Anti-contiguity: A Theory of Wh- Prosody 9780197509739

Citation preview

Anti-​contiguity

ii

oxford studies in comparative syntax Richard S. Kayne, General Editor Comparative Approaches to Adpositional Phrases Edited by Jacopo Garzonio and Silva Rossi

Japanese Syntax in Comparative Perspective Edited by Mamoru Saito

Locality and Logophoricity: A Theory of Exempt Anaphora Isabelle Charnavel

Cross-​Linguistic Studies of Imposters and Pronominal Agreement Edited by Chris Collins

The Linker in the Khoisan Languages Chris Collins

Locality Edited by Ian Roberts and Enoch Aboh

The Syntactic Variation of Spanish Dialects Edited by Angel J. Gallego

Aspects of Split Ergativity Jessica Coon

Questions of Syntax Richard S. Kayne

Variation in Datives Edited by Beatriz Fernández and Ricardo Etxepare

Exploring Nanosyntax Edited by Lena Baunaz, Liliane Haegeman, Karen De Clercq, and Eric Lander Dravidian Syntax and Universal Grammar K. A. Jayaseelan and R. Amritavalli

Adverbial Clauses, Main Clause Phenomena and Composition of the Left Periphery: The Cartography of Syntactic Structures, Volume 8 Liliane Haegeman

Functional Heads: The Cartography of Syntactic The Morphosyntax of Portuguese and Spanish in Structures, Volume 7 Latin America Edited by Mary A. Kato and Francisco Ordóñez Edited by Laura Brugé, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro, Cecilia Poletto Deconstructing Ergativity: Two Types of Ergative Discourse-​Related Features and Functional Languages and Their Features Projections Maria Polinsky Silvio Cruschina Rethinking Parameters Mapping the Left Periphery: The Cartography of Edited by Luis Eguren, Olga Fernández-​ Syntactic Structures, Volume 5 Soriano, and Amaya Mendikoetxea Edited by Paola Benincà and Nicola Munaro Argument Licensing and Agreement The Grammar of Q: Q-​Particles, Wh-​Movement, Claire Halpert and Pied-​Piping The Cartography of Chinese Syntax: The Seth Cable Cartography of Syntactic Structures, Volume 11 Comparisons and Contrasts Edited by Wei-​Tien Dylan Tsai Richard S. Kayne Beyond Functional Sequence: The Cartography of Mapping Spatial PPs: The Cartography of Syntactic Structures, Volume 10 Syntactic Structures, Volume 6 Edited by Ur Shlonsky Edited by Guglielmo Cinque and Luigi Rizzi The Architecture of Determiners The Syntax of Ellipsis: Evidence from Dutch Thomas Leu Dialects A Comparative Grammar of Borgomanerese Jeroen van Craenenbroeck Christina Tortora Scrambling, Remnant Movement and Chinese Syntax in a Cross-​linguistic Perspective Restructuring in West Germanic Edited by Y.-​H. Audrey Li, Andrew Simpson, Roland Hinterhölzl and W.-​T. Dylan Tsai Functional Heads: The Cartography of Syntactic Functional Structure from Top to Toe: The Structures, Volume 7 Cartography of Syntactic Structures, Volume 9 Edited by Laura Brugé, Anna Cardinaletti, Edited by Peter Svenonius Giuliana Giusti, Nicola Munaro, Cecilia Poletto Micro-​Syntactic Variation in North American English Edited by Raffaella Zanuttini and Laurence R. Horn

Anti-​c ontiguity A Theory of wh-​ Prosody Jason Kandybowicz

1

iv

1 Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and certain other countries. Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America. © Oxford University Press 2020 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above. You must not circulate this work in any other form and you must impose this same condition on any acquirer. Library of Congress Cataloging-​in-​Publication Data Names: Kandybowicz, Jason, author. Title: Anti-​contiguity : a theory of wh-​prosody /​Jason Kandybowicz. Description: New York : Oxford University Press, 2020. | Series: Oxford studies comparative syntax series | Includes bibliographical references and index. Identifiers: LCCN 2020015059 (print) | LCCN 2020015060 (ebook) | ISBN 9780197509739 (hardback) | ISBN 9780197509746 (paperback) | ISBN 9780197509760 (epub) Subjects: LCSH: Prosodic analysis (Linguistics) |Grammar, Comparative and general—​Syntax. |Grammar, Comparative and general—​Phonology. | West African languages—​Syntax. | West African languages—​Phonology. Classification: LCC P224 .K36 2020 (print) | LCC P224 (ebook) | DDC 414/​.6—​dc23 LC record available at https://​lccn.loc.gov/​2020015059 LC ebook record available at https://​lccn.loc.gov/​2020015060 9 8 7 6 5 4 3 2 1 Printed by Marquis, Canada Hardback printed by Bridgeport National Bindery, Inc., United States of America

For John and Lillian Kandybowicz. None of this would have been possible without them.

L I S T O F   A B B R E V I AT I O N S U S E D I N   N O N -​C I T E D  D ATA

C Complementizer cldet Clausal determiner comp Complementizer cop Copula dem Demonstrative dist Distal foc Focus marker fut Future tense H High tone ι Intonational Phrase infin Infinitive marker L Low tone LH Rising tone (e.g., L+H) M Mid tone φ Phonological Phrase pl Plural prf Perfect pres Present tense prox Proximal prt Particle pst Past tense q Question marker rel Relative clause marker sg Singular top Topic marker ω Prosodic word

ACKNOWLEDGMENTS

I owe great debt of gratitude to very many people for helping me realize the completion of this book. Thanks first and foremost to all the native speaker consultants I worked with over the years, both in West Africa and in the United States:  Daniels Ananey Adonae; Seth Arthur; Emmanuel Baidoo; Mark Nsekou Denteh; Matthew Donkor; Simon Fofie; Cecelia Gyameah; Abdul Kadir Kawu; Ahmadu Ndanusa Kawu; Alhaji Usman Kawu; Hadizat Nnǎkó Kawu; Sulaiman Ilorin Kawu; Elizabeth Kolo; Joseph Agyei Korboe; Abubakar Bello Mohammed; David Opoku; and Peter Owusu-​Opoku. For invaluable logistical support in Ghana, I thank Prof. Kofi Agyekum; Prof. Akosua Anyidoho; Mark Dundaa and the Ghana Institute for Linguistics, Literacy & Bible Translation (GILLBT); Mr. Kwame Opoku; and Mrs. Mary Opoku. For helpful feedback on earlier versions of the material in this manuscript, I am grateful to the following colleagues: Byron Ahn; Sam Al Khatib; Athulya Aravind; Michael Barrie; Jason Bishop; Bronwyn Bjorkman; Dianne Bradley; Leston Buell; Chris Collins; Amy Rose Deal; Michael Diercks; Marcel den Dikken; Gorka Elordieta; Carolina Fraga; Robert Freidin; Claire Halpert; Larry Hyman; Sharon Inkelas; Peter Jenks; Allard Jongman; Sun-​Ah Jun; Laura Kalin; Ivona Kučerová; Line Mikkelson; Michael Marlo; Norvin Richards; Sharon Rose; Joey Sabbagh; Ken Safir; Lisa Selkirk; Joan Sereno; Harold Torrence; Lisa Travis; Tonjes Veenstra; Coppe von Urk; Michael Wagner; Jie Zhang; Maria-​Luisa Zubizarreta; and three anonymous Linguistic Inquiry reviewers from 2011–​2014 who helped me improve the substance and presentation of the material in Chapter 2. Some of the material in this book was developed and presented in the fall of 2018 in my “Wh-​Prosody” seminar at the Graduate Center, City University of New York. I thank the following students from that class for their insightful comments, questions, and valuable input:  Enas Albasiri;

xvi

André Eliatamby; Carolina Fraga; Yeonju Lee-​Sikka; Armando Tapia; and Matt Stuck. The content in this book was presented in various stages of development at a number of conferences and talks. I’m grateful to the following audiences for their comments and suggestions:  Annual Conference on African Linguistics 42 and 44; 2018 Annual Meeting of the Linguistic Society of America; Berkeley Linguistic Society 40; 2011 DFW Metroplex Linguistics Conference; The Graduate Center, City University of New York; 20th International Congress of Linguists; McGill University; North East Linguistics Society 48; Princeton Symposium on Syntactic Theory 1; University of Kansas; University of Texas at Arlington; and West Coast Conference on Formal Linguistics 29. Last, but certainly not least, I want to thank the many helpful people on the OUP side of things. Thanks to Richie Kayne for supporting this project and facilitating its publication. Thanks to Hallie Stebbins, Meredith Keffer, and Macey Fairchild for their vital editorial assistance. And finally, thanks to two anonymous Oxford University Press reviewers for their excellent comments and suggestions. While I have made an attempt to incorporate their feedback and address their concerns, any and all errors, oversights, and inconsistencies are my own.

[ xiv ] Acknowledgments

Map of Languages Covered in This Book

CHAPTER 1

Introduction

S

ometimes “syntax” is syntax. Sometimes “syntax” is phonology. This book is a case in point. In recent work, Richards (2010, 2016)  has advanced a novel theory of the distribution of wh-​items. On his theory, prosody and syntax are derivationally entangled, interacting and conspiring to determine whether a given interrogative expression will move or remain in-​situ. Richards’s approach represents a departure from the standard generative conception of the architecture of grammar in which the relationship between syntax and phonology is assumed to be unidirectional. According to the classical view, the properties of phonological structures in certain domains are either determined or at least constrained by syntactic properties. This conception precludes causal interactions in the opposite direction, that is, derivations in which phonological considerations drive phenomena at the syntactic level. For Richards, however, prosodic structure building occurs much earlier in the derivation than typically assumed (i.e., before Spell-​ Out) and exerts a causal influence on syntactic operations like movement. Richards proposes a universal PF well-​formedness condition that requires wh-​items to prosodically phrase with the complementizers (null or overt) that mark their scope. When prosody and syntax build structures in which the two items are prosodically contiguous (i.e., contained within a single macro Phonological Phrase), wh-​movement is unnecessary and wh-​ in-​ situ results. When this manner of contiguity is not possible, wh-​ movement becomes the vehicle enabling prosodic grouping with C. If Richards’s characterization of wh-​movement is correct, there is a very real sense that

Anti-contiguity. Jason Kandybowicz, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780197509739.001.0001

2

some of what we have narrowly labeled as “syntax” is just as much a matter of phonology/​prosody. Richards’s work has inspired a new wave of research dedicated to exploring the link between wh-​syntax and prosody. It is in this spirit that this book is framed. This book is comparative in nature and draws empirically on wh-​phenomena in five West African languages:  Krachi (North Guang:  Ghana), Bono (Tano:  Ghana), Wasa (Tano:  Ghana), Asante Twi (Tano: Ghana), and Nupe (Benue-​Congo: Nigeria). In Chapter 2, I show that facts concerning the syntax and prosody of wh-​items in Krachi (a) argue forcefully against Richards’s conception of the entanglement of prosody and syntax and (b) demonstrate that Richards’s PF constraint forcing wh-​ and C to be prosodically contiguous cannot be correct for all languages. These conclusions arise through a careful examination of wh-​ in-​situ in embedded contexts, a syntactic environment that Richards largely does not consider. In place of Richards’s proposal, I develop a competing theory of wh-​prosody in Chapter 3, which although in opposition to Richards’s theory in detail, is kindred in spirit to Richards’s basic idea—​I propose that the distribution of wh-​is indeed regulated by a prosodic wellformedness condition at PF, but that it is an anti-​contiguity constraint. I  propose that Universal Grammar constrains wh-​items from forming prosodic constituents with overt complementizers at the level of Intonational Phrase—​in a nutshell, wh-​and overt C must not phrase together contiguously within an Intonational Phrase. In what follows, I briefly outline the book’s core proposal and results. Chapter  2 focuses entirely on the syntax and prosody of wh-​ in-​situ in Krachi. In this language, wh-​in-​situ is available both in root contexts and in embedded complement clauses. Under Richards’s theory, embedded in-​situ wh-​items must find themselves in a prosodically contiguous structural relation with matrix C under a single enveloping Phonological Phrase. I show, however, that in Krachi this is not possible without violating the Strict Layer Hypothesis and invoking discontinuous prosodic constituents. This is so because in Krachi, embedded complementizers introduce obligatory Intonational Phrase boundaries. This means that wh-​ in-​situ-​containing embedded TPs in the language are mapped prosodically as independent Intonational Phrases. As a result, there is no way to group embedded wh-​ items with matrix C under a single non-​discontinuous Phonological Phrase while respecting the Prosodic Hierarchy. Richards’s Contiguity-​theoretic approach to wh-​in-​situ is therefore fatally flawed, at least for languages like Krachi. Chapter  3 develops my Anti-​ contiguity wh-​prosody proposal. The theory is born from an investigation of the distribution of wh-​ in-​situ in

[ 2 ] Anti-contiguity

four closely related Tano languages (Krachi, Bono, Wasa, and Asante Twi). In all four languages, wh-​in-​situ is possible, but only in two of them is wh-​in-​situ permitted in embedded contexts. Krachi and Bono both allow wh-​in-​situ in embedded complement clauses, while Wasa and Asante Twi do not. I argue that the generalization underpinning this variation in Tano concerns a prohibition on wh-​items phrasing with overt C at the level of Intonational Phrase (ι). I call this constraint the anti-​contiguity of wh-​ and c (ACWC). I show that the ability of a wh-​item to appear in-​situ in these languages correlates with the prosodic status of its immediately containing clause. Embedded complement clauses are parsed as independent ι units in Krachi and Bono, but not in Wasa and Asante Twi. Thus, ι boundaries divide C from embedded interrogatives in Krachi and Bono, preventing the items from forming a prosodic constituent at the level of ι. Conversely, no such boundaries intervene between embedded C and wh-​ in Wasa and Asante Twi, yielding prosodic mappings in which the items phrase together. Consequently, embedded wh-​in-​situ is prosodically licit in Krachi and Bono, but not in Wasa and Asante Twi. In this way, the Tano pattern of wh-​in-​situ variation described earlier reduces to a minimal difference in how syntactic structures are externalized via prosodic mapping. In Chapter 4, I provide additional support for the ACWC theory of wh-​ prosody by showing how the proposal advanced in Chapter 3 explains two surprising and mysterious asymmetries of Nupe wh-​syntax. One asymmetry concerns the distribution of wh-​in-​situ in the language. Nupe is not a wh-​in-​situ language, but like English, it allows non-​superior wh-​items to remain in-​situ in multiple wh-​questions in root clause contexts. However, in embedded clauses (which are obligatorily introduced by overt C) the second wh-​item of a multiple wh-​question may not appear in-​situ. In this way, Nupe is similar to Wasa and Asante Twi—​all three languages permit wh-​in-​situ in matrix clauses, but disallow it in embedded domains. The second asymmetry explored in this chapter concerns the availability of embedded non-​interrogative focus and the impossibility of embedded interrogative focus. In Nupe, neither partial wh-​movement nor embedded wh-​questions are possible—​where other languages would tolerate partial wh-​movement, Nupe employs long-​distance wh-​movement into the matrix clause; where other languages would employ indirect wh-​ question syntax, Nupe deploys non-​interrogative relative clause structures. I argue that both of these asymmetries are explainable in terms of the ACWC’s ban on contiguous wh-​phrasing with overt C at the ι level. To establish this, I  show that like Wasa and Asante Twi, overt embedded C does not introduce an ι boundary in Nupe. That is, Nupe multi-​clausal structures may contain a number of nested Phonological Phrases, but only one ι—​the

Introduction 

[ 3 ]

4

entire utterance. As a consequence, no ι boundary insulates the focused embedded wh-​item from the overt embedding complementizer, therefore running afoul of the ACWC. The two Nupe wh-​asymmetries, then, represent additional cases where what initially appears to be “syntactic” on the surface, turns out more accurately to be prosodic in nature. Here, too, we find that sometimes “syntax” is phonology. Chapter  5 concludes the book by considering the ACWC in a crosslinguistic context. Can the proposal be successfully applied to derive asymmetries in wh-​in-​situ distribution beyond the West African languages considered in Chapters 2–​4? Chapter 5 considers how well the ACWC fares in other languages with wh-​in-​situ asymmetries. I  concentrate on thirteen languages from genetically diverse groupings (Romance, Bantu, and Indo-​Aryan languages, among others) and consider the implications of these case studies for the formulation of the ACWC. Two discoveries are made. One, the ACWC is equipped to explain a diverse range of wh-​ in-​situ asymmetries that it was not designed to account for (and it does so, in my opinion, in a rather satisfying way). Two, the formulation of the ACWC from Chapters 3–​4 must be revised to allow for parameterization. This book is intended to represent the first step in articulating and developing the Anti-​contiguity prosodic constraint described earlier. As with any new theory, a number of questions remain (e.g., Why does the constraint exist? Can it be derived from anything deep relating to language design? Is it universal or merely a parameter, active in some, but not necessarily in other/​most languages?). It is not the goal of this book to answer these bigger questions. Rather, the broad goal of the book is to stimulate the research program begun by Richards’s investigation of the prosodic influences on wh-​distribution. The narrow goal is to develop, motivate, and explore the precise formulation of the ACWC. For this reason, my intention was for this book to be compact and pithy—​something to spawn new research, invite challenges, and initiate projects that will either confirm or disconfirm the claims presented within.

[ 4 ] Anti-contiguity

CHAPTER 2

Prosodic Entanglement and the Anti-​contiguity of Wh-​  and C

2.1. PROSODIC ENTANGLEMENT AND THE DISTRIBUTION OF WH-​ IN-​S ITU IN KRACHI

The relationship between syntax and phonology is often assumed to be unidirectional. On this view, the properties of phonological structures in certain domains are either determined or at least constrained by syntactic properties. This conception precludes causal interactions in the opposite direction, that is, operations in which phonological considerations drive phenomena at the syntactic level. While in a large part conceptual, this view finds support in a robust and continually growing body of empirical evidence (e.g., Clements 1978; Zwicky & Pullum 1986; Zwicky & Kaisse 1987; Pullum & Zwicky 1988; Guasti & Nespor 1999; Déchaine 2001; Pak 2008; Kahnemuyipour 2009; Selkirk 2011; Heath & McPherson 2013; Cheng & Downing 2016; among many others) and crucially, a paucity of evidence to the contrary. Richards (2010, 2016)  challenges this conception of the syntax-​ phonology interface, proposing that wh-​in-​ situ phenomena (among others—​see Richards 2016 for other cases) furnish evidence for an architecture in which at least some syntactic behavior is underpinned by phonology.1 Richards’s idea is that certain phonological computations enter the derivation early (i.e., before Spell-​Out) and thus exert a causal influence on certain narrow syntactic operations. For Richards, prosodic mapping is one such computation. Accordingly, syntax and prosody are derivationally entangled in this ontology—​the two proceed largely in Anti-contiguity. Jason Kandybowicz, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780197509739.001.0001

6

parallel, with mappings between syntactic and prosodic structures firmly established prior to the transfer of those structures to the interfaces. The basis for Richards’s (2010) proposal lies in his claim that the distribution of wh-​items in a language (e.g., moved vs. in-​situ) can be predicted largely on the basis of the prosodic properties of those expressions. This chapter addresses Richards’s prosodic entanglement theory by applying it directly to in-​situ wh-​distribution in Krachi,2 an endangered North Guang language related to Akan that is spoken in the Volta region of central eastern Ghana. Wh-​ expressions3 in this language may appear both in-​situ and in left peripheral focus positions with no apparent interpretative/​information structural difference, as illustrated in (1) for a D-​linked wh-​expression and again in (2) for the full inventory of wh-​items in the language. (1)

a.

ɔkyι wυ ɛ-​mɔ bwatɛ wυ mυmυ woman the pst-​kill chicken the which ‘Which chicken did the woman slaughter?’

b.

Bwatɛ wυ mυmυ yι ɔkyι chicken the which foc woman ‘Which chicken did the woman slaughter?’

wυ the

ɛ-​mɔ pst-​kill

This dual distribution characterizes all Krachi interrogative expressions, with the exception of nanι ‘why’. Unlike every other wh-​item in the language (2a, c, e), ‘why’ may not appear clause-​internally (2g). Instead, it must surface in the left periphery (2h), where it is interpreted as a reason operator4 (Kandybowicz & Torrence 2011). Similar facts obtain in Akan (Torrence & Kandybowicz 2015) and many other languages (e.g., Kitharaka [Muriungi 2005]; Bakweri [Marlo & Odden 2007]; Lubukusu [Wasike 2007]; Zulu [Buell 2011]; Ikalanga [Letsholo  2011]; Chinese [Lin  1992]; English [Hornstein 1995, Thornton 2008, Stepanov & Tsai  2008]; Italian [Rizzi  2001]; Korean and Japanese [Ko 2005]; Persian [Karimi 2005]; Romanian [Shlonsky & Soare 2011]; and New Testament Greek [Kirk 2012], among others). (2)

a.

Nsɛ ɛ-​mɔ bwatɛ wυ who pst-​kill chicken the ‘Who slaughtered the chicken?’

b. Nsɛ yι ɔ-​mɔ bwatɛ who foc pst.agr-​kill chicken ‘Who slaughtered the chicken?’

[ 6 ] Anti-contiguity

wυ the

c.

ɔkyι wυ ɛ-​mɔ nɛ woman the pst-​kill what ‘What did the woman slaughter?’

d. Nɛ yι ɔkyι wυ ɛ-​mɔ what foc woman the pst-​kill ‘What did the woman slaughter?’ e. ɔkyι

wυ

ɛ-​mɔ

bwatɛ

wυ

nfrɛ/​kɛmιkɛ/​nɛnɛ

woman the pst-​kill chicken the where/​when/​how ‘Where/​when/​how did the woman slaughter the chicken?’ f.

Nfrɛ/​kɛmιkɛ/​nɛnɛ yι ɔkyι wυ ɛ-​mɔ bwatɛ where/​when/​how foc woman the pst-​kill chicken ‘Where/​when/​how did the woman slaughter the chicken?’

g. *ɔkyι wυ woman the

ɛ-​mɔ pst-​kill

bwatɛ chicken

wυ the

wυ the

nanι why

h. Nanι yι ɔkyι wυ ɛ-​mɔ bwatɛ wυ why foc woman the pst-​kill chicken the ‘Why (for what reason) did the woman slaughter the chicken?’

Similar facts hold in embedded domains5, as illustrated in the following. (3)

a. Kofi ɛ-​g yιrι fɛɛ ɔkyι wυ ɛ-​mɔ nɛ Kofi pst-​say comp woman the pst-​kill what ‘What did Kofi say that the woman slaughtered?’ b. Kofi ɛ-​g yιrι fɛɛ ɔkyι wυ ɛ-​mɔ bwatɛ wυ nɛnɛ Kofi pst-​say comp woman the pst-​kill chicken the how ‘How did Kofi say that the woman slaughtered the chicken?’ (Matrix scope reading) ‘How did the woman slaughter the chicken, according to Kofi?’ (Embedded scope reading) c.

* Kofi ɛ-​g yιrι fɛɛ ɔkyι wυ Kofi pst-​say comp woman the

ɛ-​mɔ pst-​kill

bwatɛ wυ nanι chicken the why

P r o s o di c E n ta n g l e m e n t  

[ 7 ]

8

d. Nanι yι Kofi ɛ-​g yιrι fɛɛ ɔkyι wυ ɛ-​mɔ bwatɛ wυ why foc Kofi pst-​say comp woman the pst-​kill chicken the ‘Why did Kofi say that the woman slaughtered the chicken?’ (Matrix scope reading) ‘Why did the woman slaughter the chicken, according to Kofi?’ (Embedded scope reading)

In what follows, I show that although Richards’s theory of prosodic entanglement is sufficient to derive the distribution of matrix wh-​ in-​situ from the prosodic mapping of DPs in the language, it is too restrictive to derive the distribution of wh-​in-​situ in embedded complement clauses. To show this, I first establish the prosodic status of embedded complement clauses in the language. I then demonstrate that Richards’s proposal incorrectly predicts the unavailability of wh-​in-​situ in this domain. To account for the distribution of wh-​in-​situ in Krachi, therefore, it isn’t necessary to assume that prosodic information is entangled with the narrow syntactic derivation. On the basis of this consideration, I  conclude that syntactic-​ prosodic entanglement does not universally underpin all aspects of wh-​ syntax, as envisioned by Richards. In addition to requiring phono-​syntactic entanglement, Richards’s theory of wh-​licensing also requires interrogative expressions and the complementizers that mark their scope to form a single contiguous sub-​Intonational Phrase prosodic constituent. As we will see, long-​distance wh-​in-​situ prosody in Krachi furnishes a direct argument that the relationship between wh-​and C is not always one of contiguity, but that in some languages, at least, it can be one of anti-​contiguity. The remainder of the chapter is organized as follows. In section 2.2, I  introduce Richards’s proposal in more detail and consider the problem of main clause wh-​distribution in Krachi with respect to it. Despite initial success in this domain, I demonstrate in section 2.3 that this approach fails to derive the distribution of wh-​in-​situ in Krachi embedded clauses. From there, I explore the theoretical consequences of Krachi embedded wh-​ in-​ situ for Richards’s contiguity-​theoretic approach to wh-​prosody. I conclude in section 2.4 with a summary and closing remarks.

2.2. A CONTIGUITY-​T HEORETIC APPROACH TO KRACHI WH-​ IN-​S ITU 2.2.1. Overview of Richards’s (2010, 2016) Proposal

Richards’s two works on wh-​licensing differ in certain minor details and it will be useful to clarify those differences before moving forward.

[ 8 ] Anti-contiguity

Richards (2010) adopts the Edge Alignment theory of prosodic mapping (Clements 1978; Selkirk 1984, 1986, 1996; Nespor and Vogel 1986; Chen 1987; Selkirk and Shen 1990; Truckenbrodt 1995, 1999; among others), according to which prosodic constituents are constructed by mapping certain (but not always both) syntactic boundaries (constituent edges) onto prosodic boundaries.6 In later work, Richards (2016) adopts Selkirk’s (2009, 2011)  Match theory, which maintains that prosodic structure is largely isomorphic to syntactic structure and that both left and right syntactic edges are mapped onto prosodic boundaries (although typically only one of these edges is prosodically active/​transparent). Richards’s two works assume slightly different versions of the Prosodic Hierarchy as well. The Uttering Trees framework (Richards 2010) adopts a version of the Prosodic Hierarchy in which Prosodic Words group together to form Minor Phrases, Minor Phrases combine to form Major Phrases, Major Phrases group together to form Intonational Phrases, and Intonational Phrases combine to form Utterances (Selkirk 1986, 1995). By contrast, the Contiguity Theory framework (Richards 2016) dispenses with the distinction between Major and Minor Phrases, replacing the two with a single category—​the Phonological Phrase (ϕ). In what follows, I will present and discuss Richards’s theory of wh-​licensing in terms of the Contiguity Theory framework, replacing references to Edge Alignment theory with compatible Match theoretic terminology and references to Minor/​Major Phrase with Phonological Phrase. Richards (2010, 2016)  proposes a universal PF condition regulating wh-​constructions, according to which a wh-​item/​phrase and its scopally-​ associated complementizer must form a single contiguous sub-​Intonational Phrase prosodic constituent. (4)

Richards’s (2016: 84–​85) PF condition on wh-​prosody: a. Given a wh-​ phrase α and a complementizer C where α takes scope, α and C must be dominated by a single ϕ, within which α is contiguity-​prominent. b. α is contiguity-​prominent within ϕ if α[’s prosodically active edge] is adjacent to ϕ’s prosodically active edge.

In Richards 2010, this condition is stated in a less precise way. The essence of Richards’s (2010) proposal is that a wh-​phrase and a complementizer where wh-​takes scope must be separated by as few ϕ boundaries as possible. When syntax and prosody collaborate to build structures satisfying this PF condition (recall from the previous section that Richards assumes prosodic information is available throughout the syntactic cycle), wh-​movement is unnecessary and wh-​ in-​situ obtains. When the

P r o s o di c E n ta n g l e m e n t  

[ 9 ]

01

appropriate phrasing of wh-​and C cannot be achieved, wh-​ movement becomes obligatory, repositioning the interrogative item closer to C for prosodic coupling. One way that syntax and prosody can conspire to respect (4) without resorting to movement is to directly manipulate the ϕ domain. Richards assumes a mechanism of extended Phonological Phrase formation in which the ϕ domain of an interrogative can be extended by way of an algorithm (grouping) that takes an existing Phonological Phrase boundary as one of its bounding edges (for his purposes, C) and the interrogative as its opposite edge.7 In this way, Richards’s domain-​building algorithm can yield larger (recursive) Phonological Phrase constituents. (5) Richards’s (2016: 85) grouping algorithm for constructing extended ϕ domains: Given a wh-​ phrase α and a C with which α is in a probe-​goal relation, create a ϕ that dominates C and α.

Together with the notion of contiguity-​prominence (4b), Richards’s mechanism of grouping ensures that the extended ϕ created to satisfy the PF condition on wh-​prosody (4a) will have the wh-​phrase at one of its edges. Richards (2010) calls these extended ϕ structures “wh-​domains.” In Richards’s framework, two conditions determine whether a wh-​ item can be phrased with C independent of movement:  (a) whether the interrogative’s left or right syntactic edge is prosodically active and (b) the position of C in the linear order. When a wh-​ item’s prosodically active ϕ edge and associated C fall on opposite sides, the algorithm in (5) can produce the required prosodic grouping satisfying contiguity-​prominence, thus facilitating wh-​in-​situ. Successful phrasing of this sort can happen in one of two ways according to Richards: (a) C may precede a wh-​item with a prosodically active right ϕ edge (e.g., (6a)), as in Chichewa),8 or (b) C may follow a wh-​item with a prosodically active left ϕ edge (e.g., (6b)), as in Japanese). (6)

a. 👍 C . . . wh-​)ϕ b. 👍 ϕ(wh-​ . . . C

When an interrogative’s prosodically active ϕ edge and corresponding C fall on the same side of the wh-​item, however, the requisite phrasing can obtain only if leftward movement to the opposite side of C occurs or some prosodically altruistic displacement takes place resulting in a structure where wh-​appears in a position adjacent to C. For Richards, Tagalog & English (7a) and Basque & Georgian (7b) exemplify these scenarios, respectively.

[ 10 ] Anti-contiguity

(7)

a. 👎 C . . . ϕ(wh-​ b. 👎 wh-​)ϕ . . . C

Thus, for Richards, the distinction between wh-​movement and wh-​ in-​situ is a consequence of how wh-​ domains are created. Crucially, this can vary from language to language.

2.2.2. Contiguity-​T heoretic Analysis of Krachi Wh-​ in-​s itu

Richards’s theory is largely based on case studies of languages in which the distribution of wh-​items is uniform; either all interrogative constituents front or they all appear in-​situ. On his analysis, this state of affairs follows as a consequence of the fact that the prosodically active edges of DPs in these languages are for the most part uniform (i.e., for all DPs, either the left edge or the right edge is prosodically active across-​the-​board). Nonetheless, distributional asymmetries like those in Krachi (which are not dealt with in Richards 2010)  can be made to follow from Richards’s system. An approach in this vein would appeal to a prosodic asymmetry in the active edge-​marking of Krachi interrogative DPs. More specifically, given the facts in (2), the structural requirements in (6–​7), and the fact that Krachi has clause-​initial complementizers, as shown in (8), an analysis along these lines might hypothesize that unlike all other interrogative DPs in the language, whose right ϕ edges are prosodically active, Krachi ‘why’ is exceptional in that its ϕ is prosodically active on the left edge. (8)

a. Kofi ɛ-​g yιrι fɛɛ ɔkyι wυ ɛ-​mɔ nɛ Kofi pst-​say comp woman the pst-​kill what ‘What did Kofi say that the woman slaughtered?’ b. Kɛ Kofi ɛ-​mɔ bwatɛ wυ yɛ wa wι ŋwaŋwa st comp Kofi pst-​kill chicken the it do 1 .sg strange/​surprising ‘That Kofi slaughtered the chicken is surprising (to me).’

Thus, under a Richards-​style Contiguity-​theoretic analysis, all non-​ ‘why’ interrogatives in the language would have their active ϕ edges and corresponding complementizers on opposite sides (as in (6a)), giving rise to contiguity-​prominent phrasing with C and obviating the need for movement. In the case of ‘why’, however, the DP’s prosodically active ϕ edge and associated C would occur on the same (i.e., left) side of the interrogative (as in (7a)), forcing internal merge into a high (pre-​C)

P r o s o di c E n ta n g l e m e n t  

[ 11 ]

21

peripheral position to satisfy (4). As I show in what follows, these facts are indeed borne out in the language, furnishing initial support for Richards’s theory.

2.2.2.1. Prosodic Evidence 2.2.2.1.1. The Prosody of Non-​interrogative DPs

In Krachi, the right edge of a non-​interrogative DP is ϕ-​active. The remainder of this section provides evidence for this claim. The prosodic data presented in this chapter were elicited naturalistically with native speakers of the Krachi community in Kete Krachi, Ghana, and acoustically analyzed months later in the United States. Although my analysis is based on data from several speakers, the prosodic data presented in this chapter were produced by just two male Krachi native speakers. The speaker with the higher pitch range was in his mid-​to-​late 30s, while the speaker with the lower pitch range was in his late 40s at the time of recording. When active, the presence of a Phonological Phrase boundary in the language is signaled by a low boundary tone (L%), which has certain phonetic and phonological consequences. Some of these consequences include F0 depression and suppression of otherwise productive tone sandhi processes like upstep. Let’s first consider F0 lowering. In Krachi, determiners appear final in the DP. (9)

a. i-​g yo pee pl-​yam all ‘all yams’ b. ɔkyι wυ woman the ‘the woman’

Determiners with lexically high or rising tones are realized with low falling F0 contours when appearing phrase-​finally in the DP. Such low falling contours are exemplified in (10b–​c) for the determiner pee ‘all’. (10a) shows that ‘all’ is realized with a fairly flat high pitch when it occurs outside a DP in isolated citation form, establishing that it is lexically high tone-​bearing. (In the pitch tracks that follow,9 I mark the prosodically active Phonological Phrase edge with a parenthesis-​adjacent ϕ symbol.)

[ 12 ] Anti-contiguity

(10) a. pee ‘all’

b. i-​g yo pl-​yam ‘all yams’

c.

pee)ϕ all

ι-​wυrι pee)ϕ pl-​book all ‘all books’

P r o s o di c E n ta n g l e m e n t  

[ 13 ]

41

The definite determiner wυ behaves in a similar fashion. (11a) shows that when isolated in citation form, wυ surfaces with a high rising F0, confirming that it bears a rising tone lexically. (11b) illustrates that when occurring at the right edge of a DP, the item emerges with a falling contour. The same example also demonstrates the effect that right edge positioning has on the determiner ‘all’, which as in (10b–​c) once again surfaces with a depressed/​falling F0. Because we observe F0 lowering on both determiners in this example, we can conclude that the tonal lowering effects associated with the right edges of DPs are in fact fully general and not dependent on whether the item occurs in utterance-​final position. (11) a. wυ

‘the’

b. ɔkyι wυ)ϕ ɛ-​dιkɛ i-​g yo pee)ϕ woman the pst-​cook pl-​yam all ‘The woman cooked all yams quickly.’

bireŋ quickly

I have demonstrated that determiners with lexical high/​rising tones are realized with low falling F0 contours when appearing phrase-​finally inside the DP. This fact is evidence for the presence of a Phonological Phrase-​final [ 14 ] Anti-contiguity

L% and consequently for the claim that the right edge of DP is prosodically active. The presence of an active right ϕ boundary associated with DPs is also detectable via the suppression of otherwise productive tone sandhi processes like upstep, which I discuss next. In Krachi, the second in a series of consecutive high tones undergoes an F0 boost (Snider 1990). This is exemplified in (12) by the tonal realization of the item gyɔrι ‘smooth’. (12) gyɔrι ‘smooth’

This sandhi process of upstep, however, is suppressed in DP-​final position. Compare (13a), where ‘smooth’ appears at the right edge of the null-​headed DP,10 with (13b), where the definite determiner appears phrase-​finally following the adjective. Only in the former structure is upstep overridden. In (13a), the adjective is realized with a low falling F0 characteristic of items appearing at the right edge of DP. In (13b), upstep applies to the adjective as expected and the lexically rising tone-​bearing definite article surfaces with a significantly depressed F0. (13) a. kυfυrɛ gyɔrι)ϕ egg smooth ‘a smooth egg’

P r o s o di c E n ta n g l e m e n t  

[ 15 ]

61

b. kυfυrɛ gyɔrι egg smooth ‘the smooth egg’

wυ)ϕ the

These facts provide additional evidence for the presence of ϕ-​final L% in the language and thus for the claim that the right edge of DP is prosodically active in Krachi. In addition to these considerations, we can furnish evidence that the left edge of the Krachi DP is not always prosodically marked. I offer three arguments to this effect, all of which take on a similar character. The nature of these arguments is as follows. Suppose that left edges of DPs were always prosodically marked/​active and that ϕ recursion is kept to a minimum. Then, an immediately preceding DP-​external item would fall at the end/​right edge of the preceding Phonological Phrase,11 an environment characterized by L% effects as previously shown. If a lexically high or rising tone-​bearing item were to surface with a depressed or falling F0 contour in this position, the pattern would be consistent with the presence of a right edge ϕ boundary and hence an immediately following left edge DP Phonological Phrase boundary. However, if the lexically high/​rising tone-​ bearing item were to retain its underlying tonal specification and surface without F0 depression, it would suggest that the item is prosodically phrasing with the following DP and thus is not situated at the juncture between two Phonological Phrases. In what follows, we’ll consider the tonal realizations of three immediately preceding DP-​external items that end with high or rising tones: the preposition yɛ ‘with’, the clausal coordinator yι, and the demonstrative kɛnιŋ. If the left edge of a DP were consistently prosodically marked, the F0 realization of an immediately preceding preposition should be depressed. In this light, consider the preposition yɛ ‘with’, which as illustrated in (14) is realized with a rising high F0 pattern in citation form.

