 1992. Theory and Description in Generative Syntax: A Case Study in West Flemish. Cambridge University Press. Haegeman, Liliane & Marjo Van Koppen. 2011. ‘Complementizer Agreement and the Relation Between C° And T°.’ Manuscript, Ghent University and University of Utrecht. Hale, Ken & Jay S. Keyser. 1993. ‘On Argument Structure and the Lexical Expression of Syntactic Relations.’ In The View From Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. K. Hale & J. S. Keyser, 53–109. Cambridge, MA: MIT Press. Halle, Morris & Alex Marantz. 1993. ‘Distributed Morphology and the Pieces of Inflection’. In The View From Building 20, ed. K. Hale & J. S. Keyser, 111–76. Cambridge, MA: MIT Press. Halle, Morris & Jean-Roger Vergnaud. 1987. An Essay on Stress. Cambridge, MA: MIT Press. Hankamer, Jorge & Ivan Sag. 1976. ‘Deep and Surface Anaphora.’ Linguistic Inquiry 7: 391–428. Harves, Stephanie. 2002. ‘Where Have All the Phases Gone? (Non-)Defective Categories and Case Alternations in Russian’. In Formal Approaches To Slavic Linguistics 10: The Second Ann Arbor Meeting, ed. J. Toman, 97–118. Ann Arbor, MI: Michigan Slavic Publications. Heycock, Caroline. 1995. ‘Asymmetries in Reconstruction.’ Linguistic Inquiry 26: 547–70. Hiraiwa, Ken. 2001. ‘Multiple Agree and the Defective Intervention Constraint in Japanese.’ In Proceedings of HUMIT 2000, 67–80. Cambridge, MA: MIT Working Papers in Linguistics. Hiraiwa, Ken. 2005. ‘Dimensions of Symmetry in Syntax: Agreement and Clausal Architecture’. Doctoral dissertation, MIT. Holmberg, Anders. 2001. ‘The Syntax of Yes and No in Finnish.’ Studia Linguistica 55: 141–75. Hornstein, Norbert. 1994. ‘An Argument for Minimalism: the Case of AntecedentContained Deletion.’ Linguistic Inquiry 25: 455–80. 1995. Logical Form. Oxford: Blackwell. Hornstein, Norbert & Jairo Nunes. 2002. ‘On Asymmetries between Parasitic Gap and Across-the-Board Constructions.’ Syntax 5: 26–54. Hornstein, Norbert, Jairo Nunes & Kleanthes K. Grohmann. 2005. Understanding Minimalism. Cambridge University Press. Hornstein, Norbert & Amy Weinberg. 1981. ‘Case Theory and Preposition Stranding.’ Linguistic Inquiry 12: 55–91. Horvath, Julia. 1997. ‘The Status of “Wh-Expletives” and the Partial Wh-Movement Construction of Hungarian.’ Natural Language and Linguistic Theory 15: 509–72.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

215 [207–222] 23.11.2013 10:45AM

References

215

Huang, Cheng-Teh James. 1982. ‘Logical Relations in Chinese and the Theory of Grammar.’ Doctoral dissertation, MIT. 1993. ‘Reconstruction and the Structure of VP: Some Theoretical Consequences.’ Linguistic Inquiry 24: 103–38. Iatridou, Sabine & Ivy Sichel. 2011. ‘Negative DPs, A-Movement, and Scope Diminishment. Linguistic Inquiry 42: 595–629. Inèenlikèj, Petr I. & Vladimir Nedjalkov. 1973. ‘Glagoly čuvstva v čukotskom jazyke.’ Lingvističeskie issledovanija 1972, 175–203. Ishihara, Shinichiro. 2003. ‘Intonation and Interface Conditions.’ Doctoral dissertation, MIT. 2007. ‘Major Phrase, Focus Intonation, Multiple Spell-Out (Map, FI, MSO).’ The Linguistic Review 24: 2–3. Jackendoff, Ray. 1977. X-bar Syntax: A Study of Phrase Structure. Cambridge, MA: MIT Press. Jaysaseelan, K. A. 2001. ‘IP-Internal Topic and Focus Phrases.’ Studia Linguistica 55: 39–75. Jeong, Youngmi. 2006. ‘The Landscape of Applicatives.’ Doctoral dissertation, University of Maryland. Johnson, Kyle. 2000. ‘Few Dogs Eat Whiskas or Cats Alpo.’ In UMOP 23: Issues in Semantics and its Interface, ed. K. Kusumoto and E. Villalta, 59–82. Amherst, MA: GSLA. 2009. ‘Gapping Is Not (VP-)Ellipsis.’ Linguistic Inquiry 40: 289–328. Kahnemuyipour, Arsalan. 2003. ‘Syntactic Categories and Persian Stress.’ Natural Language and Linguistic Theory 21: 333–79. 2004. ‘The Syntax of Sentential Stress.’ Doctoral dissertation, University of Toronto. 2005. ‘Towards a Phase-Based Theory of Sentential Stress.’ In Perspectives on Phases, ed. M. McGinnis & N. Richards, 125–46. Cambridge, MA: MIT Working Papers In Linguistics. Kasai, Hironobu. 2007. ‘Multidominance in Syntax.’ Doctoral dissertation, Harvard University. Katz, Jerrold J. & Paul Martin Postal. 1964. An Integrated Theory of Linguistic Descriptions. Cambridge, MA: MIT Press. Kayne, Richard S. 1980. ‘Extensions of Binding and Case-Marking.’ Linguistic Inquiry 11: 75–96. 1981. ‘On Certain Differences between French and English.’ Linguistic Inquiry 12: 349–72. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. 1999. ‘Prepositional Complementizers as Attractors.’ Probus 11: 39–73. 2004. ‘Prepositions as Probes’. In Structures and Beyond, ed. A. Belletti, 192-X. Oxford University Press. 2010. ‘Why Are There No Directionality Parameters?’ Manuscript, New York University. Kennedy, Christopher. 1997. ‘Antecedent-Contained Deletion and the Syntax of Quantification.’ Linguistic Inquiry 28: 662–88. Khomitsevich, Olga. 2007. ‘Dependencies across Phases: From Sequence of Tense to Restrictions on Movement.’ Doctoral dissertation, Utrecht: LOT. Kinyalolo, Kasangati Kikuni Wabongambilu. 1991. ‘Syntactic Dependencies and the SPEC–Head Agreement Hypothesis in Kilega.’ Doctoral dissertation, UCLA.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

