University of Calgary PRISM: University of Calgary's Digital Repository
Graduate Studies The Vault: Electronic Theses and Dissertations
2017 From phonology to syntax — and back again: Hierarchical structure in Irish and Blackfoot
Windsor, Joseph W.
Windsor, J. W. (2017). From phonology to syntax — and back again: Hierarchical structure in Irish and Blackfoot (Unpublished doctoral thesis). University of Calgary, Calgary, AB. doi:10.11575/PRISM/26235 http://hdl.handle.net/11023/4161 doctoral thesis
University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. Downloaded from PRISM: https://prism.ucalgary.ca
UNIVERSITY OF CALGARY
From phonology to syntax — and back again: Hierarchical structure in Irish and Blackfoot
by
Joseph W. Windsor
A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
GRADUATE PROGRAM IN LINGUISTICS
CALGARY, ALBERTA
SEPTEMBER, 2017
© Joseph W. Windsor 2017
Signature page UNIVERSITY OF CALGARY FACULTY OF GRADUATE STUDIES The undersigned certify that they have read, and recommend to the Faculty of Graduate Studies for acceptance, a thesis entitled “From phonology to syntax — and back again: Hierarchical structure in Irish and Blackfoot” submitted by Joseph W. Windsor in partial fulfilment of the requirements for the degree of Doctor of Philosophy.
______Supervisor, Dr. Darin Flynn, University of Calgary
______Co-Supervisor, Dr. Elizabeth Ritter, University of Calgary
______Dr. Carlos de Cuba, Queens College, New York
______Dr. Mary Grantham O’Brien, University of Calgary
______Dr. Murray McGillivray, University of Calgary
______Dr. Ryan T. Bennett, University of California, Santa Cruz
______Date
ii
Abstract
The interface between phonology and syntax is a tool that can be used to provide additional evidence for study in one grammatical component or the other. Through understanding how these components interact, one can use syntactic constituent structure to control for prosodic confounds in experimentation. Conversely, one can use phonological evidence to decide between competing syntactic analyses. In ideal cases, phonological and syntactic evidence can be used in tandem, allowing converging evidence to reinforce a hypothesis. In this dissertation, I undertake three case studies to highlight: i. how a knowledge of syntactic constituent structure can increase control over prosodic variables and enable more efficient phonological research; ii. how an understanding of prosodic constituent structure can be used to motivate an underlying syntactic structure at spell-out and enable analysis of morphosyntactic features and operations before spell-out; and, iii. how the use of phonological and syntactic study in tandem can help rule out competing analyses. The first case study utilizes an analysis of syntactic constituent structure to control for different levels of prosodic prominence. The analysis of prominence made possible by syntactic assumptions allows the establishment of a hypothesis into the origins of a stress- shift phenomenon in one dialect of Irish. The second case study correlates observable sound alternations to prosodic boundaries and morpho-syntactic categories in Blackfoot. The analysis of prosodic structure facilitates the formation of a hypothesis about suffixation that is suggested to be the result of syntactic agreement, rather than head-movement operations. The third case study uses the phonological and syntactic analyses from both of the preceding studies and applies those findings to analyze the prosodic and syntactic constituency of demonstratives in both Irish and Blackfoot. A hypothesis towards a common structure for nominal expressions in the two languages is suggested, despite obvious surface differences in realization. Finally, predictions based on that hypothesis are made with questions for future cross-linguistic research. Each of the case studies examined herein contribute to the over-arching goal of the dissertation: To understand how cross-component evidence can provide additional insight and research tools towards a specific problem in one grammatical component or the other.
ii
Preface
Portions of this dissertation have been previously published, or accepted for publication. Evidence for the Demonstrative Phrase in Chapter 3 appeared in Windsor, Joseph W. (2016a) “Prosodic evidence for the syntactic constituency of demonstratives in Irish” in the Proceedings of the 2016 Annual Conference of the Canadian Linguistic Association. The hypothesis into the origins of Munster Irish stress shift in Chapter 3 will appear in Windsor, Joseph W., Stephanie Coward, & Darin Flynn (to appear) “Disentangling stress and pitch accent in Munster Irish” in Proceedings of the West Coast Conference on Formal Linguistics 35. Portions of Chapter 4 on laryngeal specification in Blackfoot previously appeared in Windsor, Joseph W. (2016b) “Contrast, phonological features, and phonetic implementation,” in Calgary Working Papers in Linguistics 29. The Generalized Linear Model analysis of Blackfoot aspiration and coalescence from Chapter 4, and the Phase Theory analysis of Blackfoot demonstratives in Chapter 5, appeared in Windsor, Joseph W. (2017) “Predicting prosodic structure by morphosyntactic category: A case study of Blackfoot” in Glossa: A Journal of General Linguistics 2. Portions of Appendix I on the Universal Spine Hypothesis appeared in Windsor, Joseph W. (2015) “Review of: Martina Wiltschko 2014. The universal structure of categories: Towards a formal typology” on Linguistlist.org (October 30) and will appear in Windsor, Joseph W. (to appear) “Blackfoot demonstratives, referentiality, and association with the syntactic spine” in Proceedings of the Workshop on Structure and Constituency in Languages of the Americas 21. Independent fieldwork data collection was conducted under Ethics Certificate number REB15-0049, issued by the University of Calgary Conjoint Faculties Research Ethics Board for the project “The demonstrative phrase: Prosodic and syntactic evidence from Irish and Blackfoot” on May 11, 2015, and from the University of Ulster Research Governance Filter Committee for the same project, on May 8, 2015. Language consultants for this dissertation were paid $22.50/hour in Canada, and the roughly equivalent €15/hour in Ireland. This research was partially supported by the Social Sciences and Humanities Research Council of Canada and its Michael Smith Foregin Study Supplement. Fin
iii
Acknowledgments
In the acknowledgements to my MA thesis, I remarked about the inside joke within the Linguistics section about the student who specialized in nothing (literally). Now, I have gone on to be the student who studies “this and that” – again, literally. While this became a fun inside joke between members of the Linguistics section that reduced my research to two words, it was precisely funny because of how broad the scope of research around those two words became. My PhD research began with a narrow focus on one element of syntactic structure, demonstratives, but became very broad in the fact that I was trying to balance evidence from multiple grammatical components: the syntax and the phonology, with additional supporting evidence from phonetics, and the interpretation of statistical tests to support that evidence. The final product presented in this dissertation was a careful balance of each of these different ways of looking at linguistic evidence that span multiple sub-disciplines. I know that striking a balance between different literatures so that it was accessible to specialists in each of the various sub-disciplines was difficult, and so I can only imagine how difficult it must have been to supervise. Darin Flynn has been my mentor in linguistics throughout most of my training. His encouragement and enthusiasm has been unwavering from day one – when I first proposed working on Irish lenition. I simply cannot say enough good things about having Darin as a supervisor – from the seven-hour skype sessions to painstakingly go over each of my dissertation revisions to the random Star Trek memes and terrible (read: wonderful) linguistic puns. He has struck an incredible balance as a mentor, a colleague, and a friend who has never kept his enjoyment of my research a secret. Perhaps it was inevitable, but looking at Irish lenition patterns forced me to look at the dreaded area of morphosyntax – an area that I was not at all comfortable in when I started dipping my toes into those waters. My co-supervisor Betsy Ritter came on to my committee in the latter half of my PhD career to add expertise in exactly this area. Betsy was able to identify areas where I was already standing on solid ground, and those where I needed to strengthen my understanding. She worked incredibly hard with both myself and Darin to make sure that the final product showcased my individual strengths as a phonologist, while still tackling the problems I was most interested in at the interface with syntax. Further, it
iv was because of a graduate seminar with her that I first became interested in nominal syntax, and demonstratives in particular. The credit for making syntax not so scary to me, though, goes to Carlos de Cuba, one of my committee members. Carlos taught my first graduate syntax class and made it all start to make sense. He guided my research on parallelism between the nominal and verbal domains, and made sure that my various assumptions were well researched and founded – often over several pints, and eventually an imported liquor and prolonged discussion of musical tastes. I look forward to crossing paths at some future conference and going for another pint(s)! My other committee member, Mary O’Brien, once said to me that one of the most beneficial skills you can gain during a graduate degree is a knowledge of statistics. Although statistics did not come easily to me either, I thought that that was good advice, and I’m glad I took it. Mary was instrumental in making sure I checked my assumptions and let the data lead my analysis. This was an important part of my growth as a graduate student, and a lesson that I am very glad to have learned. My final defense of this dissertation was a very pleasurable experience that involved excellent questions and discussion with all of my committee members, but also my internal- external examiner, Murray McGillivray and my external examiner, Ryan Bennett. I am likewise grateful to both of them in helping me get to this final version – Murray in his early comments and steering from my candidacy exam (on which he was also an examiner), and the expertly informed post-defense comments provided by Ryan to strengthen the final version of this document. Of course, I cannot forget to also thank my neutral chair, Susanne Carroll who not only requested to take on that role, but who kept everything marching along smoothly and on time. Each and every member of my examining committee had a positive influence on this work, as well as helping to shape me into a researcher who is driven by the data rather than a commitment to a hypothesis. However, they were not alone in that respect. There are perhaps more colleagues than I can name here who also deserve credit for helping me get across this finish line. One additional professor that I would like to explicitly thank here is Steve Winters – my phonetics and statistics prof. Steve was incredibly helpful in teaching me many of the tools I needed to tackle the research for my dissertation – from basic
v phonetic skills in Praat to coding and interpretation of results in R. Even when I wasn’t directly a student in one of Steve’s classes, he still made time to answer questions and provide guidance. Turning to other members of the school who are directly responsible for me completing this degree in one form or another, the first people I have to thank are Francey Pisicoli and Biljana Arnautovic – two of our administrators. Francey and Biljana aren’t linguists, but their sheer excitement in seeing one of our students succeed is palpable. Francey and Biljana have always gone out of their way to make sure that we —the students— have every tool available to us to make sure we succeed. Additionally, I would like to explicitly thank the graduate students I had the pleasure of sharing an office with throughout my degree – including those who have already graduated and gone on to other things. While all of my fellow graduate students deserve praise for supporting me, coming out to dry-run presentations, sending me citation information where relevant, and discussing problems over a pint at the LDL, there are a few colleagues in particular that I want to single out as significant contributors to my degree: Lindsay Hracs, Kelly Burkinshaw, and Blake Lewis. Friends, colleagues, sounding boards, problem solvers, and fellow nerds and gamers. What would grad school be like without the friendships you form along the way? This also extends to the friends and colleagues I made during my study-abroad term in Ulster, Raffi Folli, Frances Kane, Jacopo Romoli, Christina Sevdali, Allison Henry, Athanasia Asyllogistou, and Sorcha Henderson. I have one more group of people to thank for standing with me and supporting me down this road, but before I do so, I cannot forget to thank the people who physically assisted in this research: My Blackfoot teachers (AKA language consultants), Piitaikiihtsipiimi, Aistanskiaki, Issapoikoan, kii Ainootaa; my Irish participants, Darina, Muireann, Siobhán, agus Caitríona; and, my research assistants, Maggie Bonsey, Katherine Gerke, and Stéphanie Chicoine. Additionally, I want to thank professors Gearóid Ó Domagáin (Ulster University) and Máire Ní Chiosáin (University College Dublin) for their contributions to the Irish stimuli used in this dissertation. Can I take the academic shortcut to thank my family and say: “see acknowledgements in Windsor (2012)?” I don’t think I can. But I also don’t think that I can thank my family any better than I did in my MA thesis; their support hasn’t changed, but perhaps their
vi excitement has grown. My family has stood behind me every step of the way – although I think they found it much easier to support me in my journey to become a linguist rather than a professional wrestler or some other childhood fantasy. (For the record, I think this suits me better anyway.) Their support has always had many forms – long distance mechanical consultations, care packages (containing Pictou Co. pizza sauce, Brothers’ peperoni, and other necessities of home), conversations where I try to explain what I’m doing with my life or where they update me on what’s happening at home, and of course the occasional bit of financial support. It might not seem like much to some, but like my grandfather once told me while teaching me to swap a transmission in my old K-car: “I can’t help you with school, but I can make sure you get there,” and that makes all the difference. I save the last major thank you for the person who has seen every up and down of this thesis process – Stephanie. She has supported me in every way possible since I started writing. She has endured endless marathons of Star Trek while I sat on my computer hammering out a draft and ignoring the world around us. She has endured the late nights where I just had to figure out one more table, or just needed to analyze one more form. She offered guidance when I was tearing my hair out trying to get a linear model to actually run (and taught me about collinearity among other confounds and pitfalls in the world of statistics). She listened to all the linguistics jargon, the babbling away in some other language (occasionally including Klingon), and picked up a D20 to be part of my non- academic life as well. Simply put, she made sure I survived these last two years of grad school and had a great life while I was doing so. The only way I know to properly thank you is to say I love you (and that yes, I’ll grammar check your dissertation when the time comes next year). There are so many more people that deserve an individual thank you that I just can’t fit into this section – it’s probably too long as it is. But all my friends (Liam, Franco, Brittany, Sarah3, Rob, Dani, Jacq…) who were there along the way, fixing cars, grabbing pints, chucking dice, making me laugh: here’s to you as well!