[ 16 ] Anti-contiguity

(14) yɛ ‘with’

When yɛ immediately precedes a DP, however, the item fails to show evidence of F0 lowering. Instead, the item surfaces with a rising F0 contour approximating its citation form realization. This is shown in (15). (15) yɛ ɔsιkan)ϕ with knife ‘with a knife’

This lack of F0 lowering on ‘with’ suggests the absence of an L% at the juncture of P and DP and thus, the absence of a ϕ boundary at the left edge of the following DP. If left edges of DPs were always prosodically marked, we would expect to find F0 depression on clausal coordinators immediately preceding subject DPs. As shown in (16), the coordinator yι is realized with a rising F0 pattern in citation form.

P r o s o di c E n ta n g l e m e n t  

[ 17 ]

81

(16) yι ‘and’

However, when immediately preceding subject DPs, the item fails to exhibit F0 lowering. Instead, it surfaces with the expected rising F0 pattern, as shown in the following, suggesting that it phrases prosodically with the following DP. (17) Kofi ɛ-​dιkɛ kugyo wυˇ )ϕ yι ɔkyι wυ)ϕ Kofi pst-​cook yam the and woman the ‘Kofi cooked the yam and the woman ate a plantain.’

e-​g yi brɔdιɛ12)ϕ pst-​eat plantain

Once again, the absence of F0 lowering on ‘and’ suggests the absence of an L% before the DP subject of the coordinated clause and thus the absence of a right edge of a preceding Phonological Phrase. This indicates that there is no ϕ boundary at the left edge of the subject DP in the coordinated clause.

[ 18 ] Anti-contiguity

Our final argument that the left edge of the Krachi DP is not always marked concerns the fact that when a demonstrative ending on a high tone immediately precedes the left edge of a DP, there is once again no F0 depression to indicate the presence of an L% that would delineate the right edge of a preceding/​separate Phonological Phrase. Consider the tonal realization of the Krachi demonstrative kɛnιŋ in citation form. (18) kɛnιŋ

‘dem’

The tonal realization of the demonstrative does not deviate from this pattern when linearly adjacent to a DP, suggesting that there is no prosodic juncture dividing the two items. This is shown in (19). (19) kɛnιŋ a-​nyιnι nι)ϕ dem pl-​man the.prox ‘these men’

P r o s o di c E n ta n g l e m e n t  

[ 19 ]

02

Thus, unlike the right edges of DPs, where the available evidence suggests the presence of active ϕ boundaries, there is no evidence that the left edges of non-​interrogative DPs are uniformly marked in Krachi.13

2.2.2.1.2. The Prosody of Non-​‘why’ Interrogative DPs

Having established the presence of active right ϕ edges in non-​interrogative Krachi DPs, we consider next the prosody of in-​situ non-​‘why’ wh-​ items. As might be expected given their DP/​XP status, all such expressions in the language pattern like non-​interrogative DPs with respect to L% effects at their right edges, suggesting that they too are prosodically active on the right edge. In what follows, I  present an assortment of pitch tracks demonstrating the comparable realizations of Krachi’s non-​‘why’ single interrogative expressions with respect to prosodic behavior at the right edge. We begin with nfrɛ ‘where’, which as shown in (20), surfaces in isolated citation form with an L-​H tonal sequence. (20) nfrɛ

‘where’

When it surfaces in an in-​situ wh-​question, however, the citation-​final high tone is realized with a significantly lowered F0, suggesting the presence of L% at the right edge of the XP and thus an active right ϕ edge, as in the non-​interrogative DPs examined in the previous section. (21) ɔkyι wυ)ϕ ɛ-​mɔ bwatɛ wυ)ϕ nfrɛ)ϕ woman the pst-​kill chicken the where ‘Where did the woman slaughter the chicken at night?’

[ 20 ] Anti-contiguity

kɛnyɛsɔ night

‘Who’ exhibits a comparable effect in the language. Although in citation form, the item nsɛ is realized with a final high tone (22a), it surfaces with a low falling F0 in an in-​situ wh-​question (22b). (22) a. nsɛ

‘who’

b. nsɛ)ϕ ɛ-​mɔ bwatɛ wυ)ϕ who pst-​kill chicken the ‘Who slaughtered the chicken?’

P r o s o di c E n ta n g l e m e n t  

[ 21 ]

2

Likewise, the expression kɛmιkɛ ‘when’ is realized with a stable final H tone in citation form (23a), but is considerably depressed and realized with a falling tone in the context of an in-​situ wh-​question (23b). (23) a. kɛmιkɛ ‘when’

b. ɔkyι wυ)ϕ ɛ-​mɔ bwatɛ wυ)ϕ kɛmιkɛ)ϕ bireŋ woman the pst-​kill chicken the when quickly ‘When did the woman slaughter the chicken quickly?’

The presence of L% and prosodic activity at the right edge of an in-​situ interrogative DP can also be detected by way of the item nɛ ‘what’. As (24a) shows, ‘what’ bears a rising tone in isolated citation form. However, when it appears clause-​internally in a wh-​question, the underlying rising tone is flattened and what surfaces instead is a low F0 plateau (24b).

[ 22 ] Anti-contiguity

(24) a. nɛ ‘what’

b. ɔkyι wυ)ϕ ɛ-​mɔ nɛ)ϕ ndiye woman the pst-​kill what yesterday ‘What did the woman slaughter yesterday?’

Lastly, consider the tonal behavior of nɛnɛ ‘how’, which in citation form surfaces with a flat F0 (25a). In the context of an in-​situ wh-​question, however, nɛnɛ is realized with a falling F0 contour characteristic of the effect of a prosodically active right ϕ edge in the language (25b). (25) a. nɛnɛ

‘how’

P r o s o di c E n ta n g l e m e n t  

[ 23 ]

42

b. ɔkyι wυ)ϕ ɛ-​mɔ bwatɛ wυ)ϕ nɛnɛ)ϕ woman the pst-​kill chicken the how ‘How did the woman slaughter the chicken at night?’

kɛnyɛsɔ night

To conclude this section, I remind the reader why these results are relevant to the discussion on Richards’s Contiguity-​theoretic entanglement approach to wh-​licensing. Richards’s model predicts wh-​in-​situ for a language with a non-​adjacent left-​side C like Krachi if and only if its wh-​ items are prosodically active on their right edges. This is precisely what has been shown for non-​‘why’ wh-​expressions in Krachi.

2.2.2.1.3. The Prosody of ‘Why’

Despite the fact that ‘why’ expressions in Krachi are restricted to peripheral positions (2g–​h) and thus do not generally interact with material to their left,14 there is evidence that ‘why’ is exceptional in having a prosodically active left edge.15 This evidence comes from sluicing in coordinate structures. Consider a structure such as the one in (26), where a full clause with wh-​ in-​situ is conjoined with a sluiced clause via the coordinator yι. (Recall from (16) that this coordinator bears an underlying rising tone). (26) [CP ɔkyι wυ ɛ-​mɔ bwatɛ wυ nɛnɛ] yι [CP nanι] woman the pst-​kill chicken the how and why ‘How did the woman slaughter the chicken? And why?’

Like many Kwa languages, Krachi employs distinct connectives depending on the size of the coordinated constituents. The rising tone-​ bearing connective yι is used for clausal coordination, while the low tone-​bearing item yɛ is used to coordinate sub-​sentential constituents

[ 24 ] Anti-contiguity

like DPs and PPs. As such, the phonological/​tonological realization of the coordinator is evidence that examples like (26) involve true sluicing (clausal coordination + ellipsis), as opposed to mere coordination of wh-​  DPs. What is the tonal realization of the clausal coordinator in this construction? Recall the considerations deployed in section 2.2.2.1.1 to argue that the left edges of non-​interrogative DPs are prosodically ϕ-​inert. If the left edge of ‘why’ were ϕ-​active in this construction, the coordinator would consequently occupy the right edge of a preceding ϕ domain and its tonal realization would thus be expected to be influenced by L%. However, if ‘why’ were ϕ-​inert on its left edge, the coordinator would be expected to phrase prosodically with the sluiced wh-​item and realize its underlying rising tonal contour due to a lack of right ϕ edge adjacency. As illustrated in the pitch track in (27), clausal coordinators surface with falling F0 contours when appearing before sluiced ‘why’ (compare the realization of ‘and’ in (17) with its realization in (27)). This suggests that the prosodic structure of ‘why’ is characterized by ϕ activity on its left edge and thus that the clausal coordinator occupies a Phonological Phrase-​final position in constructions like (26). (27) ɔkyι wυ)ϕ ɛ-​mɔ bwatɛ wυ)ϕ nɛnɛ yι)ϕ woman the pst-​kill chicken the how and ‘How did the woman slaughter the chicken? And why?’

ϕ(nanι why

The prosodic edge activity of ‘why’ is demonstrably exceptional in the language. This is exemplified by the fact that when preceding other sluiced adjunct wh-​expressions like ‘when’ (28a) and ‘how’ (28b), expressions that are prosodically ϕ-​active on their right edges as previously shown, the clausal coordinator surfaces with its lexically specified rising contour. This suggests the absence of a prosodic juncture between the clausal coordinator

P r o s o di c E n ta n g l e m e n t  

[ 25 ]

62

and the sluiced wh-​item in these cases (i.e., when the sluiced interrogative is a non-​‘why’ wh-​item). It also confirms that it is ‘why’, and not the sluicing construction itself, that is responsible for the ϕ boundary in (27). (28) a. ɔkyι wυ)ϕ ɛ-​mɔ nɛ)ϕ yι kɛmιkɛ)ϕ woman the pst-​kill what and when ‘What did the woman slaughter? And when?’

b. ɔkyι wυ)ϕ ɛ-​mɔ nɛ)ϕ yι nɛnɛ)ϕ woman the pst-​kill what and how ‘What did the woman slaughter? And how?’

The contrast between the tonal realizations of the coordinators in these three examples suggests that yι phrases prosodically with ‘when’ and ‘how’, but not ‘why’. This follows from the fact that the left edge of ‘why’ is uniquely ϕ-​active, separating the sluiced wh-​item from the coordinator by a Phonological Phrase boundary. Together, the results presented in sections 2.2.2.1.2 and 2.2.2.1.3 seem to confirm the analytical aspects of Richards’s approach sketched at the

[ 26 ] Anti-contiguity

beginning of section 2.2.2. There is a clear and detectable prosodic asymmetry in the edge-​activity of Krachi interrogatives:  ‘why’ is prosodically active on its left edge, whereas all other wh-​XPs are ϕ-​active on their right. Because non-​‘why’ interrogatives are right-​active in the language, they may occur in-​situ and successfully phrase prosodically with clause-​initial C in keeping with the Contiguity-​theoretic condition in (4). However, because ‘why’ is left-​active, it cannot form an in-​situ contiguity-​prominent phrasing relation with C and must therefore occupy a peripheral syntactic position in order to properly phrase with it.

2.2.2.2.  Ref inement

As mentioned at the outset of section 2.2.2, we have been pursuing a Contiguity-​theoretic entanglement-​based account of Krachi’s ‘why’–​non-​ ‘why’ in-​situ asymmetry that exploits an asymmetry in ϕ-​activity. I have argued that the left edge of ‘why’ is exceptionally ϕ-​active and that the active ϕ boundaries of all other wh-​items in the language fall on the item’s right edges. Coupled with the various facets of Richards’s theory outlined in section 2.2.1, this edge-​marking difference made it possible to derive the fact that only ‘why’ is precluded from surfacing clause-​internally and must instead surface in the left periphery. In older approaches to the syntax-​ prosody interface such as Edge Alignment theory, a particular edge of a prosodic category is taken to align across-​the-​board with a specific edge of a particular syntactic category. Thus, under such an approach it is non-​standard to assume that a particular XP like ‘why’ can be assigned a prosodic structure distinct from all other XPs and hence be exempt from the general prosodic phrasing algorithm operative in the language. If ‘why’ has different prosodic properties from all other wh-​words, as I have argued, this must somehow fall out from its syntax under this theory. More specifically, in virtue of being a sub-​ sentential XP, ‘why’ would receive a default mapping with a prosodically active right ϕ edge. Its special ϕ-​active left edge would have its source in a syntactic property not shared with the other wh-​XPs in the language. In what follows I  discuss the respects in which ‘why’ is syntactically exceptional. Reinhart (1998) accounts for the interpretation of wh-​in-​situ by invoking choice functions (functions that apply to non-​empty sets and yield members of those sets). In order for a choice function to apply, though, its domain must include a set of individuals. In Reinhart’s analysis, wh-​adverbials like ‘why’ are semantically distinct from other wh-​operators in that their domains crucially

P r o s o di c E n ta n g l e m e n t  

[ 27 ]

82

lack sets of individuals over which the choice function can range. Along these same lines, Szabolcsi & Zwarts (1993) argue that wh-​adverbials like ‘why’ range over propositions, not individuals. Therefore, since choice functions are what enable wh-​expressions to be interpreted in-​situ and given that they cannot apply to the denotations of propositional wh-​adverbials, Reinhart (1981, 1998) concludes that wh-​items like ‘why’ must be base-​generated in the left periphery.16 Rizzi (2001) and Shlonksy & Soare (2011) reach a similar conclusion. They posit a high left peripheral base merge position for ‘why’ and thus no movement from within the clausal mid field. Evidence for a high base-​generated position for ‘why’ in Krachi comes from the fact that the item can precede/​co-​occur with non-​interrogative focus fronted constituents, but not follow them. This is illustrated in the paradigm that follows, both for locally (29a) and long-​distance focused (29c) constituents. (29) a. Nanι bwatɛ wυ yι ɔkyι wυ ɛ-​mɔ why chicken the foc woman the pst-​kill ‘Why did the woman slaughter THE CHICKEN (as opposed to, say, the goat)?’ b. *Bwatɛ chicken c.

wυ the

(nanι) yι (nanι) ɔkyι wυ why foc why woman the

ɛ-​mɔ pst-​kill

Nanι bwatɛ wυ yι Kofí ɛ-​g yιrι fɛɛ ɔkyι wυ ɛ-​mɔ why chicken the foc Kofi pst-​say comp woman the pst-​kill ‘Why did Kofi say that the woman slaughtered THE CHICKEN?’  (Wide-​scope reading) ‘Why did the woman slaughter THE CHICKEN, according to Kofi?’  (Narrow-​scope reading)

Strikingly, none of the other wh-​operators in the language may co-​occur with peripheral focused constituents, as shown in (30). These facts are strikingly similar to comparable patterns in Italian reported by Rizzi (2001). (30) a. *Nsɛ who

bwatɛ wυ yι ɔ-​mɔ chicken the foc 3rd.sg-​kill.pst

b. *Nfrɛ/​kɛmιkɛ/​nɛnɛ where/​when/​how

bwatɛ chicken

wυ the

yι ɔkyι foc woman

wυ the

ɛ-​mɔ pst-​kill

Krachi’s ‘why’ is thus syntactically special in that unlike other wh-​ items in the language, it is native to the left periphery and occupies a position higher than the landing site accessed by focus fronted constituents. Perhaps nanι’s special active left edge is connected to this syntactic

[ 28 ] Anti-contiguity

exceptionality. If it is merged in Spec, CP, that is, higher than FocusP and to the left of C, then having a prosodically active left edge would ensure Contiguity with C in keeping with (4)  in a way that would not be independently possible if its prosodically active edge happened to be on the right.17 This is mere speculation, but nonetheless consistent with Richards’s framework. The results presented in this section illustrate the fact that Richards’s entanglement approach is sufficient to not only account for the existence of main clause wh-​ in-​situ in Krachi, but also derive the attested ‘why’–​ non-​‘why’ in-​situ asymmetry. In other words, the syntactic and prosodic properties of main clause wh-​in-​situ in Krachi strongly support Richards’s Contiguity-​theoretic entanglement theory. In the next section, however, I  show that the predictions of Richards’s approach break down when embedded domains are taken into consideration.

2.3. THE BREAKDOWN OF CONTIGUITY THEORY IN KRACHI EMBEDDED CLAUSES

In developing his theory, Richards does not consider wh-​domain formation in subordinate clauses. While his approach appears successful when applied to the domain of root clauses, it fares considerably worse when dealing with embedded contexts. In this section, I show that when applied to embedded wh-​in-​situ in Krachi, Richards’s approach breaks down and makes incorrect predictions. My critique of Richards’s proposal will center on his claim that wh-​ in-​ situ languages can create a prosodic structure for wh-​questions in which the wh-​phrase and the corresponding complementizer are contiguously wrapped under a single extended Phonological Phrase (5). I will show that embedded complement clauses in Krachi induce major prosodic domain breaks separating the root clause from the embedded domain and thus that embedding imposes quite a number of major prosodic boundaries between the interrogative and the matrix scope-​delimiting complementizer. Despite this state of affairs, wh-​in-​situ is robustly available in these contexts. More damaging to Richards’s proposal, however, I  show that embedded TP complement clauses in Krachi are prosodically mapped as Intonational Phrases, one of the largest prosodic constituents in the Prosodic Hierarchy. Therefore, unless we give up the notion of strictly layered hierarchical prosodic structure, it will not be possible for an embedded wh-​item to form an extended ϕ with a main clause complementizer across an intervening Intonational Phrase boundary. Taking his proposal at face value, then,

P r o s o di c E n ta n g l e m e n t  

[ 29 ]

03

Richards’s system wrongly predicts the impossibility of embedded wh-​ in-​ situ in Krachi, a consequence of the fact that no application of (5) in this domain will satisfy (4). 2.3.1. The Prosodic Status of Krachi Embedded Complement Clauses

In this section, I argue that embedded complement clauses in Krachi are prosodically mapped as Intonational Phrases (ι). This finding is of independent theoretical value given the debate initiated by Downing (1970) over whether only root clauses are mapped onto ι constituents. In the context of this section, however, the finding plays a decisive role in illustrating the breakdown of Richards’s approach to wh-​in-​situ in the domain of Krachi embedded clauses. 2.3.1.1. Detecting Intonational Phrase Boundaries

In section 2.2.2.1, I established that the right edge of the ϕ constituent in Krachi, when active, is signaled by way of a low boundary tone. The active right boundaries of Intonational Phrases in the language are also marked by L%. Consider the following pitch track. (31) Kɛ Kofi ɛ-​kya-​υ)ι ɛ-​bɔ ŋwaŋwa)ι comp Kofi pst-​dance-​cldet pres-​cop strange ‘That Kofi danced is strange.’

Structurally, (31) contains a sentential subject, cross-​linguistically a structure regularly parsed as an obligatory ι. The macro prosodic structure of the sentence in (31) thus consists of two ι constituents: the CP subject and the predicate phrase. Observe that the final item in each ι (ε-​kya-​υ ‘danced’

[ 30 ] Anti-contiguity

and ŋwaŋwa ‘strange’) is realized with a low/​falling F0 pattern, marking the constituent’s active right edge. Other phonetic correlates of right ι edge activity in the language can be identified, distinguishing active ϕ edges from active ι edges. These include the presence of pauses and (partial) pitch reset. The presence of a pause immediately following the first ι unit in (31) is evident in the lapse in articulation between the items ε-​kya-​υ and ε-​bɔ. Partial pitch reset or upstepping can be detected in (31) following the pause; however, because most of the items in ι2 are lexically low tone bearing, the effect is subtle and easy to miss. The pitch track in (32) more clearly exemplifies pitch reset following the right boundary of ι1. Notice that the low tone-​bearing final syllables of odum ‘heart’ and ε-​fwι ‘boil’ at the right edge of the utterance are upstepped, that is, realized with higher F0s than that of the final syllable of the L%-​bearing item ε-​kya-​υ ‘danced’ at the right edge of ι1. (32) Kε Kofi ε-​kya-​υ)ι me odum ε-​fwι)ι st comp Kofi pst-​dance-​cldet 1 .sg heart pst-​boil ‘That Kofi danced angered me (i.e. made my heart boil).’

The pitch track in (32) also exemplifies the other phonetic correlates of Krachi ι edge activity previously discussed. The presence of L% can be detected in the low/​falling F0 values at the right edges of the two Intonational Phrases and a clear prosodic break divides the sentential subject (ι1) from the predicate (ι2). 2.3.1.2. Embedded Complement Clauses as Intonational Phrases

Having established the phonetic signature of prosodically active right ι edges in Krachi, we can proceed to evaluate the prosodic status of embedded complement clauses in the language. Prosodic considerations reveal that these

P r o s o di c E n ta n g l e m e n t  

[ 31 ]

23

structures are parsed as Intonational Phrases. Evidence for this characterization comes from the following observations. One, the lexically high tone-​ bearing complementizer fɛɛ surfaces with a low/​falling F0, indicating the presence of a low boundary tone and thus a prosodically active right edge. Two, a significant pause separates the complementizer from the embedded subject. And three, partial pitch reset/​upstepping affects the F0 range of tones in the embedded clause immediately following the complementizer. Together, these facts suggest that the prosodically active right edge dividing the matrix clause from the embedded clause is an ι boundary. This prosodic behavior is illustrated in the pitch track that follows. (33) Fɛ kwarɛ fι-​g yι18 fɛɛ)ι ɔkyι wυ ɛ-​mɔ bwatɛ 2nd.sg collect 2nd.sg-​eat comp woman the pst-​kill chicken ‘You believe that the woman slaughtered the chicken.’

wυ)ι the

The ι status of complement clauses in Krachi is unaffected by the presence of in-​situ interrogatives. The following data confirm that embedded complement clauses harboring in-​situ wh-​items are also prosodically mapped as Intonational Phrases. Two pitch tracks exemplifying the ι status of wh-​ internal embedded complement clauses are presented in (34). The data showcase clausal embedding under different bridge verbs (‘believe’ in (34a) and ‘know’ in (34b)), illustrating that the ι status of the embedded clause is independent of the embedding predicate. The three acoustic correlates of ι phrasing discussed earlier (L%, pause, and partial pitch reset/​upstepping) are clearly observable in each pitch track. (34) a. Fɛ kwarɛ fι-​g yι fɛɛ)ι ɔkyι wυ 2nd.sg collect 2nd.sg-​eat comp woman the ‘What do you believe that the woman slaughtered?’

[ 32 ] Anti-contiguity

ɛ-​mɔ pst-​kill

nɛ)ι what

b. Ama nyi fɛɛ)ι Kwame ɛ-​mɔ nɛ)ι Ama know comp Kwame pres-​kill what ‘What does Ama know that Kwame slaughters?’

2.3.2. Consequences for Contiguity Theory

Richards (2010) claims that wh-​ in-​situ structures involve interrogatives that are separated from C by as few prosodic boundaries as possible. But how many prosodic boundaries count as too many? And which boundaries are relevant? Richards does not explicitly state an upper limit on the number of intervening boundaries, but he does specify that intervening Minor Phrase boundaries matter in such a calculation. What about other intervening boundaries—​boundaries such as ϕ or ι? Despite these unanswered questions, it is clear that the existence of one or two intervening ϕ boundaries is insufficient to block wh-​domain formation in his system. Consider (21), repeated here as (35), with active ϕ boundaries indicated.

P r o s o di c E n ta n g l e m e n t  

[ 33 ]

43

(35) ɔkyι wυ)ϕ ɛ-​mɔ bwatɛ wυ)ϕ nfrɛ)ϕ kɛnyɛsɔ woman the pst-​kill chicken the where night ‘Where did the woman slaughter the chicken at night?’

If we concern ourselves solely with Phonological Phrases formed from XPs and consider no other prosodic mappings in matrix contexts, the right ϕ boundary of the in-​situ interrogative nfrɛ in (35) will be separated from the phonetically null left edge C head by two ϕ boundaries (shaded). In this case, wh-​domain formation must be able to proceed because wh-​ in-​situ is possible. In the case of embedded wh-​in-​situ, however, other (higher) categories of prosodic boundaries will intervene between wh-​and matrix C. In virtue of the fact that embedded TPs are parsed as Intonational Phrases in the language, a structure like (34b) (repeated here as (36)) will impose two intervening active ϕ boundaries and at least one active ι boundary between the items looking to phrase together under a single wh-​ domain. (36) [CP C Ama)ϕ nyi fɛɛ)ι [TP Kwame)ϕ ɛ-​mɔ nε)ϕ )ι Ama know comp Kwame pres-​kill what ‘What does Ama know that Kwame slaughters?’

Because minimization of intervening prosodic boundaries is not quantified or formalized in his earlier framework, we cannot say whether or not Richards’s (2010) system would predict the grammaticality of structures like (36) solely on the basis of his claim that in-​situ wh-​ items should be separated from C by as few prosodic boundaries as possible. But there is a sense in which Richards’s (2016) system, when applied to data like (36), must make a negative (and therefore, incorrect) prediction. To see this, begin by considering the pitch track for (36) shown in (34b) with an eye to discerning whether any acoustic correlates of wh-​ domain formation can be detected. In this structure, there is simply no phonetic evidence that wh-​and matrix C are wrapped in a single extended Phonological Phrase—​the two items occupy domains characterized by distinct F0 ranges and are furthermore separated from one another by a major prosodic break.19 On the contrary, the two items inhabit separate Intonational Phrases—​that is, they are anti-​contiguous with respect to the matrix ϕ immediately dominating C.  The only prosodic domain containing both embedded wh-​and matrix C is the Utterance unit. But this is not the constituent that wh-​domain formation is defined on in Richards’s system.

[ 34 ] Anti-contiguity

For Richards, wh-​licensing requires the formation of a single ϕ constituent housing both wh-​and C (see (4a) and (5)). That is, languages will be able to leave wh-​in-​situ just in case they have ϕ boundaries placed in such a way as to be able to use the procedure in (5) to create larger ϕ phrases containing both the wh-​phrase and the complementizer in satisfaction of (4). If, following Richards, we assume that prosodic constituents are hierarchically structured, potentially recursive, and strictly layered in the sense of Selkirk (1984, 1995), that is, able to embed only units of equal or lower levels (i.e., Phonological Phrases may contain other Phonological Phrases or structures lower on the Prosodic Hierarchy, but not higher constituents, as in (37) (adapted from Richards 2010:  150)), then Richards-​style wh-​ϕ domain formation can apply if and only if no prosodic category higher than ϕ intervenes between wh-​ and C.

ι

(37)

ϕ ϕ ω

ϕ ω

ω

ϕ ω

ω

ω

Given the version of the Prosodic Hierarchy assumed by Richards in both his 2010 and 2016 frameworks (as discussed in section 2.2.1), this means that intervening Intonational Phrase boundaries will block wh-​ domain formation because building an extended Phonological Phrase in such a case would involve constructing a constituent not exclusively built from ϕ (or smaller) subconstituents. Put another way, such a mapping would violate the Strict Layer Hypothesis of Selkirk (1984, 1995).20 (38) Strict Layer Hypothesis A prosodic category of level n (e.g., ι) cannot be dominated by a category of level n–​1 (e.g., ϕ).

In the case of Krachi embedded complement clauses, embedded wh-​ and matrix C are anti-​contiguous because the presence of an intervening Intonational Phrase boundary guarantees that any extended ϕ containing embedded wh-​and matrix C will violate the Strict Layer Hypothesis. Using Richards’s grouping algorithm, we’d derive the following problematic structure.

P r o s o di c E n ta n g l e m e n t  

[ 35 ]

63

ϕ

(39)

ι Cmatrix... Cembedded

ι ...wh-...

Effectively, this means that in a system like Richards’s, wh-​domain formation (and hence, wh-​in-​situ) should be more prevalent in matrix contexts than in embedded contexts because in embedded domains intervening higher-​level prosodic boundaries can possibly emerge. In the case of Krachi embedded complement clauses, then, because (5) will not yield prosodically licit wh-​domains, Richards’s system must wrongly predict that wh-​ in-​situ will be unavailable in these contexts. In Krachi, however, there seems to be no constraint on how deeply embedded an in-​situ interrogative can be, as shown in (40). (40) a. Kofi ɛ-​g yɛnι fɛɛ Ama ɛ-​nyi nɛ Kofi pres-​think comp Ama pres-​know what ‘What does Kofi think that Ama knows?’ b. Kofi ɛ-​g yɛnι fɛɛ Ama ɛ-​nyi fɛɛ Kwame ɛ-​mɔ nɛ Kofi pres-​think comp Ama pres-​know comp Kwame pst-​kill what ‘What does Kofi think that Ama knows that Kwame slaughtered?’ c.

Kofi ɛ-​g yɛnι fɛɛ Ama ɛ-​nyi fɛɛ Kwame Kofi pres-​think comp Ama pres-​know comp Kwame ɛ-​kwarɛ  ɔ-​g yι   fɛɛ  Gifty  ɛ-​mɔ  nɛ rd pres-​collect  3 .sg  comp Gifty pst-​kill  what ‘What does Kofi think that Ama knows that Kwame believes that Gifty slaughtered?’

This suggests that neither minimization of intervening prosodic boundaries nor extended Phonological Phrase formation (i.e., grouping) are necessary or sufficient conditions governing wh-​in-​situ distribution, as in Richards’s framework. To account for the distribution of wh-​in-​situ in Krachi, therefore, it isn’t necessary to assume that prosodic information drives/​is entangled with the narrow syntactic derivation as in Richards’s approach. To do so would lead to incorrect predictions in the case of embedded domains, as we’ve seen.21 However, we must conclude that whatever relation in-​situ wh-​bears to matrix C, it is one of anti-​contiguity (39), at least in languages like Krachi.

[ 36 ] Anti-contiguity

2.4. CONCLUSION

In virtue of its potential empirical reach and wide-​ ranging theoretical implications, Richards’s (2010, 2016)  analysis of wh-​licensing and phono-​syntactic entanglement invites us to extend the system’s coverage by applying it to languages it was not designed to handle in the name of confirming and testing the limits of the theory. This is what I have done in this chapter. We have found that while initially successful in one domain, the theory falters in another. Richards’s model predicts wh-​in-​situ for a language with a left-​side C like Krachi if and only if (a) its wh-​items have prosodically active right ϕ boundaries and b) no prosodic boundaries larger than ϕ intervene between wh-​and its scope-​delimiting C, allowing for the formation of a special extended prosodic constituent called a “wh-​ domain” wrapping the two items. In the case of matrix clauses, this condition is met when applied to all interrogative XPs except ‘why’, leading predictably to the existence of wh-​in-​situ for those expressions in that domain. This result is rather unsurprising given that Richards focuses entirely on root clause wh-​phenomena in developing his theory. When applied to the domain of embedded clauses, however, the existence of intervening ι boundaries in Krachi leads to the incorrect prediction that wh-​in-​situ should be unavailable in non-​root contexts in the language. Although we find no phonetic evidence for the existence of a wh-​domain grouping together embedded wh-​and matrix C in these constructions, embedded wh-​in-​situ is nonetheless available in the language (and robustly so). This means that in Krachi, wh-​and matrix C, contrary to what Richards claims, are in fact prosodically anti-​contiguous (at least at the level of ϕ). More generally, a system like Richards’s predicts that wh-​domain formation (and hence, wh-​ in-​situ) should be more prevalent in root contexts than in embedded ones because intervening major prosodic boundaries like ι can emerge when clausal embedding occurs. In the case of Krachi, this prediction fails to pan out as well. I see this as a function of the fact that Richards’s system is based entirely on root wh-​ phenomena, providing a cautionary lesson that both main clause and embedded contexts need to be equally taken into consideration when developing accounts of wh-​prosody, licensing, and distribution. For Richards, prosodic considerations like the ones outlined in this chapter influence the outcome of the narrow syntactic derivation. In this way, Richards’s approach employs an unconventional architecture in which phonology and syntax are derivationally entangled. We have found no evidence for this relationship within the domain of Krachi interrogative syntax. That is to say, I have argued that it isn’t necessary to assume that prosodic information is present within the narrow syntactic cycle in order

P r o s o di c E n ta n g l e m e n t  

[ 37 ]

83

to account for the distribution of wh-​in-​situ in Krachi. I conclude, therefore, that phono-​syntactic entanglement does not universally underpin all aspects of wh-​syntax, as proposed by Richards. We have reached two important conclusions in this chapter that will influence the development of the theory of wh-​prosody I will propose in the next chapter. The first is that wh-​and C fail to universally group together contiguously under any non-​trivial (i.e., sub-​Utterance level) prosodic constituent. This is what I mean by the “anti-​contiguity of wh-​ and C.” This conclusion encourages us to consider the possibility that at least in some languages, the prosodic well-​formedness of wh-​may well have an anti-​Richards character, one in which wh-​and C resist phrasing together at some relevant level of prosodic phrasing. The second guiding conclusion is that it isn’t necessary to assume that prosodic considerations influence the outcome of narrow syntactic operations. We will thus be able to hold to the industry-​standard assumption that syntax proceeds free of influence from phonology/​prosody. In the next chapter, I will develop a theory where certain syntactically mysterious asymmetries in wh-​in-​situ distribution are explained in purely prosodic terms following the narrow syntactic cycle.

[ 38 ] Anti-contiguity

CHAPTER 3

An Anti-​contiguity Approach to Tano in-​situ Interrogative Distribution

3.1. INTRODUCTION

Wh-​in-​situ is a pervasive feature of Tano interrogative syntax (Torrence & Kandybowicz 2012, 2013, 2015; Kandybowicz & Torrence 2013), yet the Tano languages differ from one another in subtle ways with respect to the distribution of in-​situ interrogative expressions. As we observed in Chapter 2, Krachi allows all wh-​expressions apart from ‘why’ to appear in-​ situ in both main and embedded complement clauses. Bono, a Central Tano language of the Akan group (Williamson & Blench 2000), tolerates wh-​ in-​ situ in both main and embedded complement clauses, but draws the line at subject interrogatives and ‘why’ expressions in both domains. In Wasa and Asante Twi, two Tano languages closely related to Bono, the distribution of wh-​in-​situ in main clauses is identical to Bono’s profile. However, whereas Bono allows both non-​subject and non-​‘why’ in-​situ interrogatives in embedded complement clauses, Wasa and Asante Twi systematically disallow all instances of wh-​in-​situ in embedded domains. What accounts for this variation? In this chapter, I argue that the distribution of wh-​in-​situ is as much a matter of prosody as it is a matter of syntax/​ semantics, following in the spirit of recent work by Richards (2010, 2016), as discussed in Chapter 2. While considerations at the syntax-​semantics interface surely play a prominent role in determining the distribution of certain in-​situ interrogatives both in Tano and cross-​linguistically, I restrict my attention to a case where syntactic and semantic factors appear immaterial in the licensing of wh-​in-​situ. Unlike Richards (2010, 2016), who claims that Anti-contiguity. Jason Kandybowicz, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780197509739.001.0001

04

at PF wh-​items and their scope-​delimiting complementizers must form a contiguous Phonological Phrase, my claim is that in some languages wh-​ items are prohibited from forming prosodic constituents with any overt (but crucially, not silent) C at the level of Intonational Phrase. Put another way, prosody enforces a strict anti-​contiguity relation between wh-​and C, as argued in the previous chapter. Under this analysis, the ability of a wh-​ item to appear in an in-​situ position correlates with the prosodic status of its immediately containing clause. I  show that unlike in Wasa and Asante Twi, embedded TP complement clauses are parsed as separate Intonational Phrases in Krachi (as argued in Chapter 2) and in Bono. As such, ι boundaries divide C from embedded interrogatives in Krachi and Bono, preventing the two from forming a prosodic constituent at the ι level. Conversely, no such boundaries intervene between embedded C and wh-​in Wasa and Asante Twi, yielding prosodic mappings in which the items phrase together under a single ι. Consequently, embedded wh-​in-​situ is prosodically licit in Krachi and Bono, but not in Wasa and Asante Twi. In this way, the Tano pattern of wh-​in-​situ variation just described reduces to a difference in how narrow syntactic structures are externalized at PF by way of prosodic mapping. This chapter is organized as follows. Section 3.2 concretizes the variation in Tano wh-​in-​situ patterns described earlier by establishing the basic syntactic facts. Section 3.3 motivates a prosodic approach to deriving the variable distribution of wh-​in-​situ in Tano by calling into question the adequacy of a purely syntactic/​semantic analysis. In section 3.4, I present the analysis, grounding my claims in the observable prosodic differences dividing one class of Tano languages from the other with respect to the status of embedded clauses. The chapter concludes in section 3.5 with a summary and brief closing remarks. 3.2. THE DISTRIBUTIONAL VARIATION OF WH-​ IN-​S ITU IN TANO

This chapter examines the distribution and prosodic conditions regulating in-​situ interrogative expressions in two non-​island domains:  root and embedded complement clauses. Future work will expand on these results by extending coverage to other embedded domains, such as non-​complement clauses (of which there are several varieties in each language). 3.2.1. The Distribution of Wh-​ in-​s itu in Krachi

The distributional facts regarding wh-​in-​situ in Krachi were presented in Chapter 2. I reprise the presentation of those facts here for the benefit of the reader. [ 40 ] Anti-contiguity

With the exception of nanι ‘why’ (1d), which must be focused and appear clause-​peripherally (1e) (Kandybowicz and Torrence 2011), main clause wh-​ expressions in Krachi may surface clause-​internally. As the following data1 illustrate, in-​situ interrogatives in the language may appear on the left edge of the clause, at the right edge of the clause, or in a non-​edge position. (1)

a. Nsɛ ɛ-​mɔ bwatɛ wυ (ndiye) who pst-​kill chicken the yesterday ‘Who slaughtered the chicken (yesterday)?’ b.

ɔkyι wυ ɛ-​mɔ nɛ (ndiye) woman the pst-​kill what yesterday ‘What did the woman slaughter (yesterday)?’

c.

ɔkyι wυ ɛ-​mɔ bwatɛ wυ nfrɛ/​kɛmιkɛ/​nɛnɛ (ndiye) woman the pst-​kill chicken the where/​when/​how yesterday ‘Where/​when/​how did the woman slaughter the chicken (yesterday)?’

d.

*ɔkyι wυ ɛ-​mɔ bwatɛ wυ woman the pst-​kill chicken the

nanι (ndiye) why yesterday

e. Nanι yι ɔkyι wυ ɛ-​mɔ bwatɛ wυ (ndiye) why foc woman the pst-​kill chicken the yesterday ‘Why (for what reason) did the woman slaughter the chicken (yesterday)?’