216

216 [207–222] 23.11.2013 10:45AM

References

Kiparsky, Paul. 1998. ‘Partitive Case and Aspect’. In The Projection of Arguments, ed. M. Butt & W. Gueder. Stanford: CSLI Publications. Kiss, Katalin É. 1987. Configurationality in Hungarian. Dordrecht: Reidel. Koopman, Hilda 2000. ‘Prepositions, Postpositions, Circumpositions, and Particles: The Case of Dutch PPs’. In The Syntax of Specifiers and Heads: Collected Essays of Hilda J. Koopman, ed. H. Koopman, 204–60. London: Routledge. Kratzer, Angelika & Elisabeth Selkirk. 2007. ‘Phase Theory and Prosodic Spell-Out: The Case of Verbs.’ The Linguistic Review 24: 93–135. Laenzlinger, Christopher. 2005. ‘French Adjective Ordering: Perspectives on DP-Internal Movement Types.’ Lingua 115: 645–89. Lahne, Antje. 2008. ‘Where There Is Fire There Is Smoke: Local Modelling of SuccessiveCyclic Movement.’ Doctoral dissertation, University Of Leipzig. Landau, Idan. 1999. ‘Elements of Control.’ Doctoral dissertation, MIT. 2000. Elements of Control: Structure and Meaning in Infinitival Constructions Dordrecht and Boston: Kluwer. 2003. ‘Movement out of Control.’ Linguistic Inquiry 34: 471–98. 2008. ‘Two Routes of Control: Evidence from Case Transmission in Russian.’ Natural Language and Linguistic Theory 26: 877–924. 2010. The Locative Syntax of Experiencers. Cambridge, MA: MIT Press. Larson, Richard K. 1985. ‘Quantifying into NP.’ Manuscript, Stony Brook University. 1988. ‘Light Predicate Raising.’ MIT Lexicon Project Working Papers 27. Cambridge, MA: MIT Center for Cognitive Science. Larson, Richard & Robert May. 1990. ‘Antecedent Containment or Vacuous Movement: Reply to Baltin.’Linguistic Inquiry 21: 103–22. Lasnik, Howard. 1992. ‘Case and Expletives: Notes Toward a Parametric Account.’ Linguistic Inquiry, 23: 381–405. 1998. ‘Exceptional Case Marking: Perspectives Old And New.’ In Formal Approaches To Slavic Linguistics: The Connecticut Meeting 1997, ed. Z. Bošković, S. Franks & W. Snyder, 187–211. Ann Arbor: Michigan Slavic Publications. 1999. ‘Chains of Arguments.’ In Working Minimalism, ed. S. D. Epstein & N. Hornstein, 189–215. Cambridge, MA: MIT Press. Lebeaux, David. 1990. ‘Relative Clauses, Licensing, and the Nature of the Derivation.’ In NELS 20: Proceedings of the 20th Annual Meeting of the North East Linguistic Society, ed. J. Carter, R.-M. Dechaine, B. Philip & T. Sherer, 318–32. Amherst, MA: GLSA. 1991. ‘Relative Clauses, Licensing, and the Nature of the Derivation.’ In Syntax and Semantics 25: Perspectives on Phrase Structure, ed. S. Rothstein, 209–39. New York: Academic Press. 1994. ‘Where does binding theory apply?’ MS, University of Maryland, College Park. 2009. Where Does Binding Theory Apply? Cambridge, MA: MIT Press. Legate, Julie Anne. 2002. Phases in Beyond Explanatory Adequacy. Manuscript. 2003. ‘Some Interface Properties of the Phase.’ Linguistic Inquiry 34: 506–16. 2011. ‘Under-Inheritance.’ Handout of a paper presented at NELS 42, University of Toronto. 2012. ‘The size of Phases.’ In Phases, ed. A. Gallego, 233–50. Berlin: De Gruyter. Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity: At the Syntax–Lexical Semantics Interface. Cambridge, MA: MIT Press.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

217 [207–222] 23.11.2013 10:45AM

References

217

Lobeck, Anne C. 1995. Ellipsis: Functional Heads, Licensing, and Identification. New York: Oxford University Press. Lutz, Uli, Gereon Müller & Arnim von Stechow. 2000. Wh-Scope Marking. Amsterdam and Philadelphia: John Benjamins. Macdonald, Jonathan Eric. 2006. ‘The Syntax of Inner Aspect.’ Doctoral dissertation, Stony Brook University. Mahajan, Anoop Kumar. 1989. ‘Agreement and Agreement Phrases.’ In Functional Heads and Clause Structure, ed. I. Laka & A. K. Mahajan, 217–52, Cambridge, MA: MIT Working Papers in Linguistics. 1996. ‘Wh-Expletives and the Syntax of Partial Wh-Movement.’ In Papers On Scope Marking, ed. U. Lutz & G. Müller, 163–77. University of Tübingen / University of Stuttgart. Manetta, Emily. 2010. ‘Wh-Expletives in Hindi-Urdu: The vP Phase.’ Linguistic Inquiry 41: 1–34. Marantz, Alec. 1993. ‘Implications of Asymmetries in Double Object Constructions.’ In Theoretical Aspects of Bantu Grammar, ed. S. Mchombo, 113–50. Stanford: CSLI Publications. 2007. ‘Phases and Words.’ In Phases in the Theory of Grammar, ed. H. Choe, 191–220. Seoul: Dong In. Martin, Roger Andrew. 1996. ‘A Minimalist Theory of PRO and Control.’ Doctoral dissertation, University Of Connecticut. Marušič, Franc Lanko. 2005. On Non-Simultaneous Phases. Stony Brook University. 2009. ‘Non-Simultaneous Spell-Out in the Clausal and Nominal Domain.’ In Interphases: Phase-Theoretic Investigations of Linguistic Interfaces, ed. K. K. Grohmann, 151–81. Oxford University Press. Matushansky, Ora. 2000. ‘The Instrument of Inversion: Instrumental Case and Verb Raising in the Russian Copula’. In WCCFL 19: Proceedings of the 19th West Coast Conference on Formal Linguistics, ed. R. Billerey & B. Lillehaugen, 101–15. Somerville, MA: Cascadilla Press. 2005. ‘Going through a Phase’. In Perspectives on Phases, ed. M. McGinnis & N. Richards, 157–81. Cambridge, MA: MIT Working Papers in Linguistics. 2008. ‘Predication: A Case Study’. In Studies in Formal Slavic Linguistics: Contributions from Formal Description of Slavic Languages 6.5, ed. F. Marusic & R. Zaucer, 213–39. Frankfurt am Main: Peter Lang. May, Robert 1977. ‘The Grammar of Quantification.’ Doctoral dissertation, MIT. 1985. Logical Form: Its Structure and Derivation. Cambridge, MA: MIT Press. McCloskey, James. 1998. ‘The Prosody of Quantifier Stranding under Wh-Movement in West Ulster English.’ Manuscript. 2000. ‘Quantifier Float and Wh-Movement in an Irish English.’ Linguistic Inquiry 31: 57–84. 2001. ‘The Morphosyntax of WH-Extraction In Irish.’ Journal of Linguistics 37: 67–100. 2002. ‘Resumption, Successive Cyclicity, and the Locality of Operations.’ In Derivation and Explanation in the Minimalist Program, ed. S. D. Epstein & T. Daniel Seely, 184–226. Malden, MA: Blackwell. McDaniel, Dana. 1989. ‘Partial and Multiple Wh-Movement.’ Natural Language and Linguistic Theory 7: 565–604.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