FIN
vii
Dedication
To my teachers of Irish and Blackfoot. You shared your language and your wisdom with me, and made this entire endeavour possible. Go raibh míle maith agaibh go léir Nitsíniiyi’taki
FIN
viii
Table of Contents
Signature page ...... ii Abstract ...... ii Preface ...... iii Acknowledgments ...... iv Dedication ...... viii Table of Contents ...... ix List of figures and illustrations ...... xi List of Tables ...... xi List of abbreviations and symbols ...... xii Epigraph ...... xiv Prologue ...... xv 1. Introduction ...... 1 1.1. Structure of the thesis ...... 6 2. Interface theory ...... 8 2.1. The prosodic hierarchy ...... 8 2.2. Edge-based alignment...... 11 2.3. Wrap-XP ...... 15 2.4. Match Theory ...... 19 3. Syntactically informed phonology in Irish ...... 27 3.1. The problem: Prominence assigned by the ω or φ? ...... 29 3.2. The syntactic structure of Irish at spell-out ...... 29 3.2.1. The syntax of the Irish clause ...... 30 3.2.2. Syntactic structure of the Irish nominal expression ...... 32 3.3. Matching the syntactic output ...... 35 3.4. The phonetics of Irish prominence...... 43 3.4.1. Methodology ...... 44 3.4.1.1. Participants ...... 44 3.4.1.2. Stimuli and elicitation...... 45 3.4.1.3. Analysis...... 48 3.4.2. Results ...... 50 3.4.3. The phonetics of accent ...... 53 3.4.4. The phonetics of stress ...... 54 3.4.4.1. Vowel length ...... 56 3.5. Phonological interpretation of the results ...... 59 3.5.1. Munster Irish stress shift ...... 61 3.5.2. Edge repulsion and accent alignment ...... 64 3.6. A hypothesis into the origins of Munster Irish stress shift ...... 66 3.7. Summary and discussion ...... 69 4. Phonologically informed syntax in Blackfoot ...... 71 4.1. Prosodic categories and phonetic correlates ...... 71 4.1.1. Study 1: Right edge aspiration ...... 72 4.1.1.1. Methodology ...... 74 4.1.1.2. Results...... 76 4.1.2. Laryngeal features in Blackfoot: voicing versus aspiration ...... 77 4.1.3. Study 2: Intersyllable vowel coalescence and sandhi phenomena ...... 82
ix
4.1.3.1. Vowel coalescence ...... 84 4.1.3.2. Methodology ...... 86 4.1.3.3. Results...... 87 4.2. Uncovering the underlying syntactic structure ...... 90 4.2.1. Phonological representation of Blackfoot affixes ...... 95 4.2.2. A hypothesis on Blackfoot suffixation ...... 97 4.3. Summary and discussion ...... 103 4.3.1. On the Mirror Principle ...... 104 4.4. Interim conclusion ...... 106 5. The demonstrative in Irish and Blackfoot ...... 108 5.1. Demonstratives ...... 109 5.1.1. Demonstratives in Irish and Blackfoot – background information ...... 110 5.2. Irish demonstratives at the interface ...... 118 5.2.1. Previous syntactic analyses ...... 121 5.2.1.1. Dem as the right branching sister of DP ...... 121 5.2.1.2. Recursive DPs ...... 123 5.2.2. Phonological and syntactic evidence for Irish DemP structure ...... 126 5.2.2.1. Adnominal modifiers and lenition ...... 129 5.2.2.2. Recursive syntactic labels and φ-matching ...... 131 5.2.2.3. Summary ...... 133 5.3. Blackfoot demonstratives at the interface ...... 134 5.3.1. Previous syntactic analysis: DemP in Spec,DP ...... 136 5.3.2. Linearization of Blackfoot adjectives and articles ...... 138 5.4. Cross-linguistic generalizability ...... 143 5.4.1. Co-existence with proper nouns ...... 144 5.4.2. Modeling prosodic constituency of a DP-dominating DemP in Blackfoot ... 146 5.4.2.1. Well-formedness and the Blackfoot φ ...... 150 5.4.3. Summary ...... 152 6. Conclusion ...... 156 6.1. Summary ...... 156 6.2. Predictions and limitations ...... 160 6.2.1. Predicted typology ...... 160 6.2.2. Limitations...... 164 6.2.3. Consequences of theoretical assumptions ...... 167 6.3. Conclusion ...... 171 7. References ...... 172 8. Appendix I: The Universal Spine Hypothesis ...... 186 9. Appendix II: Elicitation sentences ...... 193 9.1. Elicitation sentences used in Irish prominence study (Chapter 3) ...... 193 9.2. Pitch accent distribution ...... 210 9.3. Elicitation targets used in Blackfoot study 1 (Chapter 4) ...... 211 9.4. Elicitation targets used in Blackfoot study 2 (Chapter 4) ...... 212
x
List of figures and illustrations
Figure 1: Participant information (image: Google) ...... 45 Figure 2: Pitch accents by condition (bláthanna ‘flowers’ as produced by a CI speaker) 49 Figure 3: Vowel duration by stress ...... 57 Figure 4: Vowel length of /i/ by syllable stress ...... 59 Figure 5: Alignment of stress and accent in CI ...... 60 Figure 6: Misalignment of stress and accent in MI ...... 60 Figure 7: Final aspirated vowel ...... 74 Figure 8: Aspiration by position ...... 76 Figure 9: Average durations in Blackfoot ...... 80 Figure 10: Final vowel aspiration: Figure 11: Final consonant aspiration ...... 81 Figure 12: Aspiration by position ...... 82 Figure 13: Phonological effect by boundary ...... 88 Figure 14: Demonstrative inflection template (Bliss 2013: 13) ...... 113
List of Tables
Table 1: Distribution of phrasal accents by dialect and condition ...... 51 Table 2: Chi-squared analysis of tokens in different conditions ...... 52 Table 3: Summary of phonetic predictors of prominence ...... 61 Table 4: Compounding in Irish (Bennett 2012: 209) ...... 62 Table 5: Stress-Accent alignment patterns ...... 65 Table 6: Summary of phonetic predictors of prominence ...... 66 Table 7: Blackfoot nominal inflection (Bliss 2013: 30) ...... 74 Table 8: Prosodic boundary cues by morphosyntactic category ...... 89 Table 9: Blackfoot demonstrative root interpretations (adapted from Frantz 2017) ...... 112 Table 10: Agreement patterns in Irish nominal expressions ...... 113
xi
List of abbreviations and symbols
Syntactic index A(P) Adjective (Phrase) DIM Diminutive C(P) Complementizer (Phrase) DIR Direct D(P) Determiner (Phrase) DIST Distal demonstrative Dem(P) Demonstrative (Phrase) DUR Durative I(P) Inflectional (Phrase) FEM Feminine K(P) Case (Phrase) FUT Future tense Link(P) Linking (Phrase) GEN Genitive case n(P) Lite/Little Noun (Phrase) IMP Imperative N(P) Noun (Phrase) IMPF Imperfective Num(P) Number (Phrase) INTNS Intensifier Φ(P) Phi (Phrase) INV Inverse Σ(P) Polarity (Phrase) INVIS Invisible P(P) Preposition (Phrase) MASC Masculine Spec Specifier position MOVG Moving T(P) Tense (Phrase) NEG Negation v(P) Lite/Little Verb (Phrase) NOM Nominative case V(P) Verb (Phrase) NON^REF Non-referential X(P) Syntactic projection OBV Obviative reference tracking X0 Minimal syntactic projection O.T. Other Time XMax Maximal syntactic projection PASS Passive
xii
Phonological index C Consonant V Vowel * Constraint violation ! Critical constraint violation Winning candidate Incorrect prediction ι Intonational phrase φ Phonological phrase ω Prosodic word Σ Metrical (stress) foot σ Syllable μ Mora (timing unit) Non-minimal (recursive) φMax Maximal phonological phrase φNon-Min phonological phrase <…> Othographic representation /…/ Phonemic representation […] Surface representation ⟦… ⟧ Phonetic representation Low tone, Left edge, or Light H High tone or Heavy syllable L syllable R Right edge # Word edge
Other symbols and abbreviations CI Connemara Irish PCat Phonological Cateogry dB Decibels (amplitude) PF Phonological Form GLM Generalized Linear Model π Phonological content Hz Hertz (pitch) Σ Semantic content LCC Lexical Category Condition SCat Syntactic Category LDA Linear Discriminate Analysis SLH Strict Layer Hypothesis LF Logical Form UoL Unit of Language MI Munster Irish USH Universal Spine Hypothesis ms Milliseconds * Ungrammatical
xiii
Epigraph
Not yIjeg!kapla not yIjegh! Qapla’! ‘Never surrender! Success!’ —Advice from Marc Okrand, p.c.
When there is a gap between one’s real and one’s declared aims, one turns, as it were, instructively to long words and exhausted idioms, like a cuttlefish squirting out ink. —George Orwell
[You] begin to write… Why shouldn’t you? Why should you fear? Carpenters don’t build monsters, after all; they build houses, stores and banks. They build some of wood a plank at a time and some of brick a brick at a time. You will build a paragraph at a time, constructing these of your vocabulary and your knowledge of grammar and basic style. As long as you stay level-on-the-level and shave even every door, you can build whatever you like—whole mansions, if you have the energy… [and there is a] rationale for building entire mansions of words. —Stephen King: On Writing
xiv
Prologue
My father is a carpenter. When I was quite young, he taught me how to use simple tools like a measuring tape, hammer, and hand saw. One of the first lessons I remember learning was in relation to sawing through a 4x4 post with a hand saw. I would measure out the desired length; set my thumb beside the blade; and, pull back along the line several times to start a groove before starting the slow back and forth motion to get through that impossible three and a half inches of pressure- treated spruce, just trying to keep my saw plumb and get a straight cut. What became immediately obvious to me was that the first inch or so of cutting was far easier (in terms of required muscle) than every subsequent inch of depth through the wood. My solution to this problem was to rotate the post and start cutting that first inch from another side, and then another, and finally the fourth, so that I only ever had to cut through that first inch. My father watched me make this all-too-common mistake and let me discover that, when I got to the fourth face of the post, my saw cuts invariably did not line up. With every rotation of the post, my saw would have a slightly different angle, no matter how much I tried to keep it straight and plumb. The result always looked more like a spiral pattern gradually descending the post rather than an attempt to cut a length off the end. As a carpenter, I learned a lot from the misalignment of those saw cuts: They taught me about the tools, the technique, and what to do with mismatched edges. After learning those lessons and getting pieces that did line up, I was then able to learn about putting them together to build a stable structure. In this respect, the interface between syntax and phonology is much like the carpentry I learned as a youth: When elements do not line up perfectly, you learn about the factors that cause the misalignment on one side or the other; but, when you understand why mismatches occur, you can control for them and start learning about the larger structure. Much like what I learned about carpentry, beginning with cutting that 4x4 post, this dissertation details a methodology for using the phonology-syntax interface as a diagnostic tool, which allows evidence from cross-component analysis to come to bear on a problem in one grammatical component or another – so long as mismatches are accounted for. Fin
xv
Introduction | 1
1. Introduction1 To date, research on the relationship between phonology and syntax has essentially focused on structural mismatches at the interface of these two components of grammar. Bennett et al. (forthcoming) summarize the field by stating, “[w]hen we study how phonological representations align with syntactic and semantic representations, we principally study imperfect parallelism.” This general approach has been successful in establishing the phonological conditions behind non-isomorphism at the interface, but it overlooks the importance of perfect parallelism between syntactic and phonological structures. I contend that isomorphism can also contribute to our understanding of the interface, and further, it provides a diagnostic tool for both phonological and syntactic research. There are several ways in which isomorphic structures between the syntax and phonology can serve to improve our understanding of any given language. When isomorphism is found, syntactic evidence can inform a phonological study, and conversely, phonological evidence can inform a syntactic analysis. In the best cases, phonological and syntactic evidence can reinforce each other to provide a clearer picture of a structure under investigation. A first example of how the interface can be utilized for diagnostic evidence comes from competing analyses of the structure of Celtic demonstratives. One analysis of demonstratives in Celtic claims that they are low within the syntactic structure, as the external argument of nP, as depicted in (1) below (Roberts to appear, cf. Brugè 2002).
1 Data in this dissertation that do not appear with citation information are my own. Every attempt was made to confirm these data with native speakers of the relevant languages. Therefore any mistakes in these data are mine.
Introduction | 2
(1) Structure of the low demonstrative (cf. Brugè 2002; Roberts to appear) a. an leabhar nua seo faoi teangeolaíocht DEF book new DEM under linguistics ‘this new book on linguistics’ b.
This analysis stems from data such as (2), which illustrates the position of the demonstrative seo ‘this’ relative to (what is argued to be) a nominal PP complement, [faoi teangeolaíocht]PP ‘about linguistics’: (2) Linear relationship of demonstratives and PPs in Irish an leabhar (*seo) nua seo faoi teangeolaíocht (*seo) DEF book PROX new PROX under linguistics PROX ‘this new book on linguistics’ (adapted from Brugè 2002: 41, corrections my own) Brugè (2002) and Roberts (to appear) argue for a low demonstrative based on the observation that the only grammatical position for the demonstrative element in constructions such as that in (1) is following the adjective but preceding (what is argued to be) the complement PP, and on the assumption that the noun must obligatorily raise in Celtic (either N-to-Num or NP-to-NumP according to Roberts). A competing analysis comes from McCloskey (2004) and Adger (2013). They argue, for Irish and Scottish Gaelic respectively, that the Celtic demonstrative is realized high within the nominal spine, dominating the DP. Part of the evidence for this alternative analysis comes from the same constructions that Brugè and Roberts analyze. However,
Introduction | 3
Adger argues that elements like demonstratives and the Scottish Gaelic universal quantifier uile ‘all’ must scope over a DP as in the nominal expression provided in (3), the structure for which is provided in (4). (3) Linear relationship of demonstratives, quantifiers, and PPs in Scottish Gaelic na dealbhan mòra ud uile de Mhàiri the pictures big that all of Màiri ‘all those big pictures of Màiri’ (4) Structure of the high demonstrative in Scottish Gaelic (Adger 2013: 110)
PP uile de Mhàiri def ud
na dealbhan mòra The structures in (1) and (4) above illustrate that very similar syntactic evidence can be used to argue for very different representations. Although the scope evidence provided in McCloskey (2004) and Adger (2013) is convincing, researchers in favour of the low demonstrative, as in (1), could potentially argue that the scope effects are achieved through covert, post-syntactic movement at LF. In sum, the analyses behind (1) and (4) are both viable based on the available syntactic evidence. Fortunately, these two analyses can be teased apart and tested phonologically. In Windsor (2011, 2012, 2014a, 2016a), I detail one test for phonological constituency in Irish using lenition patterns, which are based on an underlying morphosyntactic structure. Lenition is a consonant mutation that causes a predictable alternation of the initial consonant of a prosodic word (ω) based on hierarchical relations – see Windsor (2012) for a full discussion. In what follows, and elsewhere in this thesis, I assume —following my earlier work on lenition cited above— that when any two ωs are parsed together into a single higher prosodic unit (i.e., not separated by a prosodic phrase (φ) boundary), lenition can be observed on the initial consonant of the second word.2
2 The distribution of pitch accents in Chapter 3 provide additional support that this generalization works in the contexts explored in this thesis. The pitch accent data converge on the same prosodic structures used to model lenition in Irish nominal expressions (as in 5 and 6, below). Because the lenition diagnostic is supported by the pitch accent data, none of the conclusions presented in this dissertation rely solely on the treatment of lenition as a prosodically defined alternation (see Green 2007, 2008 or Gorrie 2011 for a morpho-
Introduction | 4
One example of the lenition diagnostic can be observed when a feminine noun (lenited itself if it co-occurs with a definite article) causes lenition on all modifying adjectives within the same phonological phrase. The relevant alternations are highlighted in the underlying and surface representations below: (5) Adjective lenition in Irish a. tá [an chloch ghorm mhór]DP ag rolladh /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ ə ɾol̘a/ [t̪ɑː ən xl̪ ɑx ɣɔɾm woːɾ ə ɾol̘a] be DEF stone blue big PRT roll.VN ‘The big blue stone is rolling.’
b. tá [an chloch ghorm]DP mór /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ/ [t̪ɑː ən xl̪ ɑx ɣɔɾm moːɾ] be DEF stone blue big ‘The blue stone is big.’ The data in (5a, b) show that when an adjective is within a DP (Merged as an adjunct to nP as represented in (1) above), the initial consonant of the adjective is lenited. This is the result of the phonological structure: When several ω are parsed into a single higher prosodic unit (e.g., a φ), the right edge of a ω causes lenition of the initial consonant of the subsequent ω, provided that those boundaries are not separated by an interceding φ boundary (Windsor 2011, 2012, 2014a, 2016a, see also Carnie 1991). According to Elfner (2012, 2015), DP is realized as a φ, and non-binary adjectives are realized as ωs in Irish. Thus, we can explain why the adjective ‘big’ is realized with lenition in (5a) where it is inside the DP, but without lenition (no change to the initial consonant) in (5b) when it is outside the DP. A very simple prediction follows from these observations: If the demonstrative is in the same syntactic position as adjectives in Irish (adjoined to nP à la Roberts to appear), lenition will be observed on a demonstrative Merged into the sentence in (5a); if, however, the demonstrative dominates the DP (à la Adger 2013), no lenition will be observable on the demonstrative:
phonological treatment of lenition). This assumption about lenition captures the majority of grammatical environments where the mutation occurs; however, the role of grammatical gender, and how grammatical gender affects the phonological component are not yet fully understood. Additional discussion of this assumption is provided in §6.2.3.
Introduction | 5
(6) Lenition and demonstratives in Irish adjective strings tá an chloch ghorm mhór sin/*shin ag rolladh /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ sʲɪn ə ɾol̘a/ [t̪ɑː ən xl̪ ɑx ɣɔɾm woːɾ sʲɪn/*hɪn ə ɾol̘a] be DEF stone blue big DIST PRT roll.VN ‘That big blue stone is rolling.’ The data in (6) show that it is ungrammatical to lenite the demonstrative in this type of construction (*/sʲɪn/ → [hɪn]). This fact indicates that the demonstrative is high within the nominal structure, as argued by McCloskey and Adger, and provides independent, phonological evidence against Brugè and Roberts’ view that they are low in Celtic. The basic interface structure depicting this analysis is provided in (7) (cf. McCloskey 2004; Windsor 2016a): (7) The interface structures of Celtic arguments modified by demonstratives an chloch ghorm sin DEF stone blue DIST ‘that blue stone’ (Irish) a. Morphosyntactic structure b. Prosodic structure φ ← DemP
φ ← DP ω ← Dem0 sin ω ← D0 ω ← A0 ghorm an chloch
Because the adjective ghorm ‘blue’ is parsed into a single φ containing the article and noun, the adjective is lenited as seen above in (6). However, because the demonstrative is separated from the adjective by the right boundary of the φDP, the adjective cannot cause lenition of the initial consonant of the demonstrative.3 The preceding discussion is a first example of how the syntax-phonology interface can be utilized as a diagnostic tool. In this case, I briefly outlined how phonological evidence can be used to decide between opposing syntactic analyses. The chapters below examine interface structures of nominal elements, including demonstratives, in Irish as well as in
3 Note that in the construction [PP P ( ó )ω Dem (shin)ω ]φ ‘from then’ where the preposition ó and the demonstrative sin are parsed into a single φ, the initial consonant of the demonstrative is lenited from /sj/ to [h] showing that the lack of lenition in (6) and (7) is not due to the category of the demonstrative, but rather its position in the prosodic structure.
Introduction | 6
Blackfoot, an Algonquian language. These comparisons demonstrate bi-directionality of the interface as a diagnostic tool. Just as phonology can help motivate a syntactic analysis (as above), evidence from syntax can drive a phonological investigation. Moreover, syntactic and phonological evidence can be used in concert, in unrelated languages (Irish and Blackfoot), to facilitate cross-linguistic comparison. The individual case studies undertaken in this dissertation make contributions to theoretical phonology and syntax, as well as to our understanding of the grammar of both Irish and Blackfoot. The larger goal of this work, however, is to highlight how the syntax- phonology interface can be used as a research tool; how attention to the interface can enhance component-internal study; how it can provide additional evidence towards answering a given empirical question in either component; and, how it can be used to test predictions of competing analyses, allowing the researcher to reject one or more analyses when it is difficult to do so based on component-internal evidence alone.