The same wh-​expressions that are permitted clause-​internally in matrix contexts are permissible in comparable positions in embedded complement clauses. This is shown in (2). Once again, in-​situ nanι ‘why’ is restricted. (2)

a. Kofi ɛ-​g yιrι fɛɛ nsɛ ɛ-​mɔ bwatɛ wυ (ndiye) Kofi pst-​say comp who pst-​kill chicken the yesterday ‘Who did Kofi say slaughtered the chicken (yesterday)?’ b. Kofi ɛ-​g yιrι fɛɛ ɔkyι wυ ɛ-​mɔ nɛ (ndiye) Kofi pst-​say comp woman the pst-​kill what yesterday ‘What did Kofi say that the woman slaughtered (yesterday)?’ c.

Kofi ɛ-​g yιrι fɛɛ ɔkyι wυ ɛ-​mɔ bwatɛ wυ Kofi pst-​say comp woman the pst-​kill chicken the nfrɛ/​kɛmιkɛ/​nɛnɛ (ndiye) where/​when/​how  yesterday ‘Where/​when/​how did Kofi say that the woman slaughtered the chicken (yesterday)?’

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 41 ]

24

*Kofi ɛ-​g yιrι fɛɛ ɔkyι wυ ɛ-​mɔ bwatɛ wυ nanι (ndiye) Kofi pst-​say comp woman the pst-​kill chicken the why yesterday

d.

3.2.2. The Distribution of Wh-​ in-​s itu in Bono

Most wh-​expressions in Bono may appear in-​situ; however, there is an additional exception that renders Bono more restrictive than Krachi. This manifests itself in the form of a subject–​non-​subject asymmetry, a widespread phenomenon observed both within African languages and beyond (Bokamba 1976; Maxwell 1981; Green & Jaggar 2003; Muriungi 2005; Sabel & Zeller 2006; Potsdam 2006; Zentz 2016). In addition to its restriction on ‘why’ (3e), Bono disallows subject interrogatives from appearing clause-​internally (3a). Both items must occur clause-​ peripherally in focus positions (3b, f). The remaining data in (3)  show that by contrast, object (3c) and non-​‘why’ adjunct wh-​items (3d) can appear in-​situ in main clauses, either at the right edge of the clause or in a non-​edge position. (3)

a. *Hwae sae (nra) who dance.pst yesterday b. Hwae ne sae (nra) who foc dance.pst yesterday ‘Who danced (yesterday)?’ c.

Bema kɛ̃ kum abe (nra) man the kill.pst what yesterday ‘What did the man slaughter (yesterday)?’

d. Bema kɛ̃ kum akoko kɛ̃ ahı˜fa/​dabe/​sɛ (nra) man the kill.pst chicken the where/​when/​how yesterday ‘Where/​when/​how did the man slaughter the chicken (yesterday)?’ e. *Bema kɛ̃ man the f.

kum akoko kɛ̃ kill.pst chicken the

senti why

(nra) yesterday

Senti ne bema kɛ̃ kum akoko kɛ̃ (nra) why foc man the kill.pst chicken the yesterday ‘Why did the man slaughter the chicken (yesterday)?’

[ 42 ] Anti-contiguity

We find a similar distribution in embedded complement clauses—​subject interrogatives (4a) and ‘why’ expressions (4e) are barred from appearing in-​situ, but all other wh-​expressions may surface in-​situ in embedded positions. (4)

a. *Wo dwene sɛ hwae kum akoko kɛ̃ (nra) 2nd.sg think comp who kill.pst chicken the yesterday b.

Hwae ne wo dwene sɛ kum akoko kɛ̃ (nra) nd who foc 2 .sg think comp kill.pst chicken the yesterday ‘Who do you think slaughtered the chicken (yesterday)?’

c.

Wo dwene sɛ bema kɛ̃ kum abe (nra) nd 2 .sg think comp man the kill.pst what yesterday ‘What do you think that the man slaughtered (yesterday)?’

d. Wo dwene sɛ bema kɛ̃ kum akoko kɛ̃ ahĩfa/​dabe/​sɛ nd 2 .sg think comp man the kill.pst chicken the where/​when/​how (nra) yesterday ‘Where/​when/​how do you think that the man slaughtered the chicken (yesterday)?’ e.

*Wo dwene sɛ bema kɛ̃ kum akoko kɛ̃ senti 2nd.sg think comp man the kill.pst chicken the why

(nra) yesterday

f.

Senti ne wo dwene sɛ bema kɛ̃ kum akoko kɛ̃ (nra) nd why foc 2 .sg think comp man the kill.pst chicken the yesterday ‘Why do you think that the man slaughtered the chicken (yesterday)?’

3.2.3. The Distribution of Wh-​ in-​s itu in Wasa

With respect to its distribution in main clauses, wh-​in-​situ patterns identically in Wasa and Bono. That is to say, we find both subject–​object interrogative asymmetries and ‘why’–​non-​‘why’ adjunct asymmetries. The following data reveal that subject interrogatives (5a) and ‘why’ adverbial interrogatives (5e) may not surface clause-​internally in root contexts (they must be focused (5b, f)), but other wh-​items may.

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 43 ]

4

(5)

a.

*Hwae who

saaye (ɛndra) dance.pst yesterday

b. Hwae na saaye (ɛndra) who foc dance.pst yesterday ‘Who danced (yesterday)?’ c.

Bɛrɛma no kum ɛdiɛn (ɛndra) man the kill.pst what yesterday ‘What did the man slaughter (yesterday)?’

d.

Bɛrɛma no kum akoko no ɛhĩfa/​mmɛrɛ bɛn/​sen (ɛndra) man the kill.pst chicken the where/​time which/​how yesterday ‘Where/​when/​how did the man slaughter the chicken (yesterday)?’

e. *Bɛrɛma no kum akoko no adiɛnti (ɛndra) man the kill.pst chicken the why yesterday f.

Adiɛnti na bɛrɛma no kum akoko no why foc man the kill.pst chicken the ‘Why did the man slaughter the chicken (yesterday)?’

(ɛndra) yesterday

With respect to its distribution in embedded complement clauses, Wasa and Bono part ways. Bono permits embedded clause-​internal non-​subject and non-​‘why’ interrogatives (4c–​d). Wasa bans all interrogatives from appearing in-​situ in non-​echo question embedded contexts. In this regard, Wasa behaves like French (Cheng 1997; Bošković 1998; Reglero 2005) in restricting wh-​in-​situ from crossing tensed clause boundaries. The data in (6) highlight the fact that wh-​items that are available clause-​internally in root contexts (5c–​d) are disallowed in clausal complements. (6)

a. *Wo dwene sɛ bɛrɛma no kum ɛdiɛn (ɛndra) nd 2 .sg think comp man the kill.pst what yesterday b.

ɛdiɛn na wo dwene sɛ bɛrɛma no kum (ɛndra) what foc 2nd.sg think comp man the kill.pst yesterday ‘What do you think that the man slaughtered (yesterday)?’

c.

*Wo dwene sɛ bɛrɛma no kum akoko no ɛhĩfa/​mmɛrɛ 2nd.sg think comp man the kill.pst chicken the where/​time bɛn/​ sen  (ɛndra) which/​ how  yesterday

[ 44 ] Anti-contiguity

d. ɛhĩfa/​mmɛrɛ bɛn/​sen

na wo dwene sɛ bɛrɛma no kum where/​time which/​how foc 2nd.sg think comp man the kill.pst akoko   no  (ɛndra) chicken  the     yesterday ‘Where/​when/​how do you think that the man slaughtered the chicken (yesterday)?’

3.2.4. The Distribution of Wh-​ in-​s itu in Asante Twi

The distribution profile of wh-​in-​situ in Asante Twi mirrors that of Wasa. In main clauses, subject wh-​items (7a) and ‘why’ (7e) must appear clause-​externally in focused positions (7b, f). All other interrogative expressions (7c, d) may appear inside the clause in either edge or non-​edge positions. (7)

a. *Hwan who

bɔɔ hit.pst

Ama (ɛnora) Ama yesterday

b. Hwan na ɔ-​bɔɔ Ama (ɛnora) who foc 3rd.sg-​hit.pst Ama yesterday ‘Who hit Ama (yesterday)?’ c.

Ama bɔɔ hwan (ɛnora) Ama hit.pst who yesterday ‘Who did Ama hit (yesterday)?’

d. Ama bɔɔ Kofi ɛhĩfa/​(ɛ)berɛ bɛn/​sɛn (ɛnora) Ama hit.pst Kofi where/​time which/​how yesterday ‘Where/​when/​how did Ama hit Kofi (yesterday)?’ e. *Kwadwo bɔɔ Kwadwo hit.pst f.

Ama adɛn Ama reason

Adɛn nti na Kwadwo reason why foc Kwadwo ‘Why did Kwadwo hit Ama?’

nti why

bɔɔ hit.pst

(Saah: 1988:20)

Ama Ama

(Saah: 1988:20)

As is the case with Wasa, Asante Twi restricts all interrogatives from appearing in clausal complements in non-​echo questions. The following data show that wh-​items that are available clause-​internally in matrix contexts (7c–​ d) are incapable of occurring in embedded complement clauses and must be fronted into the main clause left periphery.

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 45 ]

64

(8)

a. *Wo dwene 2nd.sg think

sɛ comp

Ama Ama

bɔɔ hit.pst

b. Hwan na wo dwene sɛ Ama who foc 2nd.sg think comp Ama ‘Who do you think that Ama hit yesterday?’

hwan who

(ɛnora) yesterday

bɔɔyɛ (ɛnora) hit.pst yesterday

c.

*Wo dwene sɛ Ama saa ɛhĩfa/​(ɛ)berɛ bɛn/​sɛn (ɛnora) 2nd.sg think comp Ama dance.pst where/​time which/​how yesterday

d.

ɛhĩfa/​(ɛ)berɛ bɛn/​sɛn na wo dwene sɛ Ama saayɛ nd where/​time which/​how foc 2 .sg think comp Ama dance.pst (ɛnora) yesterday ‘Where/​when/​how do you think that Ama danced (yesterday)?’

3.2.5. Delimiting the Empirical Scope of the Chapter

To briefly recap, all four Tano languages under investigation allow wh-​ in-​ situ in main clauses, but restrict ‘why’ from appearing clause-​internally. Additionally, Bono, Wasa, and Asante Twi all prohibit in-​situ subject interrogatives. The bulk of the variation, however, takes place in the embedded domain. All wh-​items that may independently appear in-​situ in root contexts may also appear in-​situ in complement clauses in both Krachi and Bono. Wasa and Asante Twi, on the other hand, systematically exclude wh-​in-​situ in embedded complement clauses. Table 3.1 summarizes.

Table 3.1   DISTRIBUTION OF WH-​ IN-​S ITU IN KRACHI, BONO, WASA , AND ASANTE TWI

krachi

bono

wasa

asante twi

✓

×

×

×

✓

✓

✓

✓

‘why’ in-​situ

×

×

×

×

wh-​ in-​situ

✓

✓

×

×

subject wh-​ in-​situ (main clauses) non-​subject wh-​ in-​situ (main clauses)

(embedded clauses)

[ 46 ] Anti-contiguity

In the remainder of this chapter, I will focus my inquiry on deriving the variable distribution of non-​subject and non-​‘why’ in-​situ interrogatives in root and embedded complement clauses (i.e., the shaded rows in Table 3.1). My reason for this is that a growing body of research has converged on the conclusion that restrictions on the distribution of in-​situ subject wh-​ items and ‘why’ interrogatives are crosslinguistically robust and plausibly syntactic/​semantic in nature (see Green & Jaggar 2003; Potsdam 2006; and Sabel & Zeller 2006, among others, for approaches to restrictions on in-​situ subject interrogatives; and Reinhart 1998; Rizzi 2001; Shlonsky & Soare 2011; and Torrence & Kandybowicz 2013 for accounts of the prohibition on ‘why’ in-​situ, as discussed in Chapter 2). This of course raises the question of whether restrictions on the distributions of other in-​situ interrogatives can likewise be grounded in syntactic/​semantic considerations. In the next section, I consider whether such forces are at play, but conclude that the factors shaping the distribution of Tano’s non-​subject and non-​‘why’ in-​ situ interrogatives cannot be purely syntactic/​semantic in nature.2 As a result, I pursue an analysis in which prosodic considerations play a role in constraining the positions that interrogatives may occupy in the interior of the clause, in the spirit of Richards (2010, 2016), but with a different implementation.

3.3. MOTIVATING A PROSODIC APPROACH TO TANO IN-​S ITU INTERROGATIVE DISTRIBUTION

Excluding subjects and ‘why’ expressions, all four languages under investigation permit wh-​in-​situ in root clauses. Because embedded domains introduce restrictions on the acceptability of wh-​in-​situ in Tano, we must probe these contexts to uncover the conditions that constrain the distribution of in-​situ interrogatives. In this section, I consider one influential approach to the licensing of (embedded) in-​situ interrogative items that appeals to the syntax-​semantics interface. I show, however, that this approach makes incorrect predictions with respect to Tano embedded interrogative syntax, motivating a non-​ syntactic/​ semantic approach to embedded in-​situ wh-​ distribution.3 The syntactic/​semantic approach to in-​situ wh-​ licensing I am referring to is actually a family of proposals, each differing slightly in their technical implementation, but sharing the core idea that in-​situ wh-​ items are semantically licensed via the formation of a syntactic dependency between wh-​ and a (potentially null) Q operator (Cheng 1991; Beck 1996; Hagstrom 1998; Pesetsky 2000; Cable 2010, among others). For some,

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 47 ]

84

this dependency is achieved via binding and for others, it is mediated by an agreement relation. Either way, a language will allow wh-​in-​situ if two conditions are met: one, the language has a dedicated Q operator (whether overt or null); and two, wh-​is accessible to Q. Applied to the languages currently under investigation, the approach would offer the following analysis of in-​situ interrogative distribution. To account for the fact that all four languages admit wh-​in-​situ (at least in matrix clauses), it must be the case that all four languages have Q particles. And to account for the asymmetrical distribution of wh-​in-​situ in embedded clauses, it would have to be the case that embedded wh-​is accessible to Q in Krachi and Bono (facilitating embedded wh-​in-​situ), but not in Wasa and Asante Twi (thereby blocking embedded wh-​ in-​situ). As for the first claim, there is sufficient evidence that all four languages have Q particles, whether overt or null. Evidence for null Q particles comes from the existence of (simple) partial wh-​focus movement, in which a silent matrix Q operator marks the matrix scope of the partially moved interrogative that surfaces in the embedded clause. In these cases, despite the fact that the interrogatives do not surface in the main clause, the utterances are interpreted as true main clause questions and not embedded or indirect questions. These interpretations suggest the presence of silent matrix Q operators. This is illustrated in (9) for Krachi and Bono. (9)

Krachi a. Kofi ɛ-​g yιrι fɛɛ nɛ yι Ama ɛ-​mɔ Kofi pst-​say comp what foc Ama pst-​kill ‘What did Kofi say that Ama slaughtered?’ Bono b. Wo dwene sɛ abe ne bema kɛ̃ 2nd.sg think comp what foc man the ‘What do you think that the man slaughtered?’

kuye kill.pst

Evidence for overt Q comes in the form of clause-​final particles deployed in the formation of polar questions. This is illustrated in (10) for Krachi and Asante Twi. Similar facts obtain in Bono and Wasa. (10) Krachi a. ɔkyι wυ ɛ-​mɔ bwatɛ wυ ɛɛ woman the pst-​kill chicken the q ‘Did the woman slaughter the chicken?’

[ 48 ] Anti-contiguity

Asante Twi b. Kofi dɔ Ama anaa Kofi love Ama q ‘Does Kofi love Ama?’

As for the second claim (i.e., embedded wh-​is accessible to Q in Krachi and Bono, but not in Wasa and Asante Twi), a prediction is made. If Q is unable to non-​locally bind/​agree with an embedded wh-​item in Wasa and Asante Twi, then neither Wasa nor Asante Twi should allow partial wh-​focus movement to a position below embedded C because otherwise, the matrix scope of the partially moved wh-​item in the embedded clause would be unaccounted for. This prediction is borne out in Asante Twi. As the following data illustrate, partial wh-​focus movement is unavailable in the language.4 (11) a. *Wo dwene sɛ hwan 2nd.sg think comp who b. *Wo kaa nd 2 .sg say.pst c.

sɛ comp

dɛn what

na ɔ-​bɔɔ Ama foc 3rd.sg-​hit.pst Ama na Kofi foc Kofi

*Wo nim sɛ ɛhĩfa/​(ɛ)berɛ bɛn 2nd.sg know comp where/​time which

diiyɛ eat.pst na Kofi foc Kofi

saayɛ dance.pst

Adding strength to the claim’s accurate prediction is the fact that the phenomenon of embedded non-​interrogative focus, where binding by matrix Q is not an issue, is not constrained in this way in the language. The following data reveal that short focus movement to an embedded peripheral position is attested in Asante Twi. Thus, it is not the case that the landing site of partial wh-​focus movement is unavailable. (12) a. Wo dwene sɛ Kofi na ɔ-​bɔɔ Ama 2nd.sg think comp Kofi foc 3rd.sg-​hit.pst Ama ‘You think that it’s KOFI who hit Ama.’ b. Wo kaa sɛ nkonya no 2nd.sg say.pst comp cake the ‘You said that it’s THE CAKE that Kofi ate.’ c.

na foc

Kofi Kofi

diiyɛ eat.pst

Wo nim sɛ ɛnora na Kofi saayɛ 2nd.sg know comp yesterday foc Kofi dance.pst ‘You know that it’s YESTERDAY that Kofi danced.’

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 49 ]

05

The prediction, however, fails to hold in Wasa. Despite the fact that in-​ situ interrogatives are restricted from appearing in embedded complement clauses, partial wh-​focus movement is available in the language.5 All Wasa wh-​items may undergo partial focus movement, regardless of thematic status. (13) a.

Wo dwene sɛ bɛrɛma bɛn na o-​kum akoko no nd rd 2 .sg think comp man which foc 3 .sg-​kill.pst chicken the ‘Which man do you think slaughtered the chicken?’

b. Wo dwene sɛ ɛdiɛn na bɛrɛma nd 2 .sg think comp what foc man ‘What do you think that the man slaughtered?’ c.

no the

kumiye kill.pst

Wo dwene sɛ ɛhĩfa/​adiɛnti na bɛrɛma no kum akoko no nd 2 .sg think comp where/​why foc man the kill.pst chicken the ‘Where/​why do you think that the man slaughtered the chicken?’

Note that Wasa partial wh-​focus movement is “simple” in the sense of Fanselow’s (2006) typological characterization—​the partially-​moved interrogative is unaccompanied by an overt Q particle in the clause where it takes scope (i.e., the root clause). The availability of partial wh-​ focus movement in spite of the absence of embedded wh-​in-​situ is unexpected for another reason. According to Fanselow’s (2006) Generalization S2, if a construction is grammatical with simple partial movement, it can also be constructed with the wh-​phrase in-​situ. Wasa, therefore, represents a clear counterexample to Fanselow’s Generalization, as it allows simple partial movement of any interrogative item, yet prohibits those items from surfacing clause-​internally in the embedded domain. Let’s return to the implications of the syntactic/​semantic approach’s failed prediction in the case of embedded wh-​licensing in Wasa. Because the partially moved wh-​item takes matrix scope, as revealed by the interpretations in (13), matrix Q must somehow non-​locally bind/​agree with the moved embedded interrogative in the spell-​out domain of the embedded C phase. But if this dependency is available under partial movement, why is it not available when an interrogative remains in-​situ?6 The syntactic/​semantic approach provides no satisfying answer to this analytical dilemma, leading to the reasonable conclusion that in actuality, embedded in-​situ interrogatives are in fact bound by matrix Q in Wasa. I conclude, therefore, that the principal active force constraining the distribution of Tano’s non-​subject/​non-​‘why’ embedded in-​situ interrogatives

[ 50 ] Anti-contiguity

is not purely syntactic/​semantic in nature. Consequently, I maintain that approaching the problem from the decidedly opposite direction (i.e., from a prosodic perspective, as in the spirit of Richards 2010, 2016, for example) is reasonably justified.

3.4. PROSODIC ANALYSIS OF TANO IN-​S ITU INTERROGATIVES

Focusing on embedded complement clauses, the primary locus of variation with respect to the distribution of Tano in-​situ interrogatives, I  demonstrate in this section that a wh-​item’s ability to appear in an in-​situ position correlates with the prosodic status7 of its immediately containing clause. More specifically, I show that in languages where embedded clauses are realized as Intonational Phrases, embedded wh-​in-​situ is available. In those languages where embedded domains do not have the status of independent Intonational Phrases, embedded wh-​in-​situ is restricted. I then exploit this finding to formulate an empirical generalization and develop a prosodic analysis that accounts for the distributional variation across the four languages as well as the asymmetry between root and embedded clause wh-​ in-​situ. This section is organized as follows. After a brief discussion of the methodology employed in the forthcoming prosodic analysis, I  examine the prosodic status of embedded complement clauses in each of the four Tano languages under investigation. Prosodic data consisting of pitch tracks will be presented, allowing us to examine fundamental frequency, pauses/​breaks and their duration, and pitch/​register reset in the embedded domains of each language. The section concludes with the resulting prosodic analysis.

3.4.1.  Methodology

The prosodic data presented in this chapter were collected over a period of four years from 2010 to 2014 in Ghana. A total of eight native speakers supplied the data. Two speakers of each language were recorded and twelve total hours of recordings were obtained, approximately three hours of recording for each language. The resulting database from which the tokens presented in this chapter were drawn consists of hundreds of recorded productions. Although prosodic data were collected from two speakers per language, the pitch tracks presented for each language in this chapter come

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 51 ]

25

from a single native speaker. I chose to limit the pitch track data in this way to ensure that prosodic data within each language is directly comparable. The Krachi pitch tracks that appear in this chapter were produced by a male speaker in his 30s. The Bono prosodic data come from a female in her 40s. The Wasa pitch tracks used in this chapter represent the speech of a male in his 40s. And the Asante Twi prosodic data featured in this chapter were produced by a male in his mid-​20s. Only structures independently determined to be grammatical from prior elicitation sessions were presented to speakers during recording sessions. Speakers produced material one sentence at a time (i.e., in unconnected speech) and were asked to produce their utterances “naturally,” that is, avoiding overly rapid, slow, careful, or deliberate deliveries. Recordings obtained in this manner were then presented to the other native speaker consultant of the language to judge the naturalness of the production. Only those productions deemed “natural” sounding by native speakers were considered in the resulting analysis. Speakers who rejected certain tokens as unnatural cited overly rapid delivery, missing pauses/​breaks, insufficient pause length, and lack of fluid delivery in their justifications. Efforts were made to produce the highest quality recordings possible. However, all recordings were made in the field. Thus, recording conditions were often suboptimal and despite efforts to minimize background noise, the sounds of the environment occasionally intruded into parts of the recordings. For this reason, certain pauses/​breaks appear noisier than would be expected and minor pitch tracking errors sometimes occurred. The pitch tracks presented in this chapter were chosen because they displayed the least amount of background noise and pitch tracking errors. In the analysis that follows, I rely on the existence of prosodic breaks and pitch/​register reset, among other considerations, to diagnose the existence of prosodic phrase boundaries. In doing so, I will make the following assumptions. One, pauses greater than 100 milliseconds constitute major prosodic breaks/​breath group boundaries and thus diagnose the divide between two Intonational Phrases. Pauses less than 100 milliseconds, on the other hand, do not constitute true breaks and thus do not indicate the presence of a major prosodic category boundary. This 100ms threshold was chosen because it represents a greater value than the average durations of non prosodically-​motivated pauses such as those resulting from stop closures, which tend to range from 20–​87ms for fast speech and 24–​97ms for slow speech (Crystal & House 1988:  1555). Two, I  will posit pitch/​ register reset when I  observe low-​toned syllables surfacing at or above the pitch level of low-​toned syllables appearing at the right edge of the preceding putative Intonational Phrase boundary, an environment that

[ 52 ] Anti-contiguity

induces F0 lowering effects in the four languages, as will be demonstrated. Both durational and pitch measurements were made using Praat (Boersma & Weenink 2017).

3.4.2. Prosodic Status of Krachi Embedded Complement Clauses

The prosody of Krachi embedded complement clauses was presented in Chapter 2. I reprise that presentation here for the benefit of the reader. The right edges of phrasal prosodic constituents in Krachi are tonally marked and detectable via a number of salient phonetic cues. Kandybowicz & Torrence (2012) show that Phonological Phrases in the language are right edge-​marked by way of low boundary tones (L%), as discussed in Chapter 2. In the same way, the right boundaries of Intonational Phrases in the language are marked by L%. This is illustrated in the following pitch track by the depressed fundamental frequency (F0) of the (clausal) determiner υ, which bears a lexical rising tone in the language (see Chapter 2 for evidence). In ι-​final position, the determiner fails to exhibit a rising F0 contour and instead surfaces with a low tone realization. (14) (Kε Kofi ε-​kya-​υ)ι comp Kofi pst-​dance-​cldet ‘That Kofi danced is strange.’

(ε-​bɔ ŋwaŋwa)ι pres-​cop strange

Structurally, (14) contains a sentential subject, crosslinguistically a structure regularly mapped as a separate ι constituent. The macro prosodic structure of the sentence thus consists of two ι units:  the CP subject and the predicate phrase. Observe that the final items in each ι (ε-​kya-​υ ‘danced’ and ŋwaŋwa ‘strange’) are realized with a low falling F0 pattern, demarcating the constituent’s right edge.

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 53 ]

45

Other phonetic correlates of right edge ι-​marking in the language can be identified. These include the presence of pauses and pitch/​register reset in natural speech. The presence of a pause immediately following the first Intonational Phrase in (14) is evident in the lapse in articulation between the items ε-​kya-​υ and ε-​​bɔ. The duration of the pause is considerable, namely, 236.6 milliseconds. In addition, pitch reset can be detected in (14) following the pause. Observe that the low tone-​bearing items that follow the break have greater F0 values than the L%-​bearing clausal determiner8 at the right edge of ι1. The pitch track in (15) perhaps more clearly exemplifies register reset following the right boundary of ι1. Notice that the low tone-​bearing second syllables of odum ‘heart’ and ε-​fwι ‘boil’ are realized with higher F0 values than that of the low tone-​bearing item kya-​υ ‘danced’ at the right edge of ι1. (15) (Kε Kofi ε-​kya-​υ)ι (me odum ε-​fwι)ι st comp Kofi pst-​dance-​cldet 1 .sg heart pst-​boil ‘That Kofi danced angered me (i.e. made my heart boil).’

The pitch track in (15) also exemplifies the other phonetic correlates of Krachi ι-​marking previously discussed. The presence of L% can be detected in the low falling F0 values at the right edges of the two Intonational Phrases and in the lack of a rising tone realization of the clausal determiner at the right edge of ι1. Additionally, a substantial 590.4ms prosodic break divides the sentential subject (ι1) from the predicate (ι2) in natural speech. Having established the phonetic correlates of right edge ι marking in Krachi, we can proceed to evaluate the prosodic status of embedded complement clauses in the language. Prosodic considerations reveal that these TP structures are parsed as independent Intonational Phrases.9 Evidence for this characterization comes from the following observations. One,

[ 54 ] Anti-contiguity

the lexically high tone-​bearing complementizer fɛɛ surfaces with a low tone, indicating the presence of L%, the right edge ι boundary tone. This differentiates Krachi from languages like Luganda (Pak 2008) and Japanese (Ishihara 2014), where embedded clauses also form separate ι constituents, but C phrases instead with the embedded clause, not with the matrix verb as in Krachi. Two, a significant pause separates the complementizer from the embedded subject in natural speech. In (16), the duration of this break is 534.4 milliseconds. And three, pitch reset affects the F0 range of tones in the embedded clause immediately following the complementizer. This prosodic behavior is illustrated in the following pitch track. (16) (Fɛ kwarɛ10 fι-​g yι11 fɛɛ)ι (ɔkyι wυ ɛ-​mɔ bwatɛ 2nd.sg collect 2nd.sg-​eat comp woman the pst-​kill chicken ‘You believe that the woman slaughtered the chicken.’

wυ)ι the

The prosodic status of complement clauses in Krachi is unaffected by the presence of in-​situ interrogatives. The following data confirm that embedded complement clauses harboring in-​situ wh-​items are also parsed as Intonational Phrases. Two pitch tracks exemplifying the ι status of wh-​ internal embedded complement clauses are presented in (17). The data showcase clausal embedding under different bridge verbs (‘believe’ (17a) and ‘know’ (17b)), illustrating that the ι status of the embedded clause is independent of the embedding predicate. The three acoustic correlates of ι phrasing discussed earlier are clearly observable in each pitch track:  a 207.2ms break separates ι1 from ι2 in (17a) and a 363.8ms pause divides the two ιs in (17b); both complementizers surface with a low falling F0 instead of their lexically specified high tones; and pitch reset has occurred in both post-​complementizer domains, as revealed by the fact that low tone-​ bearing items in these domains surface with F0 values either at or above

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 55 ]

65

the pitch level of low-​toned syllables appearing at the right edge of the preceding Intonational Phrase. (17) a. (Fɛ kwarɛ fι-​g yι fɛɛ)ι (ɔkyι wυ 2nd.sg collect 2nd.sg-​eat comp woman the ‘What do you believe that the woman slaughtered?’

ɛ-​mɔ pst-​kill

nɛ)ι what

b. (Ama nyi fɛɛ)ι (Kwame ɛ-​mɔ nɛ)ι Ama know comp Kwame pres-​kill what ‘What does Ama know that Kwame slaughters?’

3.4.3. Prosodic Status of Bono Embedded Complement Clauses

The phonetic correlates of right edge ι marking in Krachi appear to be an areal feature of the Tano languages. Bono, Wasa, and Asante Twi all employ L%, prosodic breaks, and pitch reset to mark the right edge of an Intonational Phrase. The Bono datum in (18) illustrates the prosodic behavior of the

[ 56 ] Anti-contiguity

sentential subject construction, which as previously discussed, commonly induces ι phrasing of the clausal subject crosslinguistically. The presence of L% is signaled by the fact that items at the right edge of each ι surface with falling/​depressed F0s. The 201.4ms pause separating the sentential subject from the matrix predicate phrase is also observable in (18). (18) (Sɛ Kofi kum akoko kɛ̃)ι (yɛ Ama comp Kofi kill.pst chicken the do Ama ‘That Kofi slaughtered the chicken surprised Ama.’

nwanwa)ι strange/​surprise

Although the pitch track in (18) does not appear to manifest F0 reset following the ι1 boundary in any obvious way, the presence of register reset accompanying ι phrasing can be clearly observed in the following structure involving clausal complementation. In addition to the boosted/​reset F0 range in the embedded clause, the pitch track in (19) confirms that embedded complement clauses in Bono are parsed as separate Intonational Phrases. The presence of L% on the lexically High-​ bearing complementizer sɛ and the 422.7ms prosodic break that follows indicate the presence of an ι boundary dividing the matrix clause from the embedded clause. As was noted for Krachi, Bono complementizers phrase with the main clause, not the embedded clause. Once again, this differentiates Bono (like Krachi) from languages such as Luganda (Pak 2008)  and Japanese (Ishihara 2014), where embedded clauses are also realized as independent ι constituents, but where C phrases instead with embedded TP. (19) (Wo dwene sɛ)ι (mmema kɛ̃ be-​kum 2nd.sg think comp man.pl the 3rd.pl-​kill.pst ‘You think that the men slaughtered the chicken.’

akoko chicken

kɛ̃)ι the

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 57 ]

85

As was the case with Krachi, the prosodic status of Bono complement clauses is unaffected by the presence of in-​situ interrogatives. In what follows, I demonstrate that embedded clauses containing different in-​situ wh-​items also have the status of Intonational Phrases in the language. Two pitch tracks exemplifying the ι status of wh-​internal embedded complement clauses are presented in (20). The data once again show that the ι status of the embedded clause is independent of the wh-​item contained within. The acoustic correlates of ι phrasing discussed earlier are observable in each pitch track:  a 288.9ms break separates ι1 from ι2 in (20a) and a shorter 186.2ms pause divides the ι constituents in (20b); both complementizers surface with low falling F0s rather than their lexically specified high tones; and pitch reset has clearly occurred in the post-​complementizer domain, as revealed by the fact that low tone-​bearing items in this domain surface with F0 values either at or above the pitch level of low-​toned syllables appearing at the right edge of the preceding domain. (20) a. (Wo dwene sɛ)ι (bema kɛ̃ kum 2nd.sg think comp man the kill.pst ‘What do you think that the man slaughtered?’

[ 58 ] Anti-contiguity

abe)ι what

b. (Wo dwene sɛ)ι (bema kɛ̃ kum akoko kɛ̃ 2nd.sg think comp man the kill.pst chicken the ‘When do you think that the man slaughtered the chicken?’

dabe)ι when

These facts demonstrate a correlation between the prosodic status of embedded/​complement clauses and the possible occurrence of wh-​ in-​situ within that domain. In Bono, as we observed in the previous subsection with Krachi, embedded complement clauses are parsed as Intonational Phrases to the exclusion of the embedding complementizer and embedded wh-​in-​situ is permitted. In the next two subsections, we will observe the reverse pattern. Comparable domains in Wasa and Asante Twi are prosodically integrated with the main clause. That is, they do not have the status of Intonational Phrases, nor do they tolerate wh-​in-​situ within them.

3.4.4. Prosodic Status of Wasa Embedded Complement Clauses

Comparable phonetic correlates of right edge ι marking (i.e., L%, pause and pitch reset) exist in Wasa. This is demonstrated in (21) by way of a structure involving modification by a reason adjunct. The resulting utterance is parsed into two Intonational Phrases that do not correspond to the sentence’s two major syntactic constituents. In this construction, the lexically high tone-​bearing complementizer sɛ at the right edge of ι1 surfaces with a depressed/​falling F0, as we’ve seen before in the other languages. Following the complementizer is a considerable 497.2ms pause. The range of F0 values following the break is also clearly reset.

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 59 ]

06

(21) (Nti ɛsan sɛ)ι (ɛkwan de Kofi nɪnti o-​kum akoko no)ι because ɛsan comp hunger take Kofi therefore 3rd.sg-​kill.pst chicken the ‘Because Kofi was hungry, he slaughtered the chicken.’

Unlike Krachi and Bono, Wasa embedded complement clauses do not have the prosodic status of Intonational Phrases.12 That is, they are prosodically integrated with the main clause, as is crosslinguistically common (e.g., Bàsàá [Hamlaoui & Szendrői  2017]; Catalan [Feldhausen  2010]; Durban Zulu [Cheng & Downing  2007]; English [Downing  1970]; Hungarian [Hamlaoui & Szendrői  2017]; Swedish [Myrberg  2010]; Turkish [Kan  2009]; and Xhosa [Jokweni  1995], among many others). As illustrated in (22), the prosodic status of a Wasa complement clause is revealed by the tonal realization of complementizer sɛ, whose lexical high tone is not overridden by a low boundary tone in this construction. (Contrast the F0 realization of the complementizer in (22) with that in (21).) Furthermore, if the post-​C domain in this construction constituted an independent Intonational Phrase, we would expect to find an accompanying prosodic break and pitch reset. However, as (22) shows, there is neither a pause separating the complementizer from the embedded subject, nor does pitch reset occur in the embedded domain. Instead, we find a continuous pattern of F0 declination persisting from the main clause into the embedded clause. (22) (Wo dwene sɛ mɛrɛma no be-​kum akoko no)ι 2nd.sg think comp man.pl the 3rd.pl-​kill.pst chicken the ‘You think that the men slaughtered the chicken.’

[ 60 ] Anti-contiguity

Once again, these facts illustrate the correlation between the prosodic status of embedded complement clauses and the availability of wh-​ in-​situ across the Tano languages. Unlike Krachi and Bono, these domains fail to be mapped prosodically as Intonational Phrases. And unlike Krachi and Bono, these domains may not support in-​situ wh-​ expressions. (23) a. *Wo dwene 2nd.sg think

sɛ bɛrɛma no comp man the

kum ɛdiɛn (ɛndra) kill.pst what yesterday

b. *Wo dwene sɛ bɛrɛma no kum akoko no ɛhĩfa/​mmɛrɛ 2nd.sg think comp man the kill.pst chicken the where/​time bɛn/​ sen   (ɛndra) which/​how   yesterday

3.4.5. Prosodic Status of Asante Twi Embedded Complement Clauses

As with all the Tano languages surveyed in this chapter, L%, pause, and pitch reset mark the right edge of ι in Asante Twi. I illustrate this via the clausal coordination structure in (24), which furnishes an opportunity to simultaneously observe the right edges of two Intonational Phrases in the language. In this construction, the presence of L% at the right edges of the ι constituents can be detected by way of its F0 lowering effects on the lexically high tone-​bearing quantifier bi that surfaces at the right edge of both constituents. The two clauses are set off by pauses flanking the coordinator na. The duration of the first break is 450.9 milliseconds, while the second is slightly longer at 520.7ms. Additionally, the range of F0 values in the clause following the coordinator is reset.

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 61 ]

26

(24) (ɔkraman bi kuu ɔkraman bi)ι (na) (o-​kuu ɔkraman bi)ι dog some kill.pst dog some and 3rd.sg-​kill.pst dog some ‘A dog killed a dog and (s)he killed a dog.’

Like Wasa, embedded complement clauses in Asante Twi are prosodically integrated with the main clause.13 As shown in (25) in two different sentences, the lexically rising tone-​bearing complementizer sɛ faithfully surfaces with a rising F0 (following a brief initial fall), indicating the absence of a subsequent low ι boundary tone. The remaining phonetic correlates of right edge ι marking are absent in both examples as well. We fail to find a prosodic break separating the complementizer from the embedded clause in both sentences. Moreover, pitch reset does not occur in the embedded domain of either production. Rather, we observe a continuous pattern of F0 declination from the main clause into the embedded clause. (25) a. (ɛ bɛ tumi a-​yɛ sɛ me dɔɔ Ama)ι rd st 3 .sg fut able prf-​do comp 1 .sg love.pst Ama ‘It is possible that I loved Ama.’