218

218 [207–222] 23.11.2013 10:45AM

References

McGinnis, Martha. 2001. ‘Variation in the Phase Structure of Applicatives.’ Linguistic Variation Yearbook 1: 105–46. McGinnis, Martha & Norvin Richards (eds.). 2005. Perspectives on Phases. Cambridge, MA: MIT Working papers in Linguistics. Megerdoomian, Karine. 2000. ‘Aspect and Partitive Objects in Finnish.’ In WCCFL 19: Proceedings of the 19th West Coast Conference on Formal Linguistics, ed. R. Billerey & B. D. Lillehaugen, 316–28. Somerville, MA: Cascadilla Press. Merchant, Jason. 2001. The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis. Oxford University Press. 2008. ‘An Asymmetry in Voice Mismatches in VP-Ellipsis and Pseudogapping.’ Linguistic Inquiry 39: 169–79. 2013. ‘Diagnosing Ellipsis.’ In Diagnosing Syntax, ed. L. Cheng & N. Corver, 537–42. Oxford University Press. Migdalski, Krzysztof. 2006. ‘The Syntax of Compound Tenses in Slavic.’ Doctoral dissertation, Tilburg University. Miyagawa, Shigeru. 1997. ‘Against Optional Scrambling.’ Linguistic Inquiry 28: 1–25. 2003. ‘A-movement Scrambling and Options without Optionality. In Word Order and Scrambling, ed. S. Karimi, 177–200. Oxford: Blackwell. 2011. ‘Genitive subjects in Altaic and Specification of Phase.’ Lingua 21: 1265–82. Miyagawa, Shigeru & Takae Tsujioka. 2004. ‘Argument Structure and Ditransitive Verbs in Japanese.’ Journal of East Asian Linguistics 13: 1–38. Moro, Andrea. 1997. The Raising of Predicates: Predicative Noun Phrases and the Theory of Clause Structure. Cambridge University Press. 2000. Dynamic Antisymmetry. Cambridge, MA: MIT Press. Müller, Gereon. 2004. ‘Phrase Impenetrability and Wh-Intervention.’ In Minimality Effects in Syntax, ed. A. Stepanov, G. Fanselow & R. Vogel, 289–325. Berlin: De Gruyter. 2011. Constraints on Displacement: A Phase-Based Approach. Amsterdam and Philadelphia: John Benjamins. Munn, Alan. 1999. ‘First Conjunct Agreement: Against a Clausal Analysis.’ Linguistic Inquiry 30: 643–68. Muysken, Pieter. 2008. Functional Categories. Cambridge University Press. Nespor, Marina & Irene Vogel. 1986. Prosodic Phonology. Dordrecht: Foris. Newman, Stanley. 1946. ‘On the Stress System of English.’ Word 2: 171–87. Nissenbaum, Jonathan. 2000. ‘Investigations in Covert Movement.’ Doctoral dissertation, MIT. Nunes, Jairo. 1995. ‘The Copy Theory of Movement and the Linearization of Chains in the Minimalist Program.’ Doctoral dissertation, University Of Maryland. Oehrle, Richard T. 1976. ‘The Grammatical Status of the English Dative Alternation.’ Doctoral dissertation, MIT. Ouali, Hamid. 2008. ‘On C-To-T Phi-Feature Transfer: The Nature of Agreement and Anti-Agreement In Berber’. In Agreement Restrictions, ed. R. D’alessandro, G. H. Hrafnbjargarson & S. Fischer, 159–80. Berlin: Mouton de Gruyter. Pak, Marjorie. 2008. ‘The Postsyntactic Derivation and its Phonological Reflexes.’ Doctoral dissertation, University Of Pennsylvania. Pereltsvaig, Asya. 2007. ‘The Universality of DP: A View from Russian.’ Studia Linguistica 61: 59–94.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

219 [207–222] 23.11.2013 10:45AM

References

219

2007. Copular Sentences in Russian: A Theory of Intra-Clausal Relations. Dordrecht: Springer. Pesetsky, David. 1998. ‘Some Optimality Principles of Sentence Pronounciation’. In Is the Best Good Enough? Optimality and Competition In Syntax, ed. P. Barbosa, D. Fox, P. Hagstrom, M. McGinnis & D. Pesetsky, 337–84. Cambridge, MA: MIT Press. 1991 ‘Zero Syntax: vol. 2: Infinitives.’ Unpublished manuscript, MIT. 1995. Zero Syntax: Experiencers and Cascades. Cambridge, MA: MIT Press. Pesetsky, David & Esther Torrego. 2001. ‘T-To-C Movement: Causes And Consequences’. In Ken Hale: A Life In Language, ed. M. Kenstowicz, 355–426. Cambridge, MA: MIT Press. 2006. ‘Probes, Goals and Syntactic Categories.’ In Proceedings of the 7th Annual Tokyo Conference on Psycholinguistics, ed. Y. Otsu, 25–60. Tokyo: Hituzi Syobo. 2007. ‘The Syntax of Valuation and the Interpretability Of Features’. In Phrasal and Clausal Architecture: Syntactic Derivation and Interpretation. In Honor of Joseph E. Emonds, ed. S. Karimi, V. Samiian & W. K. Wilkins, 262–94. Amsterdam and Philadelphia: John Benjamins. Peterson, David A. 2007. Applicative Constructions. Oxford University Press. Picallo, M. Carme. 1991. ‘Nominals and Nominalizations in Catalan.’ Probus 3: 279–316. Plessis, Hans Du. 1977. ‘Wh Movement in Afrikaans.’ Linguistic Inquiry, 8: 723–6. Polinsky, Maria & Eric Potsdam. 2001. ‘Long-Distance Agreement and Topic in Tsez.’ Natural Language & Linguistic Theory, 19: 583–646. Postal, Paul M. 1972. ‘On Some Rules that Are Not Successive Cyclic.’ Linguistic Inquiry 3: 211–22. Preminger, Omer. 2011. ‘Agreement as a Fallible Operation.’ Doctoral Dissertation, MIT. Progovac, Ljiljana. 1998. ‘Determiner Phrase in a Language without Determiners.’ Journal of Linguistics 34: 165–79. Pylkkänen, Liina. 2008. Introducing Arguments. Cambridge, MA: MIT Press. Quinoli, Carlos. 2008. ‘Anaphora by Phase.’ Syntax 11: 299–329. Rackowski, Andrea & Norvin Richards. 2005. ‘Phase Edge and Extraction: A Tagalog Case Study.’ Linguistic Inquiry 36: 565–99. Radkevich, Nina. 2010. ‘On Location: The Structure of Case and Adpositions.’ Doctoral dissertation, University of Connecticut. Rappaport, Gilbert C. 2001 ‘Extraction from Nominal Phrases in Polish and the Theory of Determiners.’ Journal of Slavic Linguistics 8: 159–98. Rausch, Philip, Frank Burchert & Ria De Bleser. 2005. ‘Parallels in the Breakdown of CP And DP-Internal Movement Processes in Agrammatism: A Preliminary Case Study.’ Brain And Language 95: 129–30. Reinhart, Tanya & Eric Reuland. 1993. ‘Reflexivity.’ Linguistic Inquiry 24: 657–720. Richards, Marc. 2004. ‘Object Shift and Scrambling in North and West Germanic: A Case Study in Symmetrical Syntax.’ Doctoral dissertation, University of Cambridge. 2007. ‘Dynamic Linearization and the Shape of Phases.’ Linguistic Analysis 33: 209–37. 2008. ‘On Feature Inheritance: An Argument from the Phase Impenetrability Condition.’ Linguistic Inquiry 38: 563–72. 2011. ‘Deriving the Edge: What’s in a Phase?’ Syntax 14: 74–95. Richards, Norvin. 2001. Movement in Language: Interactions and Architectures. Oxford University Press. 2012. ‘Lardil “Case Stacking” and the Timing of Case Assignment.’ Syntax 16: 42–76.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