1.1. Structure of the thesis In this dissertation, I provide a rationale for treating the syntax-phonology interface as a diagnostic tool to inform research in either component of grammar. Specifically, after describing the theoretical framework used to model the interface, I provide a series of case studies which demonstrate the effectiveness of this diagnostic tool. The dissertation consists of six chapters, briefly described below. Chapter 2 introduces Match Theory (Selkirk 2009 et seq), the unifying theoretical framework behind this dissertation. I describe the empirical and theoretical background leading up to Match Theory, the mechanics of the theory, and the principal questions it has been heretofore concerned with. In Chapter 3, I develop an in-depth analysis of Irish prosodic structure with a focus on how manipulating the syntactic structure allows the researcher to isolate prominence at different levels of the prosodic hierarchy (namely the prosodic word and phrase). I show that by paying special attention to the underlying syntax, prosodic structure can be controlled for, and specific phonetic correlates of word-level and phrase-level prominence can be uncovered. The phonetic understanding of prominence is then used to motivate a diachronic change that results in a stress shift phenomenon in one dialect of Irish.
Introduction | 7
Ultimately, then, a diachronic understanding of phonology is made possible by harnessing syntactic structure at spell-out. In Chapter 4, I explore the phonetic correlates of Blackfoot prosodic structure with a focus on how the placement of prosodic boundaries can inform a syntactic analysis of argument structure. I show that isolating phonetic correlates of different prosodic boundaries allows the researcher to compute the underlying syntactic structure of different nominal elements (syntactic heads, phrases, and bound morphemes). This phonological evidence supports an analysis of Blackfoot morphosyntactic structure that utilizes agreement rather than cyclic head movement to realize agglutinating structure. In Chapter 5, I apply the evidence from Chapters 3 and 4 to a specific syntactic question: What is the syntactic constituency of demonstratives in Irish and Blackfoot? No consensus in the syntactic literature has yet been reached on this question with reference to either Celtic or Algonquian languages, let alone whether the demonstrative category represents a cross-linguistic universal or not. In that chapter, I use the prosodic evidence established in Chapter 3 to posit a position high in the nominal spine for demonstratives in Irish. Building from the evidence for a demonstrative projection that dominates DP in Irish, and the evidence of Blackfoot nominal constituency in Chapter 4, I explore the possibility of a similar construction in Blackfoot and discuss the advantages of adopting that analysis. Chapter 6 concludes by recapping the evidence that shows how the interface between the two grammatical components may be used as a diagnostic tool. I outline a predicted typology of demonstrative constructions based on the discussion in Chapter 5 and outline problems for that analysis, which provide directions for future work in this research program.
Match Theory | 8
2. Interface theory The syntax-phonology interface has been the subject of intense inquiry. This research program became prominent in 1986 with Beckman & Pierrehumbert (1986), Ladd (1986), Nespor & Vogel (1986 [2007]), and Selkirk (1986), quickly followed by Halle & Vergnaud (1987), and Gussenhoven (1988). See Shattuck-Hufnagel & Turk (1996) and Selkirk (2011) for an overview. These authors, and subsequent literature on the topic, suggest that the ideal relationship between the syntactic and phonological components is one of isomorphism, where the structure in one component mirrors that of the other component. Isomorphism between the two components was thus concluded to be the default, unless some mitigating factor (e.g., a phonological constraint) causes a mismatch. This conclusion spawned several other topics of inquiry in the syntax-phonology interface: How can phonology reference a syntactic output? Is the output matched by direct or indirect reference? What are the factors that cause mismatches? Are multiple representations of an underlying syntax possible in prosodic constituent structure? In this chapter, I review the commonly posited prosodic hierarchy. I then discuss which levels of that hierarchy are thought to be influenced by an underlying syntactic structure. Finally, I present three theories of the interface: edge-based alignment (Selkirk 1986), wrapping (Truckenbrodt 1999), and matching (Selkirk 2009, 2011). These theories are presented using key data from Tohono O’odham, which figures prominently in the above cited literature. This discussion ultimately leads to the choice of theory for the present study, Match Theory (Selkirk 2009, 2011; Elfner 2012, 2015).
2.1. The prosodic hierarchy When work on the interface between syntax and phonology became prevalent, there were three primary models of the prosodic hierarchy containing different levels of representation. Shattuck-Hufnagel & Turk (1996) provide the following diagram to highlight the similarities and differences between the representations assumed by Nespor & Vogel (1986 [2007]), Selkirk (1986), and Beckman & Pierrehumbert (1986).4
4 Beckman & Pierrehumbert state that other representations, such as those assumed by Halle & Vergnaud (1987) or Gussenhoven (1988), are not significantly different from the ones schematized below.
Match Theory | 9
(8) Comparison of prosodic hierarchies (Shattuck-Hufnagel & Turk 1996: 206) Nespor & Vogel Selkirk Beckman & Pierrehumbert
Utterance (Utterance)
Inton. Phrase Inton. Phrase Full Inton. Phrase
Phon. Phrase Major Phrase Intermed. Inton. Phrase
--- Minor Phrase Accentual Phrase
Clitic Group ---
Prosodic Word Prosodic Word
Foot Foot
Syllable Syllable
Mora
As can be seen in (8), the hierarchy proposed by Beckman & Pierrehumbert (1986) is relatively minimal, and the hierarchies proposed by Nespor & Vogel (1986 [2007]) and Selkirk (1986) are quite similar to one another. As we will see throughout the rest of this chapter, a maximum of three levels of the prosodic hierarchy are concerned with the output from syntax, in the range between the prosodic word and the intonational phrase. I suggest, here, that there is reason to largely adopt a paired down version of the Nespor & Vogel (1986 [2007]) hierarchy, eliminating the levels Clitic Group and Utterance. Selkirk (2011: 437) provides the following list of prosodic category types in what she calls the “commonly posited prosodic hierarchy.” I adapt that hierarchy to include the symbols to indicate the stress (or metrical) foot and the syllable: (9) The commonly posited prosodic hierarchy5 Intonational Phrase (ι) Phonological Phrase (φ) Prosodic Word (ω) Foot (Σ) Syllable (σ)
5 Note that in this dissertation, Σ is used for the metrical foot, syntactic polarity, and semantic content due to common conventions in the respective literatures. These notations are independent of one another, and will be elucidated in text wherever a potential confusion may occur.
Match Theory | 10
The primary consideration that goes into positing the representations in (9) is ensuring that each level is motivated by empirical evidence that suggests a new level of representation rather than a recursive category. The early research on the interface assumed Selkirk’s (1981) Strict Layer Hypothesis: (10) The strict layer hypothesis (Selkirk 2011: 437) A constituent of category-level n in the prosodic hierarchy immediately dominates only a (sequence) of constituents at category-level n-1 in the hierarchy According to the strict layer hypothesis, whenever there is evidence for boundary effects at multiple levels of prosodic representation, each of those levels of representation must be unique. Therefore, under that hypothesis, a ω could not be nested inside another ω, nor could a φ be nested inside another φ. This restriction resulted in a theoretical distinction between the prosodic word and the clitic group, as well as between minor and major phonological phrases. However, with the advent of Optimality Theory (Prince & Smolensky 1993 [2004]), the strict layer hypothesis was reconceived as a set of violable constraints such as NONRECURSIVITY (NONREC):
(11) NONRECURSIVITY (NONREC) (Selkirk 1995) Any two prosodic categories that are not disjoint in extension are identical in extension: there can be no Ci that dominates Cj where Ci = Cj Reformulating this aspect of the strict layer hypothesis as the violable constraint that prohibits recursive structure in (11) created a predicted typology of two types of languages: Some languages allow recursive structure to be created, faithful to the syntactic structure at spell-out, whereas other languages prohibit recursive structure, creating a mismatch (see Truckenbrodt 1999 on Kimatuumbi and Chichewa). Relevant to the present discussion is the notion that some languages permit recursive structure. Recursion allows us to collapse the prosodic word and clitic group, as well as the major and minor phonological phrases, as these distinctions are not empirically motivated, i.e., there is no evidence to suggest that they are separate prosodic representations rather than recursive structures. The reduction of prosodic categories discussed above leaves us with (minimally) the prosodic word, the phonological phrase, and the intonational phrase. The decision not to use the three categories provided by Beckman & Pierrehumbert (1986) is somewhat arbitrary, since it is the type of representation that matters, not the label. However, I suggest that the motivation for which labels to adopt is based on what each of these categories do:
Match Theory | 11
If the accentual phrase distributes pitch accents, a category below this level is needed in order to assign word stress, and only a single category above the accent-assigning category is needed to assign intonational properties. Thus, in the hierarchy given in (9), the ω provides word stress, the φ provides phrasal accents, and the ι provides the full intonational structure. Although sub-ω prosodic representations are still posited, only the three categories discussed above will be shown to be influenced by syntactic structure at the interface; this focus on the ω, φ, and ι will be motivated as we examine three theories of the interface based on a particular language whose data has been re-evaluated with each revision to interface theory.6
2.2. Edge-based alignment Some of the earliest work on the interface between syntax and phonology utilized the alignment of the edge of a particular prosodic category to the edge of a particular syntactic category (Selkirk 1986, 1995). Thus, a constraint could align the right edge of a phonological phrase (φ) with the right edge of a syntactic phrase (XP), allowing a degree of isomorphism between the syntactic and prosodic structures. This effect was achieved through Generalized Alignment (Selkirk 1986, 1995; McCarthy & Prince 1993): (12) Generalized Alignment (Truckenbrodt 1999: 223) Where Cat1, Cat2 are prosodic, morphological, or syntactic categories and Edge1, Edge2 ∈ {Right, Left}: ALIGN(Cat1, Edge1; Cat2, Edge2) ↔ For each Cat1 there is a Cat2 such that Edge1 of Cat1 and Edge2 of Cat2 coincide So, for example, the constraint in (13) can be used to achieve an output where the right edge of a φ is aligned with the right edge of an XP: (13) Align-XP,R: Align(XP, R; P, R) (Truckenbrodt 1999: 223) For each XP there is a P such that the right edge of XP coincides with the right edge of P [where “P” = φ] One benefit of assuming an edge-based alignment theory of the interface is that it allows for multiple prosodic elements to be parsed into a single φ. Selkirk (1995) argues that only XPs containing lexical heads (e.g., nouns, verbs, and adjectives) would be visible to the constraints responsible for creating prosodic structure, but not XPs headed by
6 Ito & Mester (2010) argue that, specifically because sub-ω prosodic constituents are not motivated by syntactic structure, they cannot be recursive as the supra-ω categories are (cf. Bennett 2012, 2013).
Match Theory | 12 functional heads (e.g., determiners). This was later captured in the Lexical Category Condition (LCC) of Truckenbrodt (1999), who considers what influence lexical elements, functional elements, and empty elements in the syntactic structure (as well as their projections) might have on prosodic structure: (14) Lexical Category Condition (Truckenbrodt 1999: 226) Constraints relating syntactic and prosodic categories apply to lexical syntactic elements and their projections, but not to functional elements and their projections, or to empty syntactic elements and their projections7 The LCC provides a rationale for the prosodic structure of a DP being a single φ since only the extended projection of the N0 (the lexical head) would be visible to the phonology. So, a D0 could force the φ to extend in order to incorporate all pronounceable material, but the DP itself could not serve as a reference point for phonological structure creation under this theory. This is illustrated in (15) with a sentence from Tohono O’odham, an Uto-Aztecan language spoken in Arizona and the Sonoran Desert of Mexico:
7 See Windsor 2012 for an opposing discussion of empty elements in phonological representation.
Match Theory | 13
(15) The interface in Tohono O’odham (adapted from Truckenbrodt 1999: 230) a. Morphosyntactic structure Wákial ‘at g wísilo cépos cowboy AUX DET calf branded ‘The cowboy branded the calf.’
b. Distribution of phrasal tones ( H LL)φ (L HHH H L )φ Wákial ‘at g wísilo cépos cowboy AUX DET calf branded ‘The cowboy branded the calf.’ Hale & Selkirk (1987) use the distribution of tones in (15b) to analyze the prosodic phrasing in Tohono O’odham and suggest that each φ is characterized by a specific tonal pattern with zero or more L-tones followed by one or more H-tones, followed again by one or more L-tones. The H-tones begin with the first vowel of the primary stressed ω in the φ, and end with the last vowel of the word associated with ω-level stress. The last H-tone is followed by an L-tone, and all remaining vowels in the φ also receive L-tones. Hale & Selkirk (1987) argue that the relevant edge for the analysis of tones in Tohono O’odham is the right edge, therefore the right edge of every lexical XP is realized as the right edge of a φ. The complication that Hale & Selkirk (1987) observe, which sets Tohono O’odham apart from other languages studied, is that not all right lexical XP boundaries trigger the alignment of the right edge of a φ. For example, with the strict application of the alignment constraint in (13), we would predict that a φ-boundary would separate the object DP g wísilo from the following verb cépos, but that is not borne out. To account for this
Match Theory | 14 peculiarity, Hale & Selkirk suggest that φs are not aligned to the right edge of every lexical XP, but only those that are not lexically governed: (16) Phrasing sensitive to lexical government (Truckenbrodt 1999: 234) Align the {left/right} edge of each syntactic category SCat with a prosodic category PCat, where a language may choose a. SCat = XP, or b. SCat = XP so long as it is not lexically governed, (other options for SCat omitted here) and PCat = p-phrase [(φ)] (other options for PCat omitted here) For Hale & Selkirk (1987), being lexically governed means that an XP is contained in the maximal projection of a lexical head. So, in (15a), the object DP is lexically governed by (contained within) the maximal projection of the lexically-headed VP, and thus, no right φ- boundary is aligned with the right edge of the object DP. The notion of lexical government playing a role in determining phonological phrasing captured the empirical findings, but ran into some trouble with the Indirect Reference Hypothesis (Inkelas 1989): (17) Indirect Reference Hypothesis (Truckenbrodt 1999: 221) Phonological rules refer to only prosodic constituent structure This hypothesis maintains that phonological rules/constraints cannot make direct reference to syntax and provides a more restrictive theory of the interface – not letting new phonological constraints be generated for each different morphosyntactic category or feature.8 Therefore, through indirect reference, an H-tone (in the case of Tohono O’odham) cannot be inserted on the first stressed syllable of a VP, for example, but only on the first stressed syllable of a φ – a phonological constituent rather than a syntactic constituent. If this hypothesis is correct, then how would the phonological component know whether an XP was lexically governed in the syntax? Further, how would the phonological component know if a head were lexical or functional? In an attempt to resolve some of these issues,
Truckenbrodt (1999) introduces the constraint WRAP-XP, which is the subject of the next section.
8 The necessity of this assumption is further elucidated in §5.2.2.2, where I show that an interface constraint that matches maximal syntactic projections as φs cannot tell the difference between a recursive XP (in the case of adjunction) and a DP in Spec,DP. This is specifically due to the fact that the phonological component is not sensitive to syntactic indices, as predicted by a theory of indirect reference.
Match Theory | 15
2.3. WRAP-XP Rather than appealing to constraints that could be influenced by syntactic relationships (i.e., lexical government), Truckenbrodt (1999) argues in favour of a violable WRAP-XP constraint that interacts with structural markedness constraints (e.g., NONREC as discussed above, or *STRUC à la Prince & Smolensky 1993 [2004]) and alignment constraints:
(18) WRAP-XP (Truckenbrodt 1999: 228) Each XP is contained in a phonological phrase
Much like earlier edge-based alignment theories, WRAP-XP is still subject to the LCC (see 14 above), but does not rely on lexical government. The LCC is argued to not violate the indirect reference hypothesis because functional categories are merely invisible to the phonological component during structure building processes. Another advantage of WRAP- XP is that both edges of a φ can be predicted, rather than just the left or right edge (as in Selkirk’s edge alignment theory).