[ 62 ] Anti-contiguity

b. (Yaw kaa sɛ Kofi Yaw say.pst comp Kofi ‘Yaw said that Kofi hit Ama.’

bɔɔ hit.pst

Ama)ι Ama

When paired with the Wasa findings, the prosodic status of embedded complement clauses in Asante Twi completes the negative correlation between lack of Intonational Phrase parsing and unavailability of wh-​ in-​situ that was illustrated in the previous subsection. Just as in Wasa, embedded complement TPs do not have the status of Intonational Phrases in Asante Twi, nor do they permit the occurrence of in-​situ interrogative expressions. (26) a. *Wo dwene sɛ nd 2 .sg think comp

Ama Ama

bɔɔ hwan hit.pst who

(ɛnora) yesterday

b. *Wo nim sɛ Ama saa ɛhĩfa/​(ɛ)berɛ bɛn/​sɛn (ɛnora) 2nd.sg know comp Ama dance.pst where/​time which/​how yesterday

3.4.6.  Analysis

Over the course of the last four subsections, I  have demonstrated that there is variation in the prosodic realization of Tano embedded TP complement clauses. These domains are parsed as Intonational Phrases in certain languages, but not in others. In doing so, I have shown that there is a correlation between the prosodic status of a clause and its ability to host in-​situ interrogative expressions. In those languages that map embedded TP clauses onto ι constituents, embedded wh-​in-​situ is permissible. Languages like Krachi and Bono meet this description. In those languages that do not realize TP clausal complements as Intonational Phrases (i.e.,

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 63 ]

46

Wasa and Asante Twi), the embedding of in-​situ interrogatives is prohibited. Table 3.2 summarizes this correlation. In this section, I propose an empirical generalization that accounts for the distributional variation of wh-​in-​situ in Tano and underpins this correlation. After applying this generalization to the distribution of partial wh-​movement in Tano in the following subsection, I revise it and present it in its final form in subsection 3.4.8. The generalization I  propose is prohibitional in nature and has an anti-​contiguity character: wh-​items may not form prosodic constituents with overt complementizers at the level of Intonational Phrase. Put another way, my (preliminary) claim is that wh-​and overt C may not phrase together under a single ι. I  will refer to this generalization as the anti-​ contiguity of wh-​and c (ACWC), harkening back to the conclusion reached in Chapter 2 that Richards-​style contiguity of wh-​and C is simply not tenable, at least for the Tano languages considered in this book. (27) anti-​contiguity of wh-​and c (preliminary) No ι may contain both overt C and wh-​.

The ACWC stands in diametrical opposition to Richards’s (2010, 2016) Contiguity-​theoretic proposal, in which wh-​in-​situ is licensed precisely when a wh-​item is able to phrase with its scope-​marking complementizer at PF. A number of other differences between Richards’s theory and the analysis put forth earlier are worth noting. For Richards, the requisite phrasing of C and wh-​takes place at the Phonological Phrase level, not the Intonational Phrase. Additionally, the generalization formulated in (27) restricts wh-​items from prosodically grouping with any and all overt complementizers, whereas Richards’s proposal concerns the phrasing of wh-​items with scope-​marking interrogative complementizers, which are not always overtly realized. The ACWC directly accounts for the variable distribution of wh-​ in-​situ in Tano. In languages like Krachi and Bono, where TP embedded complement

Table 3.2   CORREL ATION OF EMBEDDED CL AUSE PROSODIC STATUS AND AVAIL ABILITY OF WH-​ IN-​S ITU IN TANO

krachi

bono

wasa

asante twi

embedded clause = ι

✓

✓

×

×

wh-​ in-​situ

✓

✓

×

×

(embedded clauses)

[ 64 ] Anti-contiguity

clauses are mapped as independent Intonational Phrases, embedded in-​ situ wh-​constructions satisfy (27) in virtue of the fact that an ι boundary will always intervene between overt C and the embedded wh-​item. The presence of this intermediate ι boundary will thus prevent the two items from phrasing together under a single Intonational Phrase. The following representations illustrate. (28) Krachi a. (Ama nyi fεε)ι (Kwame ɛ-​mɔ nε)ι Ama know comp Kwame pst-​kill what ‘What does Ama know that Kwame slaughtered?’ Bono b. (Wo dwene sε)ι (bema kɛ̃ kum nd 2 .sg think comp man the kill.pst ‘What do you think that the man slaughtered?’

abe)ι what

In contrast, embedded in-​ situ wh-​constructions in languages like Wasa and Asante Twi violate the ACWC in virtue of the fact that embedded TP complement clauses in these languages fail to be mapped as Intonational Phrases. As such, no ι boundary separates C from the embedded wh-​item, yielding prosodic mappings in which the items phrase together under a single macro clausal Intonational Phrase when the root clause is spelled-​out.14 The representations in (29) exemplify this state of affairs. (29) Wasa a. *(Wo dwene sε bɛrɛma no kum εdiεn)ι 2nd.sg think comp man the kill.pst what Intended: ‘What do you think that the man slaughtered?’ Asante Twi b. *(Wo dwene sε Ama bɔɔ hwan)ι nd 2 .sg think comp Ama hit.pst who Intended: ‘Who do you think that Ama hit?’

If we assume that the ACWC holds unconditionally in Wasa and Asante Twi, the prohibition on embedded wh-​in-​situ in these languages follows directly.15 In addition to accounting for the variation in Tano embedded wh-​ in-​ situ distribution, the ACWC proposal also captures something more fundamental, namely, the asymmetry between root and embedded clause

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 65 ]

6

wh-​in-​situ. In the Tano languages, it is not possible to drop or omit the complementizer in embedded contexts. Thus, the ACWC correctly predicts that only some instances of embedded wh-​in-​situ will be prosodically well-​ formed and that this will depend on the prosodic status of the mapped embedded clause (as discussed earlier). By contrast, because C is not phonetically realized in Tano matrix contexts, the ACWC accounts for the fact that wh-​in-​situ (of semantically appropriate items) is widely available (i.e., distributionally unrestricted) in Tano root clauses. The ACWC is vacuously satisfied in these contexts. Thus, the ACWC generalization covers significant empirical ground. It allows us to reduce the variation in Tano embedded wh-​in-​situ distribution to a more basic difference concerning the way narrow syntactic structures are externalized at PF by way of prosodic mapping. Furthermore, it sheds light on why wh-​in-​situ is more prevalent in root domains than in embedded contexts.

3.4.7. Extending the Analysis: Deriving the Distribution of Partial Wh-​ Movement in Tano

Recall from section 3.4.4 that embedded complement clauses in Wasa do not have the status of independent Intonational Phrases at PF (see (22)). In keeping with the ACWC generalization formalized in (27), it follows that wh-​in-​situ is unavailable in embedded domains in the language. However, another consequence of the analysis is that all wh-​ items, whether in-​situ or moved, are expected to be restricted from appearing inside Wasa complement clauses. On the surface, then, it would appear that the ACWC generalization is stated too restrictively because a limited class of interrogative expressions may in fact surface within Wasa embedded clauses. Recall from section 3.3 that partial wh-​focus movement is attested in the language. The data originally presented in (13) are repeated here. (30) a. Wo dwene sɛ bɛrɛma bɛn na o-​kum akoko no 2nd.sg think comp man which foc 3rd.sg-​kill.pst chicken the ‘Which man do you think slaughtered the chicken?’ b. Wo dwene sɛ ɛdiɛn na bɛrɛma 2nd.sg think comp what foc man ‘What do you think that the man slaughtered?’

[ 66 ] Anti-contiguity

no the

kumiye kill.pst

c.

ɛhĩfa/​ na bɛrɛma no kum akoko no adiɛnti 2nd.sg think comp where/​why foc man the kill.pst chicken the ‘Where/​why do you think that the man slaughtered the chicken?’

Wo

dwene sɛ

If the post-​complementizer strings in (30) fail to be parsed into independent ι constituents, then the preceding structures would stand in violation of the ACWC and partial wh-​focus movement would wrongly be predicted to be ungrammatical in the language. In this section, I address this issue, applying the current preliminary analysis (unmodified) to account for both the existence of partial wh-​focus movement in Wasa and the non-​existence of partial wh-​focus movement in Asante Twi. In doing so, I develop a general account of the variation in availability of partial wh-​focus movement across the Tano languages. Although complement clauses in Wasa fail to achieve independent ι status, prosodic pathways exist in the language for inducing ι phrasing in embedded domains. These pathways may be exploited to smuggle wh-​expressions into otherwise unacceptable domains. One such pathway involves the creation of ι constituents through focus movement. In Wasa, focused constituents (in the case relevant to this discussion, partially moved wh-​DPs) are prosodically realized as independent Intonational Phrases. The following pitch track demonstrates that partially moved wh-​expressions in the language (and their associate focus markers) are set off from the remainder of the clause by way of flanking ι boundaries. Observe that the lexically high tone-​bearing complementizer sɛ surfaces with a low/​falling F0 pattern when appearing before the partially moved wh-​expression. As before, this indicates the presence of a right ι boundary. The 135ms pause that follows the complementizer is also consistent with the introduction of a major prosodic boundary in this position. The relative brevity of this break is due to the fact that this utterance was produced at a slightly higher rate of speech than the other productions considered in this chapter. In addition, the F0 range of the focused constituent is clearly boosted, indicating register reset associated with ι phrasing. The presence of an additional right L% ι boundary closing off the partially moved wh-​expression and focus marker is signaled by the L realization of the lexically high tone-​bearing focus marker na, the subsequent 166.5ms pause that follows the item, and the existence of pitch reset in the remainder of the utterance.

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 67 ]

86

(31) (Wo dwene sɛ)ι (ɛdiɛn na)ι (bɛrɛma 2nd.sg think comp what foc man ‘What do you think that the man slaughtered?’

no kumiye)ι the kill.pst

The phenomenon of focus movement feeding ι phrasing is crosslinguistically robust—​in languages closely related to Wasa, focus movement also induces ι mapping. In what follows, I show that focused partially moved wh-​ items in Bono also form Intonational Phrase constituents at PF. However, unlike Wasa, the focus marker does not phrase with the item in focus. The familiar indicators of ι phrasing in Tano are readily observable in the following example:  the lexically high-​bearing complementizer surfaces with a low/​ falling F0 pattern before the partially moved wh-​expression indicating the presence of a right ι boundary; a major prosodic break of 518.9ms follows; pitch reset is observed within the focused constituent; the wh-​ constituent is separated from the remainder of the clause by an additional pause of 426.8 milliseconds; and the F0 range of the material following this second break is clearly boosted/​reset. (32) (Wo dwene sɛ)ι (mmema benie)ι (ne be-​kum akoko kɛ̃)ι 2nd.sg think comp man.pl which foc 3rd.pl-​kill.pst chicken the ‘Which men do you think slaughtered the chicken?’

[ 68 ] Anti-contiguity

In this way, the existence of partial wh-​focus movement in Wasa does not conflict with the ACWC generalization despite the fact that embedded complement clauses in the language do not independently achieve ι status. Because focus fronting induces ι phrasing, a partially moved wh-​ expression will always be separated from the complementizer by an ι boundary in accordance with (27).16 Viewed in this light, the absence of partial wh-​ focus movement in Asante Twi (see (11), repeated here as (33)) remains puzzling. (33) a. *Wo dwene sɛ 2nd.sg think comp

hwan who

b. *Wo kaa sɛ dɛn 2nd.sg say.pst comp what c.

na foc

ɔ-​bɔɔ Ama 3rd.sg-​hit.pst Ama

na Kofi diiyɛ foc Kofi eat.pst

*Wo nim sɛ ɛhĩfa/​(ɛ)berɛ bɛn nd 2 .sg know comp where/​time which

na Kofi foc Kofi

saayɛ dance.pst

The unavailability of partial wh-​focus movement in the language is mysterious for another reason. As demonstrated in (12) and repeated in (34), short focus movement of non-​interrogative constituents is permitted. Thus, the restriction on partial wh-​focus movement in Asante Twi cannot simply be due to the fact that embedded clauses lack the requisite syntactic position to host the partially moved wh-​item (i.e., FocusP). (34) a. Wo dwene sɛ Kofi na ɔ-​bɔɔ Ama nd rd 2 .sg think comp Kofi foc 3 .sg-​hit.pst Ama ‘You think that it’s KOFI who hit Ama.’ b. Wo kaa sɛ nkonya no na Kofi diiyɛ nd 2 .sg say.pst comp cake the foc Kofi eat.pst ‘You said that it’s THE CAKE that Kofi ate.’ c.

Wo nim sɛ ɛnora na Kofi saayɛ nd 2 .sg know comp yesterday foc Kofi dance.pst ‘You know that it’s YESTERDAY that Kofi danced.’

I propose that the presence of partial wh-​focus movement in both Wasa and Bono and its absence in Asante Twi derives from an important difference among the Tano languages concerning the prosodic realization of (embedded) focused constituents. Whereas (embedded) constituent focus induces ι phrasing in Wasa and Bono, (embedded) constituent focus does not result in the creation of a special ι domain in Asante Twi. Consider the

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 69 ]

07

prosodic realization of the embedded focused item ɛnora in the following pitch track. (35) (Wo dwene sɛ ɛnora na barima 2nd.sg think comp yesterday foc man ‘You think that it’s YESTERDAY that the man ate.’

no the

diiyɛ)ι eat.pst

Unlike focused constituents in Wasa and Bono, (embedded) foci in Asante Twi are not separated off from the utterance by way of pauses or prosodic breaks. A continuous and regular trend of F0 declination persists in (35), indicating the absence of pitch reset. The prosodic realization of the complementizer also indicates the absence of a clause-​medial ι domain. Because its lexical rising tone is not overridden by a low boundary tone, we can conclude that the complementizer and focused constituent immediately following it phrase together under a single Intonational Phrase. In this way, the absence of partial wh-​focus movement in the language can be understood as a direct consequence of the prohibition expressed by the ACWC.17 In addition, the variation in availability of partial wh-​focus movement across the Tano languages can be attributed to language-​specific differences in whether or not displaced foci are realized as Intonational Phrases18. 3.4.8. Ref ining the ACWC

As hinted at in note 17, the ACWC in its current formulation is too strong. (36) anti-​contiguity of wh-​and c (preliminary) No ι may contain both overt C and wh-​.

Consider the case of long distance wh-​movement (e.g., (4b, 4f, 6b, 6d, 8b, 8d)). Recall that none of the Tano languages under investigation allows complementizer drop. This means that depending on how a language [ 70 ] Anti-contiguity

prosodifies its embedded clauses, the moved wh-​expression and C run the risk of phrasing together at the level of Intonational Phrase in violation of the ACWC. In the case of Krachi, where embedded C marks the right edge of ι effectively dividing the main clause from the embedded TP, the highest copy of the moved wh-​item will mark the left edge of ι and embedded C will demarcate its right edge. This is illustrated in (37). (37) Krachi a. Nɛ yι Ama nyi fɛɛ Kwame ɛ-​mɔ what foc Ama know comp Kwame pst-​kill ‘What does Ama know that Kwame slaughtered?’ b. (Nε

yι

Ama

fεε)ι

nyi

(Kwame

ε-​mɔ)ι

In the case of a language like Asante Twi, where embedded C does not mark the juncture of two ι constituents, a similar scenario obtains. The moved wh-​occurrence and the embedded C phrase together at the level of Intonational Phrase in virtue of the fact that structures such as these are prosodically realized as single ι constituents. This is shown in the following example. (38) Asante Twi a. Hwan na wo dwene sɛ who foc 2nd.sg think comp ‘Who do you think that Ama hit?’ b.

(Hwan

na

wo

dwene

sε

Ama Ama

bɔɔyɛ hit.pst

Ama

bɔɔyε)ι

The ACWC, therefore, wrongly predicts long distance wh-​movement to be ungrammatical in Krachi and Asante Twi. What about in Bono and Wasa? Recall from the previous section that displaced foci in Bono and Wasa are prosodically realized as independent ι units. Consequently, long distance extracted wh-​items in both languages will be shielded from C in satisfaction of the ACWC. In these cases, the ACWC makes the right cut. The following structures illustrate. (39) a. Bono (wh-​)ι (foc S V C)ι (. . . wh)ι b. Wasa (wh-​ foc)ι (S V C . . . wh)ι

This work-​around will not enable long distance wh-​movement structures in Krachi and Asante Twi to escape violation of the ACWC in its current Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 71 ]

27

formulation because in neither language are displaced foci prosodically realized as independent ι constituents. The current formulation of the ACWC is therefore too restrictive. Long distance wh-​movement in Krachi and Asante Twi reveals that wh-​ and overt C may in fact phrase together at the level of ι, but only when wh-​occupies the left edge of ι. When wh-​and overt C are ι-​mates and wh-​ occupies the right edge of ι (40a) or a non–​left peripheral position (40b), wh-​in-​situ is disallowed. (40) Asante Twi a. *(Wo dwene 2nd.sg think b. *(Wo

sε comp

dwene sε

2nd.sg think

comp

Ama bɔɔ Ama hit.pst

hwan)ι who

Ama

bɔɔ

hwan ɛnora)ι

Ama

hit.pst

who

yesterday

The generalization seems to be that overt C and “non-​prominent” wh-​ items are banned from phrasing together within a single Intonational Phrase, where “prominent” here means “situated at the left edge of ι.” I therefore revise my formulation of the ACWC as follows. (41) anti-​contiguity of wh-​and c (revised) a. No ι may contain both overt C and a non-​prominent wh-​. b. An occurrence is prominent if it occurs at the left edge of ι.

Folding in the concept of prominence draws an important parallel between this anti-​contiguity approach to wh-​licensing and Richards’s (2016) contiguity theoretic proposal (repeated here). (42) Richards’s (2016: 84–​85) condition on wh-​prosody: a. Given a wh-​ phrase α and a complementizer C where α takes scope, α and C must be dominated by a single φ, within which α is contiguity-​prominent. b. α is contiguity-​prominent within φ if α is adjacent to φ’s prosodically active edge.

Under both approaches, the structural notion of prominence plays a key determining role in wh-​licensing. For Richards, an occurrence counts as prominent if it is aligned with any prosodically active edge. Under my proposal, only reference to the left edge of a domain determines whether a constituent is prosodically prominent. The reformulated ACWC in (41) successfully derives the cases incorrectly predicted to be ungrammatical under the older version. Long distance

[ 72 ] Anti-contiguity

wh-​movement in Krachi (37a) and Asante Twi (38a) are now consistent with the ACWC—​although wh-​and overt C phrase together under ι in these cases, wh-​occupies a prominent position, thereby averting violation of (41). The revised ACWC also provides an explanation of why embedded wh-​items in Asante Twi must undergo long distance movement. Because displaced foci do not receive independent ιP status in the language, short wh-​movement to the left periphery of the embedded clause will not place wh-​in a prominent position (as it does, for example, in Wasa). Only long distance wh-​movement to the matrix left periphery will ensure that wh-​ escapes the clutches of the ACWC in Asante Twi because it is only in this ι-​initial position that wh-​becomes prominent.

3.5. CONCLUSION

An important discovery facilitated by the Minimalist paradigm shift is that many phenomena once thought to be purely syntactic in nature turn out instead to have more to do with the grammatical subsystems that interface with and impose well-​formedness conditions on syntactic representations. The findings and conclusions presented in this chapter accord nicely with this position. I have argued that variation in the distribution of non-​subject and non-​‘why’ wh-​in-​situ in four Tano languages is interface-​driven and ultimately prosodic in nature. My argument for this conclusion was based on an asymmetry in the availability of embedded wh-​in-​situ and partial wh-​focus movement in Wasa that suggests that non-​syntactic/​semantic factors play an active role in constraining the distribution of non-​subject/​ non-​‘why’ wh-​items. Further support came from the cross-​Tano observation that the ability of a wh-​item to appear in an in-​situ position strongly correlates with the prosodic status of its immediately containing clause. In Krachi and Bono, where complement clauses have the prosodic status of Intonational Phrases, all in-​situ interrogatives available in main clauses are also available in embedded domains. In Wasa and Asante Twi, however, where complement clauses do not have the status of Intonational Phrases, embedded wh-​in-​situ is restricted. To account for this correlation, I proposed the anti-​contiguity of wh-​and c (ACWC) generalization banning overt C and non-​prominent wh-​items from phrasing together within a single Intonational Phrase. Because Tano root clauses lack overtly realized complementizers, the ACWC accounts for the fact that wh-​ in-​situ (of semantically appropriate items) is distributionally unrestricted in matrix domains at PF. However, because the realization of C is obligatory in Tano embedded contexts, the ACWC also accounts for the fact that certain

Ta n o i n - s i t u I n t e r r o g at i v e Di s t r i b u t i o n  

[ 73 ]

47

instances of embedded wh-​in-​situ will be restricted. Those TP complement clauses that are mapped as Intonational Phrases are able to support wh-​in-​situ, as in Krachi and Bono, because under such a prosodic mapping wh-​and C are separated by an ι boundary. Conversely, those embedded clauses that are not mapped as Intonational Phrases cannot harbor in-​situ interrogatives, as in Wasa and Asante Twi. Due to the fact that the ACWC references overt complementizers, it makes an interesting typological prediction with respect to languages that manifest wh-​in-​situ asymmetries of the sort investigated in this chapter. Although there will exist many languages allowing wh-​in-​situ in root, but not embedded clauses (e.g., Wasa and Asante Twi), languages allowing wh-​ in-​situ in embedded, but not main clause contexts should either be exceedingly rare or nonexistent. I know of no counterexamples to this generalization. This prediction stems from the fact that embedded clauses are far more likely to contain overt exponents of C than matrix clauses. Therefore, satisfying the ACWC will always be more difficult in embedded domains than in root contexts. To the extent that the ACWC successfully accounts for the distributional variation in Tano wh-​in-​situ patterns, the prosodic approach advanced in this chapter seems encouraging. One deeper issue that remains difficult to reckon with, however, is the question of why a generalization like the ACWC would hold in the first place. What interface principle or design feature would underlie such a prohibition?19 And assuming one exists, could such a generalization be motivated for languages outside the Tano group? I’ll hold off on speculating on the former question until the book’s conclusion. Regarding the latter question, further support for the ACWC will come in the next chapter, when it is pressed into service outside the Tano phylum to explain two puzzling wh-​asymmetries in the Nupe language. Regarding the former issue, our present ignorance in this domain does not, I believe, diminish or discredit the discovery of generalizations like the ACWC. Other proposals claiming that wh-​in-​situ is prosodically licensed (for at least some subset of interrogative expressions in a given language) face a similar challenge. Richards’s (2010, 2016) proposal, for instance, is difficult to ground in PF interface pressures or optimal design considerations. This seems to be the current state of affairs when it comes to research on PF well-​formedness conditions more generally. This may be the case because there simply is no deep explanation for the existence of certain PF well-​formedness conditions/​constraints. Or, perhaps more likely, it may be rooted in the fact that our understanding of the syntax-​ phonology interface is currently underdeveloped. It is my hope that the research reported in this chapter stimulates further development into this burgeoning field of inquiry to close the gap between our understanding of what we observe and why we observe it. [ 74 ] Anti-contiguity

CHAPTER 4

An Anti-​contiguity Approach to Nupe Interrogative Distribution

4.1. INTRODUCTION

My goal in this chapter is to deploy the Anti-​contiguity generalization developed in Chapter 3 (the ACWC) to give a prosodic account of two puzzling wh-​asymmetries in Nupe, a central Benue-​Congo language of Nigeria. I  consider how Richards’s (2010, 2016)  contiguity theoretic account of wh-​prosody would attempt to capture the relevant facts and conclude that an ACWC approach fares better. I conclude therefore, that Nupe wh-​ asymmetries furnish striking additional evidence for the Anti-​contiguity theory proposed in this book, as the ACWC was not originally formulated to derive the Nupe facts presented in this chapter. The chapter is organized as follows. I begin in section 4.2 by laying out the two wh-​asymmetries in detail. In section 4.3, I apply Richards’s contiguity theory to the Nupe data and argue against its feasibility. Section 4.4 is the core of the chapter. There I provide an Anti-​contiguity prosodic analysis by (i) discussing how Intonational Phrase boundaries can be detected in Nupe, (ii) establishing the prosodic status of Nupe embedded clauses, and (iii) using these findings to derive the two wh-​asymmetries via the ACWC. Section 4.5 concludes the chapter with some brief remarks that segue into the final chapter.

Anti-contiguity. Jason Kandybowicz, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780197509739.001.0001

67

4.2. TWO MYSTERIOUS WH-​  ASYMMETRIES 4.2.1. Asymmetry 1

Nupe1 is a wh-​ focus language. Interrogative items are fronted to a special left peripheral focus position and the focus marker, which will not be relevant to our discussion, appears clause-​finally. As the following set of examples illustrates, interrogative and non-​interrogative foci are in complementary distribution. That is, they occupy the same dedicated focus position—​the two XPs compete for a single peripheral focus position and may not co-​occur regardless of their order. In other words, Nupe allows one focused expression (interrogative or not) per clause. (1)

a. Ké Nànǎ si _​_​_​ tsúwó o what Nana buy.pst yesterday foc ‘What did Nana buy yesterday?’ b. Kèké Nànǎ si _​_​_​ tsúwó o bike Nana buy.pst yesterday foc ‘Nana bought THE BIKE yesterday.’ c.

Tsúwó Nànǎ si kèké _​_​_​ yesterday Nana buy.pst bike ‘Nana bought the bike YESTERDAY.’

d. *Ké tsúwó what yesterday e. *Tsúwó ké yesterday what

Nànǎ si _​_​_​ _​_​_​ Nana buy.pst Nànǎ si _​_​_​ _​_​_​ Nana buy.pst

o foc

o foc o foc

As might be expected, embedded constituents (both arguments and adjuncts) can be focused by moving to a left peripheral position in the subordinate clause. The displaced focal item occupies a position below C, which is obligatory in the language and may not be dropped under any discourse context. (2)

a. Musa kpe *(gànán) kèké Nànǎ si _​_​_​ tsúwó o Musa know.pst comp bike Nana buy.pst yesterday foc ‘Musa knew that Nana bought THE BIKE yesterday.’ b. Musa gbín-​gàn *(kó) tsúwó Nànǎ si kèké _​_​_​ Musa lose-​say.pst comp yesterday Nana buy.pst bike ‘Musa asked whether Nana bought the bike YESTERDAY.’

[ 76 ] Anti-contiguity

o foc

The asymmetry I am interested in exploring concerns the fact that although non-​interrogative XPs may undergo embedded focus movement (2a–​b), embedded wh-​items may not appear in embedded left peripheral focus positions (3). That is, despite the availability of an embedded focus position, Nupe does not allow embedded questions. (3)

a. *Musa kpe *(gànán) ké Nànǎ si _​_​_​ tsúwó Musa know.pst comp what Nana buy.pst yesterday Intended: ‘Musa knew what Nana bought yesterday.’

o foc

b. *Musa gbín-​gàn *(kó) kánci Nànǎ si kèké _​_​_​ Musa lose-​say.pst comp when Nana buy.pst bike Intended: ‘Musa asked when Nana bought the bike.’

o foc

To express the intended meanings in (3), Nupe resorts to non-​interrogative relative clause formation (4). Put another way, the work-​around is to avoid using an embedded wh-​  item. (4)

a. Musa kpe ejan na Nànǎ si _​_​_​ tsúwó na Musa know.pst thing rel Nana buy.pst yesterday prt ‘Musa knew what (lit. ‘the thing that’) Nana bought yesterday.’ b. Musa gbín-​gàn kámi na Nànǎ si kèké _​_​_​ na Musa lose-​say.pst time rel Nana buy.pst bike prt ‘Musa asked when (lit. ‘the time that’) Nana bought the bike.’

Unable to form indirect questions, the language does not tolerate partial wh-​movement either. As (5a) shows, a partially moved wh-​item may not appear in an embedded clause. The embedded interrogative must undergo long-​distance movement to the matrix left periphery (5b). (5)

a. *Musa gàn gànán ké Nànǎ si _​_​_​ tsúwó o Musa say.pst comp what Nana buy.pst yesterday foc Intended: ‘What did Musa say Nana bought yesterday?’ b. Ké Musa gàn gànán Nànǎ si _​_​_​ tsúwó o what Musa say.pst comp Nana buy.pst yesterday foc ‘What did Musa say Nana bought yesterday?’

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 77 ]

87

4.2.2. Asymmetry 2

Unlike the Tano languages considered in Chapter 3, Nupe is not a wh-​ in-​ situ language (6a, c). However, like English, it allows wh-​in-​situ in multiple question contexts when a superior wh-​expression has undergone movement to the left periphery. This is shown in (6b, d). (6)

a. *Musa si ké Musa buy.pst what b. Zě si ké o who buy.pst what foc ‘Who bought what?’ c.

*Musa Musa

á le kánci prf sleep when

d. Zě á le kánci o who prf sleep when foc ‘Who has slept when?’

The asymmetry of interest here concerns the fact that wh-​in-​situ in multiple questions is limited to root contexts. As illustrated in (7), embedded wh-​in-​situ is ungrammatical. (7)

*Musa kpe gànán zě si Musa know.pst comp who buy.pst Intended: ‘Musa knew who bought what.’

ké what

o foc

The ungrammaticality of (7)  is not surprising given the unavailability of embedded questions in the language (3). In other words, one might attribute the ill-​formedness of (7) to the placement of wh1-​‘who’ in the left periphery of the embedded clause. Therefore, to demonstrate the restriction on embedded wh-​in-​situ in multiple questions, we must move the initial wh-​item out of the embedded domain, leaving the lower interrogative in its base position. As shown in (8) for different embedding verbs and complementizers, the result is always the same. (8)

ké what

a. *Zě Musa who Musa

kpe gànán _​_​_​ know.pst comp

u-​si 3rd.sg-​buy.pst

b. *Zě Musa who Musa

gbín-​gàn lose-​say.pst

u-​si ké o rd 3 .sg-​buy.pst what foc

[ 78 ] Anti-contiguity

kó _​_​_​ comp

o foc

One reaction to the impossibility of the structures in (8) might be to attribute ungrammaticality to a Comp-​trace effect. However, as Kandybowicz (2009) shows, long distance subject extraction across embedded C in the language is in fact possible when a resumptive pronoun cliticizes to the embedded verb (as in the case of the u-​ occurrences in (8)). In fact, the structures in (8)  become fully acceptable if the in-​situ wh-​items are replaced with non-​interrogative expressions, as in (9). (9)

a. Zě Musa kpe gànán _​_​_​ u-​si who Musa know.pst comp 3rd.sg-​buy.pst ‘Who did Musa know bought the bike?’

kèké bike

b. Zě Musa gbín-​gàn kó _​_​_​ u-​si kèké who Musa lose-​say.pst comp 3rd.sg-​buy.pst bike ‘Who did Musa ask whether they bought the bike?’

o foc

o foc

This effectively reveals the asymmetry between matrix (6b,d) and embedded (8) wh-​in-​situ in the language.

4.2.3. Generalizing the Asymmetries

Asymmetry 1 concerns the fact that moved wh-​items may not surface in embedded contexts. Asymmetry 2 concerns the fact that in-​situ wh-​ items may not surface in embedded contexts. Putting these facts together, we arrive at the following generalization: Nupe wh-​phrases occur only in root environments. The question now is why.

4.3. UNDER THE LENS OF CONTIGUITY THEORY

Recall Richards’s (2016) contiguity theoretic approach to wh-​ distribution. (10) Given a wh-​ phrase α and a complementizer C where α takes scope, α and C must be dominated by a single ϕ, within which α is contiguity prominent (i.e., adjacent to a prosodically active edge of ϕ). (Richards 2016: 84–​85)

In the cases at hand, the relevant contiguity relations concern wh-​ and embedded C, not the null matrix complementizer. This is because in the constructions we are investigating the offending wh-​items are attempting to take narrow scope. Therefore, to account for the Nupe asymmetries in (3) and (8), Richards could exploit two features of (10):

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 79 ]

08

(11) a. Wh-​& embedded C are not dominated by a single ϕ. b. Wh-​& embedded C are dominated by a single ϕ, but wh-​is not adjacent to an active edge of ϕ.

The line of approach in (11a) is unpromising because, as I will show in the next section, embedded C does not prosodically divide the embedded clause from the main clause. In other words, wh-​and embedded C do in fact phrase together at the relevant prosodic level. In the case of (3), Richards’s own mechanism of contiguity adjunction (see (12)), which applies under condition of prosodic adjacency, would guarantee that wh-​ and embedded C are dominated by a single ϕ even if a ϕ boundary were to separate C from the moved wh-​  item. (12) contiguity adjunction (richards 2016: 108) Take a pair of adjacent prosodic nodes and make one of them a daughter of the other.

The analytical direction in (11b) is also unpromising. This has to do with the fact that in the case of (8), any appeal to wh-​not appearing adjacent to an active edge of ϕ will wrongly rule out wh-​in-​situ in root contexts (e.g., (6b, d)). This is because in both (8) and (6b, d) the relevant wh-​items occupy identical syntactic and prosodic positions within their immediately containing clauses. I believe that an Anti-​contiguity approach to these Nupe asymmetries fares better than a contiguity theoretic one. In the next section, I show that not only can both asymmetries be straightforwardly derived by appealing to the ACWC, but that my anti-​contiguity proposal makes an important correct prediction about Nupe that could not have been made under the contiguity theoretic model. In a nutshell, I’ll argue that Nupe patterns just like Wasa and Asante Twi from Chapter 3 in that overt embedded C does not introduce an ι boundary, yielding prosodic mappings in which non-​ prominent embedded wh-​items and C are ι-​mates in direct violation of the ACWC.

4.4. UNDER THE LENS OF ANTI-​C ONTIGUITY THEORY

As in the previous chapter, I begin this section with a brief discussion of the methodology employed in the upcoming prosodic analysis. I  subsequently develop an approach for Intonational Phrase boundary detection in Nupe. Armed with this tool, I  then investigate the prosodic status of

[ 80 ] Anti-contiguity

embedded clauses in the language, arguing that like Wasa and Asante Twi, the embedded TP domain does not represent an independent ι constituent in Nupe. The section concludes with a discussion of how the ACWC elegantly derives the two Nupe wh-​asymmetries and makes a rather striking prediction that is borne out in the language.

4.4.1.  Methodology

Four native speakers participated in this study. Two participants were recorded and a total of two hours of recordings were obtained. The other two participants listened to the recorded utterances and provided judgments about the naturalness of those productions. Speakers produced material one sentence at a time in unconnected speech and were asked to produce their utterances “naturally,” avoiding deliveries that were too fast, slow, or careful sounding. Recordings were then presented to the other consultants to judge the naturalness of the productions. Only productions judged “natural” sounding were considered in the subsequent analysis. Consultants who rejected certain tokens as unnatural cited overly rapid delivery, missing pauses/​breaks, insufficient pause length, and lack of fluid delivery in their justifications. These objections were strikingly similar to the ones voiced by the Tano language consultants from Chapter 3. As was the case in Chapter 3, although prosodic data were collected from two Nupe speakers, the pitch tracks presented in this chapter come from a single native speaker. The choice to limit the Nupe pitch track data in this way was made to ensure that the prosodic facts presented are directly comparable. The Nupe pitch tracks found in this chapter were produced by a male in his mid-​50s. As in Chapter  3, I  rely on the existence of prosodic breaks and pitch/​ register reset, among other phonological considerations, to diagnose the existence of Intonational Phrase boundaries. In doing so, I once again make the following two assumptions. (See the Methodology section in Chapter 3 for a justification of the first of these assumptions.) One, pauses greater than 100 milliseconds constitute major prosodic breaks/​ breath group boundaries and therefore reveal the juncture between two Intonational Phrase constituents. Pauses less than 100 milliseconds, on the other hand, do not constitute true breaks and thus do not indicate ι junctures. Two, I will posit pitch/​register reset when I observe syllables of tone T surfacing at or above the F0 level of T-​toned syllables appearing either before or at the right edge of the preceding putative Intonational Phrase boundary. Both durational and pitch measurements were made using Praat (Boersma & Weenink 2017).

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 81 ]

28

4.4.2. Detecting Nupe ι Boundaries

I rely on three indicators to detect the presence of ι boundaries in Nupe. Two of these are phonetic correlates and the other indicator is phonological in nature. The three ι boundary signposts are: (i) pauses/​prosodic breaks, (ii) F0 register reset, and (iii) the suspension of regular and productive phonological processes. Let’s examine each in turn. Consider the prosodic realization of a moved/​extraposed clause in Nupe, a domain that is sometimes parsed as an independent Intonational Phrase cross-​linguistically (Downing 1970; Hsu 2015; Hamlaoui & Szendrői 2017; among others). (13) (Gànán Màmŭ ba nakàn)ι (zà comp Mamu cut.pst meat person ‘That Mamu cut the meat, everyone knew.’

ndondo every

kpe)ι know.pst

The ι status of extraposed clauses in Nupe is clearly indicated by way of prosodic breaks and pitch reset. A  398ms pause separates the displaced clause from the root clause and register reset can be observed by comparing the F0 realization of the initial post-​break item zà ‘person’ with the final syllable of nakàn ‘meat’ occurring at the right edge of the first prosodic domain, both low tone-​bearing items. Because the L tone on ‘person’ is realized at a significantly higher F0 value than the L tone on the final syllable of ‘meat’, we can confidently assess that pitch reset has occurred, in this case, a consequence of the moved clause phrasing as an independent ι. We observe similar indicators in left dislocation (topicalization) structures in the language, another domain that is often realized as a special separate Intonational Phrase unit crosslinguistically (see, for example,

[ 82 ] Anti-contiguity

Frascarelli 2000 on Italian). In the following structure, a single word direct object has been topicalized via movement2 to the left periphery and a third person pronoun wu:n occurs lower in the clause in the object’s base thematic position, as in clitic left dislocation structures. (14)

(Kèké)ι (mi bike 1st.sg ‘As for the bike, I rode it.’

tú ride.pst

wu:n 3rd.sg

o)ι top

A 149.9ms break divides the topic from the remainder of the clause, indicating the presence of two ι constituents in this construction. In addition, we clearly observe pitch reset after the break. Compare the F0 realization of the H tone-​bearing tú ‘ride’ following the break with that of the second (H tone-​bearing) syllable of kèké ‘bike’ before the break. Because the F0 value of ‘ride’ is greater than the F0 value of the second syllable of ‘bike’, we can conclude the pitch reset has occurred.3 Another construction that allows us to observe these two phonetic indicators of ι boundary marking in the language is the conditional construction. The antecedents of Nupe conditional clauses seem to constitute independent ι constituents, as indicated by the 166.4ms break dividing the antecedent from the consequent and the register reset that occurs in the main clause in the following pitch track. To observe pitch reset in (15), compare the realization of the L tone on ‘Gànǎ’ with the significantly lower realization of the L tone on the final syllable of ‘meat’ at the right edge of the conditional clause. (15) (Musa gá è ba nakàn)ι (Gànǎ à pa eci)ι Musa if pres cut meat Gana fut pound yam ‘If Musa is cutting the meat, then Gana will pound the yam.’