220

220 [207–222] 23.11.2013 10:45AM

References

Richardson, Kylie. 2007. Case and Aspect in Slavic. Oxford University Press. Ritter, Elizabeth. 1992. ‘Cross-Linguistic Evidence for Number Phrase.’ Canadian Journal of Linguistics Revue Canadienne de Linguistique 37: 197–218. 1993. ‘Where’s Gender?’ Linguistic Inquiry, 24: 795–803. Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris. Roberts, Ian & Anders Holmberg. 2010. ‘Introduction: Parameters in Minimalist Theory.’ In Parametric Variation: Null Subjects in Minimalist Theory, ed. T. Biberauer, A. Holmberg, I. Roberts & M. Sheehan, 1–57. Cambridge University Press. Rooryck, Johan. 1997. ‘On the Interaction between Raising and Focus in Sentential Complementation.’ Studia Linguistica 51: 1–49. Ross, John Robert. 1967. ‘Constraints on Variables in Syntax.’ Doctoral dissertation, MIT. Rouveret, Alain. 2012. ‘VP Ellipsis, Phases and the Syntax of Morphology.’ Natural Language And Linguistic Theory 30: 897–963. Rudin, Catherine. 1988. ‘On Multiple Questions and Multiple WH Fronting.’ Natural Language and Linguistic Theory 6: 445–501. Rutkowski, Paweł. 2002 ‘The Syntax of Quantifier Phrases and the Inherent versus Structural Case Distinction.’ Linguistic Research 7: 43–74. Rutkowski, Paweł & Ljiljana Progovac. 2005. ‘Classification Projection in Polish and Serbian.’ In Formal Approaches to Slavic Linguistics 2004, ed. S. Franks et al., 289–99. Ann Arbor: Michigan Slavic Publications. Sabbagh, Joseph. 2007. ‘Ordering and Linearizing Rightward Movement.’ Natural Language and Linguistic Theory 25: 349–401. Sag, Ivan A. 1976. ‘Deletion and Logical Form.’ Doctoral dissertation, MIT. Saito, Mamoru. 2003. ‘Derivational Approach to the Interpretation of Scrambling Chains.’ Lingua 113: 481–518. Saito, Mamoru, T. H. Jonah Lin & Keiko Murasugi. 2008. ‘“N”-Ellipsis and the Structure of Noun Phrases in Chinese and Japanese.’ Journal of East Asian Linguistics 17: 247–71. Samuels, Bridget. 2011. ‘A Minimalist Program for Phonology.’ In The Oxford Handbook of Linguistic Minimalism, ed. C. Boeckx, 574–94. Oxford University Press. Sato, Yosuke. 2011. ‘P-Stranding under Sluicing and Repair by Ellipsis: Why Is Indonesian (Not) Special?’ Journal Of East Asian Linguistics 20: 339–82. 2012. ‘Successive Cyclicity at the Syntax-Morphology Interface: Evidence from Standard Indonesian and Kendal Javanese.’ Studia Linguistica 66: 32–57. Sauerland, Uli. 2005. ‘DP Is Not a Scope Island.’ Linguistic Inquiry 36: 303–14. Sauerland, Uli & Paul Elbourne. 2002. ‘Total Reconstruction, PF Movement, and Derivational Order.’ Linguistic Inquiry 33: 283–319. Scheer, Tobias 2012. ‘Chunk Definition in Phonology: Prosodic Constituency vs. Phase Structure.’ In Modules and Interfaces, ed. M. Bloch-Trojnar & A. Bloch-Rozmej, 221–53. Lublin: Wydawnictwo KUL. Scott, Gary-John. 2002. ‘Stacked Adjectival Modification and the Structure of Nominal Phrases.’ In Functional Structure in DP and IP, ed. G. Cinque, 91–120. Oxford University Press. Seely, T. Daniel. 2006. ‘Merge, Derivation C-Command and Subcategorization in a LabelFree Syntax.’ In Miminalist Essays, ed. C. Boeckx, 182–220. Amsterdam and Philadelphia: John Benjamins. Seidl, Amanda Hallie. 2000. ‘Minimal Indirect Reference: A Theory of the Syntax– Phonology Interface.’ Doctoral dissertation, University of Pennsylvania.