Still looking at Tohono O’odham, Truckenbrodt highlights how WRAP-XP can handle the data without appealing to lexical government. He provides the following tableau for the structure in example (15a) above: (19) Tableau for Tohono O’odham structure in (15) (Truckenbrodt 1999: 235) [[wakial]NP1 ‘at g [wisilo]NP2 cepos]VP]IP WRAP-XP ALIGN-XP,R a. ( )φ NP1! NP2 b. ( )φ ( )φ NP2
Truckenbrodt suggests that the LCC prevents WRAP-XP from applying to the IP layer, but
WRAP-XP does not prevent ALIGN-XP,R from providing a φ-boundary after NP1 (which would result in the unsuccessful candidate in 19a). According to the constraints utilized in that analysis, the winning candidate in (19b) is one that assigns the right φ-boundary after the VP and another one after NP1. Left φ-boundaries are not specified in the constraints used in these analyses (i.e., ALIGN-XP,L is low-ranked, and therefore largely inactive), and so the extraneous pronounceable material from the functional heads are grouped in with the 0 VPφ. Grouping the material from the I with the NP1φ would cause a violation of the alignment constraint, and so it is not considered in Truckenbrodt’s discussion. In other words, WRAP-XP does not prevent additional material from being parsed into a φ, it only penalizes candidates that do not contain all material within the lexical XP within a single φ.
Match Theory | 16
One question that may arise from the discussion of the tableau in (19) is why does NP2 not receive its own φ within the VPφ? Truckenbrodt contends that WRAP-XP would be violated if Align-XP,R forced additional boundaries to appear inside the VPφ because the material from the VP would then be split between two φs. He shows how this works for the transitive interrogative in (20) using the tableau provided in (21). (20) A Tohono O’odham interrogative structure (adapted from Truckenbrodt 1999: 232) a. Morphosyntactic structure Na-t g wákial g wísilo cépos? Q-AUX DET cowboy DET calf brand ‘Did the cowboy brand the calf?’
b. Distribution of phrasal tones ( L HHH HHH H L )φ Na-t g wákial g wísilo cépos? Q-AUX DET cowboy DET calf brand ‘Did the cowboy brand the calf?’ (21) Tableau for Tohono O’odham structure in (20) (Truckenbrodt 1999: 234) n-at [g [wakial]NP1 g [wisilo]NP2 cepos]VP WRAP-XP ALIGN-XP,R
a. ( )φ NP1 NP2 b. ( )φ ( )φ VP! NP2 c. ( )φ ( )φ VP! NP1 d. ( )φ ( )φ ( )φ VP!
As can be seen in the structure provided in (20a), the lexical XP headed by the verb contains both DPs. According to Truckenbrodt, giving the DPs their own φ structures violates
WRAP-XP by causing part of the material within the VP to be realized in two separate φs, as in (21d). Therefore, the winning candidate is the one in (21a) which has all of the VP
Match Theory | 17
material wrapped in a single φ.9 It should be noted here that although the NONREC constraint is not used in the tableaux in (19) and (21), it is a crucial part of the WRAP-XP theory. NONREC must be ranked higher than WRAP-XP in order to prevent the representation (( NP1 )φ (NP2)φ VP )φ from being the prosodic organization of the structure in (20). Truckenbrodt continues to elaborate the WRAP-XP theory using further case studies from Kimatuumbi and Chichewa to show how the relative ranking of the NONREC constraint can either facilitate or eliminate the possibility of recursive structures.
Although WRAP-XP was shown to successfully handle the Tohono O’odham data above without relying on ad hoc stipulations or violating the LCC, a potential problem for the hypothesis is that it may over-predict possible typological variations. Selkirk (2011:
462) makes a comparison of the typological variants predicted by WRAP-XP compared to her 2011 theory, which I explore in the following section. The comparison that Selkirk offers is based on SOV structures in German embedded clauses and the distribution of pitch accents: (22) Pitch accents in German SOV embedded clauses (Selkirk 2011: 462)10 a. …weil María die neuen Gesétze studiert because M. art new law-s study-pres:3s ‘because Maria is studying the new laws’
b. …weil [ [ María]DP VP[ [die neuen Gesétze ]DP studiert ]VP ]
c. Match(XP,φ)/ Match(φ,XP):…weil φ( María )φ φ( φ( die neuen Gesétze)φ studiert )φ
d. Wrap-XP and Align-R XP: … weil φ( María )φ φ( φ( die neuen Gesétze)φ studiert )φ
e. Wrap-XP and Align-L XP: *… weil φ( María )φ φ( die neuen Gesétze studiert )φ
9 It should be noted that the Tohono O’odham syntactic structure assumed in these prosodic analyses were constructed prior to Larson’s (1988) VP-shell hypothesis (i.e., prior to the assumption that subjects are Merged in Spec,vP). I do not try to update the analyses presented in this literature review for current syntactic theory. However, because vP is a functional projection, interface theories that assume the LCC would contend that the vP is the extended projection of the lexical V0 and so there would be no meaningful change to the prosodic analysis presented. (See Truckenbrodt 1999: §1.4 for further discussion relative to the formalization of the LCC.) 10 Matching constraints alluded to in this example will be elucidated in the following section. At present, it is only important that the typological predictions are highlighted.
Match Theory | 18
According to Selkirk (2011), the combination of WRAP-XP and ALIGN-L XP produce an incorrect (i.e., unattested) result: (22e). This incorrect result is not in and of itself detrimental to the theory, as language-specific constraint rankings are expected. Rather, the problem is the fact that the pattern in (22e) is not attested in any language (Kahnemuyipour
2004, 2009), although it is predicted by WRAP-XP.
A further criticism of the WRAP-XP hypothesis is its use of what Selkirk (2011) calls
Strict Layer Hypothesis (SLH) constraints (Selkirk 1981, 1986, 1995) like NONREC to ban recursive prosodic structure: (23) The strict layer hypothesis (Selkirk 2011: 437) A constituent of category-level n in the prosodic hierarchy immediately dominates only a (sequence of) constituents at category-level n-1 in the hierarchy: ι
φ φ φ
ω ω ω ω ω ω
Selkirk (2011: 464) provides the following typology that an interaction of WRAP-XP and generalized alignment predicts with a highly ranked NONREC(φ) constraint. She argues that these non-recursive φ organizations should be evidenced in natural languages if
NONREC is a necessary component of phonological grammar:
(24) φ-domain structures generable from syntactic input [NP NP V]VP a. ( NP NP verb)φ by NONREC(φ) » WRAP-XP » ALIGN-R/L XP b. (NP)φ (NP)φ verb by NONREC(φ) » ALIGN-R XP » WRAP-XP c. (NP)φ (NP verb)φ by NONREC(φ) » ALIGN-L XP » WRAP-XP
Conversely, if a theory of syntax-phonology correspondence contends that recursion is basic (i.e., expected based on the output of hierarchical syntactic phrases) and does not make use of SLH constraints like NONREC, only one structure can be computed from the above constituents (assuming a structure such as that in (20) from Tohono O’odham):
(25) φ-domain structures generable from syntactic input [NP NP V]VP without NONREC ( (NP)φ (NP)φ verb )φ The analysis provided above from Truckenbrodt (1999) for Tohono O’odham crucially relies on a non-recursive structure that wraps the entire clause in a single φ. The assumption that the phonological component can actively prevent recursion does not survive into Match Theory, the theory detailed in the next section. Match Theory will be shown to improve on
Match Theory | 19
Wrap-XP in two regards: It adequately explains the same data while constraining the predicted typological variants to only those patterns found in natural languages, and it reduces the theoretical machinery, thus providing a more advantageous theory.
2.4. Match Theory Match Theory (Selkirk 2009, 2011) is a theory of indirect reference that computes prosodic structure by matching it to various types of syntactic structures visible to the phonological component after spell-out. The three primary constraints used to derive prosodic structure from syntactic structure are:
(26) MATCHWORD (Selkirk 2011: 439) A word in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it ω, in phonological representation
(27) MATCHPHRASE (Selkirk 2011: 439) A phrase in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it φ, in phonological representation
(28) MATCHCLAUSE (Selkirk 2011: 439) A clause in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it ι, in phonological representation In other words, according to the constraints above, a minimal projection (X0) is matched in the prosodic representation as a ω; a maximal projection (XP) is matched in the prosodic representation as a φ; and, a clause (CP)11 is matched in the prosodic representation as an ι. Using these constraints, we can predict prosodic structure from the spell-out of syntactic structure. The prosodic structure will be isomorphic to the syntactic structure providing that there are no mitigating factors in the form of higher ranking phonological constraints, which I discuss below. The following Irish sentence in (29) illustrates near-isomorphic structures between the two grammatical components.12 As can be seen in this example, a transitive VSO sentence with branching arguments is matched in prosodic constituent structure as a series of recursive φs:
11 Bennett et al (2016) suggest that the syntactic representation that is matched as an ι is the clausal projection that has the potential to express illocutionary force, possibly a functional head within Rizzi’s (1997) C layer. 12 Note that the adjective phrases in (29) are realized as ωs rather than φs. This mismatch between the syntax and prosody is explained in Chapter 3 where I discuss Irish syntax-prosody matching in greater detail.
Match Theory | 20
(29) Matching prosodic structure to syntactic structure in Irish (Elfner 2015: 1179) 13 díolfaidh leabharlannaí dathúil bláthanna áille sell.FUT librarian handsome flower.PL beautiful.PL ‘The handsome librarian will sell beautiful flowers.’ a. Morphosyntactic structure b. Prosodic structure ΣPa φa ← ΣP
ω φb ← TP Vi TPb díolfaidh DP → φc φd,e← vP/VP/DP DPc …d ω ω ω ω N A leabharlannaí dathúil bláthanna áille
ti DPe
N A In the structures provided in (29), we can see the recursive φ structure that Selkirk (2011) argues is basic (i.e., expected from the hierarchical layering of syntactic structure).14 Each of the syntactic XPs (ΣP, TP, DP, etc.) is matched as a φ in the prosodic structure. Like
WRAP-XP, Match Theory ignores syntactic structure that contains no pronounceable material; thus, φd,e can hypothetically be matched from either the vP, VP, or object DP in this example, whereas the verb has undergone successive head-movement outside of the vP/VP, leaving the object DP as the only pronounceable material under those XPs. Further, we can see that the N0 and the A0 in each of the DPs has been matched as a ω (see fn. 12 above). Because this example does not contain an embedded clause, the structure in (29b) can be parsed into a single ι, matched from a CP that dominates the ΣP (not shown in Elfner’s example cited here). If this example contained multiple clauses, each clause would be matched as an ι in accordance with MATCHCLAUSE in (28). In an ideal case, the constraints in (26-28) would cause total isomorphism between the syntactic and phonological representations. However, because Match Theory is a theory of indirect reference, other higher-ranking constraints may cause non-isomorphism between
13 The structure provided in this example is fully fleshed out and motivated in Chapter 3. 14 ΣP (Laka’s (1990) Polarity Phrase) is a projection below CP that McCloskey (1996, 2009) argues is where the verb ultimately raises to in Irish. Like previous interface research in Irish, I assume this syntactic representation in the Irish studies to be presented in this dissertation. See §3.2.1 for further discussion of the assumed syntactic structure of the Irish clause.
Match Theory | 21 the two representations. A prime example of non-isomorphism between the syntactic and phonological components is the case of post-syntactic pronoun postposing in Irish. There is a well-known phenomenon of object pronoun postposing that takes place in Modern Irish (Chung & McCloskey 1987; Duffield 1995; Adger 1997, 2007; Doyle 1998; McCloskey 1999; Mulkern 2003, 2011; Elfner 2011; Bennett et al. 2016) whereby a prosodically light (i.e., monomoraic) object pronoun moves rightward from its spell-out position (the position of the element at the output of syntax to the phonological component) between the subject and any adjuncts, to follow one of the adjunct phrases. This postposing is illustrated in (30) where a light object pronoun é ‘it’ can be realized in multiple positions: (30) Surface positions of a light object pronoun in Irish15 léigh Liam (é) ar an traein (é) aréir (é) PST-read Liam (it) on DEF train (it) last.night (it) ‘Liam read it on the train last night.’ In the sentence given in (30), the light object pronoun é ‘it’ can appear in its spell-out position to the right of the subject Liam (as a verbal complement), or following either of the two adjunct phrases, [ar an traein]PP or [aréir]PP. Bennett et al. (2016: 172) diagram postposing using the example in (31). (31) Light object pronoun postponement in Irish (Bennett et al. 2016: 172)
[ V DP Pron XP YP ZP ] In attempting to explore what syntactic operations might be responsible for postposing, Bennett et al. provide a particularly problematic example (given in 32) where the light object pronoun mé ‘me’ seems to undergo rightward movement to follow the verb in a small clause complement of the perception verb chíonn ‘saw’. However, it is not obvious where the attachment site of this pronoun after the postposing might be. Bennett et al. conclude that there is no principled syntactic explanation for the rightward movement of a light object. If this were to be analyzed syntactically, using the phrase in (32), we would have to assert that the object pronoun had lowered into an unspecified position in the middle of a small clause – an untenable conclusion:
15 See footnotes in Elfner (2011: 18) regarding generational differences in the acceptability of realizing the pronoun in this example in all positions.
Match Theory | 22
(32) Pronoun postponement example (Bennett et al. 2016: 181) má chíonn tú ___ ag troid mé le ridirí if see.PRES you PROG fight me with knights ‘if you see me fighting with knights…’
By contrast, if this phenomenon is examined prosodically, we can make sense of this seemingly illicit movement operation. The data (33) (adapted from Elfner (2011: 18) to show focal prominence and prosodic phrasing) shows how the sentence (33a) can be used to respond to the various questions in (33b-d). The prosodically prominent (i.e., focused) phrase is bolded in each of the answers given in (33b-d). (33) Light object pronoun postposing and focal prominence a. léigh Liam (é) ar an traein (é) aréir (é) read.PST Liam it on DEF train it last.night it ‘Liam read it on the train last night.’ b. Q: Cé léigh é ar an traein aréir? ‘Who read it on the train last night?’ A: VP[léigh Liam é]φ PP[ar an traein]φ AP[aréir]φ c. Q: Cá léigh Liam é aréir? ‘Where did Liam read it last night?’ A: VP[léigh Liam]φ PP[ar an traein é]φ AP[aréir]φ d. Q: Cén uair léigh Liam é ar an traein? ‘When did Liam read it on the train?’ A: VP[léigh Liam]φ PP[ar an traein]φ AP[aréir é]φ Elfner (2011) and Bennett et al. (2016) —assuming Match Theory— analyse this and similar patterns using constraints called STRONGSTART and EQUALSISTERS which penalize
Match Theory | 23 prosodically weak objects when they appear in initial position of a φ, or when they are hierarchically sister to a higher prosodic category respectively. Bennett et al. analyse the light object pronoun as being a syllable (σ) which is sister to a φ and argue that postposing that syllable to cliticize to the right edge of a (prosodically prominent) φ provides the optimal repair to the STRONGSTART violation. This phonological repair strategy causes the prosodic structure to be unfaithful to the structure it is matching by altering the linear order of the string.
Match Theory has one key difference from WRAP-XP: WRAP-XP utilizes SLH constraints such as NONREC to penalize recursive structure. Selkirk (2011) argues that the inclusion of such constraints predicts more variation than is found in the attested typology evidenced in natural languages. Selkirk’s argument makes the previously discussed case study of Tohono O’odham a critical proving ground. Recall from the previous section that Truckenbrodt’s (1999) analysis of the Tohono O’odham interrogative in example (20-21)
(repeated in 34-35 for convenience) used NONREC to wrap the entire clause in a single φ: (34) Structure of the Tohono O’odham interrogative – repeated from (20) (Truckenbrodt 1999: 232) a. Morphosyntactic structure Na-t g wákial g wísilo cépos? Q-AUX DET cowboy DET calf brand ‘Did the cowboy brand the calf?’
b. Distribution of phrasal tones ( L HHH HHH H L )φ Na-t g wákial g wísilo cépos? Q-AUX DET cowboy DET calf brand ‘Did the cowboy brand the calf?’
Match Theory | 24
(35) Tableau for Tohono O’odham structure in (20) (Truckenbrodt 1999: 234) n-at [g [wakial]NP1 g [wisilo]NP2 cepos]VP WRAP-XP ALIGN-XP,R
a. ( )φ NP1 NP2 b. ( )φ ( )φ VP! NP2 c. ( )φ ( )φ VP! NP1 d. ( )φ ( )φ ( )φ VP!