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 83 ]

48

Coordinated clauses (argued by Downing (1970) to be obligatorily parsed into what we would now call separate Intonational Phrases in English) further demonstrate these two ι boundary indicators and reveal a third means of detecting ι junctures. Consider the following pitch track, where a 358.5ms pause divides the two clauses and pitch reset is indicated by way of a slightly boosted F0 value for the perfect marker á in the second clause as compared to its value in the first clause. (16) (Musa á nakàn ba)ι (u ma á cènkafa du)ι Musa prf meat cut.pst 3rd.sg and prf rice cook.pst ‘Musa has cut the meat and he has cooked the rice.’

This example affords us a third way to detect ι boundaries in the language. This third indicator is phonological in nature. In Nupe, regressive vowel height assimilation and hiatus resolution via glide formation are regular and productive phonological processes (Kawu 2000). These processes, however, are ι-​bounded, meaning that the operations apply only to items within the same Intonational Phrase (not that all words within ι share the same height specification, in the case of vowel height assimilation). [ 84 ] Anti-contiguity

The following examples illustrate how these processes work ι-​internally. Consider first regressive height assimilation via the sentential subject construction. In this configuration, when the clause-​initial complementizer is followed by a vowel-​initial subject with mismatched height features, the low vowels on the complementizer regressively assimilate to match the high vowel on the subject. (17) ι-​internal regressive assimilation /​gànán + u/​ → [gùnún u] (Gùnún u si dòkò mafi rd comp 3 .sg buy.pst horse please.pst ‘That s/​he bought a horse pleased Musa.’

Musa)ι Musa

Next, consider hiatus resolution. Nupe disallows heterosyllabic sequences and remedies offending outputs by converting the initial (non-​low) vowel into glide. We observe ι-​internal hiatus resolution in simple instances of object wh-​movement with pronominal subjects, which are always vowel-​initial in the language. Consider the case in (18), where the word-​final vowel on the fronted wh-​item and the vowel on the following pronominal subject form a hiatus that is resolved by means of glide formation on the first vowel. (18) ι-​internal hiatus resolution via glide formation /​ké + u/​ → [kj û] (Kj û ko _​_​_​ o)ι what 3rd.sg grind.pst foc ‘What did s/​he grind?’

Returning to (16), the presence of an ι boundary separating the coordinated clauses can also be detected phonologically now. The putative ι boundary dividing the clausal conjuncts is signaled by the fact that regressive assimilation is blocked. The verb in conjunct1 surfaces as [ba], not *[bu]. (19) (Musa á nakàn ba/​*bu)ι (u ma á cènkafa du)ι (Compare with (17)) rd Musa prf meat cut.pst 3 .sg and prf rice cook.pst

In this way, we have three tools at our disposal for detecting ι boundaries in Nupe: pauses/​prosodic breaks; F0 register reset; and the suspension of regular and productive phonological processes such as regressive vowel height assimilation and glide formation. The last illustrative example we will consider demonstrates all three. In German, certain adverbial clauses are realized as separate Intonational Phrases (Truckenbrodt 2005). Likewise, some adverbial clauses in Nupe are mapped as independent prosodically non-​integrated ι constituents. N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 85 ]

68

‘Because’ clauses are one such case.4 (See note 4 for others.) As shown in the following pitch track, this is revealed by the presence of a significant prosodic break (394.3ms) and pitch reset. (20) (Ebó

na Musa á

nakàn ba

because rel Musa prf meat

na)ι (Nànǎ á

ze

ewùn be

wu:n yin)ι

cut.pst prt Nana prf turn. pst anger with 3rd.sg prt

‘Because Musa had cut the meat, Nana was angry at him.’

Further evidence for the ι boundary dividing the ‘because’ clause from the root comes from the fact that regressive assimilation across the two clauses is blocked when the matrix clause contains a pronominal subject. The example in (21) illustrates. (21) (Ebó na Musa á nakàn ba na/​*nu)ι (u á ze because rel Musa prf meat cut.pst prt 3rd.sg prf turn.pst ewùn be wu:n yin)ι anger with 3rd.sg prt ‘Because Musa had cut the meat, s/​he was angry at him.’    (Compare with (17))

4.4.3. The Prosodic Status of Nupe Embedded Clauses

Four exponents of C comprise the Nupe complementizer system, each of which introduces a semantically distinct clause type (Kandybowicz 2008; Kawu 1990, 1999). Clausal complements whose propositional content is asserted/​presupposed are headed by gànán. (22) Musa wo gànán Nànǎ á nakàn Musa hear.pst comp Nana prf meat ‘Musa heard that Nana has cooked the meat.’

[ 86 ] Anti-contiguity

du cook.pst

Clauses whose propositional content is not asserted involve the two-​part particle ke  .  .  .  na. Ke precedes the embedded clause and na follows it. Neither particle may be omitted. (23) U bè ke Màmŭ má nakàn du na rd 3 .sg seem.pst comp Mamu know.pst meat cook.pst prt ‘It seemed that Mamu knew how to cook meat.’

Indirect (non-​wh-​) questions are introduced by the complementizer kó. (24) Musa gbín-​gan

kó

Màmŭ

á

nakàn du

Musa lose-​say.pst comp Mamu prf meat ‘Musa asked whether Mamu had cooked the meat.’

cook.pst

Lastly, relative clauses also employ a two-​part particle na . . . na in which the identical na components flank the relative clause. As before, neither piece may be dropped. (25) Musa si nakàn na Màmŭ du Musa buy.pst meat comp Mamu cook.pst ‘Musa bought the meat that Mamu cooked.’

na prt

Regardless of the choice of complementizer, overt C does not induce ι phrasing of its clausal complement. As the following pitch tracks demonstrate, all Nupe embedded clauses are prosodically integrated into the main clause ι. Consider first gànán-​headed clausal complements of V. (26) Gànán-​headed clausal complement of V (Musa wo gànán Nànǎ á nakàn du)ι Musa hear.pst comp Nana prf meat cook.pst ‘Musa heard that Nana has cooked the meat.’

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 87 ]

8

Here we observe no pause after C and no F0 register reset. The absence of pitch reset is visible in the continuous F0 downtrend associated with decreasing air pressure in the absence of a major prosodic break. If we modify the example to include a pronominal embedded subject, we find that the vowels on C regressively assimilate in height to the vowel on the following subject. This is consistent with there being no ι boundary between C and the embedded subject to block assimilation. This is illustrated in (27). (27) (Musa wo gùnún/​*gànán u á nakàn du)ι rd Musa hear.pst comp 3 .sg prf meat cook.pst ‘Musa heard that s/​he has cooked the meat.’

We can also consider gànán-​headed clausal complements of N in this light. The pitch track in (28) reveals a similar prosodic structure. (28) Gànán-​headed clausal complement of N (A wo labarí gànán Màmŭ ba nakàn)ι 3rd.pl hear.pst news comp Mamu cut.pst meat ‘They heard the news that Mamu cut the meat.’

As with gànán-​headed clausal complements of V, we find no prosodic break either preceding or following C. We fail to observe pitch reset as well. And like the case previously considered, there is no suspension of regressive assimilation when the embedded subject is pronominal, as shown in (29). (29) (A wo labarí gùnún/​*gànán u ba nakàn)ι 3rd.pl hear.pst news comp 3rd.sg cut.pst meat ‘They heard the news that s/​he cut the meat.’

[ 88 ] Anti-contiguity

Now consider clausal complements of V introduced by ke . . . na. As is clear in the following pitch track, there is no prosodic break associated with this structure. (30) Ke . . . na-​headed clausal complement of V (U bè ke Màmŭ má nakàn 3rd.sg seem.pst comp Mamu know.pst meat ‘It seemed that Mamu knew how to cook meat.’

du na)ι cook.pst prt

Additionally, there is no F0 register reset associated with this structure. The H tone on má ‘know’ surfaces either at or below the F0 value of preceding M tone-​bearing syllables (u and ke). Lastly, we fail to observe our third ι boundary indicator in this construction. As revealed by the following datum, glide formation can and does occur across the C/​TP boundary, consistent with there being no ι break in this position. (31) (U bè kj/​*ke u má nakàn 3rd.sg seem.pst comp 3rd.sg know.pst meat ‘It seemed that s/​he knew how to cook meat.’

du na)ι cook.pst prt

The third embedded clause type we need to consider is the kind headed by kó. The same facts play out here as well. We observe no prosodic break between the main and embedded clauses and there is a clear absence of F0 register reset in the embedded domain. The following pitch track shows that once again, the embedded clause is prosodically integrated into the main clause, at least at the level of ι. (32) Kó-​headed clausal complement of V (Musa gbín-​gan kó Màmŭ á nakàn Musa lose-​say.pst comp Mamu prf meat ‘Musa asked whether Mamu had cooked the meat.’

du)ι cook.pst

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 89 ]

09

Additional supporting evidence for the ι-​level prosodic integration of kó-​ headed embedded clauses with main clauses comes from the fact that productive processes such as hiatus resolution via glide formation regularly occur across the C/​TP boundary in these constructions. Because glide formation is not suspended, we have evidence for the absence of an ι boundary dividing the main from the embedded clause. (33) (Musa gbín-​gan kw/​*kó û5 á nakàn du)ι rd Musa lose-​say.pst comp 3 .sg prf meat cook.pst ‘Musa asked whether s/​he had cooked the meat.’

The final embedded clause type to consider is the adnominal (i.e., non-​adverbial (see note 4)) relative clause construction, which in Nupe is introduced by the discontinuous two-​ part relative complementizer na . . . na. Unlike the case of adverbial clauses with relative clause syntax, which are prosodically disintegrated from the main clause (e.g., ‘because’ clauses (20), ‘although’ clauses (note 4, example (i)), and ‘when’ clauses (note 4, example (ii)), Nupe adnominal relative clauses are prosodically integrated constituents. As in the other instances of clausal embedding we’ve examined in this section, no prosodic break or F0 register reset is observed. That is, adnominal relative clauses and all other Nupe embedded clause types turn out to be highly uniform prosodically. The following pitch track shows that once again, the embedded domain is prosodically integrated into the matrix domain at the level of the Intonational Phrase. (34) Na . . . na-​headed relative clauses (Musa si nakàn na Màmŭ Musa buy.pst meat comp Mamu ‘Musa bought the meat that Mamu cooked.’

[ 90 ] Anti-contiguity

du na)ι cook.pst prt

In the presence of pronominal subjects of relative clauses, the clause-​initial C particle na can and does regressively assimilate in height to the vowel of the following subject. Once again, this is fully consistent with the absence of an ι boundary in this position. (35) (Musa si nakàn nu/​*na u Musa buy.pst meat comp 3rd.sg ‘Musa bought the meat that s/​he cooked.’

du na)ι cook.pst prt

We have seen in this section that Nupe embedded complement clauses are prosodically uniform. Regardless of which C introduces the subordinate clause, embedded domains are prosodically integrated into the main clause at the ι level. Therefore, because C does not induce ι phrasing of its complement, the macro prosodic structure of Nupe embedded clauses is identical to the macro prosodic structure of Wasa and Asante Twi embedded clause structures: (36) (TP . . . v . . . C TP . . . v . . .)ι

4.4.4. Deriving the Two Asymmetries via the ACWC

Armed with the prosodic discoveries laid out in the previous section, we can now derive Nupe’s two mysterious wh-​asymmetries from the ACWC (repeated here). (37) anti-​contiguity of wh-​and c (acwc) a. No ι may contain both overt C and a non-​prominent wh-​. b. An occurrence is prominent if it occurs at the left edge of ι.

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 91 ]

29

A direct consequence of the fact that overt Nupe complementizers do not induce ι phrasing of their complements is that C and its clausal complement are rendered prosodically contiguous. Given the ACWC, it follows that wh-​items in the language are generally restricted from appearing in embedded contexts,6 whether moved (3) (i.e., Asymmetry 1) or in-​situ (8) (i.e., Asymmetry 2) because neither moved nor in-​situ wh-​ expressions in embedded environments count as “prominent” in the eyes of the ACWC, as neither align with the left edge of an ι constituent. (38) macro prosodic structure of embedded wh-​constructions in Nupe 👎 (s v C . . . wh-​ . . .)ι

For this reason, if a wh-​item associated with an embedded clause is to be used at all, it must typically7 undergo long distance movement to the root left periphery (5b), where it occupies a prosodically prominent position at the left edge of the matrix ι. The root vs. embedded wh-​in-​situ asymmetry (Asymmetry 2) is a consequence of the fact that matrix C is obligatorily null in Nupe. Because the ACWC polices the prosodic relationship between wh-​items and overt complementizers, it cannot be violated by wh-​in-​situ in the matrix domain in monoclausal structures (6b,d). It can, however, be violated in the matrix domain in biclausal structures and this fact is fully predicted by the ACWC.8 The relevant construction involves a ditransitive predicate that selects both a goal and clausal complement. In this structure, a non-​prominent in-​situ wh-​expression in the matrix goal position will surface in the same ι as the overt C that introduces the embedded clause. Outputs like these are correctly eliminated by the ACWC. (39) *(Zě ta yà zě gànán Màmŭ du nakàn who say.pst give.pst who comp Mamu cook.pst meat Intended: ‘Who told whom that Mamu cooked the meat?’

o)ι foc

The ACWC, therefore, derives the fact that wh-​in-​situ is permissible in monoclausal multiple questions (6b,d) but not in multiclausal multiple questions, whether in root (39) or embedded positions (8). The interrogative vs. non-​interrogative embedded focus asymmetry (Asymmetry 1) is the consequence of an additional factor. In some related Niger-​Congo languages, focused constituents are realized as ι constituents (Kandybowicz 2017). We observed precisely this phenomenon in Chapter 3, when we detected that focus-​fronted XPs in Wasa and Bono are parsed as independent ι domains. If the short-​moved wh-​items in (3) were realized as independent ι constituents, violation of the ACWC would be averted

[ 92 ] Anti-contiguity

because an ι boundary would separate wh-​from C, thereby guaranteeing anti-​contiguity. The following representation illustrates. (40) 👍 (s v1 c)ι (wh-​)ι (s v2  . . . )ι

Nupe, however, is not a language that realizes focus-​ fronted XPs as prosodically independent ι constituents. As shown in (41), focused expressions are neither flanked by prosodic breaks, nor do they induce F0 register reset. (41) (Musa kpe

gànán

kèké

Nànǎ

si

tsúwó

o)ι

Musa know.pst comp bike Nana buy.pst yesterday foc ‘Musa knew that Nana bought THE BIKE yesterday.’

Furthermore, if the embedded subject were pronominal and a hiatus were to result from the focusing of ‘bike’, as in the following datum, its final syllable would undergo glide formation provided that no ι juncture intervenes between the hiatal vowels. Suspension of glide formation in this context is unattested. This is fully consistent with the ι integration of focused constituents and the clauses that precede and follow them. (42) (Musa kpe gànán kèkj/​*kèké û si tsúwó o)ι Musa know.pst comp bike 3rd.sg buy.pst yesterday foc ‘Musa knew that s/​he bought THE BIKE yesterday.’

The ι integration of focused constituents is fully general in the language and is not dependent on the syntactic status of the focused expression, the embedding verb, or complementizer choice. In what follows, I present two additional pitch tracks demonstrating the prosodic status of Nupe foci. Whereas (41) involved focusing an argument under the embedding verb ‘know’ and complementizer gànán, (43a) features a focused adjunct under

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 93 ]

49

the complex embedding verb ‘ask’ and complementizer kó. In (43b), an argument is focused once again, but with a different embedding verb (‘seem’) and a different complementizer (ke . . . na). Note that in both pitch tracks (as in (41)) there is no prosodic break associated with the focused XP and there is no subsequent F0 register reset. The evidence is thus strong that Nupe does not realize its focused constituents as independent Intonational Phrases. (43) a. (Musa gbín-​gàn kó tsúwó Nànǎ si kèké Musa lose-​say.pst comp yesterday Nana buy.pst bike ‘Musa asked whether Nana bought the bike YESTERDAY.’

o)ι foc

b. (U bè ke kèké Nànǎ si tsúwó o na)ι rd 3 .sg seem.pst comp bike Nana buy.pst yesterday foc prt ‘It seemed that Nana bought THE BIKE yesterday.’

Nupe is, however, a language that realizes left dislocated XP topics as independent ι constituents. Recall the prosodic evidence (i.e., prosodic break + pitch reset) presented in (14), repeated here for convenience.

[ 94 ] Anti-contiguity

(44) (Kèké)ι (mi tú wu:n o)ι bike 1st.sg ride.pst 3rd.sg top ‘As for the bike, I rode it.’

If a hiatus were to result from the left dislocation of ‘bike’, as in the following datum, its final syllable would fail to undergo glide formation due to the ι boundary following the topic. (45) (Kèké/​*kèkj)ι (u tú bike 3rd.sg ride.pst ‘As for the bike, s/​he rode it.’

wu:n o)ι 3rd.sg top

Suspension of glide formation in this context is thus fully consistent with the independent ι status of topicalized constituents in the language. This fact, in conjunction with the ACWC, gives way to a clear and testable prediction. Outputs that violate the ACWC should be repairable by way of left dislocation.9 If Nupe were to allow left dislocated topics to precede focused XPs in the embedded left periphery, C and wh-​would then become separated by an intervening ι topic boundary, circumventing violation of the ACWC. This would effectively (and exceptionally) insulate embedded wh-​from C, as schematized in (46). (46) 👍 (s v1 c)ι (top)ι

(wh-​ s v2  . . . )ι

As it turns out, topic and focus-​fronted XPs may co-​occur in the Nupe CP domain with the relation TopP >> FocP, as predicted by cartographic approaches like Rizzi 1997. The following data demonstrate that left dislocated topics may precede fronted foci (47a), but that foci may not precede displaced topics (47b).

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 95 ]

69

(47) a. Kèké kánci Musa si wu:n bike when Musa buy.pst 3rd.sg ‘As for the bike, when did Musa buy it?’ b. *Kánci when

kèké bike

Musa Musa

o foc

o top

si wu:n o o rd buy.pst 3 .sg top foc

Given this fact, the preceding prediction is fully testable. And as shown in (48b), the prediction is confirmed. Embedded questions become fully acceptable in concert with left dislocation. (48) a. *Musa gbín-​gàn kó kánci Nànǎ si kèké Musa lose-​say.pst comp when Nana buy.pst bike Intended: ‘Musa asked when Nana bought the bike.’

o foc

b. ✓Musa gbín-​gàn kó kèké kánci Nànǎ si wu:n o o Musa lose-​say.pst comp bike when Nana buy.pst 3rd.sg foc top ‘As for the bike, Musa asked when Nana bought it.’

The amelioration of embedded wh-​dissolves, however, when the left dislocated topic moves into the matrix clause, as expected under the ACWC. Compare (48b) with the following sentence. (49) *Kèké Musa gbín-​gàn kó kánci Nànǎ si wu:n o o bike Musa lose-​say.pst comp when Nana buy.pst 3rd.sg foc top

The same repair strategy also applies to embedded wh-​in-​situ in multiple questions.10 An ungrammatical example like (7), repeated here as (50a), becomes acceptable if a topicalized expression is smuggled in between C and wh1-​  (50b). (50) a. *Musa kpe gànán zě si Musa know.pst comp who buy.pst Intended: ‘Musa knew who bought what.’ b.

ké what

o foc

✓Musa kpe gànán tsúwó zě si ké o o Musa know.pst comp yesterday who buy.pst what foc top ‘As for yesterday, Musa knew who bought what.’

As we observed in (49), this effect disappears if instead the topic surfaces in the main clause. This is so because in the root clause left periphery the

[ 96 ] Anti-contiguity

dislocated expression fails to create an intervening ι juncture between wh-​ and C. Compare (51) with (50b). (51) *Tsúwó Musa yesterday Musa

kpe gànán know.pst comp

zě si ké who buy.pst what

o o foc top

These facts are striking. Even native speakers, who have internalized a pre-​ theoretical expectation that wh-​expressions will never be found in subordinate clauses, are surprised and fascinated by facts like (48b) and (50b). The confirmation of this prediction clearly shows that there is nothing syntactically illegal about embedded question formation or embedded wh-​ in-​ situ in multiple questions in Nupe. Rather, the problem is that overt C and non-​prominent wh-​cannot occupy the same Intonational Phrase. I  take these facts to be a strong vindication of the anti-​contiguity generalization proposed in this book. It is important to note that this prediction could not have been made by either of Richards’s 2010 or 2016 proposals. For Richards’s 2010 theory, the central imperative is to minimize the number of major intervening prosodic boundaries between wh-​and C. Installing an intermediate ι boundary between the items via left dislocation would be rather counterproductive from this perspective and would be incorrectly predicted to yield prosodically undesirable structures due to the number and types of prosodic boundaries separating wh-​and C. From the perspective of Richards 2016, prosodic structures like those in (46) must be predicted to be impossible. The reason for this is that in such structures there would be no way to create a ϕ domain wrapping wh-​and C while at the same time respecting the Strict Layer Hypothesis. Because an ι boundary associated with the dislocated topic would intervene, the only sort of all-​encompassing ϕ domain that could be created would violate strict layering by creating a ϕ-​dominating ι constituent in a way reminiscent of the objection raised at the end of Chapter 2. Given that the ACWC derives the two Nupe wh-​ asymmetries in a direct straightforward way and makes correct predictions that could not have been made by its competitor, I conclude that an anti-​ contiguity approach to Nupe wh-​prosody fares better than a contiguity theoretic approach.

4.4.5. Revisiting the Generalization of the Two Asymmetries

Before concluding this chapter, I would like to briefly return to the descriptive generalization laid out in section 4.2.3:

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 97 ]

89

(52) Nupe wh-​phrases occur only in root environments.

Our discussion of the ameliorating effect of embedded left dislocation revealed that this generalization, while largely correct, is too coarse-​grained. By looking at the prosodic effects of topicalization in the last section, we discovered that certain special prosodic conditions permit the surfacing of embedded wh-​items, as predicted by the ACWC. The statement in (52), therefore, is not a true unqualified generalization, but rather represents a strong tendency. Building on the discovery that embedded left dislocation facilitates the occurrence of embedded wh-​, we can identify another special circumstance where embedded wh-​items are prosodically licensed. This involves the prosodic wh-​licensing effect of clausal coordination,11 which, as with left dislocation, is fully predicted by the ACWC. In section 4.4.2 (example (16), repeated here), we observed that coordinated clauses are prosodically independent ι constituents in Nupe. (53) (Musa á nakàn ba)ι (u ma á rd Musa prf meat cut.pst 3 .sg and prf ‘Musa has cut the meat and he has cooked the rice.’

cènkafa du)ι rice cook.pst

If we were to conjoin two embedded clauses under C, the prosodic status of the conjoined clauses would differ. Because overt C does not mark the juncture between two ι constituents in the language, the initial clausal conjunct would be prosodically integrated into the main clause and therefore phrase prosodically with C. The second clausal conjunct, however, would be parsed as an independent ι, as schematized in the following. (54) (s v C clausal conjunct1)ι (conj clausal conjunct2)ι

Because only wh-​items in conjunct one would surface as ι-​mates with overt C, the ACWC predicts that wh-​occurrences will be unable to appear in conjunct one, but may surface in conjunct two. As illustrated in (55), this prediction is confirmed. Once again, the ACWC makes the correct cut in its ability to predict asymmetries in wh-​ distribution. (55) a. *Musa kpe gànán [CONJUNCT 1 ké Nànǎ si o] [CONJUNCT 2 u Musa know.pst comp what Nana buy foc 3rd.sg ma lo dzukó] and go.pst market Intended: ‘Musa knew what Nana bought and (that) she went to the market.’

[ 98 ] Anti-contiguity

b. ✓Musa kpe gànán [CONJUNCT 1 Nànǎ lo dzukó] [CONJUNCT 2   Musa know.pst comp Nana go.pst market ké  ma    u   si     o] what and 3rd.sg buy.pst  foc ‘Musa knew that Nana went to the market and what she bought.’

In this special environment, embedded wh-​in-​situ is also allowable provided that the second clausal conjunct is a multiple wh-​ question. (56) ✓Musa kpe gànán [CONJUNCT 1 egi-​zì lo dzukó] [CONJUNCT 2 zě ma Musa know.pst comp child-​pl go.pst market who and si ké   o] buy.pst what   foc ‘Musa knew that the children went to the market and who bought what.’

4.5. CONCLUSION

To the extent that (37) adequately derives the two asymmetries addressed in this chapter, Nupe joins Krachi, Bono, Wasa, and Asante Twi in providing evidence for an anti-​contiguity theory of wh-​prosody. Thus far, all support for the ACWC has come from Niger-​Congo languages of West Africa (specifically, the Kwa and Benue-​Congo languages of Ghana and Nigeria). One lingering question remains. Can the anti-​contiguity generalization expressed by the ACWC derive similar restrictions on embedded wh-​distribution in typologically unrelated languages outside the West African context? For example, how well does the ACWC fare in deriving well-​known asymmetries and restrictions on wh-​in-​situ in languages such as French (Obenauer 1994; Mathieu 1999, 2002, 2004; Cheng 1997), Spanish (Uribe-​Etxebarria 2002; Reglero 2004), Iraqi Arabic (Wahba 1991; Simpson 2000), Bangla (Bayer 1996; Simpson & Bhattacharya 2003), Hindi-​Urdu12 (Mahajan 1990; Srivastav 1991), and Malayalam (Aravind 2018), etc.? If the ACWC alone cannot derive similar restrictions outside the Niger-​Congo family, would it make more sense to view the ACWC as a violable Optimality Theoretic-​ style constraint (perhaps undominated or else very highly ranked in the languages considered in this book) rather than a universal principle of prosodic organization? Although these are not easy questions to answer, they provide a springboard for the upcoming discussion in the next and final chapter.

N u pe I n t e r r o g at i v e Di s t r i b u t i o n  

[ 99 ]

01

CHAPTER 5

Anti-​contiguity Crosslinguistically

5.1. A CHANGE OF SCENERY

It is now time to take the show on the road. The previous chapters have demonstrated the fruitfulness of the notion of Anti-​contiguity in accounting for various wh-​asymmetries in a variety of West African languages. One wonders, however, whether the ACWC is an areal feature of the Kwa/​Benue-​Congo languages surveyed in this book or whether it holds more broadly. In this final chapter, I view wh-​asymmetries in thirteen (optional) wh-​ in-​situ languages1 from geographically disparate and unrelated families through the lens of the ACWC. Previewing our results, we’ll make two discoveries. One, the ACWC is indeed well suited to explain wh-​ in-​situ asymmetries that are typically viewed as syntactic/​semantic phenomena. And two, the final formulation of the ACWC that concluded Chapter 3 is incomplete and must be parameterized with respect to the dimension of “edge prominence.”

5.2. SOME FAMOUS (AND NOT SO FAMOUS) WH-​ IN-​S ITU ASYMMETRIES 5.2.1. Romance Languages

In this section, I’ll consider two well-​known wh-​ in-​situ asymmetries relating to distribution in root vs. embedded (complement) clauses. The languages manifesting these asymmetries are French and Spanish. In addition, I’ll also briefly consider Catalan in light of the predictions of the ACWC. Anti-contiguity. Jason Kandybowicz, Oxford University Press (2020). © Oxford University Press. DOI: 10.1093/oso/9780197509739.001.0001

5.2.1.1.  French

An oft-​noted observation about colloquial French wh-​questions is that despite the availability of optional wh-​in-​situ in root clauses2 (1a), wh-​ in-​situ is banned in embedded questions (1b) and embedded in-​situ wh-​ phrases cannot take scope across clause boundaries (1c) [Obenauer 1994; Mathieu 1999, 2002, 2004; Boeckx 2000; Bošković 2000; Cheng & Rooryck 2000; Sabel 2000; Tailleur 2013].3 (1)

a. Marie

a

embrassé

qui

(Poletto & Pollock 2015: 142)

Marie has kissed who ‘Who(m) has Marie kissed?’ b. *Pierre a demandé [tu as vu Pierre has asked you have seen Intended: ‘Peter asked who you have seen.’

qui] who

(Bošković 2000)

c. *Jean et Pierre croient [que Marie a vu qui] (Bošković 2000) Jean and Peirre believe comp Marie has seen who Intended: ‘Whom do Jean and Pierre believe that Marie has seen?’

The restriction on French embedded in-​situ wh-​items taking matrix scope (1c) may plausibly reduce to the ACWC. French complement clauses are prosodically integrated into the main clause Intonational Phrase (Hamlaoui & Szendroi 2017; Sun-​ Ah Jun, personal communication). So, just like Wasa and Asante Twi, embedded TPs are not independent Intonational Phrases in French. Consequently, all embedded in-​situ wh-​ items will (a)  count as non-​prominent given the current formulation of the ACWC and (b) phrase with C at the level of ι, in direct violation of the ACWC. The restriction on embedded wh-​in-​situ therefore follows. The ban on wh-​in-​situ in embedded questions (1b), however, cannot be captured by the ACWC (as it is currently formulated) because such constructions lack overt C heads. The explanation for this restriction may very well have a purely syntactic or informational source instead (see, for example, the proposals in Bošković 2000 or Déprez et al. 2012). The ACWC may also account for another interesting asymmetry in French wh-​syntax. In certain French dialects, C may be overtly realized in matrix questions involving wh-​ movement. (2)

Qui que tu as vu who comp you have seen ‘Who did you see?’

(Bošković 2000)

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 101 ]

02 1

However, despite the availability of wh-​in-​situ in this (matrix) environment (1a), unmoved interrogatives are expressly forbidden when C is phonetically realized. (3)

*Que tu as vu qui comp you have seen who

(Bošković 2000)

The ACWC offers a straightforward explanation of this asymmetry. Although wh-​and C are ι-​mates in (2), the ACWC is respected because the moved wh-​expression occupies a “prominent” position at the left edge of the Intonational Phrase. By contrast, the in-​situ interrogative in (3) does not occupy a position of prominence as defined by the ACWC. Consequently, wh-​and C are ι-​wrapped in direct violation of the constraint. The asymmetry therefore follows.

5.2.1.2.  Spanish

Like French, Spanish exhibits both wh-​movement (4a) and optional wh-​ in-​ situ (4b) (Uribe-​Etxebarria 2002). (4)

a. Qué compró Juan what bought Juan ‘What did Juan buy?’

(Reglero 2005)

b. Juan compró qué Juan bought what ‘What did Juan buy?’

(Reglero 2005)

Unlike French, however, Spanish in-​ situ interrogatives must appear domain-​finally. Uribe-​Etxebarria (2002) noted that Spanish in-​situ wh-​ items must appear clause-​finally, otherwise the sentence becomes considerably degraded. (5) a.

Tú le diste la guitarra a quién you cl gave the guitar to who ‘Who did you give the guitar to?’

b. */​?? Tú le diste a quién la guitarra you cl gave to who the guitar

[ 102 ] Anti-contiguity

(Reglero 2004)

(Reglero 2004)

Building on insights by Uribe-​Etxebarria (2002) regarding the placement of pauses, Reglero (2004) discovered that Intonational Phrase is the relevant domain within which in-​situ wh-​items must appear finally. Sentences like (5b) are ameliorated if there is an ι break immediately following the wh-​  item. (6)

Tú le diste a quién # la guitarra

(Reglero 2004)

There is another interesting parallel between French and Spanish wh-​ in-​situ concerning distribution in embedded complement clauses and it is directly addressed by the ACWC. Just as in French, embedded in-​situ wh-​ expressions are impossible in embedded questions (Reglero 2004). (7)

*Pedro ha preguntado [que Pedro has asked comp Intended: ‘Pedro asked who you saw.’

has you.have

visto seen

a dat

quién] who

Reglero (2005) assumes that embedded complement CPs in Spanish are parsed as independent Intonational Phrases, an assumption that is very much in line with the findings of D’Imperio et  al. (2005). If this is correct, then despite occurring ι-​finally (as required), in-​situ interrogatives in embedded questions will phrase together with overt C in violation of the ACWC. A complication arises, however, when we consider embedded wh-​ in-​situ in non-​indirect question contexts. In these environments, unmoved embedded interrogatives can take matrix scope unlike French (Uribe-​ Etxebarria 2002). This is illustrated in (8). (8)

a. Tú crees [que vendrá You think comp will.come ‘Who do you think will come?’

quién] who

b. Pedro ha dicho [que Juan ha comprado qué] Pedro has said comp Juan has bought what ‘What did Pedro say that Juan bought?’

(Reglero 2004)

(Reglero 2004)

To account for the Spanish data in (8), two moves are possible. One option would be to explain the restriction on embedded wh-​ in-​situ in embedded questions (7) by appealing to the ACWC, as in earlier examples, and claim that due to a prosodic difference in the realization of embedded clauses in non-​embedded question contexts, the ACWC is respected in (8). For example, along these lines, we might hypothesize that although the embedded CP constitutes an independent ι constituent in embedded questions (9a),

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 103 ]

0 41

it is the embedded TP that is prosodically isolated as an independent ι in long-​distance wh-​in-​situ constructions (9b), as in Krachi and Bono. This difference is illustrated in (9). (9)

a. possible prosody of Spanish embedded questions (S V )ι (C S V wh-​)ι b. possible prosody of Spanish long-​distance wh-​in-​situ constructions (S V C)ι (S V O)ι

Another approach to handling this asymmetry, one that I prefer, is to claim that the restriction on wh-​in-​situ in embedded questions is not due to the ACWC (or prosody in general), but is rather syntactic/​semantic in nature, as claimed for French by Bošković (2000) and Déprez et al. (2012). Then, to account for the availability of long-​distance wh-​in-​situ on the assumption that embedded CPs are independent ι constituents, we might speculate that the ACWC is parameterized. I will briefly pursue this idea and make reference to it in later sections. The ACWC restricts wh-​items from phrasing together with overt C at the ι level when they are non-​“prominent.” In Chapters 3 and 4, I empirically motivated a structural definition of prominence that references positions at the left edge of ι. The revised version of the ACWC presented in Chapter 3 is repeated in (10). (10) anti-​contiguity of wh-​and c (revised) a. No ι may contain both overt C and a non-​prominent wh-​. b. An occurrence is prominent if it occurs at the left edge of ι.

I propose that condition (10b) might be parameterized to also include the right edge of ι as a possible position of prominence. The idea would be that either edge of the Intonational Phrase counts as “prominent” and languages have to decide which one (or possibly both) gets referenced when the ACWC is assessed. I take this move to be a relatively natural one, given that in prosodic phonology (e.g., within a framework of generalized alignment constraints and pre-​Optimality Theoretic end-​based approaches to the syntax-​phonology interface) reference is frequently made to both the left and right edges of domains. Along these lines, the ACWC would take on the following revised form. (11) anti-​contiguity of wh-​and c (parameterized) a. No ι may contain both overt C and a non-​prominent wh-​. b. An occurrence is prominent if it occurs at either the left or right edge of ι.

[ 104 ] Anti-contiguity

Adopting (11), we might suppose that Spanish is a right-​edge prominent language. As such, the embedded in-​situ interrogatives in (8) would occupy prominent positions and therefore respect the ACWC despite phrasing together with overt C at the ι level.

5.2.1.3.  Catalan

Catalan allows wh-​in-​situ in both echo question contexts (12a) and in multiple wh-​questions (12b) (Castroviejo Miró 2006). (12) a. Que ha dit què you have said what ‘You have said WHAT?’ b. Quin pintor ha pintat quin What painter has painted what ‘Which painter has painted which picture?’

(Castroviejo Miró 2006)

quadre picture

Feldhausen (2008) shows experimentally that embedded complement clauses headed by overt C (as in (13), which follows) are prosodically set off from the main clause as independent ι constituents in Catalan, though not obligatorily. (13) a. (La Bàrbara suposa)ι (que L’àguila robà el ratolí)ι (Feldhausen 2008) the Barbara assumes comp the.eagle stole the mouse ‘Barbara assumes that the eagle stole the mouse.’ b. (La Sílvia no va mancionar)ι (que Ángxela havia the Silvia not pst mention comp Angela have comptat  les   errades)ι

(Feldhausen 2008)

counted  the mistakes ‘Silvia did not mention that Angela has counted the mistakes.’

The ACWC in its original unparameterized formulation makes a clear and concrete prediction in terms of the distribution of wh-​in-​situ in Catalan embedded clauses. Much like Nupe, the prediction is that although available in main clauses, wh-​in-​situ in embedded multiple questions is restricted, owing to the fact that the unmoved non-​left-​edge-​prominent interrogative would form an ι grouping with the overt complementizer. However, if Catalan were a right-​edge prominent language like Spanish, wh-​in-​situ in embedded multiple questions would be permitted either clause-​finally or before a major intonational break. I  currently lack the

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 105 ]

061

data to test these predictions and leave them to be determined by future research. 5.2.2. Bantu Languages

Returning to Africa but venturing beyond Kwa/​Benue-​Congo in this section, we find further potential evidence for the relevance of the ACWC in addressing embedded wh-​ in-​situ asymmetries beyond the languages featured in Chapters 3–​4. What’s more, we find additional motivation for the parameterization of the ACWC (11) discussed in section 5.2.1.2. 5.2.2.1.  Zulu

All interrogative expressions other than subjects may appear in-​situ in Zulu. Wh-​in-​situ is available in both main and embedded clauses, as illustrated in the following. (14) a. U-​bona-​ni 2sg-​see-​what ‘What do you see?’ b. Ngi-​buze [ukuthi u-​Peter 1sg-​asked comp 1a-​Peter ‘I asked what Peter bought.’ c.

(Sabel & Zeller 2006)

u-​thenge-​ni] 1a-​bought-​what

U-​cabanga [ukuthi ku-​sebenze bani] nd 2 .sg-​think comp expl-​worked who.1a ‘Who do you think worked?’