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

221 [207–222] 23.11.2013 10:45AM

References

221

Selkirk, Elisabeth. 1995. ‘Sentence Prosody: Intonation, Stress and Phrasing.’ In The Handbook of Phonological Theory, ed. J. Goldsmith, 550–69. London: Blackwell. Sener, Serkan & Daiko Takahashi. 2009. Argument Ellipsis in Japanese and Turkish. University of Connecticut, Storrs, and Tohoku University. Sigurðsson, Halldór Ármann. 2002. ‘To Be an Oblique Subject: Russian vs. Icelandic.’ Natural Language and Linguistic Theory 20: 691–724. Sportiche, Dominique. 1988. ‘A Theory of Floating Quantifiers and its Corollaries for Constituent Structure.’ Linguistic Inquiry 19: 425–49. Starke, Michal. 1995. ‘On the Format for Small Clauses.’ In Syntax and Semantics 28: Small Clauses, ed. A. Cardinaletti & M. Guasti, 237–69. New York: Academic Press. Stepanov, Arthur. 2000. ‘WH-Scope Marking in Slavic.’ Studia Linguistica 54: 1–40. Stjepanovic, Sandra. 2008. ‘P-Stranding under Sluicing in a Non P Stranding Language.’ Linguistic Inquiry 39: 179–90. Stowell, Timothy. 1981. ‘Origins of Phrase Structure.’ Doctoral dissertation, MIT. 1982. ‘The Tense of Infinitives.’ Linguistic Inquiry 13: 561–70. Stoyanova, Marina. 2008. Unique Focus: Languages without Multiple Wh-Questions. Amsterdam and Philadelphia: John Benjamins. Svenonius, Peter. 1994. ‘Dependent Nexus: Subordinate Predication Structures in English and the Scandinavian Languages.’ Doctoral dissertation, University of California Santa Cruz. 2002. ‘Case Is Uninterpretable Aspect.’ Paper presented at the Perspectives on Aspect Conference, University of Utrecht. Available at www.hum.uit.no/a/svenonius/papers/ Svenonius02CUA.pdf 2004. ‘On the Edge.’ In Peripheries: Syntactic Edges and their Effects, ed. D. Adger, C. de Cat & G. Tsoulas, 261–87. Dordrecht: Kluwer. 2007a. ‘Interpreting Uninterpretable Features.’ Linguistic Analysis 33: 375–413. 2007b. ‘The Position of Adjectives and Other Phrasal Modifiers in the Decomposition of DP.’ In Adjectives and Adverbs: Syntax, Semantics, and Discourse, ed. L. McNally & Ch. Kennedy, 16–42. Oxford University Press. Swan, Oscar. 2002. A Grammar of Contemporary Polish. Bloomington, IN: Slavica Publishers. Szabolcsi, Anna. 1983. ‘The Possessor that Ran away From Home.’ The Linguistic Review 3: 89–102. 1994. ‘The Noun Phrase.’ In Syntax and Semantics 27: The Syntactic Structure of Hungarian, ed. F. Kiefer & K. Kiss, 179–274. New York: Academic Press. Szczegielniak, Adam 2008. ‘Islands And Sluicing in Polish.’ In WCCFL 27: Proceedings of the 27th West Coast Conference on Formal Linguistics, ed. N. Abner & J. Bishop, 404–12. Somerville, MA: Cascadilla Proceedings Project. Takahashi, Masahiko. 2011. ‘Some Consequences of Case-Marking in Japanese.’ Doctoral dissertation, University of Connecticut, Storrs. Tanaka, Tomoyuki & Azusa Yokogoshi. 2010. ‘The Rise of a Functional Category in Small Clauses.’ Studia Linguistica 1–32. Tellier, Christine. 1988. ‘Universal Licensing: Implications for Parasitic Gap Constructions.’ Doctoral dissertation, University of Connecticut. Thornton, Rosalind Jean. 1990. ‘Adventures in Long-Distance Moving: The Acquisition of Complex Wh-Questions.’ Doctoral dissertation, University of Connecticut. Tomaszewicz, Barbara. 2012. ‘The Syntactic Position of Slavic “by” and its Effect on Main Clause Phenomena.’ In Linguistik Aktuell/Linguistics Today 190: Main Clause

{DESKTOP}NEW FOLDER//9781107040847RFA.3D

222

222 [207–222] 23.11.2013 10:45AM

References

Phenomena, ed. L. Aelbrecht, L. Haegeman & R. Nye. Amsterdam and Philadelphia: John Benjamins. Travis, Lisa de Mena. 2010. Inner Aspect: The Articulation of VP. Dordrecht: Springer. Uriagereka, Juan. 1988. ‘On Government.’ Doctoral dissertation, University of Connecticut. 1999. ‘Multiple Spell-Out.’ In Working Minimalism, ed. S. D. Epstein & N. Hornstein, 251–82. Cambridge, MA: MIT Press. 2012. Spell-Out and the Minimalist Program. Oxford University Press. Valois, Daniel. 1991. ‘The Internal Syntax of DP.’ Doctoral dissertation, UCLA. van Riemsdijk, Henk. 1978. A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases. Lisse: Peter De Ridder Press. 1990. ‘Functional Prepositions.’ In Unity in Diversity: Papers Presented to Simon C. Dik on His 50th Birthday, ed. H. Pinkster & I. Geene, 229–42. Dordrecht: Foris. 1998. ‘Categorial Feature Magnetism: The Endocentricity and Distribution of Projections.’ Journal of Comparative Germanic Linguistics 2: 1–48. Wachowicz, Krystyna A. 1974. ‘Against the Universality of a Single Wh-Question Movement.’ Foundations of Language 11: 155–66. Watanabe, Akira. 1992. ‘Subjacency and S-Structure Movement of Wh-in-Situ.’ Journal of East Asian Linguistics 1: 255–91. Wexler, Kenneth & Peter Culicover. 1980. Formal Principles of Language Acquisition. Cambridge, MA: MIT Press. Wiland, Bartosz. 2010. ‘Overt Evidence from Polish Left Branch Extraction for Punctuated Paths.’ Linguistic Inquiry 41: 335–47. Witkoś, Jacek. 2010a. ‘Movement Theory of Control and CP-Infinitives in Polish.’ In Movement Theory of Control, ed. N. Hornstein & M. Polinsky, 45–65. Amsterdam and Philadelphia: John Benjamins. Witkoś, Jacek. 2010b. ‘On Lack of Case on the Subject of Infinitives in Polish.’ Folia Linguistica 1: 179–237. Yadroff, Michael. 1999. ‘Formal Properties of Functional Categories: The Minimalist Syntax of Russian Nominal and Prepositional Expressions.’ Doctoral dissertation, University of Indiana. Yadroff, Michael & Steven Franks. 2001. ‘The Origin of Prepositions.’ In Current Issues in Formal Slavic Linguistics, ed. G. Zybatow, U. Junghanns, G. Mehlhorn, & L. Szucsich, 69–79. Frankfurt am Main: Peter Lang. Zagona, Karen. 1988. ‘Proper Government of Antecedentless VP In English and Spanish.’ Natural Language and Linguistic Theory 6: 95–128. Zamparelli, Roberto. 1996. ‘Layers in the Determiner Phrase.’ Doctoral dissertation, University of Rochester. Zubizarreta, Maria Luisa & Jean-Roger Vergnaud. 2005. ‘Phrasal Stress and Syntax.’ In The Blackwell Companion to Syntax, ed. M. Everaert & H. Van Riemsdijk, 522–68. Malden, MA: Blackwell. 1998. Focus, Prosody, and Word Order. Cambridge, MA: MIT Press. Zwart, C. Jan-Wouter. 1993. ‘Dutch Syntax: A Minimalist Approach.’ Doctoral dissertation, Groningen: Universiteitsdrukkerij. 1997. Morphosyntax of Verb Movement: A Minimalist Approach to the Syntax of Dutch. Dordrecht: Kluwer.