Selkirk’s Match Theory, on the other hand, predicts that, because recursion is basic, (35a) is not a possible phonological representation of (34). Match Theory predicts that this sentence should be parsed as: (( NP1 )φ (NP2)φ verb )φ. If the data from Tohono O’odham can be handled by the more restrictive Match Theory equally well, without relying on ad hoc stipulations, then it can be shown to be the more advantageous theory. Selkirk (2011: 465) proposes that the correct representation of the interrogative in (34) is as given in (36) below, with (36a) containing the original intonational transcription from Hale & Selkirk (1987) and (34b) containing the phonological representation they propose: (36) Tohono O’odham phrasing analysis (Hale & Selkirk 1987 qtd. in Selkirk 2011: 467) a. L HHH HHH H L Na-t [ [g wákial ]XP [ g wísilo ]XP cépos ]VP INTER-AUX.3.PERF ART cowboy ART calf brand.PERF ‘Did the cowboy brand the calf?’ b. L H*------H*------H* L% Na-t φ( φ(g wákial)φ φ(g wísilo)φ cepos )φ While Truckenbrodt (1999) characterized the (L)HL pattern of tones in Tohono O’odham as being the pattern of tonal assignment in a φ, Selkirk suggests, instead, that this is the distribution of tones in a maximal φ (φMax). See Itô & Mester (2013) for a motivation of this natural class: (37) (Some) natural classes of recursion-based prosodic categories (adapted from Elfner 2015: 1182)16 a. Maximal/minimal projections of φ b. Non-maximal projections of φ φ ← Maximal projection φ
φ φ Non-maximal x…x φ ← Minimal projection x…x φ projections
16 I have altered this diagram to show non-maximal projections rather than the non-minimal projections shown by Elfner 2015.
Match Theory | 25
Using this analysis, Selkirk (2011) then argues that a boundary tone (L% in 36b) can be restricted to the right edge of a φMax – thus explaining why no other boundary tones are visible in the transcription. If this analysis is correct, Selkirk would predict that if the subject were moved outside of the VP, it would be phrased separately from all VP material, and would therefore show an L% boundary tone at the right edge of the subject DP. This prediction is indeed borne out in the declarative version of the sentence in (36).17 (38) Phrasing analysis of the Tohono O’odham declarative (Hale & Selkirk 1987 qtd. in Selkirk 2011: 467) a. HLL L HHH H L [ g wákial ]XP ‘at VP[ [ g wísilo ]XP cépos ]VP ART cowboy AUX ART calf brand.PERF ‘The cowboy branded the calf.’ b. H*LL% L H*------H*L% φ(g wákial )φ ‘at φ( φ( g wisilo )φ cépos )φ Because Match Theory can equally handle the Tohono O’odham data without reliance on the NONREC constraint, which forces WRAP-XP to predict unattested typological variants, Selkirk (2011: 467) concludes that the favoured hypothesis should be: …the theory of prosodic domain formation and prosodic domain-sensitivity that includes (i) the highly restrictive universal Match theory of syntactic- prosodic constituency correspondence, (ii) a theory of domain-sensitivity in phonology which allows for domain-sensitive phenomena to be sensitive to any of the prosodic category types defined in the theory and in the general manner sketched [above], and (iii) a phonological theory of markedness constraints on prosodic structure [also as above].
I concur with Selkirk’s assessment, and therefore will also assume Match Theory for the case studies to be undertaken in this dissertation. Match Theory has been designed specifically to tackle questions such as the following: Given that isomorphism between the syntax and phonology (and recursive structure) is basic, how do we account for mismatches and what are the phonological constraints responsible for them? Several authors have explored questions like these in a variety of
17 Hale & Selkirk (1987: 154) state that: “word order is extremely flexible [in this language], and surface structures give little direct evidence as to the word order which should be regarded as basic – sentences exhibit orderings SOV, SVO, VSO, and even VOS and OSV.” In this chapter, I report the structures that previous prosodic studies in the literature have assumed. It should not be inferred from this statement that subject- raising happens in declarative sentences exclusively and that interrogatives always display predicate-internal subjects.
Match Theory | 26 languages from English (Elfner 2014), to Irish (Elfner 2012 et seq.; Bennett et al. 2016, forthcoming), Xitsonga (Lee & Selkirk 2013), South Baffin Inuktitut (Arnhold et al. to appear), Blackfoot (Windsor 2017; Windsor & Lewis to appear), Japanese (Selkirk 2009), and Lekeitio Basque (Selkirk 2011), to name a few. Thanks to many previous studies, we now understand the techniques of analysis, potential factors that influence mismatches between the grammatical components, and how to account for nonisomorphism. In the case studies detailed in this dissertation, I use Match Theory to account for misalignment factors, and to examine isomorphic examples of syntax-phonology matching. I also use empirical evidence from Irish and Blackfoot to detail techniques for using the interface as an analytical tool in motivating hierarchical structures in the syntactic and phonological components. Moreover, I use syntactic evidence to motivate phonological analysis, and phonological evidence to motivate syntactic analysis. The data explored in the following chapters highlight a way to use cross-component data as an additional source of evidence in the study of syntax or phonology. I start by providing a syntactically-motivated study of stress and accent in Irish in the following chapter. FIN
Syntactically informed phonology i n I r i s h | 27
3. Syntactically informed phonology in Irish This chapter details a study of prominence and stress shift in different dialects of Irish. I show that careful control of the underlying syntactic structure and an understanding of the prosodic interface constraints allows for the study of φ-level accent and ω-level stress in isolation. In turn, this study informs a hypothesis into the origins of Munster Irish stress- shift, whish is argued to be due to Tonal Crowding and Edge Repulsion Effects (cf. Gordon 2014). Phonological studies that investigate words in isolation suffer a potential confound. Words elicited in isolation often bear metrical stress, as well as prominence from other levels of prosody such as the prosodic phrase (φ) or intonational phrase (ι) (Hyman 1977; Gordon 2014). A striking example of this confound comes from a description of the Berber language Shilha by Applegate (1958: 9, qtd. in Hyman 1979: 44): [P]rimary or heavy stress occurs on the last vowel of the stem... [However, it] should be noted that the stress patterns referred to here apply only to utterances consisting of a single word. If the utterance contains more than a single word, the stress is reduced slightly on all vowels except those in the final word. It can be said, therefore, that primary stress occurs only at the end of an utterance.
In this instance, Applegate conflates prominence from higher prosodic units (e.g., φ or ι) with word-level stress. In an effort to avoid such interference from higher prosodic structures (e.g., boundary tones or final lengthening), many researchers utilize carrier sentences, as in the following example from a phonological study of Somali (Flynn et al. 2011): (39) Sample carrier sentence elicitation method a. Hadda i dhagayso _____ markaan dhaho now 1.SG listen when.1.SG say ‘Now listen (to) me when I say ____.’ b. Hadda i dhagayso bad markaan dhaho ‘Now listen to me when I say sea.’ c. Hadda i dhagayso dhib markaan dhaho ‘Now listen to me when I say difficulty.’ d. Hadda i dhagayso abti markaan dhaho ‘Now listen to me when I say uncle.’ e. Hadda i dhagayso darbi markaan dhaho ‘Now listen to me when I say wall.’ f. Hadda i dhagayso abaar markaan dhaho ‘Now listen to me when I say drought.’
In this study, we used the carrier sentence hadda i dhagayso ____ markaan dhaho to elicit lexemes containing target sounds in various environments. The examples in (39) were used to elicit the phoneme /b/ in various contexts: #CV (39b), VC# (39c), VCCV (39d), VCCV
Syntactically informed phonology i n I r i s h | 28
(39e), and VCV (39f). The study used a carrier sentence in an effort to circumvent independent phenomena such as final lengthening. However, we may have introduced a new confound: When we asked the consultant to repeatedly provide the same sentential frame differing only by a single lexeme, we ran the risk of introducing contrastive focus onto the target lexeme. Moreover, while we could be certain that we were not eliciting an utterance-final token, we ran the further risk that the position was initial or final in some other prosodic unit (e.g., φ or ι). Because prosodic constituency was not controlled for in that study, it was not possible to determine if any duration effects observed, for example, were the result of segmental alternations (the hypothesis being investigated in that study), or prosodic constituency. Separating prosodic levels of representation and the effects that each level introduces is important in order to avoid mis-attribution of an effect. Turning now to Irish, the language of focus in this chapter, Elfner (2015: 1173) remarks that phrasal pitch accents and word stress are almost always realized on the same syllable in Connemara Irish. She therefore suggests that stress is realized with pitch – as it is largely indistinguishable from accent. Once again, the point is, when a single lexical word is elicited, φ-level and ι-level structure are both computed, causing multiple levels of prominence to be realized on a single syllable. Thus, a single word such as cailín ‘girl’, in Connemara Irish, would have the following prosodic structure: (40) Prosodic structure of a single-word utterance
The representation in (40), consistent with Elfner’s (2015) analysis that pitch accents and word stress are realized on the same syllable (cf. Windsor et al. to appear), shows that the initial [ka] syllable of the word receives prominence from ι and φ, as well as prominence from the prosodic word (ω) and the stress foot (Σ).
Syntactically informed phonology i n I r i s h | 29
In the following sections, I detail a methodology that allows me to distinguish prominence assigned by the ω and the φ by appealing to syntactic constituent structure to control for prosodic constituency. The results of this syntactically-informed prosodic study allows for a hypothesis into the origins of a stress-shift phenomenon in Munster Irish, presented in §3.6.
3.1. The problem: Prominence assigned by the ω or φ? A major confound in studying prosodic prominence is identifying which prosodic unit is being measured.18 As stated in the previous section, the hierarchical nature of prosodic prominence causes Elfner (2015: 1173) to state that Irish “word-level stress, which falls on the initial syllable with few exceptions (e.g., Ó Siadhail 1989; Christian Brothers 2004), is thus distinguishable by pitch when it is associated with a phrase accent.” However, while it is true that the stressed syllable can be identified by the presence of a pitch accent (when in the correct prosodic environment), it is not necessarily the case that ω-level prominence is phonetically realized by a measure of pitch (Hz). In fact, Blankenhorn (1981: 244) suggests in passing that substantial insight could be gained into the nature of stress through an investigation of how pitch accent interferes with stress. In the study presented in this chapter, we use an understanding of Irish syntax, and of the constraints that match phonological representation to the syntactic output (cf. Elfner 2012, 2015; Bennett et al 2016) to accomplish three goals: To isolate stress and accent; to measure the phonetic correlates of these levels of prominence; and, to contrast the instantiations of prominence by representation and dialect.19
3.2. The syntactic structure of Irish at spell-out In order to understand the prosodic structure that assigns phrasal pitch accents in Irish, and to control for that structure so as to contrast phrasal pitch accent with word-level stress, it is first necessary to understand the syntactic constituent structure that the prosody matches.
18 In this chapter, I use the word “prominence” as a blanket term to refer to stressing or accenting from any level of the prosodic hierarchy. I use “stress” to refer to prominence at the level of the ω and “accent” (or “pitch accent”) to refer to prominence at the level of the φ. 19 The study presented in this chapter is part of a collaborative project presented in Windsor et al (to appear) with co-authors, Stephanie Coward and Darin Flynn.
Syntactically informed phonology i n I r i s h | 30
3.2.1. The syntax of the Irish clause Irish is a VSO language in which the only typical syntactic elements to precede the verb are what are frequently referred to as pre-verbal particles, which mark a clause as an interrogative or conditional, or show negation: (41) Sample Irish sentences a. tuigfidh muid Gaeilge understand-FUT 1.PL Irish ‘We will understand Irish.’ b. an dtuigfidh muid Gaeilge Q understand-FUT 1.PL Irish ‘Will we understand Irish?’ c. dá duigfeadh muid Gaeilge… if understand-COND 1.PL Irish ‘if we understand Irish…’ d. ní thuigfidh muid Gaeilge NEG understand-FUT 1.PL Irish ‘We will not understand Irish.’ There are no words for ‘yes’ or ‘no’ in Irish. Instead, the polarity of the reply to a yes/no question is indicated by the verb appearing in the positive or negative form: (42) Verbal polarity and replies to yes/no questions a. an dtuigfidh muid Gaeilge Q understand-FUT 1.PL Irish ‘Will we understand Irish?’ b. tuigfidh understand-FUT ‘Yes.’ (Lit: ‘(We) will understand.’) c. ní thuigfidh NEG understand-FUT ‘No.’ (Lit: ‘(We) will not understand.’) Data such as those in (41) and (42) show that the verb must minimally raise to T0 in order to precede the subject noun and gain tense inflection. McCloskey (2011 and references therein), however, suggest that the verb raise even higher, to the Σ0, to be inflected for the correct polarity, but not all the way to C0 where pre-verbal question particles (41b, and 42a) and complementizers (41c) are located. Importantly, negation, despite appearing as a pre-verbal particle, does not receive the same analysis. Acquaviva (2014) (citing Guilfoyle 1990, 2000; Duffield 1995; Doherty 1996; McCloskey 1996, et
Syntactically informed phonology i n I r i s h | 31 seq.; Doyle 2002) argues that although negation can surface on a higher syntactic head, such as C0 in (43), it is interpreted on the lower head, containing the verb. He provides the following example as his (11).
(43) Morphological realization versus semantic interpretation of NEG mur dtéighinn agus iad cailleadh, mhuirbhfeadh siad mé if.NEG go.COND.1.SG and 3.PL lose.VN kill.COND 3.PL 1.SG ‘If I were not to go today and they were to lose, they would kill me’ Acquaviva contends that although the negation is realized on the complementizer mur in the above example, it is actually interpreted on the verb dtéighinn because it scopes only over the first conjunct, not the second as: ‘if [it were not the case that I go] and [it were the case that they lose]…’. He provides the following structure as his example (12) to show that, although negation is realized on mur in the C0 position, the interpretation of negation applies only to the first conjunct, what he designates “clause1” in his structure, not the second conjunct, designated as “clause2”: (44) Scope of negation in coordination CP
C X
[NEG-clause1]
and [clause2] In sum, even though negation is pronounced on a higher head, a pre-verbal particle, it is semantically realized on the verb, as shown by Acquaviva’s (2014) data presented above in (43) and (44). Evidence such as this suggests that the Irish verb undergoes successive head-movement up to Σ0. These previous syntactic analyses that argue for V-raising to Σ0 are assumed in the previous prosodic matching analysis motivated by Elfner (2012 et seq.) described in §3.3. Elfner (2012 et seq.) details the matching constraints that build prosodic structure in Irish from an underlying syntactic structure at spell-out. She assumes, following Chung & McCloskey (1987) and McCloskey (1991, 1996, 2011), as discussed above, that the verb in an Irish clause raises to its spell-out position in Σ0, that the subject raises to Spec,TP, and that objects remain in situ within the VP. Example (45) provides the syntactic structure of transitive clauses in Irish assumed by Elfner, and for the studies reported in this chapter:
Syntactically informed phonology i n I r i s h | 32
(45) The syntax of the Irish clause with branching subject and object (Elfner 2015: 1179) ΣP
Vi TP
DPj ti vP N A tj ti VP
ti DP
N A Throughout this chapter, I will assume the same underlying syntactic structure of the clause as that used in Elfner (2012, 2015) (see also Bennett et al. 2016, forthcoming). However, the primary focus of the present study is on nominal expressions, so I give a more detailed analysis of those structures now.
3.2.2. Syntactic structure of the Irish nominal expression The standard linear order of syntactic elements in Irish nominal expressions has nouns preceding the adjectives that modify them, as illustrated in (46). It is widely assumed that if nouns precede adjectives, they must do so by raising above the adjective, which is left- adjoined to nP in Celtic (see Roberts to appear for discussion). However, there is disagreement as to the hierarchical position and category of the landing site for the noun. Earlier studies such as Guilfoyle (1988) argue for N-to-D raising to account for what appeared to be Construct State Nominals (similar to those in Hebrew), which ban the realization of a definite article specifying definiteness on the head noun in genitive constructions (cf. Ritter 1991; Carnie 1995; see Kane 2015 for arguments against analysing Irish as having construct state nominals). However, it has been argued that numerals, which intervene between determiners and nouns provide evidence that the N0 cannot raise all the way to D0 in Irish (see Duffield 1996 and Kane 2015 for relevant discussion):
Syntactically informed phonology i n I r i s h | 33
(46) Linear order in Irish nominal expressions a. báid móra cf. bád mór boat.PL big.PL boat big ‘big boats’ ‘(a) big boat’ b. na báid mhóra DEF.PL boat.PL big.PL ‘the big boats’ c. na cuig bhád/bháid mhóra DEF.PL five boat/boat.PL big.PL ‘the five big boats’ d. na cuig bhád/bháid deag mhóra DEF.PL five boat/boat.PL ten big.PL ‘the fifteen big boats’ As can be seen above, the typical order of nominal elements in Irish is: Article > Num > Noun > Adjective. The apparent counter example to this order appears in example (46d) where the noun appears sandwiched between the numeral five and the numeral ten, which together form the numeral 15. This ‘sandwich’ construction occurs with cardinal numbers between 11 and 19 and linearizes the noun between the positions occupied by the cardinal numeral expressing the mathematical ones and tens. This type of evidence has been used to motivate the fact that nouns must syntactically move to the Num position, but no higher. For example, Roberts (to appear), investigating nominal expressions in Welsh, argues that the structure of the Celtic DP necessarily involves noun-raising. He remains agnostic, however, as to whether this evidence suggests N0-to-Num0 movement or NP-to- Spec,NumP movement (cf. Carnie 1995). I suggest, based largely on prosodic evidence introduced in Windsor (2011, 2012), that N0 does undergo head-movement to Φ0, after first moving to n0, but then continues the successive head-movement to D0.20 In my earlier work, I argued that the N0 shows a parallel pattern of head-movement to the V0 in the clausal domain (Bennett et al. 2016), creating a complex incorporated head under D0. This movement correctly predicts a stress-shift phenomenon in definite compounds in Irish and lenition patterns concerning the definite
20 For consistency with the discussion of number in Chapters 4 and 5, I use the label ΦP for the functional projection that introduces number to a syntactic derivation, rather than NumP. See Déchaine & Wiltschko (2002) for relevant discussion on the Φ-projection, or Shlonsky (2004) on separate projections for cardinal numbers. Coming to a conclusion about the position of cardinal numbers in Irish is beyond the scope of the present analysis.