(Sabel & Zeller 2006)

(Sabel & Zeller 2006)

Eligible in-​ situ interrogatives in the language (i.e., non-​ subject wh-​ expressions) are subject to a curious restriction—​they must appear immediately after the verb. (15) a. *U-​theng-​e ingubo entsha nini 2sg.sm-​buy-​pfv 9.dress 9.new when Intended: ‘When did you buy a new dress?’ b.

(Buell 2009)

U-​yi-​theng-​e nini, ingubo entsha (Buell 2009) 2sg.sm-​9.om-​buy-​pfv when 9.dress 9.new ‘When did you buy a new dress?’ (Lit., ‘You bought it when, a new dress?’)

[ 106 ] Anti-contiguity

Cheng & Downing (2007) and Hamlaoui & Szendroi (2017) claim that there is no major prosodic break preceding sentential complements. In other words, because there is no ι boundary separating the matrix from the embedded clause, sentential complements in Zulu are prosodically integrated into the main clause Intonational Phrase. This means that in cases of embedded wh-​in-​situ (e.g., (14b–​c)), the ACWC must not be violated despite the fact that wh-​and C phrase together at the ι level.4 One way to achieve this result is to adopt a parameterization of the ACWC in which right ι edges count as “prominent,” as in Spanish. If that were the case, the availability of wh-​in-​situ in (14b–​c) would follow. But what about cases like (15a), where the in-​situ interrogative seemingly occupies the right edge of the lone ι and the result is ungrammatical? And what about cases like (15b), where wh-​ appears to be ι-​medial and the result is grammatical? For starters, the acceptability of the sentences in (15) does not directly relate to the ACWC, as neither sentence includes an overt C that the wh-​item must avoid phrasing with. The restriction does seem to be prosodic in nature, though. I propose that the so-​called immediately-​after-​the-​verb (IAV) position in Zulu is special because it marks the right edge of ι. For this reason, embedded wh-​in-​situ is tolerated provided that it occupies the IAV position because in this position of prominence, the ACWC is respected. Admittedly, I lack solid prosodic evidence that the Zulu IAV position coincides with the right edge of ι and thus leave it for future Zulu phonologists to determine.

5.2.2.2. Bàsàá

Compared to the prosodic descriptions of European languages, the prosodies of African languages are poorly understood and analyses are exceedingly rare. I have chosen to discuss the case of Bàsàá, a Narrow Bantu language spoken in the Centre and Littoral provinces of Cameroon, because it is one of only a handful of Bantu languages that has received a thorough enough prosodic analysis to test the predictions of the ACWC.5 Hamlaoui & Szendroi (2017) and Makasso, Hamlaoui, & Lee (2018) take a detailed and careful look at the properties of prosodic edges in Bàsàá and use their findings to investigate the ι status of root and non-​root clauses. They find that like Zulu, embedded clauses are prosodically integrated into the main clause ι. In other words, overt complementizers fail to divide embedded TP from matrix TP at the ι level. Bàsàá allows wh-​ in-​situ (Bassong 2010), as shown in the following.

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 107 ]

0 81

(16) a. ɲàŋgó à-β ​ í-​nùŋúl kí í ɓòm (Makasso, Hamlaoui & Lee 2018) 1.wife 1.agr-​pst2-​sell 7.what loc 7.market ‘What did the wife sell at the market?’ b. ɲàŋgó à-β ​ í-​nùŋúl hé móó (Makasso, Hamlaoui, & Lee 2018) 1.wife 1.agr-​pst2-​sell where 6.oil ‘Where did the wife sell the oil?’

Given that Bàsàá allows wh-​in-​situ, the prediction made by the non-​ parameterized ACWC is that non-​initial in-​situ interrogatives should be unavailable in embedded domains because no ι boundary separates the overt embedding complementizer from the non-​ prominent embedded interrogative. Although I lack concrete data in the case of embedded wh-​ in-​situ questions with a single wh-​expression, the evidence available in Bassong 2010 suggests that this prediction is not borne out. In multiple wh-​questions originating in embedded complement clauses, the structurally superior wh-​raises into the matrix left periphery and the second interrogative may remain in a non-​initial embedded in-​situ position. The following examples illustrate. (17) a. Njɛ̀ɛ̀ u ŋ́-​kâl [lɛ́ a ŋ-​kɛ̀ kɛ́lkíí] who you pres-​say comp subj pres-​go when ‘Who do you say will go when?’ (Bassong 2010) b. Njɛɛ u ŋ́-​hɔ´ŋɔ´l [lɛ́ mɛ who you pres-​think comp I ‘Who do you think I will ask what?’

gá-​∫at f2-​ask

kíí] what (Bassong 2010)

If the ACWC were parameterized for both left and right ι edge prominence in Bàsàá, however, the structures in (17) would be fully ACWC-​compliant and the availability of embedded wh-​in-​situ in these contexts would be accounted for. Although the embedded wh-​items in (17) would phrase prosodically with C at the level of ι, they would count as prominent due to their position at the right edge of ι. As for the moved matrix wh-​ occurrences in (17), they too would phrase together with C at the sentences’ one and only Intonational Phrase level due to the fact that embedded C does not impose an ι boundary in Bàsàá. However, if the left edge of ι also counts as a position of prominence in the language, then cross-​clausal fronted wh-​ items will also fail to violate the ACWC.

[ 108 ] Anti-contiguity

Notice that this analysis makes the prediction that only initial or final (i.e., edge-​aligned) wh-​items may co-​occur with embedded C in the language. The data in (16), coupled with the embedded wh-​in-​situ data in (17), suggest that this prediction may turn out to be incorrect. The sentences in (16) show that (matrix) wh-​in-​situ is not restricted to clausal edge positions. This gives me the impression that embedded medial wh-​ in-​situ is likely possible in the language. To account for this possibility and square such data with the ACWC, I propose that perhaps like Zulu, Bàsàá in-​situ wh-​items must appear immediately after the verb because like in Zul u, the IAV position in Bàsàá marks the right edge of ι. Although all the Bàsàá data presented in this section are consistent with an IAV analysis, I lack additional data to test this generalization. If Bàsàá parameterizes the ACWC to count both left and right edges of ι as positions of prominence and in-​situ wh-​occurrences must occupy IAV positions, then the distribution of wh-​ items in both main and embedded clauses can be readily accounted for. In this way, Bàsàá joins Spanish and Zulu in furnishing motivation to parameterize the ACWC.

5.2.2.3.  Duala

Duala is another Cameroonian Bantu language. The language is like French in exhibiting optional wh-​in-​situ (Sabel & Zeller 2006), as shown in (18). (18) a. O bodi nja moni you give who money ‘Who did you give the money to?’

(Sabel & Zeller 2006)

b. Nja o bodi no moni who you give prt money ‘Who did you give the money to?’

(Sabel & Zeller 2006)

Like French (and Spanish), Duala restricts wh-​in-​situ from occurring in embedded questions. Wh-​ex-​situ is obligatory in this environment. (19) *Na si bi [Kuo a-​andi nje] I not know Kuo he-​buy what Intended: ‘I don’t know what Kuo bought.’

(Sabel & Zeller 2006)

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 109 ]

0 1

But unlike French, this time like Spanish, the language allows long-​distance wh-​in-​situ under overt complementizers. (20) O to o pula [na Kuo a keke [wanea you aux you want comp Kuo he try bring muna-​o    nje]] child-​his  what ‘What did you want Kuo to try to bring to his child?’

(Sabel & Zeller 2006)

I do not know of any existing prosodic analyses of Duala and therefore, it is not known whether complement CPs (or the TPs they embed) are prosodically integrated into root ι constituents or are parsed as independent ι units. Furthermore, every example of embedded wh-​ in-​situ I have encountered in the language is one in which the in-​situ interrogative appears sentence-​finally at the right edge of the root ι. Assuming that this distribution is indeed typical of embedded wh-​in-​situ in Duala, the typological analyses presented in this chapter coupled with the ACWC allow us to generate clear and testable predictions about Duala wh-​ prosody. I predict one of the following two possibilities. Possibility one:  like Krachi and Bono, overt C introduces an ι boundary dividing embedded TPs from the main clauses and complementizers that embed them. In this case, regardless of the parameterization of the ACWC, embedded in-​ situ interrogatives will always be anti-​contiguous with C at the ι level, accounting for the availability of long-​distance wh-​in-​situ in the language. Possibility two:  Duala has Spanish-​like prosody insofar as embedded clauses are prosodically integrated into matrix ι constituents. In this case, Duala would have the ACWC parameterized for right ι edge prominence. Because embedded wh-​aligns with the prominent right edge of ι in this scenario, all long-​distance wh-​in-​situ occurrences would satisfy the ACWC. I  leave these predictions in the hands of capable Duala phonologists to test. As for the restriction on embedded wh-​ in-​situ in embedded questions (19), I speculate that just as in French and Spanish, the source of this restriction is syntactic/​semantic in nature rather than prosodic. 5.2.2.4.  Shona

Zentz (2016) shows that Shona is typically Bantu, in that it is an optional wh-​in-​situ language. Some examples illustrating main clause argument and adjunct wh-​in-​situ in the language follow.

[ 110 ] Anti-contiguity

(21) a. V-​aka-​teng-​er-​a ø-​Thandi chi-​i ku-​chi-​toro nezuro (Zentz 2016) 2.sm-​ta-​buyappl-​fv 1a-​Thandi 7-​what 17-​7-​store yesterday ‘What did they buy Thandi at the store yesterday?’ b. V-​aka-​teng-​er-​a ø-​Thandi ø-​rokwe ku-​chi-​toro rinhi (Zentz 2016) 2.sm-​ta-​buyappl-​fv 1a-​Thandi 5-​dress 17-​7-​store when ‘When did they buy Thandi a dress at the store?’

Shona is also typical of Bantu languages in another way:  embedded wh-​ in-​situ is possible and subordinate clause in-​situ interrogatives may take scope outside their clauses. Consider the following. (22) a. W-​ai-​fung-​a kuti v-​aka-​teng-​er-​a ø-​Thandi 2sg.sm-​ta-​think-​fv comp 2.sm-​ta-​buy-​app-​fv 1a-​Thandi chi-​i  ku-​ chi-​ toro  nezuro 7-​what  17-​7-​store    yesterday ‘What do you think that they bought Thandi at the store yesterday?’ (Zentz 2016) b. W-​ai-​fung-​a kuti v-​aka-​teng-​er-​a ø-​Thandi 2sg.sm-​ta-​think-​fv comp 2.sm-​ta-​buy-​app-​fv 1a-​Thandi ø-​rokwe  ku-​chi-​toro  rinhi 5-​dress   17-​7-​store  when ‘When do you think that they bought Thandi a dress at the store?’ (Zentz 2016)

Zentz’s dataset is rich enough to show us that wh-​in-​situ in Shona is not limited to sentence-​final (or possibly ι-​final) position. Although I lack information about the prosodic structure/​phrasing of embedded clauses in the language, I can utilize the ACWC to make a prediction about it. On the account developed in this book, sentences like (22a) will comply with the original unparameterized ACWC if embedded TPs are prosodically mapped onto independent Intonational Phrases. That is, the ACWC predicts that an ι boundary intervenes between the complementizer kuti and the embedded clause-​initial predicate complex, just as in Krachi and Bono. Lacking the means to test this prediction, I leave it to future Shona linguists to either confirm or falsify. 5.2.2.5.  Lubukusu

Lubukusu, spoken in Kenya, patterns like Shona when it comes to the distribution of wh-​in-​situ. I have included it here to showcase the prevalence

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 111 ]

2 1

of Krachi, Bono, and Shona-​type languages with respect to the prosodic realization of embedded clauses and their interaction with the ACWC. As shown in (23), Lubukusu permits wh-​ expressions to appear in-​situ in main clauses. (23) Maayi a-​a-​tekh-​a (bwaangu) si(ina) (bwaangu) (Wasike 2007) 1.mother 1.sm-​pst-​cook-​fv quickly 7.what quickly ‘What did Mother cook quickly?’

Wh-​in-​situ is also possible in embedded clauses in the language, both for arguments (24a) and adjuncts (24b). (24) a. Nafula a-​subil-​a [a-​li Wafula e-​eb-​a 1.Nafula 1.sm-​prs-​believe-​fv 1-​comp 1.Wafula 1-​steal-​fv si(ina)] (Wasike 2007) 7.what ‘What does Nafula believe Wafula stole?’ b. Nafula a-​subil-​a [a-​li Wafula e-​eb-​a sii-​tabu waae(na)] 1.Nafula 1.sm-​prs-​believe-​fv 1-​comp 1.Wafula 1-​steal-​fv 7-​book where ‘Where does Nafula believe Wafula stole the book?’

Unfortunately, all of the data I have access to regarding embedded wh-​ in-​ situ involve cases where the interrogative expression appears sentence-​ finally. Without knowing whether embedded in-​situ expressions can surface away from the right edge of the clause/​ι, we have two ways of proceeding. If they can, then Lubukusu has the properties of Krachi, Bono, and maybe Shona—​embedded TPs are phrased as independent ι constituents, thus insulating the in-​situ interrogative from the overt embedding C and thereby complying with the ACWC. If they must occur at the right edge, then perhaps Lubukusu has parameterized its ACWC so that right edges of ι constituents count as positions of prominence. I leave it for future research to determine which analysis is more satisfactory. 5.2.2.6. Kîîtharaka

Kîîtharaka, a language of Kenya, exhibits a similar pattern. Example (25) shows that main clause wh-​in-​situ is grammatical. (25) Maria a-​gûr-​ir-​e 1.Maria 1.sm-​buy-​pvf-​fv ‘What did Maria buy?’

[ 112 ] Anti-contiguity

mbi 7.what

(Abels & Muriungi 2008)

The following example demonstrates that wh-​in-​situ expressions may appear in embedded complement clauses. (26) John a-​ug-​ir-​e [Pat a-​ug-​ir-​e [Maria a-​gûr-​ir-​e mbi]] 1.John 1.sm-​say-​pvf-​fv 1.Pat 1.sm-​say-​pvf-​fv 1.Maria 1.sm-​buy-​pvf-​fv 7.what ‘What did John say Pat said Maria bought?’       (Abels & Muriungi 2008)

In addition, both argument and adjunct in-​situ interrogatives are allowed in Kîîtharaka embedded questions, as shown in the following. (27) a. N-​ti-​iji

[Munene

a-​gûr-​ir-​e

mbi]

(Muriungi 2005)

1.sg.sm-​neg-​know 1.Munene 1.sm-​buy-​pvf-​fv 7.what ‘I don’t know what Munene bought.’ b. Tu-​ri-​ama [Munene a-​ka-​aja ri] (Muriungi 2005) 1.pl.sm-​neg-​know 1.Munene 1.sm-​come-​pvf-​fv when ‘We don’t know when Munene came.’

The ACWC predicts the acceptability of all of these possibilities for the simple reason that none of the wh-​ in-​situ cases in the preceding involves the phonetic realization of a complementizer. If C is not realized at PF, then the anti-​contiguity condition captured by the ACWC is vacuously satisfied. It is not clear to me if Kîîtharaka ever realizes its complementizers overtly.6 If not, then the ACWC will always be trivially satisfied and wh-​ in-​ situ should be widely available in a diverse range of environments. If it does overtly realize its Cs, then some restrictions on the distribution of wh-​ in-​ situ may very well exist. I leave it to future research to address this issue. 5.2.3. Indo-​A ryan Languages

In this section, we’ll have a look at two famous cases involving restrictions on embedded in-​situ interrogatives in Indo-​Aryan languages. Hindi-​Urdu (Mahajan 1990; Srivastav 1991) and Bangla (Bayer 1996) are well-​known for having in-​situ interrogatives that are unable to take scope out of certain embedded clauses. The languages therefore represent instances in which the grammar prohibits wh-​in-​situ in certain embedded domains, despite allowing it in matrix environments. 5.2.3.1. Hindi-​Urdu

In Hindi-​Urdu preverbal non-​finite embedded clauses, wh-​in-​situ is available and the interrogative takes matrix scope. A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 113 ]

4 1

(28) a. Raam-​ne [Mohan-​ko kise dekhne ke liye] kahaa (Mahajan 1990) Raam-​erg Mohan-​erg whom to.see for told ‘Who did Raam tell Mohan to look at?’ b. Tum [kyaa paRhnaa] caahte you what read.inf want ‘What do you want to read?’

ho pr

(Dayal 1996)

However, when a finite embedded clause occurs postverbally (unlike non-​ finite clauses, Hindi-​Urdu does not allow finite complement clauses in preverbal position), wh-​in-​situ is impossible. The interrogative is unable to scope out of the embedded clause in this context (29a), necessitating long distance wh-​movement (29b). (29) a. *Raam-​ne kahaa [ki kaun aaya Raam-​erg said comp who come Intended: ‘Who did Raam say has come?’

hai] has

(Mahajan 1990)

b. Kaun Raam-​ne kahaa [ki aaya hai] who Raam-​erg said comp come has ‘Who did Raam say has come?’

(Mahajan 1990)

Although existing proposals in the literature attempt to derive this distributional/​scope asymmetry syntactically (Mahajan 1990; Srivastav 1991; Dayal 1996; Ouhalla 1996; Mahajan 1997; Simpson 2000; Bhatt & Dayal 2007; among others), I believe the ACWC may offer insight as well. One crucial difference between non-​finite and finite subordinate clauses in Hindi-​Urdu appears to be the presence of an overt complementizer. Non-​finite clauses lack them. Finite clauses contain them. In cases like (28), where the in-​situ interrogative appears in an embedded non-​finite clause, there is no way to violate the ACWC because there is no overt C that wh-​must avoid phrasing with. By contrast, in-​situ wh-​ items appearing inside finite embedded clauses must contend with overt C heads. Provided that no ι boundary separates C from wh-​in sentences like (29a), as in Wasa, Asante Twi, and Nupe, the ACWC derives the unacceptability of finite embedded wh-​in-​situ. By moving out of the offending configuration and into a prominent position at the left edge of the root ι (29b), however, violation of the constraint is averted. In this way, the restricted distribution of Hindi-​Urdu embedded wh-​ in-​situ follows.

[ 114 ] Anti-contiguity

5.2.3.2.  Bangla

Bangla (Bengali) is commonly regarded as an SOV wh-​ in-​situ language (Bayer 1996). Consider the following data in which an in-​situ interrogative appears in a matrix clause (30a) and in a preverbal finite complement clause with matrix scope (30b). (30) a. Jɔn kon boi-​ʈ poɽlo John which book-​cl read ‘Which book did John read?’

(Simpson & Bhattacharya 2003)

b. Jɔn [ke cole gæche] bollo John who left gone said ‘Who did John say left?’

(Simpson & Bhattacharya 2003)

Finite complement clauses are not obligatorily preverbal in the language. As the following datum shows, they may optionally occur to the right of their selecting verb. (31) Jɔn ([Meri cole

gæche]) bollo ([Meri cole gæche]) (Simpson & Bhattacharya 2003) John Mary left gone said Mary left gone ‘John said that Mary left.’

If the finite embedded clause occurs preverbally, then wh-​in-​situ is possible and wh-​may take wide scope (30b). But if it occurs postverbally, wh-​ in-​situ is unavailable (32), unless the embedding verb allows for indirect question complements. (32) *Jɔn bollo [ke cole John said who left Intended: ‘Who did John say left?’

gæche] gone

Can the ACWC make any sense of this distributional/​scope asymmetry? The answer is no and it rests on the fact that overt C plays no role in the asymmetry. Because both preverbal finite clausal complements and their postverbal counterparts can be realized without complementizers, the ACWC is vacuously satisfied in both (30b) and (32) and thus fails to differentiate among the two outputs. Unlike many of the other asymmetries we have looked at in this chapter, where what was traditionally deemed “syntax” might plausibly turn out to be described more accurately as “phonology,” I  follow Simpson

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 115 ]

6 1

& Bhattacharya (2003) in understanding this asymmetry to be purely syntactic. Simpson & Bhattacharya argue that Bangla is SVO in the base, with complement clauses originating in postverbal position (not extraposed to this position, as in many accounts of Bangla and Hindi-​Urdu). According to Simpson & Bhattacharya, wh-​items must move in Bangla and the landing site of movement is a position just below the surface subject. The derivation of a sentence like (30b), for example, involves wh-​movement that pied-​pipes the embedded CP from a postverbal position, as in analyses of languages like Basque (de Urbina 1990) and Quechua (Hermon 1985). In this way, (30b) does not involve wh-​in-​situ at all. And the ungrammaticality of sentences like (32) owes to the fact that obligatory wh-​movement does not take place. Strong evidence for this analysis comes from the fact that many Bangla speakers allow an alternative to wh-​CP raising in which only the wh-​DP raises, as in (33b). See Simpson & Bhattacharya 2003 for a more detailed analysis. (33) a. Jɔn [ke cole gæche] bollo John who left gone said ‘Who did John say left?’ b. Jɔn ke bollo [cole gæche] John who said left gone ‘Who did John say left?’

(Simpson & Bhattacharya 2003)

(Simpson & Bhattacharya 2003)

5.2.4. Iraqi Arabic and Malayalam

Rounding out our crosslinguistic survey of embedded wh-​ in-​situ asymmetries are Iraqi Arabic and Malayalam, grouped together here because although similar in many ways to the Indo-​Aryan languages explored in the previous section with respect to wh-​in-​situ, they are non-​Indo-​ Aryan languages. Much like what we saw in the previous section on Bangla, I will suggest that the well-​known asymmetries in these languages are best viewed through a syntactic lens. That is to say, I do not believe the following asymmetries fall under the purview of the ACWC.

5.2.4.1. Iraqi Arabic

In Iraqi Arabic, wh-​expressions may move or remain in-​situ. Example (34a) illustrates the well-​formedness of wh-​in-​situ, while (34b), a multiple wh-​ question, shows that both strategies may coincide in a given sentence.

[ 116 ] Anti-contiguity

(34) a. Mona shaafat meno Mona saw whom ‘Who did Mona see?’

(Wahba 1991)

b. Sheno ishtara Ali minshaan yenti li-​meno what bought Ali in.order.to give to-​whom ‘What did Ali buy to give to whom?’

(Wahba 1991)

(34b) also illustrates that in-​situ wh-​expressions may appear in embedded positions in the language. (35) furnishes another example. Mona raadat [tijbir Su’ad tisa’ad meno] Mona wanted to.force Su’ad to.help who ‘Who did Mona want to force Su’ad to help?’

(35)

(Wahba 1991)

In both (34b) and (35) the hosting embedded clause is non-​finite. When the host clause is finite, wh-​in-​ situ becomes ungrammatical (36a), necessitating wh-​movement (36b). (36) a. *Mona tsawwarat [Ali istara sheno] Mona thought Ali bought what Intended: ‘What did Mona think Ali bought?’ b.

Sheno tsawwarit Mona [Ali ishtara] what thought Mona Ali bought ‘What did Mona think Ali bought?’

(Wahba 1991)

(Wahba 1991)

The relationship between finiteness and embedded wh-​in-​situ in Iraqi Arabic closely parallels the relationship in Hindi-​Urdu: wh-​in-​situ is available in non-​finite embedded clauses (35) and blocked in finite embedded domains (36a). Because the restriction on finite embedded wh-​ in-​situ holds regardless of whether a complementizer is overtly realized, as we have seen in other languages in this chapter, it would not seem that this distributional asymmetry is in any way grounded in conditions related to anti-​contiguity. Perhaps, then, the roots of the asymmetry lie squarely in the syntax of wh-​in-​situ in the language. Simpson (2000), for example, argues that languages can vary parametrically with respect to the locality conditions on wh-​licensing and that the domain for wh-​feature checking in Iraqi Arabic is the immediate tense domain of the +Q complementizer. On this proposal, Iraqi Arabic disallows long-​distance wh-​in-​situ from finite domains because wh-​expressions inside these domains cannot be

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 117 ]

8 1

accessed for wh-​feature checking, owing to the fact that the matrix +Q C head cannot look beyond its immediately local tense domain. I suggest that something like this (and crucially not the ACWC) is what is responsible for Iraqi Arabic’s well-​known wh-​ in-​situ asymmetry. 5.2.4.2.  Malayalam

Malayalam’s principle question formation strategy involves wh-​ in-​situ (Aravind 2018). (37) a. Eethu kuTTi ee pustakam which child this book ‘Which child read this book?’

vaayi-​chu read-​perf

b. Raman eethu pustakam vaayi-​chu Raman which book read-​perf ‘Which book did Raman read?’

(Aravind 2018)

(Aravind 2018)

Similar to what we have seen in Hindi-​Urdu and Iraqi Arabic, embedded in-​ situ interrogatives in the language can take matrix scope out of non-​finite clauses. (38) Raman [eethu pustakam vaayikk-​aan] shramichu Raman which book read-​inf tried ‘Which book did Raman try to read?’

(Aravind 2018)

Despite this fact, the wh-​in-​situ strategy is unavailable in certain bi-​clausal constructions. When a Malayalam wh-​expression occurs inside the finite clausal complement of a verb like ‘know,’ which can take either interrogative or declarative complements, only the embedded question interpretation is available (39a).7 When a Malayalam wh-​expression occurs inside the finite clausal complement of a verb like ‘think,’ which can only take declarative complements, neither narrow nor wide scope readings are available (39b). In other words, in these environments, embedded wh-​ in-​situ is completely unavailable.8 Note that finite embedded clauses in the language do not appear in canonical preverbal positions, but are rather preposed to the left edge of the clause.9 (39) a. [Sita eethu pustakam vaayikk-​um ennu] Raman-​u ariyaam (Aravind 2018) Sita which book read-​fut comp Raman-​dat know ‘Raman knows which book Sita will read.’ *‘For which book x does Raman know that Sita will read x?’

[ 118 ] Anti-contiguity

b. *[Sita eethu pustakam vaayikk-​um ennu] Raman vicaarichu Sita which book read-​fut comp Raman thought *‘Raman thought which book Sita will read.’ *‘Which book did Raman think Sita will read?’           (Aravind 2018)

Aravind (2018) argues that the motivation for clausal preposing in these cases is prosodic in nature. Embedded clauses with “heavy” subjects must front in order to circumvent violation of prosodic constraints such as Strict Layering, MaxBin, and ProsProm (see Aravind 2018 for details). Notably, however, preposing is obviated when the embedded finite clause’s subject is prosodically “light” and in these cases, wh-​in-​situ is possible with wide scope interpretations. (40) Nee [enthu patt-​um ennu] vicaarikk-​unnu you what happen-​fut comp think-​prog ‘What do you think will happen?’

(Aravind 2018)

In Aravind’s informal prosodic analysis, wh-​and C phrase together as an independent ι in structures such as (39a). Therefore, sentences like (39a) and (40) suggest that contiguity of wh-​and C under ι in Malayalam is not driving the wh-​asymmetry in question. As with Iraqi Arabic, my position is that the asymmetry is ultimately syntactic in nature. I  follow Aravind (2018) in treating the restriction on embedded wh-​in-​situ in fronted finite clauses as reducing to a Minimality violation resulting from a pernicious interaction between wh-​agreement and A-​bar operations (e.g., movement) affecting embedded clauses.

5.3. CONCLUSION

This book led with the idea that sometimes “syntax” is syntax and sometimes “syntax” is phonology. Our crosslinguistic investigation into wh-​ in-​ situ asymmetries furnishes examples of each case. I sided with a number of researchers when I  claimed that certain restrictions on in-​situ wh-​ expressions in French, Spanish, and Duala embedded questions have a purely syntactic character. In addition, I echoed a number of strictly syntactic analyses of restrictions on wh-​in-​situ in the finite embedded clauses of Bangla, Iraqi Arabic, and Malayalam. In Chapters  3 and 4, I  argued that wh-​asymmetries that might appear at first blush to be narrowly “syntactic” are more accurately prosodic in nature. And I have suggested in this chapter that comparable asymmetries in French, Spanish, Zulu,

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 119 ]

201

Bàsàá, and Hindi-​Urdu may have prosodic underpinnings as well. To account for these types of asymmetries prosodically, I  developed a theory centered on the ACWC in Chapter 3. This chapter has revealed the need to further refine the ACWC by allowing for parametric variation with respect to the ι edge that serves as a position of “prominence” in a given language (11). Languages like Spanish, Zulu, and Bàsàá (unlike Krachi, Bono, Wasa, Asante Twi, and Nupe) seem to be languages in which the right edge of ι counts as “prominent” in the eyes of the ACWC. Byron Ahn (personal communication) has suggested that another dimension along which the ACWC might be parameterized is the prosodic domain that the ACWC applies to. In its current formulation, the ACWC militates against the contiguity of wh-​and overt C at the level of Intonational Phrase. Ahn suggests that new phenomena/​asymmetries may reveal the need to reference prosodic constituents such as Intermediate Phrase or Phonological Phrase when formulating the ACWC with respect to a given language. This strikes me as a very interesting and promising direction for future research. We have not seen any clear-​cut cases in this chapter where the ACWC is violated and yet the output is fully grammatical. This is most likely due to the fact that the prosodic details of some of the languages investigated here are incomplete or lack the requisite level of detail. But what if we were to discover such cases? Would their discovery undermine the plausibility of the ACWC? I hesitate to say “yes.” Once more languages are studied with the constraint in mind, it may be revealed that the ACWC is merely a tendency. In that case, one way to conceptualize the ACWC might be to regard it as a violable constraint, in the sense mentioned in passing at the conclusion of Chapter 4. I see this as another promising avenue to explore in future work. The conclusion of Chapter 3 posed a number of big questions that arise with a theory like the ACWC. The biggest of these, in my opinion, is the question of why a prosodic constraint like the ACWC would exist in the first place. On a macro level, what interface principle or design feature would underpin something like the ACWC? On a micro level, why would a prosodic constraint exist targeting overt complementizers and wh-​ expressions rather than any other set of items? Is the ACWC a reflection of something more general? Is there something special about wh-​words and overt complementizers, or could the ACWC extend to other configurations of words? From a prosodic perspective, it seems strange that a prosodic constraint would target specific words rather than generalized structural configurations. One possibility in this regard is that Anti-​contiguity is a relation that mediates the prosodic grouping of probe-​goal pairs, in the same way that Richards’s (2016) Contiguity proposal generalizes to all

[ 120 ] Anti-contiguity

probe-​goal relations mediated by Agree. Assuming overt embedded C to be an interrogative-​seeking probe in the case of a wh-​ construction (either encoded directly on C by way of an uninterpretable Q feature or inherited from some higher +Q C probe in the structure) and assuming wh-​to be its goal, perhaps the ACWC captures a prosodic restriction on probes and their goals forming a contiguous prosodic grouping. In other words, we might speculate that prosody seeks to distance probes and goals, restricting them from appearing in the same relevant prosodic domain. Could Universal Grammar contain a Contiguity parameter? That is, a principle that is set positively in the languages explored by Richards (2010, 2016)  and negatively in the languages considered in this book? These questions are part of a larger puzzle that concerns the compatibility of both Contiguity Theory and Anti-​contiguity Theory in a theory of the syntax-​phonology interface. Are both systems needed? Or is Anti-​ contiguity the only relevant relation when it comes to wh-​/​C relations at the syntax-​phonology interface? Although I have no definitive answers to these questions, I suspect that when considered against the backdrop of linguistic diversity, both systems will be needed. Although the theory of wh-​prosody advanced in this book has an “anti” contiguity signature, it should not be taken to be anti-​Richards (2010, 2016). In detail, of course, the current theory parts ways with Richards’s proposal. But in spirit, the two theories are kindred, for they both view prosody as playing a key role in the grammar of wh-​. And they are both guided by the central intuition that “syntax” IS syntax sometimes, while in some cases “syntax” is more accurately phonology.

A n t i - C o n t i g u i t y C r o s s l i n g u i s t i c a l ly  

[ 121 ]

NOTES

CHAPTER 2 1. Richards is not the first to challenge this architectural assumption. Other proposals that some sub-​domains of phonology influence syntax include Bresnan 1971; Hetzron 1972; Zec & Inkelas 1990; and Golston 1995. 2. The data presented in this chapter are based exclusively on fieldwork in Ghana conducted between the years 2010 and 2014. The examples are presented in the official Krachi orthography developed by the Ghana Institute for Linguistics, Literacy, and Bible Translation (Dundaa 2007). Because the orthography does not mark Krachi’s two level tones (high and low [Snider 1990, Adonae 2005]) or its rising contour tone, I have omitted tone marking diacritics in the data. Underlying or lexical tones of items, however, are indicated in the “tones” tiers of the pitch tracks. See note 9 for more on the representation of segments and tones in the pitch tracks that follow. 3. For the purposes of this chapter, I set aside a number of interesting topics in Krachi wh-​syntax such as islandhood and the internal structure of wh-​ items because these topics do not relate directly to the chapter’s analytical focus. For more on these topics, see Torrence & Kandybowicz 2015. 4. It is worth noting that not even purpose readings for ‘why’ are available in cases like (2d), as they are in languages like French (though, on the basis of distributional and interpretational evidence, it isn’t clear that Krachi makes a distinction between causal/​epistemic ‘why’ and purpose ‘why’). 5. Although unable to occur clause-​internally, the wh-​ item nanι may scope over embedded clauses to express indirect questions, but only from an embedded left peripheral position lower than embedded C (Torrence & Kandybowicz 2015)—​ see note 14, example (i). As such, despite its inability to occur clause-​internally, Krachi ‘why’ does not resist embedding. In this way, nanι is not comparable to how come in English. 6. In this model, the formal objects of syntax and prosody are assumed to be nearly identical (e.g., Wagner 2005; Selkirk 2006, 2009; Adger 2007; Ishihara 2007; Kratzer & Selkirk 2007, among others). 7. In addition to grouping, Richards (2016) invokes a second ϕ-​manipulating operation known as contiguity adjunction. Unlike grouping, contiguity adjunction requires linear adjacency. I will not discuss this second domain-​ changing operation, however, since it is not relevant to the discussion at hand. In the cases we consider in this chapter, non-​‘why’ wh-​items and C are never string adjacent and thus contiguity adjunction will never be applicable.

2 41

8. See Cheng & Downing 2011 for empirical problems associated with the treatment of Chichewa in Richards’s framework. 9. A note on the representation of vowels in the “words” tiers of the Krachi pitch tracks in this chapter—​due to font limitations in Praat, the orthographic vowel , which is sometimes realized as +ATR [i]‌and other times realized as –​ATR [ɪ] depending on the vowel harmony context, is represented as ; the orthographic vowel , which is sometimes realized as +ATR [u] and other times realized as –​ATR [ʊ] depending on the vowel harmony context, is represented as . A note on the representation of tone in the “tones” tiers of the pitch tracks in this book—​in these tiers, I represent lexical tone, not surface tone. In many cases, the presence of boundary tones or tone sandhi phenomena will obscure the tones indicated on the tonal tier of the pitch tracks, yielding insightful tonal patterns for our prosodic analyses. 10. Indefinite determiners in Krachi are phonetically null. 11. An anonymous reviewer points out that if recursive prosodic phrasing is allowed as in Match Theory (Selkirk 2011; Elfner 2012, 2015), then evidence for the left edge of ϕ does not necessarily constitute evidence of a right edge of the preceding constituent. For example, in the case of a structure like PP[x DP[y]‌], both structures (x)(y) and (x(y)) would be consistent with the left edge being “active” for DPs. 12. Despite appearing ϕ/​utterance-​finally, the tonal realization of brɔdιɛ in this sentence does not appear to be affected by L%. I suspect that this may be related to the fact that brɔdιɛ is a borrowed word, most likely from Akan. The same is true of the tonal realization of bireŋ in (11b). My guess is that utterance-​final position may play a role, but leave a more thorough investigation for future research. 13. One might object that the validity of the three arguments presented earlier against the existence of prosodically marked left DP edges in Krachi depends upon the prosodic status of prepositions, clausal coordinators, and demonstratives. If these items are clitics, then regardless of their syntactic structure they will depend upon the following words (or possibly adjacent higher level prosodic phrases) as their hosts and thus prosodically group together with them (Selkirk 1996). In this case, the absence of F0 lowering effects associated with these items would be entirely expected/​accounted for. For example, in the case of prepositions, it would be unlikely for such items to phrase as a separate ϕ from their DP complements because prepositions are prosodically light elements (likely clitics, but at most, minimal prosodic words). See Elfner 2012 for discussion of the role of prosodic weight in prosodic constituent formation. Some evidence that these items are not clitics comes from vowel harmony. In Krachi, the prosodic word is the domain for vowel harmony and with very few exceptions, all word-​internal vowels share the same ATR specification (Adonae 2005, Dundaa 2007). If prepositions, clausal coordinators, and demonstratives were clitics and thus formed prosodic word-​level groupings with the items immediately following them, we would expect to find variation in the surface forms of the items based on the ATR specifications of the following words (e.g., *ye ‘with’; *yi ‘and’; *kenιŋ ‘dem’). We do not, however, find variation of this sort in the phonological realization of these three items. Furthermore, if the items were procliticized we would incorrectly predict the ATR disharmonic outputs that follow to be ill-​formed, where the functional element bears [–​ATR] vowels and the linearly adjacent lexical element bears [+ATR] vowels. In fact, forms such as these are fully grammatical and widely attested in the language, as shown in the following.

[ 124 ] Notes

(i)

a. yɛ i-​g yo with pl-​yam ‘with the yams’ b. ɔkyι wυ ɛ-​dιkɛ kugyo wυ yι Kofi e-​g yi brɔdιɛ woman the pst-​cook yam the and Kofi pst-​eat plantain ‘The woman cooked the yam and Kofi ate a plantain.’

kɛnιŋ i-​g yo wυ dem pl-​yam the.dist ‘those yams’   An alternative to this line of argument would be to claim that prepositions, clausal coordinators, and demonstratives are indeed clitics that recursively group with the following prosodic word, but that the domain of vowel harmony in the language is the minimal prosodic word. In this case, the absence of ATR harmony in such groupings might be accounted for, as the preposition, clausal coordinator, or demonstrative would be expected to be opaque to outside harmonizers. 14. Partial wh-​movement is available in Krachi and does not exclude ‘why’ (Torrence & Kandybowicz 2015). As such, nanι may appear in embedded peripheral positions and thus surface to the right of/​interact with complementizers, contrary to the previous description concerning the item’s general lack of interaction with material to its left. c.