{DESKTOP}NEW FOLDER//9781107040847IND.3D

223 [223–228] 23.11.2013 10:52AM

Index

Abels, Klaus, 5, 37, 77, 89, 138, 139, 145, 146, 147, 148, 168, 173, 175 Abney, Steven P., 108, 139, 140 Abraham, Werner, 139, 140 Adger, David, 6, 7, 12, 15, 21, 47, 143, 185, 191, 192, 193, 194, 195, 196 Aelbrecht, Lobke, 64 Afrikaans, 80, 99 Agree, 8, 18, 20–22, 31, 34, 35, 38 Long Distance Agree, 37–41 Multiple Agree, 21, 122, 128 Agüero-Bautista, Calixto, 100, 101 Aissen, Judith, 115 Aldridge, Edith, 95, 97, 98 Alexiadou, Artemis, 109, 167 Anagnostopoulou, Elena, 167 antecedent contained deletion, 102–104, 107, 136 antilocality, 77, 105, 156, 173 applicatives and passive movement, 155–157 high as phases, 169 high versus low, 153–154 Attract Closest, 37, 155 Bachrach, Asaf, 182, 183, 184 Bailyn, John Frederick, 126, 127, 128, 152, 159, 179 Baker, Mark C., 123, 128, 139, 176, 208 Baltin, Mark R., 103 Bański, Piotr, 51 Bare Output Conditions, 8, 14 Bare Phrase Structure, 11, 43, 80

Barros, Matthew, 63 Barss, Andrew, 4, 66, 74, 124, 152, 200 Belletti, Adriana, 35, 36, 208 Berber, 73 Bernstein, Judy, 108, 109 Bhatt, Rajesh, 37, 38 Biberauer, Theresa, 176 binding, 74, 75, 90, 112, 142, 148, 149 Biskup, Petr, 6, 66, 142, 200, 202, 203 Bittner, Maria, 109 Bobaljik, Jonathan, 36, 70, 77, 174 Boeckx, Cedric, 1, 2, 3, 12, 37, 67, 69, 71, 73, 181 Booij, Geert, 51 Borer-Chomsky conjecture, 74, 176 Bošković, Željko, 19, 37, 40, 53, 54, 55, 62, 65, 70, 73, 77, 79, 90, 91, 115, 138, 149, 150, 171–173, 174–175, 179, 184, 199 Bowers, John, 124 Bresnan, Joan W., 190, 191 Bruening, Benjamin, 67, 82, 102, 103, 136 Bukusu, 152, 155 Bulgarian, 73 Büring, Daniel, 148 Burzio’s Generalization, 92 Butler, Jonny, 200 Cable, Seth, 95 Calabrese, Andrea, 73 Carnie, Andrew, 66, 124, 200 Carstens, Vicki, 4, 50, 51, 84, 121, 123 case Accusative, 17, 91

223

{DESKTOP}NEW FOLDER//9781107040847IND.3D

224

224 [223–228] 23.11.2013 10:52AM

Index

case (cont.) case concord, 113, 121, 174 Dative, 114 Genitive, 35, 111, 173 Instrumental versus Nominative, 126–128 Locative, 142 Nominative, 17, 19, 21, 108, 167 quirky, 35, 47 valuation, 21, 121–123, 127 Castillo, Juan Carlos, 90 Ćavar, Damir, 79 Cecchetto, Carlo, 12, 13, 67, 177 Chamorro, 97, 98, 115 Charlow, Simon, 117, 119 Cheng, Lisa, 19 Chi-Mwhi:ni:, 198 Chinese, 95 Chomsky, Noam, 1, 2, 3, 7, 8, 10, 11, 12, 13, 14, 16, 18, 19, 20, 21, 23, 28, 29, 30, 31, 32, 33, 43, 47, 48, 49, 50, 54, 62, 66, 67, 74, 80, 104, 108, 111, 112, 125, 148, 163, 168, 176, 189 Chukchee, 37, 38 Chung, Sandra, 64, 97, 98, 211 Cinque, Guglielmo, 109, 137, 139, 140, 142, 189, 191 Citko, Barbara, iv, 10, 12, 13, 22, 64, 95, 123, 126, 127, 128, 129, 158, 171, 182, 184, 211 Clause Partitioning, 203 Cole, Peter, 98 Collins, Chris, 13 complementizer agreement, 4, 50–51, 83–85 conceptual-intentional system, 8 Condition on Extraction Domains, 69 control, 88 exhaustive, 39 obligatory, 39 partial, 39 Copy Theory of Movement, 80 Corver, Norbert, 115, 172 Craenenbroeck, Jeroen van, 52, 53, 64, 145 Culicover, Peter W., 101, 102, 184, 222 cyclic linearization, 6, 185, 188 Danon, Gabi, 16, 121 Davies, William D., 114 Dayal, Veneeta, 4, 82 Den Dikken, Marcel, 3, 4, 5, 70, 100, 106, 124, 136, 138, 142, 160, 161–166, 167, 169, 199

Diercks, Michael, 4, 50, 51 Diesing, Molly, 200, 201–203 distributed deletion, 79 Distributed Morphology, 8, 104 Dobashi, Yoshihiko, 185 Dogil, Grzegorz, 51 Donati, Caterina, 12, 13 Dornisch, Ewa, 95 double objects, 30, 123, 151, 159, 164, See also applicatives DP Hypothesis, 170–171 Parametrized DP Hypothesis, 174 Universal DP Hypothesis, 171 Drummond, Alex, 138 Drury, John, 90 Dubinsky, Stanley, 114 Dutch, 144 Dyakonova, Marina, 152, 158, 159 Dziwirek, Katarzyna, 54 Elbourne, Paul, 180, 181 ellipsis, 60, 64, 65, 88, 93–94, 136, 150, 159–160, 169, 174 nominal ellipsis, 65, 120–121 pseudogapping, 88, 94 sluicing, 85, 145 swiping, 145 verb phrase ellipsis, 64, 88, 93, 102 Elordieta, Gorka, 185 Embick, David, 51 Emonds, Joseph, 116, 135, 139, 144 Enç, Mürvet, 17 Engdahl, Elisabet, 116 Epstein, Samuel D., 18, 29, 37, 50, 59, 70, 90, 199 equidistance, 47, 48, 165 exceptional case marking, 54–55, 88 expletive subject, 24, 36 extraction restriction, 98 Fanselow, Ginsbert, 79 Feature Inheritance, 3, 46–47, 49–50, 55–57, 85, 88, 133, 170, 176 and Donate, 52 and Keep, 52, 85 and Share, 52, 85 from Above, 53 from Below, 53 underinheritance, 52