Syntactically informed phonology i n I r i s h | 34 articles, detailed in Windsor (2014b). The structure of the N0 and V0 at spell-out is therefore assumed to be as represented in (47). (47) Structure of complex verbal and nominal heads in Irish21 a. Structure of the complex verb b. Structure of the complex noun ΣP DP
Σ TP D ΦP T Φ
v DPsubj vP n nP V N VP AP NP
Complement Complement The structure I assume for the present analysis is as given in (48). (48) Structure of the Irish nominal a. na bláthanna corcra DEF.PL flower.PL purple.PL ‘the purple flowers’ b. DP
Di ΦP
na bláthanna Φ nP ti AP nP
corcra n NP ti ti
As we will see in the following section, nothing in the present analysis hinges on the precise location of the incorporated nominal in Irish. What is crucial for the present analysis is the location of the demonstrative relative to the DP. I briefly discussed prosodic evidence for the position of the Celtic demonstrative in Chapter 1. There, I argued that lenition evidence showed that the demonstrative must be dominate the DP in Celtic as originally proposed by McCloskey (2004) (see also Adger 2013 and Windsor 2011, 2012, 2014a, 2016a). I will provide further motivation for that conclusion in Chapter 5, which presents a
21 The structure of the complex verb is taken from Bennett et al. (2016: 191). The structure of the complex noun is my own.
Syntactically informed phonology i n I r i s h | 35 detailed phonological and syntactic analysis of demonstratives in Irish and Blackfoot. However, at present, I will assume the structure of the Irish demonstrative provided in (49), and this structure will inform the study of prominence and stress-shift detailed in the current chapter: (49) The Irish DemP (cf. McCloskey 2004; Windsor 2016a) na bláthanna corcra seo DEF.PL flower.PL purple.PL PROX ‘these purple flowers’
Within the DP, the only thing that is important to the present discussion is the position of adjectives. Following Roberts (to appear), I assume that adjectives are adjuncts to the nP as represented in (48b) above. This is not a significant departure from Elfner’s (2015) assumption that the adjunction site of adjectives is NP, and will not affect the prosodic analysis developed here. In the next section, I detail how each of the above structures is realized in phonological representation using Match Theory (Selkirk 2009, 2011; Elfner 2012, 2015; Bennett et al. 2016, forthcoming).
3.3. Matching the syntactic output A Match theoretic analysis of Irish is provided in Elfner (2012, 2015). In those works, Elfner provides the necessary constraints for mapping the prosody from an underlying syntactic representation at spell-out. In addition to the basic matching constraints described in Chapter 2 (repeated here as 50 and 51 for convenience), Elfner argues that prosodic phrases (φ) must dominate a minimum of two prosodic words (ω) to be well formed in
Irish, according to the constraint BINARY-MINIMUM (BIN-MIN), which interacts with
MATCHPHRASE:22
(50) MATCHWORD (Selkirk 2011: 439) A word in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it ω, in phonological representation
22 Because this study investigates prominence at the levels of the word and phrase, MatchClause (as discussed in §2.4) is not relevant to the present study.
Syntactically informed phonology i n I r i s h | 36
(51) MATCHPHRASE (Selkirk 2011: 439) A phrase in syntactic constituent structure must be matched by a corresponding prosodic constituent, call it φ, in phonological representation
(52) BINARY-MINIMUM (φ) (BIN-MIN) (Elfner 2015: 1180) A φ constituent in the prosodic representation must dominate a minimum of two ωs Elfner argues that non-branching adjective phrases (APs) are matched only as ωs in Irish, not φs, due to BIN-MIN – a crucial fact for the analysis to be detailed in this chapter: (53) Unattested mapping from simultaneously maximal/minimal A to φ AP/AMax *φ | → | A ω áille áille ‘beautiful’ (Elfner 2015: 1179) (54) Attested mapping from simultaneously maximal/minimal A to ω AP/AMax | A → ω áille áille ‘beautiful’ (Elfner 2015: 1179) Utilizing the constraints in (50-52) and the syntactic structure of a transitive clause with branching subjects and objects to be matched (as in 45 above, repeated as 55a), we arrive at the predicted prosodic representation in (55b). (55) Matching prosodic structure to syntactic structure (Elfner 2015: 1179) díolfaidh leabharlannaí dathúil bláthanna áille sell.FUT librarian handsome flower.PL beautiful.PL ‘The handsome librarian will sell beautiful flowers.’ a. Morphosyntactic structure b. Prosodic structure ΣPa φa ← ΣP
ω φb ← TP Vi TPb díolfaidh DP → φc φd,e← vP/VP/DP DPc …d ω ω ω ω N A leabharlannai dathúil bláthanna áille
ti DPe
N A The example in (55) shows that the syntactic XPs (ΣP, TP, and the DPs) are matched as φs in the prosodic representation. The syntactic projections, vP and VP, are omitted from this
Syntactically informed phonology i n I r i s h | 37 structure as they have no pronounceable material in them, and it is not obvious that either of those structures creates a φ that is separate from the one created by the object DP. Concerning the nominal expressions in the structure represented in (55a), Elfner (2012, 2015) has shown that non-branching APs do not meet minimal prosodic requirements to form a φ, and instead match as a ω which combines with the ω matched from the N0 and which is parsed into the φ matched from the DP. This portion of the matching structures is illustrated in (56) below using the object DP from the example in (55) above: (56) Matching branching arguments a. Morphosyntactic structure b. Prosodic structure φ ← DP
ω ← N0/D0 ω ← A0 bláthanna áille
The structures in (56) highlight the importance of Elfner’s BIN-MIN constraint in preventing the AP from being realized as a φ dominated by the φ matched from the DP. Note that recursive prosodic structure is permitted in Irish, as can be seen in (55b). However, recursive φ structure is not attested within this DP, as per the evidence provided by Elfner to motivate the structures and the interface matching constraints. Elfner found that HL accents are realized on the primary stressed syllable in the rightmost ω in every φ. Example (57) provides the same interface structures as (55) above, but indicates the distribution of HL accents by rightward descending arrows:
Syntactically informed phonology i n I r i s h | 38
(57) The distribution of phrasal HL pitch accents (Elfner 2015: 1179) díolfaidh leabharlannaí dathúil bláthanna áille sell.FUT librarian handsome flower.PL beautiful.PL ‘The handsome librarian will sell beautiful flowers.’ a. Morphosyntactic structure b. Prosodic structure ΣPa φa ← ΣP
ω φb ← TP Vi TPb díolfaidh DP → φc φd,e← vP/VP/DP DPc …d ω ω ω ω N A leabharlannaí dathúil bláthanna áille
ti DPe
N A Additionally, Elfner discovered that LH accents are associated with the primary stressed syllable in the leftmost ω in every non-minimal φ (φNon-Min) (i.e., all but the lowest φ in a recursive φ). Example (58) provides the same structure as (57) above, but additionally indicates the distribution of LH accents by leftward descending dotted-line arrows, to highlight the contrast: (58) The distribution of phrasal LH and HL pitch accents (Elfner 2015: 1179) díolfaidh leabharlannaí dathúil bláthanna áille sell.FUT librarian handsome flower.PL beautiful.PL ‘The handsome librarian will sell beautiful flowers.’ a. Morphosyntactic structure b. Prosodic structure ΣPa φa ← ΣP
ω φb ← TP Vi TPb díolfaidh DP → φc φd,e← vP/VP/DP DPc …d ω ω ω ω N A leabharlannaí dathúil bláthanna áille
ti DPe
N A Observe that the object noun bláthanna ‘flowers’ in the representation in (58b) remains unaccented, as it is neither the rightmost daughter of a φ, nor the leftmost daughter of a
Syntactically informed phonology i n I r i s h | 39
φnon-min. Consider what the representation would be if BIN-MIN did not prevent the AP from being matched as a φ:
(59) Representation of a branching object without BIN-MIN a. Morphosyntactic structure b. Unattested Prosodic structure φ ← DP
ω ← N0/D0 φ ← AP bláthanna ω ← A0 áille
As can be seen in the representation in (59b), if the BIN-MIN constraint did not prevent non- branching φ-structure in Irish, we would predict that an LH accent (denoted by the leftward descending dotted-line arrow) would be realized on the noun bláthanna ‘flowers’. As I suggested in §3.2.2, whether the N0 raised to a functional projection below D0 (perhaps Num0 as suggested by Roberts to appear, cf. Elfner 2015) or all the way to D0 is not important for the present analysis. Now that the accent distribution has been provided, I will show how both analyses of noun raising equally handle the pitch accent data to be examined in this chapter. Under Elfner’s analysis, the definite article in Irish is not matched by a ω in phonological representation and instead, it is parsed into the same ω containing the nominal. Bennett et al. (2016: 217) provide the following representation for the ditransitive sentence in (53) below: (60) Prosodic representation of determiners I (Bennett et al. 2016: 216-217) sciob an cat an t-eireaball den luch cut.PST the cat the tail off.the mouse ‘The cat cut the tail off the mouse.’
φ
ω φ
sciob ω φ
an cat ω ω
an t-eireaball den luch Under the analysis presented in Elfner (2012, 2015) and Bennett et al. (2016), the article is not a separate ω, and accent aligns to the primary stressed syllable, which is the first syllable
Syntactically informed phonology i n I r i s h | 40 of the noun. This analysis is consistent with the empirical facts of pitch accent assignment in Irish. By contrast, the analysis of the prosodic structure of definite articles that I develop in Windsor (2011, 2014a, b) does match (some) articles as ωs in prosodic representation, and creates a recursive ω structure for articles and nouns (see below for discussion of demonstratives, which are distinct from definite articles in Irish):23 (61) Prosodic representation of determiners II (adapted from Windsor 2014a: 108) an chloch mhór sin DEF stone big DIST ‘that big stone’ a. Morphosyntactic structure b. Prosodic structure φ ← DemP
φ ← DP ω ← Dem0 sin ω ← D0 ω ← A0 mhór ω ← D0 ω ← N0 an chloch Because both analyses handle the attested pitch accent distribution patterns observed, the choice of analysis does not represent a crucial difference for the present study: Both analyses maintain that the primary stressed syllable of the ω encapsulating the article and noun is the initial syllable of the noun. Therefore, phrasal accents are still realized on the correct syllable regardless of the nominal representation assumed.24
23 “(some) articles” refers to prosodic weight – only articles which contain two mora are realized as ωs. Thus an [ən] with a nucleus and sonorant coda is realized as a ω, but the monomoraic na [nə] is not. 24 Elfner shows that prepositions are similarly irrelevant for the distribution of accents and assigns them a similar prosodic representation to articles. The following representation is adapted from Elfner (2015: 1184), to be consistent with the depictions used in this section: i. díolfaidh rúnaí dathúil bláthanna áille le daoine anamúla sell.FUT secretary handsome flowers beautiful.PL to people animated ‘A handsome secretary will sell beautiful flowers to animated people.’
Syntactically informed phonology i n I r i s h | 41
The crucial fact about the prosodic structure of representations such as those in (61) is the status of the demonstrative. Irish demonstratives, despite being closed-class elements, are uncontroversially realized as ωs in phonological representation (see Bennett et al. 2016: §8.6 for additional discussion). The syntactic organization of demonstratives in Irish was briefly described in Chapter 1; further evidence for their position as elements that dominate DP will be provided in Chapter 5 (cf. McCloskey 2004; Adger 2013; Windsor 2014a). For now, what is important about the prosodic representation of demonstratives is how they affect object nouns of transitive clauses. Consider the status of the pitch accent on the object nouns in the two prosodic representations in example (62) below:
φ ←
φ ← ΣP
ω ← V0 φ ← TP díolfaidh φ ← DP φ ← vP/VP
ω ← N0 ω ← A0 φ ← DP φ ← PP rúnaí dathúil ω ← N0 ω ← A0 ω ← N0 ω ← A0 bláthanna áille le daoine anamúla
Syntactically informed phonology i n I r i s h | 42
(62) The distribution of phrasal accents by demonstratives (Windsor et al. to appear) 25 a. LH LH HL HL ˈjɛsʲiɡ ən pʲeːnˈtʲeɾʲ ˈklaxtˠax ən ˈtakas ˈbɾɪsʲtʲə fixed the painter experienced the easel broken ‘The experienced painter fixed the broken easel.’
b. LH LH HL LH HL HL ˈjɛsʲiɡ ən pʲeːnˈtʲeɾʲ ˈklaxtˠax ən ˈtakas ˈbɾɪsʲtʲə sʲɤ fixed the painter experienced the easel broken this ‘The experienced painter fixed this broken easel.’ a'. φ ← ΣP
ω ← V0 φ ← TP ˈjɛsʲiɡ φ ← DP φ ← vP/VP/DP
ω ← D0/N0 ω ← AP ən pʲeːnˈtʲeɾʲ ˈklaxtˠax ω ← D0/N0 ω ← AP ənˈtakas ˈbɾɪsʲtʲə b'. φ ← ΣP
ω ← V0 φ ← TP ˈjɛsʲiɡ φ ← DP φ ← vP/VP/DemP
ω ← D0/N0 ω ← AP φ ← DP ω ← Dem0 ən pʲeːnˈtʲeɾʲ ˈklaxtˠax sʲɤ ω ← D0/N0 ω ← AP ənˈtakas ˈbɾɪsʲtʲə As can be seen in the representations above, when a demonstrative is added to the object noun of a transitive clause, it creates an additional φ. The recursive φ structure is created by matching a DemP that dominates DP in the syntax, as shown in example (61). The additional φ created by matching the DemP is non-minimal, and its leftmost daughter is the D+object noun. By making the D+object noun the leftmost daughter of a non-minimal φ, we force an LH accent to be realized on the primary stressed syllable of that ω. This minor manipulation of the syntactic structure that underlies the prosodic representations in (62a)
25 Transcription for these sentences represent the production by a speaker of Munster Irish. Which words these accents are realized on does not change by dialect, only the particular syllable they align with. This alignment difference is discussed in greater detail in §3.5.
Syntactically informed phonology i n I r i s h | 43 and (62b), accompanied by an understanding of the interface with PF, allows for the study of stress and accent in isolation, while maintaining the position of the target noun within the breath group. In the following section, I detail a phonetic analysis of stress and accent in Irish, which supports the hypothesized representations in (61) and (62), and also provides insight into the possible diachronic origins of a stress-shift phenomenon in Munster Irish, reported in §3.6.
3.4. The phonetics of Irish prominence In this section, I detail the phonetic analysis of an elicitation experiment conducted with four native speakers of Irish based on the representations proposed in §3.3. The results of the phonetic analysis provide evidence in support of the hypothesized phonological and syntactic representations in (61) and (62) – as well as independently supporting the results of Elfner (2012, 2015). The goal of the following study is to determine the phonetic correlates of stress (ω- level prominence) independent of accent (φ-level prominence) in two dialects of Irish with the ultimate goal of providing a phonological account of the two levels of prominence and their differences in realization between Connemara and Munster Irish. The hypothesis that underlays this study is that if the representations in (62) are correct, eliciting object nouns both with and without a demonstrative will allow me to control for pitch accents: Object nouns with a single adjective but no demonstrative will display only stress, but object nouns with a single adjective and a demonstrative will additionally display accent. Therefore, contrasting accented object nouns and accentless object nouns, if the hypothesis is not rejected based on the data elicited, will provide evidence for the phonetic correlations of these two distinct levels of prominence. As a result of disentangling stress and accent, additional insight into the phonological similarities and differences between two Irish dialects is gained by extending the analysis from Connemara Irish (CI) —the dialect area studied by Elfner— to two micro-dialects of Munster Irish (MI). Finally, the phonetic differences between CI and MI are used to motivate a hypothesis towards a possible diachronic catalyst, which may have led to a phonologized stress shift pattern in modern MI.