(i)

Fe e-​nu fɛɛ nanι yι Ama ɛ-​mɔ m-​bwatɛ 2nd.sg pst-​hear comp why foc Ama pres-​kill pl-​chicken ‘Why does Ama slaughter chickens, according to what you heard?’

In these cases, the complementizer is realized with the familiar depressed/​falling F0 pattern (see (ii)), as would be expected if ‘why’ were bounded on its left edge by a prosodic boundary, rendering the complementizer phrase-​final in the preceding Phonological Phrase domain and thus subject to lowering by a right edge L boundary tone. (ii)

Notes  [ 125 ]

261

However, because C is a phase head, its complement (which includes ‘why’) is spelled-​out separately. Thus, ‘why’ and C occupy separate prosodic domains in embedding constructions. The F0 lowering on C in this case is thus more accurately attributed to C’s position at the right edge of the preceding prosodic domain (itself owing to the fact that C is a phase head) than to the fact that nanι’s left edge is prosodically marked. As such, partial wh-​ movement constructions cannot be used to decisively establish an empirical argument for the exceptional left edge marking of nanι. 15. One might wonder whether Krachi’s expression for ‘why’ is a DP or an adverbial, pointing out the possibility that its exceptional prosody may derive from the fact that it belongs to a different category than the other wh-​items. Unfortunately, I know of no concrete morphosyntactic evidence to justify categorizing the expression as a DP and thus leave determining its categorial status for future research. 16. For other left peripheral base generation approaches to the syntax of ‘why’ see Rizzi 1990; Collins 1991; Bromberger 1992; Lasnik & Saito 1992; Ko 2005; and Stepanov & Tsai 2008; among others. 17. Technically, contiguity with C could be established in Richards’s 2016 system via the operation Contiguity Adjunction. This is precisely Richards’s account for the missing ϕ boundary on Krachi ‘why’ (Richards 2016: 111). I prefer to conceptualize things in the way presented in the text and avoid appealing to Contiguity Adjunction due to its highly stipulative nature. 18. The item kwarε . . . gyι ‘think’ is an idiosyncratic “split” verb in Krachi composed of the predicates kwarε and gyι, which in isolation bear the independent meanings ‘to collect’ and ‘to eat’, respectively. In the non-​compositional split verb construction, however, neither predicate contributes its independent lexical meaning. 19. Richards (2010: 148) leaves open the possibility that what he calls “wh-​ domains” may not always be phonetically signaled by special means in different languages. For this reason, the fact that no discernable wh-​domain is cued phonetically in cases like (34a–​b) is not sufficient evidence against his proposal. 20. An anonymous reviewer points out that strict layering in the strongest sense has been argued against in different ways since its proposal in the 1980s and that it is currently uncontroversial to assume certain deviations from strict layering such as prosodic recursion and level-​skipping, at least under certain conditions. Given that Match theory predicts that level-​reversals may be possible (see the discussion on relative clauses in Selkirk 2011), the reviewer remarks that it wouldn’t be unusual to propose that strict layering in the sense of the level-​ ordering hypothesis captured in (38) is also violable. 21. On top of these considerations, there is an additional technical concern raised by embedded wh-​in-​situ in Richards’s framework that is independent of Krachi. Given the mechanics of phase-​based derivation (Chomsky 2001) as assumed by Richards, an indeterminacy would arise in the case of the construction of a lower C phase when an in-​situ interrogative has been merged. In a Contiguity-​ theoretic framework, on the basis of syntactic and prosodic information available at any given stage of the derivation, the system must determine for any wh-​item whether or not it is necessary to move wh-​in order to satisfy the phrasing requirement in (4). However, when wh-​is merged in a lower C phase, the system does not have enough information to decide whether or not wh-​ can phrase with its scopally associated C because that item (matrix C, i.e., the wh-​ probe) has not yet entered the derivation. This has the undesirable consequence

[ 126 ] Notes

that without the requisite phono-​syntactic information/​motivation to move, an embedded wh-​item should remain positioned in the lower C phase at Spell-​Out (i.e., unmoved). This means that at the point of Transfer, the required prosodic grouping of wh-​and C will fail to obtain and the system should deem wh-​ in-​situ prosodically ill-​formed. Once again, the framework would make the incorrect prediction that embedded wh-​in-​situ is unavailable and thereby force long-​ distance wh-​ movement. CHAPTER 3 1. Like the examples presented in Chapter 2, the Krachi examples in this chapter are presented in the official orthography developed by the Ghana Institute for Linguistics, Literacy, and Bible Translation (Dundaa 2007). As the orthography does not mark Krachi’s two level tones (high and low [Snider 1990; Adonae 2005]) or its rising contour tone, I have omitted tone marking diacritics from all representations apart from those accompanying pitch tracks, where the representation of tone is relevant. To represent Bono and Wasa data, I have used a modified version of the Akan script, as I have been unable to locate written materials in either language. The Asante Twi data in this chapter have been written in conformity with the unified Akan orthography. As with Krachi, Bono, and Wasa, the Asante Twi data are non-​IPA representations and tone marking is indicated only on pitch track data. 2. This should be the expected outcome given that the four languages all have strikingly similar syntax and the constructions under consideration do not deviate semantically from language to language. 3. Perhaps the strongest arguments against the syntactic/​locality account considered in this section come later in section 3.4.6 during the discussion of the distribution of partial wh-​movement and its correlation with prosody. The considerations discussed there would be difficult for a purely syntactic approach to cope with. 4. Wh-​movement to an embedded position above C is also unavailable in Asante Twi, as exemplified in the following. (i) *Wo    kaa    dɛn  na   sɛ     Kofi  diiyɛ 2nd.sg  say.pst what foc comp Kofi eat.pst 5. Partial wh-​focus movement with a null matrix Q is also robustly attested in Krachi and Bono (see Torrence and Kandybowicz 2013, 2015 for full paradigms). However, since the coexistence of partial wh-​focus movement and embedded wh-​in-​situ in these languages is fully consistent with the claims and predictions of the syntactic/​semantic approach sketched earlier, it will not be discussed because it does not play a decisive role in the argumentation of this section. 6. One might object that these facts might be amenable to a purely syntactic/​ locality-​based account in which matrix interrogative Q in Wasa can probe as far as the C domain (i.e., the left periphery) of an embedded clause, but no further. The idea here would be that due to locality constraints like the Phase Impenetrability Condition (Chomsky 2001), a probe can only reach down as far as another probe with the same feature. This sort of analysis would surely capture the fact that a partially moved wh-​expression, but not its in-​situ counterpart, could be accessible to matrix Q, but it breaks down in cases of deeper embedding because it incorrectly predicts that a partially moved wh-​ expression will not be accessible to matrix Q if more than one clause boundary

Notes  [ 127 ]

2 81

separates the two. More concretely, in a case involving two embedded clauses, for example, such an analysis incorrectly predicts that it should be impossible for a wh-​item that originates in the most deeply embedded clause to partially move to the edge of its immediately containing clause and then stop. As the following data reveal, this is indeed possible in the language, suggesting once again that a purely syntactic/​locality-​based approach is insufficient. (i) Wo   dwene sɛ   Kofi kaa   sɛ   ɛdiɛn  na    bɛrɛma no kumiye   2nd.sg  think    comp     Kofi  say.pst  comp  what  foc  man    the  kill.pst ‘What do you think that Kofi said that the man slaughtered?’ By contrast, the prosodic analysis developed in the next section correctly predicts and accounts for the existence of partial focus movement structures like those in (i). 7. As in Chapter 2, I continue to assume the existence of the Prosodic Hierarchy (Selkirk 1984; Nespor & Vogel 1986), according to which prosodic constituents are hierarchically organized and strictly layered. I also adopt the Match theory of Selkirk (2009, 2011), which maintains that prosodic structures are built from and largely correspond to syntactic structures. Coupled with the Prosodic Hierarchy hypothesis, the Match theory constitutes a theory of prosodic mapping according to which a morphological word corresponds to a Prosodic Word (ω), a sub-​sentential syntactic phrase (XP) maps onto a Phonological Phrase (ϕ), and a clause is prosodically realized as an Intonational Phrase (ι): ι > ϕ > ω. 8. See Kandybowicz & Torrence 2019 for more on Krachi clausal determiners. 9. This finding is of independent theoretical interest given the debate initiated by Downing (1970) that only root clauses are mapped onto ι. This position has been challenged recently by a growing body of research suggesting that in addition to root domains, embedded clauses in languages such as German, Luganda, and Japanese are mapped as ι constituents (Truckenbrodt 2005; Pak 2008; Selkirk 2009; Ishihara 2014). My findings in this regard are thus consistent with these observations. 10. Recall from Chapter 2 that Krachi has a tone sandhi process in which the second in a series of consecutive high tones is upstepped (Snider 1990). For this reason, the second syllable of kwarɛ is realized with a higher F0 than the first, despite both being H-​bearing. 11. The item kwarɛ . . . gyι ‘think’ is an idiosyncratic “split” verb in Krachi composed of the predicates kwarɛ and gyι, which in isolation bear the independent meanings ‘to collect’ and ‘to eat’ respectively. In the non-​compositional split verb construction, however, neither predicate contributes its independent lexical meaning. 12. Note that this mapping constitutes a violation of Selkirk’s (2011) Match condition because a clausal syntactic constituent (i.e., the embedded TP) does not correspond to an Intonational Phrase in the prosody (see note 7 for a brief description of Match theory). This, however, is unproblematic, as the Match condition is taken to be a violable constraint in Selkirk’s (2011) framework. 13. See Kügler 2018 for an opposing analysis of the prosodic status of embedded complement clauses in Asante Twi based on competing/​conflicting data. Kügler’s (2018: 105–​106) prosodic data lead him to the conclusion that embedded complement clauses in the language are not prosodically integrated with the main clause domain—​that embedded clausal complements are in fact independent Intonational Phrases. It isn’t entirely clear to me why the data in Kügler’s study conflict with the data presented in this chapter. Perhaps the

[ 128 ] Notes

speakers I consulted and those surveyed by Kügler speak different dialects or belong to different age/​social groups. If this is the case, it might be that in these different populations, the prosodic status of embedded complement clauses is variable. I leave exploring this interesting issue to future research. 14. This analysis makes an interesting prediction for Wasa. In (21), it was established that reason adjuncts in the language (e.g., ‘because Kofi was hungry’) are prosodically divided from the main clause by an ι boundary. Because this ι boundary divides C from the material where a wh-​item might surface in this construction, the ACWC generalization in (27) predicts that although Wasa can’t have wh-​in-​situ in a complement clause, it should tolerate wh-​in-​situ in an embedded reason adjunct clause. If this were true, it would represent a strong argument against the purely syntactic/​locality-​based alternative considered in section 3.3. Unfortunately, I do not presently have the data to address this prediction and so must reserve exploring it for future research. 15. This analysis makes a very interesting and testable prediction that lies beyond the scope of the present chapter given its psycholinguistic nature. If the restriction on embedded wh-​in-​situ in languages like Wasa and Asante Twi is ultimately (accidently) related to the fact that embedded TP complement clauses in these languages fail to be prosodically realized as ι constituents, it is conceivable that Wasa and Asante Twi speakers could be “coerced” into accepting instances of embedded wh-​in-​situ if those embedded clauses were produced with an unnatural (i.e., forced) ι phrasing. In those cases, the PF outputs in question would evade violation of the ACWC, despite having artificial prosodies. Such structures might then be expected to exceptionally support embedded in-​ situ interrogatives. The analysis would also make the prediction that Kügler’s (2018) population of Asante Twi speakers, in contrast to mine (see note 13), should accept embedded wh-​in-​situ structures given the fact that embedded TP complements are independent ι constituents in their grammars. In this regard, Kügler’s work might constitute a natural experiment testing the predictions of the ACWC. The prediction/​experiment would also work in the opposite direction. By manipulating the prosody of complement clauses in languages like Krachi and Bono so that they fail to be mapped onto ι (contrary to norm), it might be possible to engineer special contexts in which speakers of these languages exceptionally reject embedded wh-​in-​situ. Findings of this sort would greatly strengthen the central claim of this chapter, namely, that the distribution of wh-​in-​situ is as much a matter of prosody as it is a matter of syntax/​semantics. I leave the pursuit of this prediction for future research. 16. This account raises an interesting prediction for Wasa. If focus fronting induces Intonational Phrasing across the board in both matrix and embedded contexts as claimed earlier, then short focusing any embedded constituent will create a prosodic context in an embedded clause where an ι boundary will be introduced, insulating wh-​from C. All things being equal, this would mean that although it would be impossible to say ‘You think that the men slaughtered what yesterday?’, it should be possible to say ‘You think that YESTERDAY, the men slaughtered what?’. I do not currently have the data to address this very interesting prediction and must therefore leave its exploration for future research. 17. We may wonder, then, about the prosodic well-​formedness of full wh-​ focus movement in languages like Asante Twi in light of the ACWC. If embedded clauses in such languages are characterized by the absence of embedded ι domains, then fully fronted wh-​items and embedded complementizers might

Notes  [ 129 ]

301

conceivably phrase together in violation of the ACWC. Why, then, is long-​ distance wh-​focus movement prosodically permissible in languages like Asante Twi? I explore this issue and subsequently refine the formulation of the ACWC in subsection 3.4.8. 18. Analogous to the psycholinguistically oriented prediction/​experiment outlined in note 15, this analysis makes the testable prediction that if focused constituents were unnaturally realized as separate Intonational Phrase units in Asante Twi, partial wh-​movement structures might then become acceptable outputs. As before, the test would be whether native speakers could be coerced into accepting partial wh-​movement constructions exclusively on the basis of manipulating the natural prosody of such structures. A positive finding in this regard would once again strengthen the contention at the heart of this chapter that in addition to syntactic/​semantic factors, prosodic considerations play a role in constraining the positions that interrogative expressions may occupy in the interior of the clause. 19. One possible approach to this question suggested to me by Chris Collins is to deny the existence of the ACWC as a PF interface condition and instead seek to explain the observed prosodic differences between the languages investigated in this chapter in purely syntactic terms. Along these lines, we might hypothesize a small but crucial syntactic difference between the left peripheries of embedded clauses in Krachi and Bono and those in Wasa and Asante Twi. In particular, suppose there is a phonetically null Interrogative phrase (call it, IntP) that projects above TP in Krachi and Bono, but not in Wasa or Asante Twi. Suppose furthermore that IntP in these languages plays some crucial role linking the matrix Q to the embedded wh-​item. The presence of IntP in Krachi and Bono would explain why embedded wh-​in-​situ is licensed in these languages, but not in Wasa and Asante Twi. It might also explain the differences in the prosodic status of embedded clauses in the languages if IntP is obligatorily parsed as an ι constituent at the syntax-​phonology interface. I leave a more rigorous exploration of this interesting analytical possibility for future research. Thanks to Chris Collins for stimulating this line of thought. CHAPTER 4 1. The Nupe data presented in this chapter come exclusively from fieldwork and are presented in the official Nupe orthography (see Madugu 1980) with added tone diacritics. high tone is indicated by way of acute accent (´). low tone is marked with a grave accent (`). Mid tones are unmarked. rising tones are represented as sequences of grave + acute marking (ˇ). 2. It is not crucial for the present discussion and forthcoming analysis that left dislocated/​topicalized constituents are moved expressions. Nothing that follows hinges on this analysis. However, evidence that left dislocated topics are moved comes from the fact that they are island-​sensitive—​the clitic pronouns that mark the thematic base positions of topics may not occur inside islands in the language. 3. The example in (14) contains an atomic left dislocated expression. When phrases are topicalized in this manner, as in the following sentence, the prosodic effects are identical. (i) (Kèké  tú)ι  (mi     woma wu:n  o)ι bike    ride  1st.sg   enjoy  3rd.sg     top ‘As for bike riding, I enjoy it.’

[ 130 ] Notes

In my recordings of this utterance, I observed prosodic breaks ranging from 149.5ms to 304.7ms separating the topicalized phrase from the main clause. Following the break, pitch reset was observed across all tokens. 4. ‘Because’ clauses in Nupe are adverbial relative clauses. The ‘because’ item heads the relative clause, which is introduced by the two-​part discontinuous morpheme na . . . na (see the description of Nupe relative clauses that follows in section 4.4.3 (example (25)). Other adverbial clauses in Nupe have a similar syntactic and prosodic structure. ‘Although’ clauses (i) and ‘when’ clauses (ii) both appear clause-​ initially and contain a relativized expression functioning as an adverbial. As the following pitch tracks illustrate, the adverbial clauses are also parsed as independent Intonational Phrases. In both constructions, the speaker’s F0 range is reset following the adverbial clause and a prominent break divides the adverbial clause from the main clause. In the ‘although’ clause example (i), the duration of the pause is 240.5ms. The break length is 234.6ms in the case of the ‘when’ adverbial (ii). (i) (Tò     na  Musa á  nakàn ba na)ι (Nànǎ á    ze although  rel Musa prf meat   cut.pst   prt  Nana prf turn.pst ewùn  be   wu:n  yin)ι anger with 3rd.sg prt ‘Although Musa had cut the meat, Nana was angry at him.’

(ii)  (Kámi  na    Musa   si     kèké   na)ι   (Nànǎ   è    pa    eci)ι time  rel  Musa   buy.pst   bike   prf     Nana   pres   pound   yam ‘When Musa bought the bike, Nana was pounding the yam.’

Notes  [ 131 ]

321

5. Although M tone-​bearing in isolation, the third person pronominal subject surfaces in this case with a falling tone. This is a consequence of glide formation. When the underlyingly H tone-​bearing vowel on the complementizer kó is converted to a glide, the H tone disassociates and reassociates rightward on the pronominal subject. As a result, the third person subject is associated with an H+M contour, which surfaces as a falling tone. See Kandybowicz 2008 for more discussion. 6. I qualify the claim that wh-​items in the language are restricted from appearing in embedded contexts because there are in fact certain special prosodic conditions that license the appearance of embedded wh-​occurrences, as predicted by the ACWC. These cases are discussed in (48) and (50) in this section and also in section 4.4.5. 7. Again, given the special cases discussed in (48) and (50) and section 4.4.5, this statement must be qualified. Under normal circumstances, though, the description is accurate. 8. Thanks to Laura Kalin (personal communication) for pointing out this prediction to me. 9. Special thanks to Carolina Fraga (personal communication) for insightful and productive discussion of this prediction. 10. In Kandybowicz 2008: 96, I reported that left dislocation in Nupe is a main clause phenomenon, with topic movement restricted both within and across embedded contexts. The data in (48b) and (50b) seem to suggest that this claim is incorrect. That is, at least for the subset of four speakers I consulted for this chapter, left dislocation does not appear to be restricted to root clauses. The judgments on which the 2008 claim were based came from different speakers than those consulted for the present research. It is entirely possible that the difference in judgments is due to inter-​speaker variation or, more likely, an error on my part when recording and analyzing judgments in 2008. 11. Thanks to Ivona Kučerová (personal communication) for bringing this prediction to my attention. 12. In a number of South Asian languages, restrictions on embedded wh-​ distribution are tied to finiteness (Mahajan 1990; Srivastav 1991; Bayer 1996; Aravind 2018; among others). The question naturally arises—​can Nupe wh-​ expressions appear in non-​finite embedded contexts as in many South Asian languages? The answer is an uninteresting “yes.” As argued by Kandybowicz & Baker (2003), Nupe non-​finite contexts are not true embedded domains. Rather, they are more akin to monoclausal serial verb constructions. These types of structures therefore involve the projection of a non-​finite particle, but not a tense layer or a CP layer. For this reason, short wh-​focus movement to an embedded position (i.e., Asymmetry 1) is impossible in non-​finite contexts simply because there is no embedded focus position to move to. Wh-​ in-​situ, by contrast, is available in non-​finite multiple question contexts, as shown in the following. This, however, is unsurprising under the ACWC given the absence of overt C in such constructions. (i)  Zě   má      ké     yin    du Who   know.how   what   infin   cook ‘Who knows how to cook what?’ (ii)  Zě   yá   yin   du   kánci Who   begin   infin   cook   when ‘Who began to cook when?’

[ 132 ] Notes

CHAPTER 5 1. It is important to note that the ACWC only applies to limit the distribution of wh-​ in-​situ in languages that independently allow for wh-​ in-​situ syntactically. Whereas Richards’s (2010, 2016) frameworks are designed to predict and account for whether a given language will allow wh-​ in-​situ on the basis of both its syntactic and prosodic properties, the ACWC is merely a prosodic well-​ formedness constraint that applies to limit the distribution of wh-​ in-​situ in languages whose syntax provide the resources to license wh-​ in-​situ. For this reason, the ACWC cannot provide an explanation for why a given language lacks wh-​ in-​situ altogether. Its explanatory power is limited to accounting for asymmetries in wh-​ in-​situ distribution in languages that independently allow wh-​expressions to remain unmoved. 2. Beginning with the work of Chang (1997), it has been recognized that moved and in-​situ French wh-​questions are not equivalent from a semantic perspective. The two varieties are not associated with the same presuppositions. If this is correct, French wh-​movement is technically not optional—​it is conditioned by semantic properties. See Hamlaoui (2010) for an account of the prosodic and discourse-​related differences between French wh-​ in-​situ and wh-​ movement questions. 3. This claim has been challenged in recent literature on the basis of attested corpus examples of embedded wh-​ in-​situ (Adli 2006; Mathieu 2009). See also Poletto & Pollock (2015): 142–​144. Despite these challenges, I will adopt the traditional position in the literature dating back to Obenauer (1994) and assume the unavailability of embedded wh-​ in-​situ in French. 4. The prosodic integration of embedded complement clauses in Zulu might lead us to predict on the basis of the ACWC that partial wh-​movement to a position below C is impossible (as in Nupe). However, as Sabel & Zeller (2006) show, partial wh-​movement to this position is indeed available. (i)

U-​cabanga [ukuthi y-​ini a-​yi-​thengile-​yo u-​Bev] 2nd.sg-​think comp cop-​what.9 rc1a-​oc9-​bought-​rs 1a-​Bev ‘What do you think Bev bought?’         (Sabel & Zeller 2006)

Facts like these do not necessarily counter-​exemplify the ACWC. If the partially moved/​focused wh-​expression were parsed as an independent ι constituent, as in Wasa, wh-​and C would not be contained within a single Intonational Phrase and anti-​contiguity would be respected. I lack information on the prosodic status of partially moved or focused expressions in the language and therefore must leave it to Zulu scholars to evaluate this particular analysis. 5. See the collection of articles in Downing & Rialland (2018) for prosodic analyses of additional Bantu languages, as well as several other non-​Bantu African tone languages. 6. The only case I am aware of where C may be overtly realized in Kîîtharaka is the relative clause construction. Depending on the correct analysis of Kîîtharaka relativizers, the item glossed as ‘that’ (û-​ra) in the following may in fact turn out to be the overt exponent of the C head or some other functional head in the left periphery.

Notes  [ 133 ]

341

Boriisi ba-​ka-​thaik-​a [mw-​amba û-​ra Peter 2.police 2.sm-​fut-​arrest-​fv 1-​thief 1-​that 1.Peter a-​ ra-​ on-​ ir-​ e         î-​ goro] 1.sm-​rec.pst-​see-​ppv-​fv  5-​yesterday ‘The police will arrest the thief that Peter saw yesterday.’ (Abels & Muriungi 2008) In the event that the relativizer is indeed an overt C head, the way to test the ACWC in Kîîtharaka is to determine whether or not wh-​ in-​situ is available inside relative clauses, as it is in other African languages (e.g., Krachi; Torrence & Kandybowicz 2015). 7. For finite clauses, Malayalam has been argued to have only clause-​bound (i.e., non-​embedded) wh-​ in-​situ (Srikumar 1992, 2007; Madhavan 2013). See Aravind 2018 for evidence that this claim is factually incorrect. 8. This characterization is simplified, but it is more or less accurate. 9. See Aravind (2018) for details about which finite embedded clauses must prepose, which are exempt, and the prosodic motivation for these movements. (ii)

[ 134 ] Notes

REFERENCES

Abels, Klaus, and Peter Muriungi. 2008. The Focus Marker in Kîîtharaka: Syntax and Semantics. Lingua 118(5): 687–​731. Adger, David. 2007. Stress and Phasal Syntax. Linguistic Analysis 33: 238–​266. Adli, Aria. 2006. French Wh-​In-​situ Questions and Syntactic Optionality: Evidence from Three Data Types. Zeitschrift für Sprachwissenschaff 25: 163–​203. Adonae, Daniels Ananey. 2005. Aspects of Kaakyi Tonology. M.Phil. thesis, University of Ghana, Legon. Aravind, Athulya. 2018. Licensing Long-​distance Wh-​in-​situ in Malayalam. Natural Language and Linguistic Theory 36: 1–​43. Bassong, Paul Roger. 2010. The Structure of the Left Periphery in Bàsàá. Master’s thesis, Université de Yaoundé. Bayer, Josef. 1996. Directionality and Logical Form. Dordrecht: Kluwer. Beck, Sigrid. 1996. Quantified Structures as Barriers for LF Movement. Natural Language Semantics 4: 1–​56. Bhatt, Rajesh, and Vaneeta Dayal. 2007. Rightward Scrambling as Rightward Remnant Movement. Linguistic Inquiry 38(2): 287–​301. Boeckx, Cedric. 2000. Decomposing French Questions. In Jim Alexander, Na-​R ae Han, and Michelle Minnick Fox (eds.), Proceedings of the 23rd Annual Penn Linguistics Colloquium, 69–​80. Philadelphia: University of Pennsylvania Working Papers in Linguistics. Boersma, Paul, and David Weenink. 2017. Praat: Doing Phonetics by Computer (Version 6.0.35). Available from: http://​www.praat.org. Bokamba, Eyamba G. 1976. Question Formation in Some Bantu Languages. Ph.D. thesis, Indiana University. Bošković, Željko. 1998. LF Movement and the Minimalist Program. In Pius Tamanji and Kiyomi Kusumoto (eds.), Proceedings of NELS 28, 43–​57. Amherst: University of Massachusetts, Graduate Linguistic Student Association. Bošković, Źeljko. 2000. Sometimes in SpecCP, Sometimes in-​situ. In Roger Martin, David Michaels, and Juan Uriagereka (eds.), Step by Step: Essays on Minimaism in Honor of Howard Lasnik, 53–​87. Cambridge, MA: MIT Press. Bresnan, Joan. 1971. Sentence Stress and Syntactic Transformation. Language 47: 257–​281. Bromberger, Sylvain. 1992. On What We Know We Don’t Know: Explanation, Theory, Linguistics and How Questions Shape Them. Chicago: University of Chicago Press.

361

Buell, Leston. 2009. Evaluating the Immediate Postverbal Position as a Focus Position in Zulu. In Masangu Matando, Fiona McLaughlin, and Eric Potsdam (eds.), Selected Proceedings of the 38th Annual Conference on African Linguistics, 166–​172. Somerville, MA: Cascadilla Proceedings Project. Buell, Leston. 2011. Zulu Ngani ‘why’: Postverbal and Yet in CP. Lingua 121: 805–​821. Cable, Seth. 2010. The Grammar of Q: Q-​Particles, Wh-​Movement and Pied Piping. New York: Oxford University Press. Castroviejo Miró, Elena. 2006. Wh-​Exclamatives in Catalan. Ph.D. thesis, Universitat de Barcelona. Chang, Lisa. 1997. Wh-​in-​situ Phenomena in French. Master’s thesis, University of British Columbia. Chen, Matthew. 1987. The Syntax of Xiamen Tone Sandhi. Phonology Yearbook 4: 109–​149. Cheng, Lisa Lai-​Shen. 1991. On the Typology of Wh-​Questions. Ph.D. thesis, Massachusetts Institute of Technology. Cheng, Lisa Lai-​Shen, and Laura J. Downing. 2007. The Prosody and Syntax of Zulu Relative Clauses. SOAS Working Papers in Linguistics 15: 51–​63. Cheng, Lisa Lai-​Shen, and Laura J. Downing. 2011. Mapping Phonology to Syntax: Evidence from Two Wh-​in-​situ Languages. Paper presented at GLOW 34, University of Vienna. Ms. Leiden University and ZAS, Berlin. Cheng, Lisa Lai-​Shen, and Laura J. Downing. 2016. Phasal Syntax = Cyclic Phonology? Syntax 19: 156–​191. Cheng, Lisa Lai-​Shen, and Johan Rooryck. 2000. Licensing Wh-​ in-​situ. Syntax 3: 1–​19. Chomsky, Noam. 2001. Derivation by Phase. In Michael Kenstowicz (ed.), Ken Hale: A Life in Language, 1–​52. Cambridge, MA: MIT Press. Clements, George N. 1978. Tone and Syntax in Ewe. In Donna Jo Napoli (ed.), Elements of Tone, Stress, and Intonation, 21–​99. Washington, DC: Georgetown University Press. Collins, Chris. 1991. Why and How Come. In Lisa Cheng and Hamida Demirdache (eds.), More Papers on Wh-​Movement: MIT Working Papers in Linguistics 15, 31–​ 45. Cambridge, MA: MITWPL. Crystal, Thomas H., and Arthur S. House. 1988. Segmental Durations in Connected-​ Speech Signals: Current Results. Journal of the Acoustic Society of America 83: 1553–​1573. Dayal, Vaneeta. 1996. Locality in Wh-​Quantification: Questions and Relative Clauses in Hindi. Dordrecht: Kluwer. Déchaine, Rose-​Marie. 2001. On the Left Edge of Yoruba Complements. Lingua 111: 83–​130. Déprez, Viviane, Kristen Syrett, and Shigeto Kawahara. 2012. Interfacing Information and Prosody: French wh-​in-​situ Questions. In Irene Franco, Sara Lusini, and Andrés Saab (eds.), Romance Languages and Linguistic Theory 2010: Selected Papers from ‘Going Romance’ Leiden 2010, 135–​153. Amsterdam: John Benjamins. D’Imperio, Mariapaola, Gorka Elordieta, Sónia Frota, Pilar Prieto, and Marina Vigário. 2005. Intonational Phrasing in Romance: The Role of Syntactic and Prosodic Structure. In Sónia Frota, Marina Vigário, and Maria João Freitas (eds.), Prosodies: With Special Reference to Iberian Languages, 59–​97. Berlin and New York: Mouton de Gruyter.

[ 136 ] References

Downing, Bruce. 1970. Syntactic Structure and Phonological Phrasing in English. Ph.D. thesis, University of Texas at Austin. Downing, Laura J., and Annie Rialland. 2018. Intonation in African Tone Languages. Berlin: Mouton De Gruyter. Dundaa, Mark. 2007. Transitional Primer: Can You Read and Write in Kaakyi? Tamale, Ghana: Ghana Insitute of Linguitics, Literacy, and Bible Translation. Elfner, Emily. 2012. Syntax-​prosody Interactions in Irish. Ph.D. thesis, University of Massachusetts, Amherst. Elfner, Emily. 2015. Recursion in Prosodic Phrasing: Evidence from Connemara Irish. Natural Language and Linguistic Theory 33: 1169–​1208. Fanselow, Gisbert. 2006. Partial Movement. In Martin Everaert, Henk van Riemsdijk, Rob Goedemans, and Bart Hollebrandse (eds.), The Blackwell Companion to Syntax (Volume 3), 437–​492. London: Blackwell. Feldhausen, Ingo. 2008. The Prosody-​Syntax Interface in Catalan. Ph.D. thesis, University of Potsdam. Feldhausen, Ingo. 2010. Sentential Form and Prosodic Structure of Catalan. Amsterdam: John Benjamins. Frascarelli, Mara. 2000. The Syntax-​Phonology Interface in Focus and Topic Constructions in Italian. Dordrecht: Kluwer. Golston, Chris. 1995. Syntax Outranks Phonology: Evidence from Ancient Greek. Phonology 12: 343–​368. Green, Melanie, and Philip J. Jaggar. 2003. Ex-​situ and in-​situ Focus in Hausa: Syntax, Semantics and Discourse. In Jacqueline Lecarme (ed.), Research in Afroasiatic Grammar II [CILT 241], 187–​213. Amsterdam: John Benjamins. Guasti, Maria Teresa, and Marina Nespor. 1999. Is Syntax Phonology-​Free? In René Kager and Wim Zonneveld (eds.), Phrasal Phonology, 73–​98. Nijmegen: Nijmegen University Press. Hagstrom, Paul. 1998. Decomposing Questions. Ph.D. thesis, Massachusetts Institute of Technology. Hamlaoui, Fatima. 2010. On the Role of Phonology and Discourse in Francilian French Wh-​ Questions. Journal of Linguistics 47: 1–​34. Hamlaoui, Fatima, and Kriszta Szendrői. 2017. The Syntax-​Phonology Mapping of Intonational Phrases in Complex Sentences: A Flexible Approach. Glossa: A Journal of General Linguistics 2(1), 55: 1–​31. Heath, Jeffrey, and Laura McPherson. 2013. Tonosyntax and Reference Restriction in Dogon NPs. Language 89: 265–​296. Hermon, Gabriella. 1985. Syntactic Modularity. Dordrecht: Foris. Hetzron, Robert. 1972. Phonology in Syntax. Journal of Linguistics 8: 251–​265. Hornstein, Norbert. 1995. Logical Form. Oxford: Blackwell. Hsu, Brian. 2015. Variation in Bangla Complementizer Order at the Syntax-​Prosody Interface. In Thuy Buy and Deniz Özyildiz (eds.), Proceedings of the 45th Meeting of the North East Linguistics Society: Volume 2, 35–​44. Amherst, MA: GLSA. Ishihara, Shinichiro. 2007. Major Phrase, Focus Intonation, Multiple Spell-​Out. The Linguistic Review 24: 137–​167. Ishihara, Shinichiro. 2014. On Match Constraints. Talk Presented at the Workshop on the Prosodic Hierarchy in a Typological Perspective. Stockholm University. Jokweni, Mbulelo. 1995. Aspects of IsiXhosa Phrasal Phonology. Ph.D. thesis, University of Illinois, Urbana-​Champaign. Kahnemuyipour, Arsalan. 2009. The Syntax of Sentential Stress. Oxford: Oxford University Press.

References  [ 137 ]

381

Kan, Seda. 2009. Prosodic Domains and the Syntax-​Prosody Mapping in Turkish. Master’s thesis, Boǧaziçi University. Kandybowicz, Jason, and Mark C. Baker. 2003. On Directionality and the Structure of the Verb Phrase: Evidence from Nupe. Syntax 6: 115–​155. Kandybowicz, Jason. 2008. The Grammar of Repetition: Nupe Syntax at the Syntax-​ Phonology Interface. Amsterdam: John Benjamins. Kandybowicz, Jason. 2009. Embracing Edges: Syntactic and Phono-​syntactic Edge Sensitivity in Nupe. Natural Language and Linguistic Theory 27: 305–​344. Kandybowicz, Jason. 2017. On Prosodic Variation and the Distribution of Wh-​in-​ situ. Linguistic Variation 17: 110–​147. Kandybowicz, Jason, and Harold Torrence. 2011. How Why is Different: Wh-​ in-​situ in Krachi. Snippets 23: 5–​6. Kandybowicz, Jason, and Harold Torrence. 2012. Krachi Wh-​in-​situ: A Question of Prosody. In Jaehoon Choi, E. Alan Hogue, Jeffrey Punske, Deniz Tat, Jessamyn Schertz, and Alex Trueman (eds.), Proceedings of the 29th West Coast Conference on Formal Linguistics, 362–​370. Somerville, MA: Cascadilla Press. Kandybowicz, Jason, and Harold Torrence. 2013. Unweaving the Interrogative Rainbow: The Interplay between Syntax, Semantics, and Prosody in Four Tano Languages. Talk Presented at the 44th Annual Conference on African Linguistics, Georgetown University. Kandybowicz, Jason, and Harold Torrence. 2019. A First Look at Krachi Clausal Determiners. In Margit Bowler, Philip T. Duncan, Travis Major, and Harold Torrence (eds.), Schuhschrift: Papers in Honor of Russell Schuh, 66–​76. Los Angeles: eScholarship. Karimi, Simin. 2005. A Minimalist Approach to Scrambling: Evidence from Persian. Berlin: Mouton de Gruyter. Kawu, Ahmadu Ndanusa. 1990. Some Instances of Sentential Complementation in Nupe. Master’s thesis, University of Ibadan. Kawu, Ahmadu Ndanusa. 1999. Wh-​Questions and Focus Constructions in Nupe. Ms. Rutgers University. Kawu, Ahmadu Ndanusa. 2000. Hiatus Resolution in Nupe. Journal of West African Languages 18: 27–​43. Kirk, Allison. 2012. Word Order Variation in New Testament Greek wh-​ Questions. In Ans M. C. van Kemenade and Nynke de Haas (eds.), Historical Linguistics 2009: Selected Papers from the 19th International Conference on Historical Linguistics, 293–​314. Amsterdam: John Benjamins. Ko, Heejeong. 2005. Syntax of ‘why’-​in-​situ: Merge into [spec, CP] in the Overt Syntax. Natural Language and Linguistic Theory 23: 867–​916. Kratzer, Angelika, and Elisabeth Selkirk. 2007. Phase Theory and Prosodic Spell-​ Out: The Case of Verbs. The Linguistic Review 24: 93–​135. Kügler, Frank. 2018. Tone and Intonation in Akan. In Laura J. Downing and Annie Rialland (eds.), Intonation in African Tone Languages, 89–​129. Berlin: Mouton De Gruyter. Lasnik, Howard, and Mamoru Saito. 1992. Move α: Conditions on Its Application and Output. Cambridge, MA: MIT Press. Letsholo, Rose. 2011. The Syntactic Distribution of Argument and Adjunct Question Word Constructions in Ikalanga. Journal of African Languages and Linguistics 32: 219–​250. Lin, Jo-​Wang. 1992. The Syntax of Zenmeyang ‘How’ and Weishenme ‘Why’ in Mandarin Chinese. Journal of East Asian Linguistics 1: 293–​331.