{DESKTOP}NEW FOLDER//9781107040847IND.3D

225 [223–228] 23.11.2013 10:52AM

Index features EPP feature, 32, 47, 69, 75, 167 feature valuation, 18, 68, 110, 111–112 interpretability versus valuation, 19–20 interpretable, 14, 15 interpretable versus uninterpretable, 14–15 uninterpretable, 3, 14, 18, 30, 31, 46–47, 60, 110, 176 Felser, Claudia, 4, 6, 67, 71, 78, 79, 80, 81, 82, 177 Finnish, 17 Fox, Danny, 6, 74, 75, 90, 100, 101, 185, 187, 188, 219 Frampton, John, 9, 19 Franks, Steven, 51, 79, 140, 184, 209 Frascarelli, Mara, 1 Frederici, Angela, 139 French, 17 Froud, Karen, 139 Gallego, Ángel J., 1, 3, 5, 12, 13, 31, 68, 69, 160, 161, 166, 167, 169, 176 Gavruseva, Elena, 115 Gengel, Kirsten, 64 German, 17, 79, 82, 108, 203 Goal, 20, 21, 22, 38, 40, 110, 122 Green, Georgia M., 151 Grimshaw, Jane, 139 Grohmann, Kleanthes, 1, 2, 3, 7, 32, 37, 69, 77, 90 Gutmann, Sam, 9, 19 Haegeman, Liliane, 4, 50, 51, 52, 53, 85 Hale, Ken, 66, 109 Halle, Morris, 8, 189 Hankamer, Jorge, 64 Harves, Stephanie, 124, 126, 129 Hermon, Gabriella, 98 Heycock, Caroline, 75, 214 Hindi-Urdu, 38, 83, 95, 96 Hiraiwa, Ken, 5, 21, 108, 122 Holmberg, Anders, 48, 64, 187, 208 Holmberg’s Generalization, 48, 187 Hornstein, Norbert, 7, 23, 32, 78, 101, 103, 138 Horstein, Norbert, 136 Horvath, Julia, 82 Hovav, Malka Rappaport, 58 Huang, Cheng-Teh James, 69, 112

225

Hungarian, 82, 95, 108, 114 Iatridou, Sabine, 181 Icelandic, 35, 36, 47, 149 Inclusiveness Condition, 12, 13 Indonesian, 97, 98 interface conditions, 8 inverse scope, 118, 174 Irish, 73, 84 Ishihara, Shinichiro, 185, 191, 198 islands, 68, 73–74, 184 Italian, 73 Jackendoff, Ray, 120, 121, 138 Japanese, 65, 152, 180, 198 Jayaseelan, Karattuparambil, 95 Jeong, Youngmi, 156 Johnson, Kyle, 64, 128 Jonas, Dianne, 36 Kahnemuyipour, Arsalan, 6, 185, 189, 190, 195–197 Kasai, Hironobu, 101 Kashmiri, 96 Katz, Jerrold, 80 Katzir, Roni, 182, 183, 184 Kayne, Richard, 43, 54, 80, 118, 138, 139, 185, 191 Kennedy, Christopher, 103 Keyser, Jay, 66 Khomitsevich, Olga, 55 Kilega, 51, 84 Kinande, 198 Kinyalolo, Kasangati Kikuni Wabongambilu, 84 Kinyarwanda, 155 Kiparsky, Paul, 17 Kiss, Katalin É., 95 Koopman, Hilda, 142 Koppen, Marjo van, 4, 85 Kratzer, Angelika, 6, 185, 191 labels, 12–13 labeling algorithm, 12 Laenzlinger, Christopher, 109 Lahne, Antje, 4, 59, 67, 70, 71, 79, 82, 97, 100 Landau, Idan, 35, 38, 39, 40, 41 Larson, Richard, 101, 102, 103, 117, 118, 119 Lasnik, Howard, 36, 54, 55, 138, 152

{DESKTOP}NEW FOLDER//9781107040847IND.3D

226 [223–228] 23.11.2013 10:52AM

226

Index

Late Vocabulary Insertion, 8 Lebeaux, David, 74, 75, 90 Left Branch Condition, 172 left branch extraction, 99, 134–135, 172–173 Legate, Julie Anne, 4, 16, 31, 52, 53, 71, 100, 103, 104, 106, 107, 155, 190, 194 Levin, Beth, 58 Lexical Array, 8, 28 subarray, 28, 29 Limburgian, 85 Linear Correspondence Axiom, 43, 80, 185 Lobeck, Anne, 64, 88, 120, 121 locality, 21, 23, 47, 69, 155 long distance wh-movement, 72–73, 75, 83, See also successive cyclic movement Lubukusu, 51 Luganda, 197 Lutz, Uli, 4, 82, 217 MacDonald, Jonathan Eric, 195 Mahajan, Anoop Kumar, 38, 83 Manetta, Emily, 4, 82, 95, 96, 97 Mapping Hypothesis, 200–201 Marantz, Alec, 8, 104, 152 Martin, Roger Andrew, 55 Marušič, Franc Lanko, 5, 42, 119, 177, 179, 180, 182 Matushansky, Ora, 5, 37, 59, 61, 66, 110, 113, 117, 119, 121, 122, 124, 126, 129, 177 May, Richard, 118 May, Robert, 67, 103, 117, 181 Mayalalam, 95 McCloskey, James, 4, 76, 77, 83, 84, 98, 99, 100, 149 McDaniel, Dana, 4, 80, 82 McGinnis, Martha, 1, 5, 151, 155, 157, 170, 197, 198 Megerdoomian, Karine, 17 Merchant, Jason, 64, 65, 85, 86, 88, 93, 94 Merge, 8 External Merge, 10–12 Internal Merge, 13–14 Parallel Merge, 10, 22 Merge over Move, 2, 23–25 Migdalski, Krzysztof, 51 Minimal Link Condition, 155 Miyagawa, Shigeru, 31, 152, 179, 198 Moro, Andrea, 132, 166

Müller, Gereon, 3, 33, 37, 59, 69, 70, 199 multidominance, 101, 183 Munn, Alan, 17 Muysken, Pieter, 139 negative polarity item licensing, 101, 113, 136 Nespor, Marina, 188 Nissenbaum, Jonathan, 101, 102 nuclear scope, 66 nuclear stress, 60, 189–191 and phases, 191–193 Nuclear Stress Rule, 189, 190, 191, 195 null subject languages, 167 Numeration, 9, 28 Nunes, Jairo, 7, 32, 90, 101 Oehrle, Richard T., 151 order preservation, 185–188 Ouali, Hamid, 52, 85 Pair List readings, 100 Pak, Marjorie, 185, 191, 197 parasitic gaps, 95, 101–102, 113, 117 partial wh-movement, 76, 81, 82, See also scope marking passives and unaccusatives and nuclear stress, 190, 193–194 as phases, 104–107 Pereltsvaig, Asya, 126, 171 Persian, 190, 196 Pesetsky, David, 6, 12, 17, 19, 20, 21, 35, 53, 55, 90, 167, 185, 187, 188, 219 Peterson, David A., 152, 155 Phase Condition, 68, 167 Phase Extension, 5, 161–166 Phase Impenetrability Condition, 3, 4, 31–34, 41, 49, 98, 130 and locality, 37–41 and phasehood diagnostics, 67–68 PIC1 versus PIC2, 33, 36–37, 47, 132, 161 Strong PIC, 33 Weak PIC, 33 Phase Sliding, 5, 161, 166–168 Picallo, M. Carme, 109 Pinker, Steven, 9 Plessis, Hans Du, 78, 80 Polinsky, Maria, 37, 38