Syntactically informed phonology i n I r i s h | 44
Finally, if the hypothesized distribution of pitch accents is empirically supported by the elicited data, this evidence will be used to support the hypothesized syntactic constituency of Irish demonstratives in (61), which are discussed in more detail in Chapter 5.
3.4.1. Methodology 3.4.1.1. Participants In order to independently confirm the results of Elfner’s (2012, 2015) study, two female native speakers of Connemara Irish from an Cheathrú Rua, County Galway —the same dialect area used in Elfner’s study— aged between 18-25 years old were recruited. Additionally, two female native speakers of Munster Irish, one from Cill na Martra, County Cork and one from an tSeanphobail, County Waterford were also recruited. The speaker from Cork was aged between 18-25 years old, and the speaker from Waterford was aged between 32-40 years old.26 All speakers reported using Irish on a daily basis, outside of academic settings. This information is summarized in Figure 1:
26 The language consultant questionnaire used in this study allowed participants to select age ranges: 18-25, 26-31, 32-40, 41-60, 61-80, or 81+.
Syntactically informed phonology i n I r i s h | 45
Figure 1: Participant information (image: Google) 3.4.1.2. Stimuli and elicitation Sentences with the basic structure of [Verb [Noun Adjective]subject [Noun Adjective]object] were created to elicit 40 target multisyllabic nouns27 in each of three conditions: Subject Condition (target noun as subject of the sentence), Object Condition (target noun as direct object of the sentence), and Demonstrative Condition (target noun as direct object, modified by a demonstrative). The sentences were designed to replicate the results of Elfner (2012, 2015), showing that subject nouns of a transitive sentence with branching subjects and objects have a pitch accent, but objects in the same sentential frame do not. Further, if adding a demonstrative reliably forced a pitch accent onto an object noun, the
27 All target nouns were at least disyllabic to allow for comparison between the stressed syllable and the immediately adjacent unstressed syllable. While the majority of target nouns contained only two syllables, 11 targets contained three syllables. Words longer than three syllables are exceedingly rare in Irish, but the target list also contained four four-syllable words, and one five-syllable word.
Syntactically informed phonology i n I r i s h | 46 representations in (61) and (62) in §3.3 would be supported by empirical evidence. Examples of each of these three conditions are presented in (63).28 (63) Example sentences in Irish stress study a. Subject condition thug an coinnlín céarach solas geal PST-give DEF candle-DIM wax light bright ‘The little wax candle gave bright light.’ b. Object condition lasfaidh an córbhuchaill óg an coinnlín naofa light-FUT DEF choir-boy young DEF candle-DIM holy ‘The young choirboy will light the holy little candle.’ c. Demonstrative condition lasfaidh an córbhuchaill óg an coinnlín naofa sin light-FUT DEF choir-boy young DEF candle-DIM holy DEM ‘The young choirboy will light that holy little candle.’ A high degree of variability was used in the sentences in an attempt to eliminate participant fatigue and list intonation. The only true minimal pairs frequently found in the elicitation set are between the object condition and demonstrative condition as that was the primary comparison under investigation – the subject condition was used as a control. Further, 40 filler sentences composed largely of seanfhocail29 ‘proverbs’ were added to the elicitation sentences to avoid list intonation that might otherwise occur because every test sentence has the same syntactic frame of [Verb [Noun Adjective] [Noun Adjective]]:
28 Note that an imperative verb with a null subject and two objects (VOO) has the same phonological representation as a VSO sentence (see Elfner 2014 for a comparison of VOO and VSO). Six VOO imperative sentences were included in the sentences used for elicitation in this study such as: i. íoc an píobaire tallannach dhá phunt airgid pay.IMP DEF piper talented two pound Stirling ‘Pay the talented piper two Stirling pounds.’
All sentences used for elicitation are provided in Appendix II. 29 Seanfhocail are common sayings in everyday Irish or idioms taught in Irish-medium education in Ireland.
Syntactically informed phonology i n I r i s h | 47
(64) Example filler sentences in Irish stress study a. éist le ceol na habhann agus gheobhaidh tú breac listen with music DEF river and get.FUT 2.SG trout ‘Listen to the sound of the river and you’ll get a trout.’ b. ní bhíonn suas gan síos NEG be up without down ‘There is no up without a down.’ c. cén fáth a bhfuil mé tuirseach i gcónaí what reason PRT be 1.SG tired in always ‘Why am I always tired?’ Stimuli were presented to the participants visually, in a randomized order, using an caighdeán oifigiúil, the official standard spelling convention.30 The participants were asked to read the sentences to themselves, and then say them as naturally as possible, as if speaking to a friend – the same methodology used in Elfner (2012, 2015). This instruction encouraged participants to use their native dialect rather than a formal register. Some evidence that this methodology was successful in eliciting naturalistic data comes from the fact that participants would frequently offer alternative pronunciations saying things like: “thit an mála daor ar an urlár salach [sl̪ ˠax]… nó [sɑːl̪ ˠax]. Déarfadh siad [sɑːl̪ ˠax] cá tú as, nó [hɑːl̪ ˠax] i mBéal Feirste” (‘the expensive bag fell on the dirty ([sl̪ ˠax]) floor (elicited sentence) or [sɑːl̪ ˠax]. They would say [sɑːl̪ ˠax] where you’re out of, or [hɑːl̪ ˠax] in Belfast’), acknowledging that they used a form particular to their dialect that I might not have been familiar with. Beyond processes like syllable elision (/sɑːl̪ ˠax/ → [sl̪ ˠax] ‘dirty’ or /pɪŋən/ → [piːn] ‘penny’ in various dialects), participants also provided the initial mutation that would be used in their dialect rather than simply reading what was presented to them. There are two initial mutations that apply to consonants in Irish: lenition and eclipsis. Lenition, as briefly outlined in Chapter 1, weakens an initial consonant such that /f/ → Ø or /b/ → [v] (for example).31 Eclipsis, on the other hand, fortifies initial consonants such that /f/ → [v] or /b/ → [m].32 Lenition is captured in the orthography by inserting
30 I am grateful to professors Gearóid Ó Domagáin (Ulster University) and Máire Ní Chiosáin (University College Dublin) for their assistance in composing these elicitation sentences in the official standard system so as not to represent any particular regional variety. 31 Weakening in Irish lenition includes: Causing labial and dorsal stops to become continuants (either fricatives or glides, depending on context); debuccalization of coronal consonants to glottal [h]; and, /f/ reduces to Ø. 32 Fortification in Irish eclipsis causes voicing of voiceless obstruents and nasalization of voiced obstruents.
Syntactically informed phonology i n I r i s h | 48 as the second letter of the word (e.g.,
3.4.1.3. Analysis Phonetic analysis was conducted in Praat (Boersma & Weenink 2016), and statistical analyses were conducted in R (R core team 2013). Target nouns were segmented out of
33 de Bhaldraithe (1977 [1992/2013]) lists go as a dialectal variant of the relative particle a in some usages.
Syntactically informed phonology i n I r i s h | 49 their carrier sentences and coded for speaker, a token identification number, and condition. The segmented targets were played separately for two phonetically trained undergraduate research assistants who provided transcriptions and judgements on which syllable was stressed (or if there was a secondary stressed syllable (see Ó Sé 2000 and Iosad 2013 for discussion)). Tokens were visually inspected for the presence or absence of a pitch accent on the stressed syllable, as depicted in Figure 2.
Figure 2: Pitch accents by condition (bláthanna ‘flowers’ as produced by a CI speaker) In the pitch traces provided in Figure 2, we can observe an LH accent on the initial (stressed) syllable of the target noun (bláthanna ‘flowers’) in the subject condition (blue line) as well as in the demonstrative condition (green line), which are contrasted with the relatively flat pitch trace of the target noun in the object condition (red line). After segmentation, the stressed syllable and the immediately adjacent following syllable were analyzed for various measures of pitch (Hz) (min, mean, max, and the range of pitch across the syllable), amplitude (dB) (the same measures as pitch), and for vowel duration (ms).34 These measurements were tabulated in an Excel spreadsheet and imported into R for statistical analysis. Of note, minimum and maximum values of pitch and amplitude were recorded in order to get the range of each of these measures across the
34 “Range” or “range across the syllable” refers to the difference in pitch or amplitude between the minimal and maximal values: Range = Maximum − Minimum.
Syntactically informed phonology i n I r i s h | 50 syllable, but were not both included in the models to avoid collinearity – since both measures were contained within range, which was included in the models.35 There were two targets that were eliminated from the analysis. The target word foclorín ‘wordlist’ was not elicited in all three conditions, and so it could not be used for comparison. The target word siompóisiam ‘symposium’ presented with an English pronunciation (with stress on the second syllable rather than aligned with the left edge of the word, and with a voiced medial fricative [z], which is not in the Irish phonetic inventory), and so it too was excluded. Additionally, tokens from the MI dialect that presented with final stress (MI stress shift is detailed in §3.5.1) were also excluded because the ω-level prominence analysis was conducted by contrasting the stressed syllable with the immediately following unstressed syllable.
3.4.2. Results The first test performed on the data was a chi-squared (χ2) test on the pitch accent data.36 In order to meaningfully utilize the elicited MI speaker data, and compare it to the elicited CI speaker data, it was first necessary to determine whether or not Elfner’s previous analysis could be extended to MI varieties.37 In other words, before looking at the data as a whole, it was first necessary to determine whether or not the distribution of accents was the same across the two dialects. Table 1 compares the number of tokens displaying the presence or absence of pitch accents on nouns in each of the three conditions between the two dialects:
35 Collinearity is a high degree of correlation between two predictor variables; this masks the effect of the two predictor variables on the outcome, potentially nullifying the underlying effect in the model. Generalized Linear Models work by assessing predictive capability of independent variables on some specified dependent variable. Collinearity can cause independent variables to assess against each other rather than the dependent variable. 36 Significance reported for all tests in this dissertation relies on the standardly assumed α of 0.05. 37 Bennett et al. (forthcoming), investigating pitch accents in northern varieties spoken in Ulster, suggest that the difference between dialects with respect to pitch accents lays in the tunes and shapes of the accents, rather than their distribution.
Syntactically informed phonology i n I r i s h | 51
Subject Object Demonstrative
Condition Condition Condition CI MI CI MI CI MI Presence of LH pitch accent on 57 45 21 15 55 40 stressed syllable Absence of LH pitch accent on 19 14 55 41 21 19 stressed syllable χ2 0.029061 0.011631 0.33296 p =0.865 =0.914 =0.564 Table 1: Distribution of phrasal accents by dialect and condition As can be seen in the table above, there is no significant difference between the two dialects with respect to accent distribution in any of the three conditions.38 This means that the data could be analyzed as a whole to determine whether or not adding a demonstrative to the object noun could reliably force a pitch accent to be realized on that noun. The next test performed on the data was a χ2 test examining the difference in accent realization between the three conditions. If Elfner’s (2012, 2015) results are replicated, we expect a significant difference between the frequency of pitch accent realization on subject nouns compared to object nouns, with the higher frequency of tokens displaying pitch accents in subject position. In order to reject the hypothesized structures in (61) and (62), we would need to find a similar difference between the subject condition and the demonstrative condition, but no significant difference between the demonstrative and object conditions. On the other hand, if there is a significant difference between the object and demonstrative conditions, but not between subject and demonstrative conditions, this will be evidence in support of the hypothesized representations in (61) and (62). The results of this test are presented in Table 2, which compares the number of tokens displaying the presence or absence of a pitch accent between the various conditions:
38 Likewise, no difference was found between the micro-dialects within MI to suggest that there was any intra-dialect variation between the Cork speaker and Waterford speaker: [χ2 = 1.866; p = 0.17].
Syntactically informed phonology i n I r i s h | 52
Subj v. Obj Subj v. Dem Obj v. Dem
Condition Condition Condition Subj Obj Subj Dem Obj Dem Presence of LH pitch accent on 102 36 102 95 36 95 stressed syllable Absence of LH pitch accent on 33 96 33 40 96 40 stressed syllable χ2 63.32 0.92 50.6192 p <0.001 =0.337 <0.001 Table 2: Chi-squared analysis of tokens in different conditions The results presented in Table 2 show that there is a significant difference in the frequency of accent realization between subjects and objects (p < 0.001), with accents being realized more frequently on subject nouns. This result shows that we are able to replicate Elfner’s (2012, 2015) findings with independent elicitation data, and provides a reason to continue to assume her analysis of pitch accent distribution based on prosodic constituency. In the results presented in Table 2, we also find that there is no significant difference in the frequency of accent realization on nouns in subject position compared to nouns in object position with a demonstrative (p = 0.337). However, there is a significant difference in the frequency of accent realization between nouns in object position with a demonstrative compared to those without a demonstrative (p < 0.001); in this comparison, accents are realized more frequently in the demonstrative condition than in the object condition. These results support the hypothesized structures provided in (61) and (62) above. The support for these hypothesized representations comes from the fact that (assuming Elfner’s previous analysis of the distribution of phrasal accents), if the DemP dominates DP, and both projections are matched as φs, then the object noun will become the leftmost daughter of a φNon-Min and thus present with a pitch accent (as in 62b). By contrast, without a demonstrative, object nouns are not hypothesized to be the leftmost daughter of a φNon- Min, and therefore not predicted to present with a pitch accent (as in 62a). These predictions, based on the assumed syntactic representation in (61), the syntax-prosody matching algorithms motivated by Elfner, and the accent distribution patterns also motivated by Elfner, are borne out by the tests reported in Table 2.
Syntactically informed phonology i n I r i s h | 53
The results of the χ2 tests presented above demonstrate that the distribution of phrasal pitch accents can largely be controlled for by altering the syntactic structure, and thus the prosodic structure, of object nouns using the presence or absence of demonstratives. Thus, the same lexeme can be elicited both with, and without a pitch accent in the same position within the breath group – allowing for the control of potential confounds such as pitch decline across an utterance. Having shown that the elicitation methodology successfully allows for control of the relevant factors, we turn to an examination of the phonetic correlates of the two levels of prominence beginning with φ-accent.
3.4.3. The phonetics of accent The hypothesis that adding a demonstrative to an object noun could force a pitch accent to be realized on that otherwise accentless noun was supported by a χ2 test, presented in the previous section. Because the methodology detailed above successfully allowed the elicitation of nouns both with and without an accent in the same position within the breath group, we can now meaningfully compare phonetic measures of pitch (Hz) to find the phonetic correlates of accent. The comparison was made using a Generalized Linear Model (GLM) with binomial distribution in R. The successful model considered the independent variables Minimum Hz (MinP), Range Hz (RangeP), and Mean Hz (MeanP) on the dependent variable of pitch accent realization.39 The GLM compared the accented syllable of an object noun in the demonstrative condition (as in 66b) to the same syllable of an accentless object noun in the object condition (as in 66a) to see which phonetic variables could reliably predict the presence/absence of a pitch accent:
39 Models for each GLM were created using all viable phonetic correlates examined in the study that did not lead to collinearity (see fn. 35 above). Models were assessed for Goodness of Fit and successively pared down until a successful model was found for each GLM, which best fit the data being examined.
Syntactically informed phonology i n I r i s h | 54
(66) Token comparators for accent a. Object condition (token 3a, CI speaker) LH LH HL HL ˈjɛsʲiɡ ən ˈpʲeːntʲeɾʲ ˈklaxtˠax [ən ˈtakəs ˈbɾɪsʲtʲə] fixed DEF painter experienced DEF easel broken ‘The experienced painter fixed the broken easel.’
b. Demonstrative condition (token 5a, CI speaker) LH LH HL LH HL HL ˈjɛsʲiɡ ən ˈpʲeːntʲeɾʲ ˈklaxtˠax [ən ˈtakəs ˈbɾɪsʲtʲə sʲɤ] fixed DEF painter experienced DEF easel broken PROX ‘The experienced painter fixed this broken easel.’ Although there was no significant difference in accent distribution between the two major dialects, there was an effect of dialect on the GLM predicting accent. The results of this GLM are therefore given for CI and MI separately.40 Phrasal prominence (accent) can be predicted in CI by an interaction of RangeP * MeanP (the range that Hz changes across the syllable and the mean Hz of the overall syllable): [G2 (1, 435) = 10.016; p = 0.001]. In MI, the interaction with MeanP is MinP * MeanP (the minimum Hz value in the syllable and the mean Hz of the overall syllable): [G2 (1, 352) = 30.714; p = 0.003]. These results show that, if we generalize across dialects, the common independent phonetic variable used in the realization of accent is the mean Hz within the syllable. However, the interaction with that variable changes between the CI dialect and the MI dialect. Next, we look at the phonetic variables used in the realization of stress.