[ 138 ] References

Madhavan, Punnapurtath. 2013. Multiple Wh-​Questions and the Cleft Construction in Malayalam. In Tonjes Veenstra and Katharina Hartman (eds.), Cleft Structures, 269–​284. Amsterdam: John Benjamins. Madugu, Isaac S. George. 1980. Nupe Orthography. Ms., University of Ibadan. Mahajan, Anoop. 1990. The A/​A-​bar Distinction and Movement Theory. Ph.D. thesis, Massachusetts Institute of Technology. Mahajan, Anoop. 1997. Rightward Scrambling. In Dorothee Beerman, David LeBlanc, and Henk van Riemsdijk (eds.), Rightward Movement, 185–​213. Amsterdam: John Benjamins. Makasso, Emmanuel-​Moselly, Fatima Hamlaoui, and Seunghun J. Lee. 2018. Aspects of the Intonational Phonology of Bàsàá. In Laura J. Downing and Annie Rialland (eds.), Intonation in African Tone Languages, 167–​193. Berlin: Mouton De Gruyter. Marlo, Michael and David Odden. 2007. The Exponence of TAM in Bakweri. SOAS Working Papers in Linguistics 15: 19–​31. Mathieu, Éric. 1999. Wh-​in-​situ and the Intervention Effect. UCL Working Papers in Linguistics 11: 441–​472. Mathieu, Éric. 2002. The Syntax of Non-​canonical Quantification: A Comparative Study. Ph.D. thesis, University College London. Mathieu, Éric. 2004. The Mapping of Form and Interpretation: The Case of Optional Wh-​Movement in French. Lingua 114(9): 1090–​1132. Mathieu, Éric. 2009. Les Questions en Français: Micro et macro-​variation. In France Martineau, Raymond Mougeon, Terry Nadasdi, and Mireille Tremblay (eds.), Le Français d’ici: Études linguistiques et sociolinguistiques de la variation, 37–​66. Toronto: GREF. Maxwell, Edith. 1981. Question Strategies and Hierarchies of Grammatical Relations in Kinyarwanda. In Proceedings of the 7th Annual Meeting of the Berkeley Linguistics Society, 166–​177. Berkeley: University of California. Muriungi, Peter Kinyua. 2005. Wh-​Questions in Kiiharaka. Studies in African Linguistics 34: 43–​104. Myrberg, Sara. 2010. The Intonational Phonology of Stockholm Swedish. (ACTA Universitatis Stockholmiensis 53 Stockholm Studies in Scandinavian Philology New Series). Ph.D. thesis, Stockholm University. Nespor, Marina, and Irene Vogel. 1986. Prosodic Phonology. Dordrecht: Foris. Obenauer, Hans-​Georg. 1994. Aspects de la syntaxe A-​barre: Effets d’intervention et mouvements des quantifieurs. Ph.D. thesis, Université de Paris VIII. Ouhalla, Jamal. 1996. Remarks on the Binding Properties of Wh-​ Pronouns. Linguistic Inquiry 27(4): 676–​707. Pak, Marjorie. 2008. The Postsyntactic Derivation and its Phonological Reflexes. Ph.D. thesis, University of Pennsylvania. Pesetsky, David. 2000. Phrasal Movement and Its Kin. Cambridge, MA: MIT Press. Poletto, Cecilia, and Jean-​Yves Pollock. 2015. Arguing for Remnant Movement in Romance. In Günther Grewendorf (ed.), Remnant Movement, 135–​178. Berlin: Walter de Gruyter. Potsdam, Eric. 2006. More Concealed Pseudoclefts in Malagasy and the Clausal Typing Hypothesis. Lingua 116: 2154–​2182. Pullum, Geoffry K., and Arnold M. Zwicky. 1988. The Syntax-​Phonology Interface. In Frederick J. Newmeyer (ed.), Linguistics: The Cambridge Survey, I –​Linguistic Theory: Foundations, 255–​280. Cambridge, UK: Cambridge University Press.

References  [ 139 ]

4 01

Reglero, Lara. 2004. A’ Dependencies in Spanish and Basque. Ph.D. thesis, University of Connecticut, Storrs. Reglero, Lara. 2005. Wh-​in-​situ Constructions: Syntax and/​or Phonology? In John Alderete, Chung-​hye Han, and Alexei Kochetov (eds.), Proceedings of WCCFL 24, 334–​342. Somerville, MA: Cascadilla Proceedings Project. Reinhart, Tanya. 1981. A Second Comp Position. In Adriana Belletti, Luciana Brandi, and Luigi Rizzi (eds.), Theory of Markedness in Generative Grammar, 518–​557. Pisa: Scuola Normale Superiore. Reinhart, Tanya. 1998. Wh-​in-​situ in the Framework of the Minimalist Program. Natural Language Semantics 6: 29–​56. Richards, Norvin. 2010. Uttering Trees. Cambridge, MA: MIT Press. Richards, Norvin. 2016. Contiguity Theory. Cambridge, MA: MIT Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, MA: MIT Press. Rizzi, Luigi. 1997. The Fine Structure of the Left Periphery. In Liliane Haegeman (ed.), Elements of Grammar, 281–​338. Dordrecht: Kluwer. Rizzi, Luigi. 2001. On the Position Int(errogative) in the Left Periphery of the Clause. In Guglielmo Cinque and Giampaolo Salvi (eds.), Current Studies in Italian Syntax: Essays Offered to Lorenzo Renzi, 287–​296. Amsterdam: Elsevier North-​Holland. Saah, Kofi. 1988. Wh-​Questions in Akan. Journal of West African Languages 18: 17–​28. Sabel, Joachim. 2000. Partial Wh-​Movement and the Typology of Wh-​ Questions. In Uli Lutz, Gereon Müller, and Arnim von Stechow (eds.), Wh-​Scope Marking, 409–​446. Amsterdam: John Benjamins. Sabel, Joachim, and Jochen Zeller. 2006. Wh-​Question Formation in Nguni. In John Mugane, John P. Hutchinson, and Dee A. Worman (eds.), Selected Proceedings of the 35th Annual Conference on African Linguistics, 271–​283. Somerville, MA: Cascadilla Proceedings Project. Selkirk, Elisabeth. 1984. Phonology and Syntax: The Relation between Sound and Structure. Cambridge, MA: MIT Press. Selkirk, Elisabeth. 1986. On Derived Domains in Sentence Phonology. Phonology Yearbook 3: 371–​405. Selkirk, Elisabeth. 1995. Sentence Prosody: Intonation, Stress, and Phrasing. In John Goldsmith (ed.), The Handbook of Phonological Theory, 550–​569. Oxford: Basil Blackwell. Selkirk, Elisabeth. 1996. The Prosodic Structure of Function Words. In James L. Morgan and Katherine Demuth (eds.), Signal to Syntax: Prosodic Bootstrapping from Speech to Grammar in Early Acquisition, 187–​214. Mahwah, NJ: Lawrence Erlbaum Associates. Selkirk, Elisabeth. 2006. Strong Minimalist Spell-​Out of Prosodic Phrases. Paper presented at GLOW 29: Workshop on Prosodic Phrasing. Universitat Autònoma Barcelona. Selkirk, Elisabeth. 2009. On Clause and Intonational Phrase in Japanese: The Syntactic Grounding of Prosodic Constituent Structure. Gengo Kenkyu Journal of the Linguistic Society of Japan 136: 35–​74. Selkirk, Elisabeth. 2011. The Syntax-​Phonology Interface. In John Goldsmith, Jason Riggle, and Alan Yu (eds.), The Handbook of Phonological Theory (2nd edition), 435–​484. Oxford: Wiley Blackwell. Selkirk, Elisabeth, and Tong Shen. 1990. Prosodic Domains in Shanghai Chinese. In Sharon Inkelas and Draga Zec (eds.), The Phonology-​Syntax Connection, 313–​ 337. Chicago: University of Chicago Press.

[ 140 ] References

Shlonsky, Ur, and Gabriela Soare. 2011. Where’s ‘why’? Linguistic Inquiry 42: 651–​669. Simpson, Andrew. 2000. Wh-​Movement and the Theory of Feature Checking. Amsterdam: John Benjamins. Simpson, Andrew, and Tanmoy Bhattacharya. 2003. Obligatory Overt Wh-​ Movement in a Wh-​in-​situ Language. Linguistic Inquiry 34: 127–​142. Snider, Keith. 1990. Tonal Upstep in Krachi: Evidence for Register Tier. Language 66: 453–​474. Srikumar, K. 1992. Question-​Word Movement in Malayalam and GB Theory. Ph.D. thesis, Osmania University, Hyderabad. Srikumar, K. 2007. Clausal Pied-​Piping and Subjacency. In Joseph Bayer, Tanmoy Bhattacharya, and M. T. Hany Babu (eds.), Linguistic Theory and South Asian Languages: Essays in Honour of K. A. Jayaseelan, 53–​69. Amsterdam: John Benjamins. Srivastav, Vaneeta. 1991. Subjacency Effects at LF: The Case of Hindi Wh. Linguistic Inquiry 22: 762–​769. Stepanov, Arthur, and Wei-​Tien Dylan Tsai. 2008. Cartography and Licensing of wh-​Adjuncts: A Cross-​Linguistic Perspective. Natural Language and Linguistic Theory 26: 589–​638. Szabolcsi, Anna, and Frans Zwarts. 1993. Weak Islands and an Algebraic Semantics for Scope Taking. Natural Language Semantics 1: 235–​285. Tailleur, Sandrine. 2013. The French Wh-​Interrogative System: Est-​ce que, Clefting? Ph.D. thesis, University of Toronto. Thornton, Rosalind. 2008. Why Continuity. Natural Language and Linguistic Theory 26: 107–​146. Torrence, Harold, and Jason Kandybowicz. 2012. Comparative Tano Interrogative Syntax: Typological and Genetic Implications. Paper Presented at the 43rd Annual Conference on African Linguistics, Tulane University. Torrence, Harold, and Jason Kandybowicz. 2013. Comparative Tano Interrogative Syntax: The View from Krachi and Bono. In Olanike Ola Orie and Karen W. Sanders (eds.), Selected Proceedings of the 43rd Annual Conference on African Linguistics, 222–​234. Somerville, MA: Cascadilla Press. Torrence, Harold, and Jason Kandybowicz. 2015. Wh-​Question Formation in Krachi. Journal of African Languages and Linguistics 36: 253–​286. Truckenbrodt, Hubert. 1995. Phonological Phrases: Their Relation to Syntax, Focus, and Prominence. Ph.D. thesis, Massachusetts Institute of Technology. Truckenbrodt, Hubert. 1999. On the Relation between Syntactic Phrases and Phonological Phrases. Linguistic Inquiry 30: 219–​256. Truckenbrodt, Hubert. 2005. A Short Report on Intonation Phrase Boundaries in German. Linguistiche Berichte 203: 273–​296. Urbina, Jon Ortiz de. 1990. Operator Feature Percolation and Clausal Pied-​Piping. In Lisa Lai-​Shen Cheng and Hamida Demirdache (eds.), Papers on Wh-​Movement, 193–​208. Cambridge, MA: MIT Working Papers in Linguistics 13. Uribe-​Etxebarria, Myriam. 2002. In-​situ Questions and Masked Movement. Linguistic Variation Yearbook 2(1): 217–​257. Wagner, Michael. 2005. Prosody and Recursion. Ph.D. thesis, MIT. Wahba, Wafaa Abdel-​Faheem Batran. 1991. LF Movement in Iraqi Arabic. In C.-​ T. James Huang and Robert May (eds.), Logical Structure and Linguistic Structure: Cross-​Linguistic Perspectives, 253–​276. Dordrecht: Kluwer. Wasike, Aggrey. 2007. The Left Periphery, Wh-​in-​situ, and A-​bar Movement in Lubukusu and other Bantu Languages. Ph.D. thesis, Cornell University.

References  [ 141 ]

421

Williamson, Kay, and Roger Blench. 2000. Niger-​Congo. In Bernd Heine and Derek Nurse (eds.), African Languages: An Introduction, 11–​42. Cambridge: Cambridge University Press. Zec, Draga, and Sharon Inkelas. 1990. Prosodically Constrained Syntax. In Sharon Inkelas and Draga Zec (eds.), The Phonology-​Syntax Connection, 365–​378. Zentz, Jason. 2016. Wh-​question Formation Strategies in Bantu. Ph.D. thesis, Yale University. Zwicky, Arnold M., and Ellen M. Kaisse (eds.). 1987. Syntactic Conditions on Phonological Rules (Phonology Yearbook 4). Cambridge, UK: Cambridge University Press. Zwicky, Arnold M., and Geoffry K. Pullum. 1986. The Principle of Phonology-​ Free Syntax: Introductory Remarks. Ohio State University Working Papers in Linguistics 32: 63–​91.

[ 142 ] References

GENERAL INDEX

Abels, Klaus, 112–​113, 134 Adger, David, 123 adjacency, 9–​10, 19, 24–​25, 72, 79–​80, 123–​124 Adli, Aria, 133 Adonae, Daniels Ananey, 123–​124, 127 adverbial clause, 85, 90, 131 Agree, 49–​50, 121 agreement, 48, 119 Ahn, Byron, 120 amelioration (prosodic), 95–​96, 98, 103 Anti-​contiguity, 2–​4, 8, 36, 38, 40, 64, 70, 72–​73, 75, 80, 91, 93, 97, 99–​100, 104, 113, 117, 120–​121,  133 Anti-​contiguity of Wh-​and C (ACWC), 3–​4, 64–​67, 69–​75, 80–​81, 91–​92, 95–​116, 118, 120–​121, 129–​130, 132–​134 Aravind, Athulya, 99, 118–​119, 132, 134 architecture of grammar, 1, 37, 121 assimilation, 84–​86, 88 ATR (Advanced Tongue Root), 124–​125   Baker, Mark C., 132 Bassong, Paul Roger, 107–​108 Bayer, Josef, 99, 113, 115, 132 Beck, Sigred, 47 Benue-​Congo, 2, 75, 99–​100, 106 Bhatt, Rajesh, 114 Bhattacharya, Tanmoy, 99, 115–​116 binding,  48–​49 Blench, Roger, 39 Boeckx, Cedric, 101 Boersma, Paul, 53, 81 Bokamba, Eyamba, 42

Bošković, Željko, 44, 101–​102, 104 boundary breath group, 52, 81 Intonational Phrase (ι), 3–​4, 29, 31–​32, 34–​35, 52, 54–​55, 57, 65, 67–​69, 74, 80–​81–​86, 88–​93, 95, 97, 107–​108, 110–​111, 114, 129 obligatory, 2, 30 Phonological Phrase (ϕ), 10, 12, 15–​18, 26, 34, 80, 126 prosodic, 2–​3, 9, 29, 33–​34, 36–​ 37, 52, 67, 97, 125 (see also boundary: Phonological Phrase [ϕ] and boundary: Intonational Phrase [ι]) syntactic, 9–​10, 89–​90, 101, 127 tensed clause, 44 tone, 12, 30, 32, 53, 60, 62, 70, 124–​125 break. See prosodic break breath group. See boundary: breath group Bresnan, Joan, 123 bridge verb, 32, 55 Bromberger, Sylvain, 126 Buell, Leston, 6, 106   Cable, Seth, 47 cartography (of clauses), 95 Castroviejo Miró, Elena, 105 Chang, Lisa, 133 Chen, Matthew, 9 Cheng, Lisa Lai-​Shen, 44, 47, 60, 99, 101, 107, 124 choice function, 27–​28 Chomsky, Noam, 126–​127 clausal determiner, 53, 128

41

clause type, 86, 89–​90 Clements, George N., 5, 9 clitic, 79, 83, 124–​125, 130 Collins, Chris, 126, 130 Comp-​trace effect, 79 complement clause, 2–​3, 8, 29–​32, 35, 39–​41, 43–​47, 50–​51, 53, 55–​67, 69, 73, 86–​87, 89, 91–​92, 100–​101, 103, 105, 107–​108, 110, 113–​116, 118, 126, 128–​129, 133 complementizer, 1–​2, 4, 8–​9, 11, 29, 32, 35, 40, 55, 57–​60, 62, 64, 66–​67, 69–​70, 72–​74, 78–​79, 85–​87, 90, 92–​94, 105, 107–​111, 113, 115, 117, 120, 125, 129, 132 complementizer drop in Nupe, 76, 87 in Tano languages, 66, 70 conditional clause, 83 contiguity, 1, 29, 64, 79, 119–​121, 126 Contiguity Adjunction, 80, 123, 126 Contiguity Prominent, 9, 11, 27, 72, 79 coordination in Asante Twi, 61–​62 in Krachi, 18, 24–​25 in Nupe, 84, 98 coordinator, 16–​17, 24–​26, 61, 124–​125 CP (prosodic status of) in Asante Twi, 61–​64, 128–​129 in Bàsàá, 107 in Bono, 56–​59, 64 in Catalan, 105 in French, 101 in Krachi, 31–​32, 34–​36, 53–​56, 64, 125–​126,  128 in Malayalam, 119 in Nupe, 82–​91, 131 in Spanish, 103–​104 in Wasa, 59–​61, 64, 129 in Zulu, 107 Crystal, Thomas H., 52   D’Imperio, Mariapaola, 103 D-​linked wh-​expression, 6, 50, 66, 68, 105, 115, 118–​119 Dayal, Vaneeta, 114 Déchaine, Rose-​Marie, 5 demonstrative, 16, 19, 124–​125 Déprez, Viviane, 101, 104 determiner, 12, 14–​15, 53, 124, 128

[ 144 ]  General Index

ditransitive construction, 92 domain prosodic, 10–​11, 25, 29, 33–​37, 51, 55, 58–​61, 63, 67, 69, 70, 72, 82, 89–​92, 97, 102–​104, 108, 120, 123, 125–​126, 128–​129 spell-​out,  50 domination (structural), 9–​10, 34–​35, 72, 79–​80, 97 Downing, Bruce, 30, 60, 82, 84, 128 Downing, Laura J., 5, 60, 107, 124, 133 DP interrogative DP prosody in Bono, 68 in Krachi, 8, 11, 20–​27 in Wasa, 67–​69 non-​interrogative DP prosody in Krachi, 11–​19, 124 null-​headed, 15, 124 Dundaa, Mark, 123–​124, 127   echo question, 44–​45, 105 edge active (transparent), 9–​12, 15–​16, 20, 23–​27, 28–​34, 37, 72, 79–​80,  124 adjacency to, 9, 25, 72, 79–​80 inert, 25 prominence, 72, 79, 91, 100, 102, 104–​105, 107–​108, 110, 120 prosodic, 9–​12, 16, 18–​20, 22–​23, 25, 27, 30–​31, 34, 52–​59, 61, 62, 71–​72, 81–​82, 91–​92, 102, 104, 107–​110, 112, 114, 120, 124–​126 syntactic, 9–​10, 14–​20, 25–​29, 34, 41–​42, 45, 83, 109, 112, 118, 124–​125,  128 Edge Alignment Theory, 9, 27 Elfner, Emily, 124 Elordieta, Gorka, 103 embedded clause, 2, 3, 7–​8, 29–​30, 31–​32, 34–​37, 39–​41, 43–​51, 53–​67, 69, 71, 73–​75, 77–​78, 80–​81, 86–​92, 98, 100–​101, 103–​119, 123, 127–​130, 133–​134 embedded context, 2, 3, 7, 29, 36–​37, 44, 66, 74, 79, 92, 132 entanglement (of syntax and prosody), 2, 5–​6, 8, 24, 27, 29, 37–​38

externalization, 3, 40, 66 extraposed clause, 82, 116   F0 boost, 12, 15, 31–​32, 57, 67–​68, 84, 128 declination, 60, 62, 70 depression/​lowering, 12, 14–​23, 25, 31–​32, 53–​55, 57–​59, 61, 67–​68, 88, 124–​126 range, 32, 34, 55, 57, 59, 61, 67–​68,  131 reset, 31–​32, 51–​52, 54–​62, 67–​68, 70, 81–​86, 88–​90, 93–​94,  131 Fanselow, Gisbert, 50 Fanselow’s Generalization (S2), 50 Feldhausen, Ingo, 60, 105 focus, 3–​4, 6, 28, 41–​43, 45, 49, 67–​70, 76–​77, 92–​94, 129–​130, 132–​133 focus movement, 48–​50, 66–​70, 73, 77, 92–​93, 95, 127–​130, 132 FocusP, 29, 69, 95 Fraga, Carolina, 132 Frascarelli, Mara, 83 Frota, Sónia, 103 fundamental frequency. See F0   Ghana Institute for Linguistics, Literacy, and Bible Translation (GILLBT), 123, 127 glide formation, 84–​85, 89–​90, 93, 95, 132 goal (of a probe), 10, 120–​121 Golston, Chris, 123 Green, Melanie, 42, 47 Grouping (in Contiguity Theory), 10, 35–​37,  123 Guasti, Maria Teresa, 5   Hagstrom, Paul, 47 Hamlaoui, Fatima, 60, 82, 101, 107–​108,  133 Heath, Jeffrey, 5 Hermon, Gabriella, 116 Hetzron, Robert, 123 hiatus (resolution), 84–​85, 90, 93, 95 Hornstein, Norbert, 6 House, Arthur S., 52 Hsu, Brian, 82

immediately-​after-​the-​verb position (IAV), 107, 109 Inkelas, Sharon, 123 interface syntax-​phonology, 5–​6, 27, 73–​74, 104, 120–​121, 130 syntax-​semantics, 39, 47, 129 internal merge. See movement inter-​speaker variation, 132 Intonational Phrase, 2–​3, 8–​9, 29–​32, 34–​35, 40, 51–​61, 63–​68, 70–​75, 80–​82, 84–​85, 90, 94, 97, 101–​104, 107, ​108, 111, 120, 128–​131,  133 IntP, 130 Ishihara, Shinichiro, 55, 57, 123, 128 island, 40, 123, 130   Jaggar, Philip J., 42, 47 Jokweni, Mbulelo, 60 Jun, Sun-​Ah, 101   Kahnemuyipour, Arsalan, 5 Kaisse, Ellen M., 5 Kalin, Laura, 132 Kan, Seda, 60 Kandybowicz, Jason, 6, 39, 41, 47, 53, 79, 86, 92, 123, 125, 127–​128, 132, 134 Karimi, Simin, 6 Kawahara, Shigeto, 101, 104 Kawu, Ahmadu Ndanusa, 84, 86 Kirk, Allison, 6 Ko, Heejeong, 6, 126 Kratzer, Angelika, 123 Kučerová, Ivana, 132 Kügler, Frank, 128–​129   Lasnik, Howard, 126 Lee, Seunghun J., 107–​108 left dislocation, 82–​83, 95–​98, 132 left periphery, 6, 11–​12, 24, 27–​28, 41–​42, 45, 49, 72–​73, 76–​78, 83, 92, 95–​96, 108, 123, 125–​127, 130, 133 Letsholo, Rose, 6 level-​skipping (in prosodic structure), 126 Lin, Jo-​Wang, 6 linear order, 10, 19, 76, 123–​124

General Index  [ 145 ]

461

Madhavan, Punnapurtath, 134 Madugu, Isaac S. George, 130 Mahajan, Anoop, 99, 113–​114, 132 main clause phenomenon, 132 Major Phrase, 9 Makasso, Emmanuel-​Moselly, 107–​108 Marlo, Michael, 6 Match (condition), 128 Match Theory, 9, 124, 126, 128 Mathieu, Éric, 99, 101, 133 MaxBin, 119 Maxwell, Edith, 42 McPherson, Laura, 5 methodology, 51–​53,  80–​81 Minimality, 119 Minor Phrase, 9, 33 moved clause. See extraposed clause movement leftwards, 10, 82–​83, 94–​98, 108, 130, 132 prosodically altruistic, 10. See also wh-​ movement Muriungi, Peter Kinyua, 6, 42, 112–​113,  134 Myrberg, Sara, 60   Nespor, Marina, 5, 9, 128 non-​finite clause, 113–​114, 117–​118,  132   Obenauer, Hans-​Georg, 99, 101, 133 Odden, David, 6 Optimality Theory, 99, 104 orthography Akan, 127 Krachi, 123, 127 Nupe, 130 Ouhalla, Jamal, 114   Pak, Marjorie, 5, 55, 57, 128 parameterization (general), 4, 117, 121 parameterization (of ACWC), 4, 100, 104–​112,  120 pause, 31–​32, 51–​52, 54–​55, 57–​61, 67–​68, 70, 81–​82, 84–​85, 88, 103, 131 Pesetsky, David, 47 PF, 1–​2, 9–​10, 40, 64, 66, 68, 73–​74, 113, 129–​130 phase, 50, 126–​127

[ 146 ]  General Index

phase head, 126 Phase Impenetrability Condition, 127 Phonological Phrase (ϕ), 1–​2, 9–​12, 14, 16, 18–​27, 29–​31, 33–​37, 40, 64, 72, 79–​80, 97, 120, 123–​126,  128 phono-​syntactic entanglement. See entanglement pitch reset, 31–​32, 51–​52, 54–​62, 67–​68, 70, 81–​86, 88–​90, 93–​94, 131 pitch track, 13–​26, 30–​34, 52–​63, 67–​68, 70, 81–​91, 93–​95, 125, 127, 131 polar question in Asante Twi, 49 in Bono, 48 in Krachi, 48 Poletto, Cecilia, 101, 133 Pollock, Jean-​Yves, 101, 133 Potsdam, Eric, 42, 47 Praat, 53, 81, 124 preposing, 118–​119, 134 preposition, 16–​17, 124–​125 Prieto, Pilar, 103 probe, 10, 120–​121, 126–​127 Prominent, 9, 11, 27, 72–​73, 79–​80, 91–​92, 97, 101–​102, 104–​105, 107–​108, 110, 114, 120, 131 prosodic break, 29, 31, 34, 51–​52, 54–​62, 67–​68, 70, 81–​83, 85–​86, 88–​90, 93–​94, 103, 106–​107, 131. See also pause prosodic condition. See prosodic constraint prosodic constituent discontinuous, 2 domain-​building, 10–​11, 29, 33–​37, 72, 97, 123 formation of, 10, 33–​37, 124 heavy, 119 integrated, 59–​60, 62, 85, 87, 89–​91, 93, 98, 101, 107, 110, 128, 133 light, 119, 124 non-​integrated (independent), 85, 90 recursive, 10, 35, 124–​125 prosodic constraint, 2–​4, 9–​10, 27, 36–​37, 40, 47, 72–​74, 98, 104, 119–​120, 128, 132–​133 prosodic grouping, 1, 10, 35, 37, 64, 105, 120, 121, 124, 127 Prosodic Hierarchy, 2, 9, 29, 35, 128

prosodic integration, 59–​60, 62, 85, 87, 89–​91, 93, 98, 101, 107, 110, 128, 133 prosodic mapping, 3, 5–​6, 8–​9, 27, 34–​35, 40, 65–​66, 68, 74, 80, 128 prosodic phonology, 104 prosodic phrasing, 27, 38, 124. See also prosodic grouping prosodic structure, 1, 9, 25, 27, 29, 30, 88, 91–​92, 111, 131 prosodic weight, 124 prosodic word, 9, 124–​125, 128 prosody, 1–​3, 5, 8–​10, 12, 20, 24, 27, 37–​40, 53, 72, 75, 97, 99, 104, 110, 121, 123, 126–​130 ProsProm, 119 Pullum, Geoffry K., 5   Q (operator), 47–​50, 117–​118, 121, 127, 130 quantifier, 12–​14,  61–​62   recursion (prosodic), 16, 126 register reset. See pitch reset Reglero, Lara, 44, 99, 102–​103 Reinhart, Tanya, 27–​28, 47 relative clause, 3, 77, 87, 90–​91, 126, 131, 133–​134 resumptive pronoun, 79 Rialland, Annie, 133 Richards, Norvin, 1–​2, 4–​6, 8–​12, 24, 26–​27, 29–​30, 33–​39, 47, 51, 64, 72, 74–​75, 79–​80, 97, 120–​121, 123–​124, 126, 133 Rizzi, Luigi, 6, 28, 47, 95, 126 Rooryck, Johan, 101 root clause, 2–​3, 29–​30, 37, 40, 43–​44, 46–​47, 50–​51, 65–​66, 73, 74, 78–​80, 82, 86, 92, 96, 98, 100–​101, 107, 110, 114, 117, 128, 132 root context. See root clause   Saah, Kofi, 45 Sabel, Joachim, 42, 47, 101, 106, 109–​110,  133 Saito, Mamoru, 126 scope asymmetry, 114–​115 embedded, 7–​8, 101, 123

marking, 8, 29, 37, 40, 48, 50, 64 matrix, 7–​8, 29, 48–​50, 101, 103, 111, 113, 115, 118–​119 narrow, 28, 79, 101, 123 out of embedded clause, 50, 111, 113–​114, 118–​119 wh-​, 1, 9, 49, 72, 79, 101, 103, 111, 113–​114, 118–​119,  123 wide, 28, 111, 115, 118–​119 Selkirk, Elisabeth, 5, 9, 35, 123–​124, 126, 128 sentential subject, 30–​31, 53–​54, 57, 85 serial verb construction, 132 Shen, Tong, 9 Shlonsky, Ur, 6, 47 short focus movement, 49, 69, 73, 92, 129, 132 Simpson, Andrew, 99, 114–​117 sluicing,  24–​26 Snider, Keith, 15, 123, 127–​128 Soare, Gabriela, 6, 28, 47 Spell-​Out, 1, 5, 50, 65, 126–​127 split verb, 126, 128 Srikumar, K., 134 Srivastav, Vaneeta, 99, 113–​114, 132 Stepanov, Arthur, 6, 126 Strict Layering (constraint), 97, 119, 126 Strict Layer Hypothesis, 2, 35, 97, 126, 128 subject–​non-​subject asymmetry,  42–​46 subordinate clause. See embedded clause Syrett, Kristen, 101, 104 Szabolcsi, Anna, 28 Szendrői, Kriszta, 60, 82, 101, 107   Tailleur, Sandrine, 101 Thornton, Rosalind, 6 tone boundary (see boundary tone) contour, 12, 14, 16–​17, 23, 25, 53, 123, 127, 132 falling, 12, 14–​16, 21–​23, 25, 31–​32, 53–​55, 57–​59, 67–​68, 125, 132 high, 12, 14–​16, 19–​21, 32, 55, 57–​61, 67–​68, 123, 127–​128, 130 L%, 15–​20, 22, 31, 54, 57, 61, 67 lexical, 12, 14–​16, 25, 31–​32, 53, 55, 57–​60, 62, 67–​68, 70, 123–​124 low, 12, 14–​15, 21–​22, 24, 30–​32, 52–​56, 58, 60, 62, 67–​68, 70, 82, 123, 127, 130

General Index  [ 147 ]

4 81

tone (cont.) mid, 130 retention, 16 rising, 12, 14–​18, 22, 24–​25, 53–​54, 62, 70, 123, 127, 130 sandhi, 12, 15, 124, 128 topicalization, 82–​83, 94–​98, 130–​132 TopP, 95 Torrence, Harold, 6, 39, 41, 47, 53, 123, 125, 127–​128, 134 TP (prosodic status of), 29, 34, 40, 54, 57, 63–​65, 71, 73–​74, 81, 89–​91, 104, 107, 128–​129 Transfer, 6, 127 Truckenbrodt, Hubert, 9, 85, 128 Tsai, Wei-​Tien Dylan, 6, 126   Universal Grammar, 1–​2, 121 upstep, 12, 15, 31–​32, 128. See also F0: boost Urbina, Jon Ortiz de, 116 Uribe-​Etxebarria, Myriam, 99, 102–​103 Utterance (prosodic constituent), 9, 14, 31, 34, 38, 124   Vigário, Marina, 103 Vogel, Irene, 9, 128 vowel harmony, 124–​125   Wagner, Michael, 123 Wahba, Wafaa Abdel-​Faheem Batran, 99, 117 Wasike, Aggrey, 6, 112 Weenink, David, 53, 81 wh-​agreement, 48, 119 wh-​CP raising, 116 wh-​domain, 29, 33–​34, 36–​37, 126 wh-​focus, 48–​50, 66–​67, 69–​70, 73, 76, 127, 129–​130, 132 wh-​ in-​situ adjunct in Asante Twi, 45–​46, 49, 63 in Bàsàá, 108 in Bono, 42–​43, 46, 59, 61 in Kîîtharaka, 113 in Krachi, 7–​8, 20–​29, 34, 41–​42, 46, 123 in Lubukusu, 112 in Nupe, 78, 132 in Shona, 111

[ 148 ]  General Index

in Wasa, 43–​46, 50 in Zulu, 106 argument in Asante Twi, 45–​46, 49, 63, 65, 72 in Bangla, 115–​116 in Bàsàá, 108 in Bono, 42–​43, 46, 48, 58, 61, 65 in Catalan, 105 in Duala, 109–​110 in French, 101–​102 in Hindi-​Urdu, 114 in Iraqi Arabic, 117 in Kîîtharaka, 112–​113 in Krachi, 6–​7, 21, 23, 26, 28, 32–​34, 36, 41, 46, 48, 65 in Lubukusu, 112 in Malayalam, 118–​119 in Nupe, 78, 132 in Shona, 111 in Spanish, 102–​103 in Wasa, 43–​46, 50, 65 in Zulu, 106 clause-​bound,  134 embedded in Asante Twi, 46, 63–​65, 72, 129–​130 in Bangla, 115–​116 in Bàsàá, 108–​109 in Bono, 43, 46, 58–​59, 64–​65, 130 in Catalan, 105 in Duala, 109–​110 in French, 101, 133 in Hindi-​Urdu, 114 in Iraqi Arabic, 117 in Kîîtharaka, 113, 133–​134 in Krachi, 7–​8, 11, 32–​34, 36, 41–​42, 46, 56, 64–​65, 123, 126–​127, 130 in Lubukusu, 112 in Malayalam, 118–​119, 134 in Nupe, 78 in Shona, 111 in Spanish, 103 in Wasa, 44–​46, 61, 64–​65, 129–​130 in Zulu, 106 long-​distance, 8, 104, 109–​110, 117 (see also wh-​ in-​situ: embedded) matrix in Asante Twi, 45–​46 in Bangla, 115 in Bàsàá, 108

in Bono, 42, 46 in Catalan, 105 in Duala, 109 in French, 101–​102 in Iraqi Arabic, 117 in Kîîtharaka, 112 in Krachi, 6–​7, 20–​26, 34, 41, 46 in Lubukusu, 112 in Malayalam, 118, 134 in Nupe, 78 in Shona, 111 in Spanish, 102 in Wasa, 44, 46 in Zulu, 106 optional, 101–​102, 109–​110 ‘why’, 7, 42–​47, 123 wh-​licensing (feature checking), 8–​9, 24, 35, 37, 47, 50, 72, 98, 117 wh-​ movement long-​distance in Asante Twi, 46, 71, 130 in Bono, 43, 71 in Krachi, 71 in Nupe, 77–​79 in Wasa, 44–​45, 71–​72 obligatory, 10, 109, 116

partial in Asante Twi, 49, 67, 69, 127, 130 in Bono, 48, 68–​69, 127 in Krachi, 125, 127 in Nupe, 3, 77 in Wasa, 49–​50, 66–​67, 69, 73, 133 wh-​prosody, 2–​3, 8–​10, 37–​38, 72, 75, 97, 99, 110, 121 wh-​ question embedded, 7–​8, 97, 103, 115, 118 indirect (see wh-​ question: embedded) multiple in Bàsàá, 108 in Catalan, 105 in Iraqi Arabic, 117 in Nupe, 3, 78, 92, 96–​97, 99, 105, 132 presuppositions of, 133 Williamson, Kay, 39   Zec, Draga, 123 Zeller, Jochen, 42, 47, 106, 109–​110,  133 Zentz, Jason, 42, 110–​111 Zwarts, Frans, 28 Zwicky, Arnold M., 5

General Index  [ 149 ]

LANGUAGE INDEX

Akan, 6, 39, 124, 127. See also Asante Twi Asante Twi, 2–​3, 39–​40, 45–​46, 48–​49, 52, 56, 59, 61–​65, 67, 69–​74, 80–​81, 91, 99, 101, 114, 120, 127–​130   Bakweri, 6 Bangla, 99, 113, 115–​116, 119 Bantu, 4, 106–​107, 109–​111, 133 Bàsàá, 60, 107–​109, 120 Basque, 10, 116 Bengali. See Bangla Benue-​Congo, 2, 75, 99–​100, 106 Bono, 2–​3, 39–​40, 42–​44, 46, 48–​49, 52, 56–​61, 63–​65, 68–​71, 73–​74, 92, 99, 104, 110–​112, 120, 127, 129–​130   Catalan, 60, 100, 105–​106 Chichewa, 10, 124 Chinese, 6   Duala, 109–​110, 119   English, 3, 6, 10, 60, 78, 84, 123   French, 44, 99, 100–​104, 109–​110, 119, 123, 133   Georgian, 10 German, 85, 128 Greek (New Testament), 6   Hindi-​Urdu, 99, 113–​114, 116–​118, 120 Hungarian, 60

Ikalanga, 6 Indo-​Aryan, 4, 113, 116 Iraqi Arabic, 99, 116–​119 Italian, 6, 28, 83   Japanese, 6, 10, 55, 57, 128   Kîîtharaka, 112–​113, 133–​134 Korean, 6 Krachi, 2, 3, 5–​6, 8, 11–​42, 46, 48–​49, 52–​61, 63–​65, 71–​74, 99, 104, 110–​112, 120, 123–​130, 134 Kwa, 24, 99–​100, 106   Lubukusu, 6, 111–​112 Luganda, 55, 57, 128   Malayalam, 99, 116, 118–​119, 134   Niger-​Congo, 92, 99 North Guang, 2, 6 Nupe, 2–​4, 74–​99, 105, 114, 120, 130–​133   Persian, 6   Quechua, 116   Romance, 4, 100 Romanian, 6   Shona, 110–​112 Swedish, 60   Tagalog, 10

521

Tano, 2–​3, 39–​40, 46–​47, 51, 61, 63–​70, 73–​74, 78, 81 Turkish, 60 Twi. See Asante Twi   Wasa, 2–​3, 39–​40, 43–​46, 48–​50, 52, 56, 59–​60, 62–​71, 73–​74, 80–​81,

[ 152 ]  Language Index

91–​92, 99, 101, 114, 120, 127, 129–​130,  133   Xhosa, 60   Zulu, 6, 60, 106–​107, 109, 119–​120,  133