{DESKTOP}NEW FOLDER//9781107040847IND.3D

227 [223–228] 23.11.2013 10:52AM

Index Polish, 16, 35, 51, 55, 79, 82, 83, 86, 87, 95, 99, 111, 113, 121, 126, 129, 134, 140, 141, 145, 149, 150, 158, 167, 172, 173, 174, 203 possessor extraction, 115 Postal, Paul, 78, 80, 101, 215 Potsdam, Eric, 37, 38 predication phrases, 124–125 extraction from, 129–131 Preminger, Omer, 9 preposition stranding, 78, 85, 86, 99, 143–144, 145–148, 168, 175 prepositions directional, 141–143, 145 functional, 140, 145 lexical, 140, 145 locative, 141–143, 145 Probe, 12, 20, 21, 22, 48, 84, 110, 122, 126, 127, 139, 206 Progovac, Ljiljana, 171 Pylkkänen, Liina, 151, 152, 153, 154, 170 quantifier float. See quantifier stranding quantifier lowering, 202 quantifier raising, 66, 67, 102, 107, 117–119, 136, 150, 181 quantifier stranding, 76–77, 90, 98, 133–134 Quinoli, Carlos, 112 Rackowski, Andrea, 97, 99 Radkevich, Nina, 5, 142, 148, 149 Ramchand, Gillian, 143 Rappaport, Gilbert C., 171 Rausch, Philip, 108 reconstruction, 4, 74–75, 90, 100–101, 106, 179 Reinhart, Tanya, 148, 149 Relator Phrase, 164 restrictive clause, 66 Reuland, Eric, 148, 149 Richards, Marc, 3, 18, 33, 36, 37, 41, 49, 50, 53, 60, 162 Richards, Norvin, 1, 73, 97, 99 Richardson, Kylie, 126 right node raising, 62, 182–184 Ritter, Elizabeth, 109 Rizzi, Luigi, 35, 73, 137 Roberts, Ian, 128 Romani, 80 Romanian, 73

227

Rooryck, Johan, 54 Ross, John Robert, 172 Rouveret, Alain, 64 Rubach, Jerzy, 51 Rubin, Edward, 126 Rudin, Catherine, 73 Russian, 17, 82, 83, 126, 152 Rutkowski, Paweł, 171 Sabbagh, Joseph, 184 Sag, Ivan, 64 Saito, Mamoru, 179, 180 Samuels, Bridget, 185 Sato, Yosuke, 86, 97, 98 Sauerland, Uli, 117, 119, 180, 181 scattered deletion, 79 Scheer, Tobias, 185 scope, 184, 198–199 scope marking, 4, 81–83, 96 Scott, Gary-John, 109 Scottish Gaelic, 143, 190 Seely, T. Daniel, 13, 18, 50, 59, 70, 90, 199 Seidl, Amanda Hallie, 197 Selkirk, Elisabeth, 6, 185, 190, 191 semantic types, 66, 199 Şener, Serkan, 65 sensorimotor system, 8 sentence fragments, 61–62 sequence of tense, 55 Serbo-Croatian, 172 short wh-movement, 95 Shortest Link, 155 Sichel, Ivy, 181 Sigurðsson, Halldór Ármann, 35, 47 Single Answer readings, 101 Slovak, 149 small clauses, 124, 131 Somali, 73 Spell-Out, 60, 61 Delayed Spell-Out, 182 Multiple Spell-Out, 1, 3, 41–43, 177, 185 Non-Simultaneous Spell-Out, 43 Spellee, 195 Spell-Out domain, 32, 161, 185, 195 Sportiche, Dominique, 77 Starke, Michal, 134 Stepanov, Arthur, 4, 82, 83

{DESKTOP}NEW FOLDER//9781107040847IND.3D

228 [223–228] 23.11.2013 10:52AM

228 Stjepanovic, Sandra, 86 Stowell, Timothy, 55, 88 Stranding Generalization, 145 Strong Minimalist Thesis, 8 subject to object raising. See exceptional case marking successive cyclic movement, 4, 32, 57, 67, 70, 75–76, 78, 80, 83, 88–91, 94 Svenonius, Peter, 5, 15, 17, 47, 109, 124 Swan, Oscar, 111 Szabolcsi, Anna, 108, 114, 115 Szczegielniak, Adam, 86 Tagalog, 97 Takahashi, Daiko, 65, 179 Tamazight Berber, 52 Tanaka, Tomoyuki, 124, 125, 131, 132, 133 Tellier, Christine, 108, 116 Thornton, Rosalind Jean, 80 Tomaszewicz, Barbara, 129 Torrego, Esther, 12, 17, 19, 20, 21, 53, 167 Transfer, 19, 30, 41–42, 177, 180 Multiple Transfer, 178, 180 Non-Simultaneous Transfer, 179, 180, 181 Simultaneous Transfer, 177 transitive expletive, 36 Travis, Lisa, 195 Tsez, 37, 38 Tsujioka, Takae, 152 Turkish, 17

Index Tzez, 38 Tzotzil, 115 Uriagereka, Juan, 1, 43, 115, 172 Valois, Daniel, 108, 109, 113, 115, 116 Value-Transfer Simultaneity, 49 van Riemsdijk, Henk, 139, 140, 143, 144 variable binding, 90, 100, 101, 106 Vergnaud, Jean-Roger, 189, 190 Vicente, Luis, 63 Vogel, Irene, 188 voice mismatches, 94 Wachowicz, Krystyna A., 73 Weinberg, Amy, 78 West Flemish, 51 West Ulster English, 76, 77, 99, 149 Wexler, Kenneth, 184 wh-agreement, 98 wh-copying, 4, 6, 76, 79, 80–81 Wiland, Bartosz, 4, 99 Wurmbrand, Susi, 70, 174 Yadroff, Michael, 139, 140 Yokogoshi, Azusa, 124, 125, 131, 132, 133 Zagona, Karen, 64 Zamparelli, Roberto, 109 Zwart, C. Jan-Wouter, 4