3.4.4. The phonetics of stress In §3.4.2, I demonstrated that the object noun of a transitive sentence modified only by a single adjective was in an accentless position based on the phonological representation. In order to find the phonetic variables that realize accent, that accentless noun was compared to an accented noun – one modified by both a single adjective and a demonstrative. In this section, I compare the stressed syllable of the accentless object noun to the immediately adjacent unstressed syllable to determine the phonetic variables that realize stress devoid of accent. Therefore, only target nouns from the object condition are examined in this section, and stressed syllables of object nouns are contrasted with their immediately
40 There was no effect of speaker between the two micro-dialects within the Munster region.
Syntactically informed phonology i n I r i s h | 55 adjacent unstressed syllables rather than nouns from one condition or dialect being contrasted with nouns in a separate condition/dialect: (67) Token comparators for stress a. Object condition (token 3a, CI speaker) LH LH HL HL ˈjɛsʲiɡ ən ˈpʲeːntʲeɾʲ ˈklaxtˠax [ən ˈtakəs ˈbɾɪsʲtʲə] fixed DEF painter experienced DEF easel broken ‘The experienced painter fixed the broken easel.’ b. Object condition (token 4l, MI speaker) LH LH HL HL ˈɣɑw.siɡ nə kɑ.ˈl̪ ʲiː.niː ˈuv.ɾə.xə l̪ ʲɛsʲ [nə hoː.ˈɡɑː.niɡʲ l̪ʲɛsʲ.ˈkʲuː.l̪ ˠə] dance.PST DEF.PL girl.PL spirited with DEF.PL young.boy.PL shy ‘The spirited girls danced with the shy young boys.’ It is important to note here again that several tokens provided by MI speakers (as exampled in 67b) displayed shifted stress, where stress was not realized on the initial syllable of the noun. In §3.5.1, I detail the phenomenon of stress shift that occurs in MI, where stress may shift to the second or third syllable because of metrical weight (Green 1996, 1997; Gussmann 1997; Thrift 1997; Ó Sé 2008; Iosad 2013). For now, it should be noted that if stress was shifted to the final syllable of the word (e.g., in páipéar ‘paper’ which was frequently realized as [bɑː.ˈpeːɾʲ] by MI speakers), it was excluded from the study since there was no subsequent syllable for comparison (as noted in §3.4.1.3). This criterion also excluded the target noun pingin ‘penny’ from use in the CI data since it was always reduced to monosyllabic [piːn] for those speakers. These exclusions result in n = 70 for CI speakers and n = 54 for MI speakers (out of a possible n = 80 for each dialect region, based on 40 target nouns produced in the accentless condition by each of two speakers). In addition to the measures of pitch reported for accent in the preceding section, syllables were also examined for amplitude (dB): Minimum dB (MinA), Mean dB (MeanA), Maximum dB (MaxA), and the range that amplitude changed across the syllable (RangeA). Additionally, vowels were measured for length (ms) (vowel length is contrastive in Irish, but most Irish dialects are analysed as being quantity-insensitive (Bennett 2012) – this is discussed in greater detail in the following two sections). The phonetic measurements were tested in a GLM in R to determine which variables could successfully predict the difference between stressed and unstressed syllables.
Syntactically informed phonology i n I r i s h | 56
As with the GLM used in the prediction of accent, there was an effect of dialect in predicting stress. The successful model for predicting stress in CI examined MaxP, RangeP, and RangeA and produced two significant interactions: RangeA * MaxP: [G2 (1, 140) = 11.8231; p < 0.001] and RangeA * RangeP: [G2 (1, 139) = 8.4895; p = 0.007]. The successful model for predicting stress in MI examined MaxP, RangeA, and MaxA and produced a significant three-way interaction: RangeA * MaxA * MaxP: [G2 (1, 107) = 11.1146; p = 0.012].41 Similar to the results presented in the previous section on accent, stress in Irish presents with a single common phonetic variable, the range of amplitude change across the syllable, which shows different interactions based on dialect. These results are summarized in §3.5, but first, I discuss why vowel length was not a useful predictor variable in any of the models used.
3.4.4.1.Vowel length At first glance, vowel length appears to be significantly different between stressed and unstressed syllables. A t-test of the difference in average vowel length in stressed and unstressed syllables would seem to back up that observation:
41 Once again, there was no effect of speaker between the micro-dialects within the Munster region.
Syntactically informed phonology i n I r i s h | 57
Figure 3: Vowel duration by stress Figure 3 depicts a t-test with Welch correction, which shows a significant difference between the average length of vowels in stressed versus unstressed syllables: [t (602.01) = 11.919; p < 0.001]. However, there are two reasons to be suspicious of this result. The first reason to doubt the result shown in Figure 3 is that this test does not consider the length of vowels in other unstressed syllables within the word. For example, the target noun ceannaitheoir [ˈkʲa.na.hoəɾʲ] ‘buyer’ was consistently realized with the initial syllable longer than the middle syllable, but with the final syllable longer than either of the preceding two, despite not being stressed or in φ-final position.42 The second reason to be suspicious of the result presented in Figure 3 is that the test presented compares mean lengths across vowel categories and does not allow for within- category comparisons. Given the tendency for schwa to appear in the second (unstressed) syllable of Irish words, it is just as likely that vowel reduction is responsible for the result
42 In the Munster Irish dialects, stress is shifted to the final syllable of this word, but the generalizations about vowel length remain true for Connemara Irish speakers who do not shift stress, nor show any indication of secondary stress on this lexeme.
Syntactically informed phonology i n I r i s h | 58 in Figure 3 as it is that vowel length is a function of stress. As can be seen in the data in (68), even if there is synchronic pressure to shorten vowels in unstressed syllables, it is not the case that an unstressed syllable must reduce to schwa. It is also not at all the case that a long vowel cannot occupy an unstressed syllable in Irish, as is the case in 105 of 400 tokens (26.25%) elicited for the stress studies presented in this chapter (all speakers): (68) Long vowels in unstressed syllables a. [ˈkɤ.nʲiːnʲ] ‘rabbit’ b. [ˈuː.da.ɾaːs] ‘authority’ c. [ˈboː.ɾiːnʲ] ‘country lane’ d. [ˈʃɛɾ.koː.ɡə] ‘sweetheart’ e. [ˈka.soː.ɡʲiːnʲ] ‘little jacket’
Given that full vowels (i.e., vowels not reduced to schwa) appear in both stressed and unstressed syllables, we can perform a within-category test to see if vowels can be classified as stressed or unstressed based on vowel length. For example, there are 116 instances of the vowel /i/ in the data elicited (all speakers). A Linear Discriminant Analysis (LDA) with cross-validation fails to correctly classify tokens as stressed or unstressed based on length in 58 of 116 tokens (50%), performing exactly at chance. An LDA is a statistical test used for pattern recognition that seeks to determine if a given variable (or combination of variables) can be used to discriminate between two or more categories. The LDA with cross-validation analyzes variance in the data to predict the classification of individual tokens based on a linear relationship with a specified variable(s). The cross-validation then compares the output of the predicted categories versus actual data labels to determine how effective the variable is in allowing the sorting of variables into categories. The results of an LDA with cross-validation predicting if a vowel belongs to a stressed syllable or not based on vowel length are shown in Figure 4. As can be seen in that figure, there is nowhere to draw a line that would separate stressed syllable tokens from unstressed syllable tokens based on length (along the y-axis). Instead, the tokens seem to be distributed almost randomly. This visual inspection is supported by the fact that the LDA performed exactly at chance, failing to correctly classify tokens in 50% of instances:
Syntactically informed phonology i n I r i s h | 59
Figure 4: Vowel length of /i/ by syllable stress This null result is perhaps unexpected based on the observation that stressed vowels seem to be longer than the immediately adjacent unstressed vowels. However, a within category analysis presented by the above LDA provides a plausible explanation as to why duration was not a variable in predicting stress in the above GLMs: The classification of a vowel as either carrying stress or not based on length is only accurate in 50% of tokens.
3.5. Phonological interpretation of the results The results of the phonetic analysis of prominence may provide some insight into the alignment of stress and accent in the two varieties of Irish: CI aligns stress and accent to the syllable at the left edge of the word, but MI displays a greater amount of misalignment between the two levels of prominence, a significant difference, as well as a stress shift phenomenon (detailed below). In Figure 5, an example is given from CI which shows that the initial syllable of údarás ‘authority’ is the stressed syllable (as indicated primarily by the increased amplitude, shown by the dotted blue line). The stressed syllable of the target noun in Figure 5 also contains the pitch peak (shown by the solid black line), indicating that stress and accent are phonetically realized on the same syllable (aligned).
Syntactically informed phonology i n I r i s h | 60
Figure 5: Alignment of stress and accent in CI By contrast, the target noun fairrige ‘sea’, as produced by an MI speaker in Figure 6, displays an amplitude peak on the first syllable, but a pitch peak on the second syllable, indicating that stress and accent are phonetically realized on separate syllables – what I refer to as misalignment in the discussion to follow:
Figure 6: Misalignment of stress and accent in MI The phonetic results from §3.4.3 and §3.4.4 are summarized in Table 3:
Syntactically informed phonology i n I r i s h | 61
Connemara Irish Munster Irish Range Hz : Mean Hz Min Hz : Mean Hz φ-prominence (accent) [G2 (1,435) = 10.016; p = 0.001] [G2 (1,352) = 30.714; p = 0.003] Range Hz : Range dB [G2 (1,139) = 8.4895; p = 0.007] Max Hz : Range dB : Max dB ω-prominence (stress) Max Hz : Range dB [G2 (1,107) = 11.1146; p = 0.012] [G2 (1,140) = 11.8231; p < 0.001] Table 3: Summary of phonetic predictors of prominence Table 3 summarizes the results of the GLM testing discussed in the previous sections. Of note is the fact that the two levels of prominence in MI rely on completely separate phonetic variables, while both levels of prominence in CI utilize Range Hz in their realization.43 The statistical analysis of these phonetic realizations provide insight into differences in the alignment of stress and accent in the two dialects, and also suggest a diachronic catalyst for a stress-shift phenomenon that characterizes the MI variety.
3.5.1. Munster Irish stress shift With very few exceptions, stress in Irish is realized on the first syllable of the ω (Ó Siadhail 1989; Windsor 2011; Bennett 2012; Elfner 2015; Bennett et al. 2016). Bennett (2012) provides what he calls a straightforward description of stress in most dialects of Modern Irish (other than MI). Bennett (2012: 202) describes the Irish stress system as a quantity- insensitive system (i.e., not conditioned by syllable weight (cf. Green 1997)) where a non- iterative, trochaic foot (Σ) is aligned with the left edge of the ω: (69) [ˈLH] (Bennett 2012: 202) a. feiceáil *[ˈfʲe.kʲɑːlʲ] ‘seeing’ b. *[fʲe.ˈkʲɑːlʲ]
(70) [ˈLHL] (Bennett 2012: 202) a. bunábhar *[ˈbu.nɑː.wəɾ] ‘raw material b. *[ bu.ˈnɑː.wəɾ]
43 An earlier version of this research (presented at the Celtic Linguistics Conference 9 in Cardiff, Wales and ACOL 2016 in Calgary, Alberta) achieved similar results using t-tests with a Bonferroni corrected α of 0.0083. The t-tests showed that the significant differences between stressed and unstressed syllables in CI were: Max Hz (p = 0.002), Mean Hz (p = 0.007), Range Hz (p = 0.001), Max dB (0.001), and Range dB (p < 0.001) – overlapping with the significant predictors of accent, Mean Hz * Range Hz (p = 0.001). In MI, the significant differences between stressed and unstressed syllables were: Max Hz (p = 0.004) and Range dB (p = 0.001) – not overlapping at all with the significant predictors of accent, Min Hz * Mean Hz (p = 0.003).
Syntactically informed phonology i n I r i s h | 62
(71) [ˈLLH] (Bennett 2012: 202) a. cogarnaíl *[ˈko.ɡəɾ.niːlʲ] ‘whispering’ b. *[ko.ɡəɾ.ˈniːlʲ]
Bennett argues that, with the exception of Munster Irish (discussed below), instances of multiple stresses come from separate ωs, as in the case of compounds or quantifier + adjective constructions, not from iterative Σ structure (see Green 2008 and Windsor 2011 for additional discussion). Bennett provides the following table to show that distinction: Word Transcription Structural parse Gloss
droch-bhád [ˈdɾox.ˈwɑːd] [ω(ˈdɾox)] [ω (ˈwɑːd)] ‘bad boat’
fíor-bhean [ˈfiːəɾ.ˈvʲæːn] [ω (ˈfiːəɾ)] [ω (ˈvʲæːn)] ‘true woman’
an-mhór [ˈaːnˈwoːɾ] [ω (ˈaːn)] [ω (ˈwoːɾ)] ‘very big’ Table 4: Compounding in Irish (Bennett 2012: 209)44 In keeping with traditional characterizations of compound stress in descriptive grammars (e.g., Ó Sé 2000), Bennett maintains that the stresses realized on the two words are equal, but Green (2008) and Windsor (2011) argue that there are stress differences based on the structure that each of the two words are parsed into. This representational difference leads to the minimal pair in (72), differentiated phonetically by stress, but phonologically by structure – a recursive ω structure as compared to a φ containing non-recursive ωs:45 (72) Stress differences based on prosodic parses a. Parsed into a recursive ω b. Parsed into a φ ‘long-eared owl’ ‘head of a cat’
One of the hallmarks of the MI variety is a phenomenon of stress shift, which causes the main stress of a word to be shifted from the left edge —where it is realized in the other
44 The words presented in Table 4 are not typically written with a hyphen, this is included by Bennett to highlight the compositional nature of these forms. 45 Green (2008: 199) notes, following Ó Sé (2000: 52), that (72a) has a secondary-primary stress pattern, but that (72b) has a primary-primary stress pattern. I depict the single primary stress in (72b) by accounting for φ-prominence as well as ω-prominence.
Syntactically informed phonology i n I r i s h | 63 dialects of the language— to either the second or third syllable, depending on weight (see Iosad 2013 and references therein for discussion):46 (73) Examples of stress application in Munster Irish (Iosad 2013: 67) Syllables Lexeme Transcription Gloss a. ĹL carraig [ˈkarɪɡʲ] ‘rock, boulder’ b. ĹLL clagarnach [ˈklaɡəɾnəx] ‘clattering’ c. ĹLLL armacach [ˈarəməkəx] ‘tender’ d. ĹLLH imleacán [ˈimʲɪlʲəkɑːn] ‘navel’ e. H́ L álainn [ˈɑːlɪnʲ] ‘nice’ f. H́LL cúramach [ˈkuːrəməx] ‘careful’ g. H́LH údarás [ˈuːdərɑːs] ‘authority’ h. H́LHL údaráiseach [ˈuːdərɑːʃəx] ‘authoritative’ i. H́LHH amparánaíocht [ˈoumpərɑːniːxt] ‘ungainliness’ j. LH́ cailín [kaˈlʲiːnʲ] ‘girl’ k. LH́L coiméadann [kʲɪˈmʲaːdən] ‘(s)he observes’ l. LH́H bithiúntaíocht [bʲɪˈhuːntiːxt] ‘villainy’ m. LLH́ ceannaitheoir [kʲanəˈhoːrʲ] ‘buyer’ n. LLH́H amadántaíocht [əməˈdɑːntiːxt] ‘foolishness’ o. HH́ díomhaoin [dʲiːˈviːnʲ] ‘idle’ p. HH́L ógánach [oːˈɡɑːnəx] ‘young man’ q. HH́H amhránaí [uːˈrɑːniː] ‘songs’
Data of the type found in (73) are explained by Ó Sé (2008: 96, qtd. in Iosad 2013: 67), who summarizes the pattern by stating: “Word stress falls on the second syllable if it is heavy, on the third syllable if it is heavy and the preceding syllables are light, and on the first syllable in all other cases.” Iosad (2013), like Green (1997), considers only long vowels (including diphthongs) capable of creating a heavy syllable for the purposes of stress assignment and suggests that coda consonants contribute no weight to the syllable.47 Iosad provides the following alternations to illustrate the fact that stress is attracted to (C)Vː syllables but not (C)VC ones:
46 Some authors have noted a possible secondary stress in MI varieties that is not present in other dialects, see Doherty (1991); Green (1997); Ó Sé (2000); or, Iosad (2013) for relevant discussion. I will not address the possibility of secondary stress here, although it is likely, based on the data examined in the present study, that they are observing a pitch accent that is not aligned with the ω stress rather than a secondary stress. 47 Ní Chiosáin (1991), on the other hand, uses evidence from vowel epenthesis to argue that sonorant codas are moraic in Irish; this is not an important distinction for the analysis presented in this chapter (cf. Ní Chiosáin 2000).
Syntactically informed phonology i n I r i s h | 64
(74) Alternations that lead to non-initial heavy syllables and stress shift (Iosad 2013: 68) a. (i) dúthaigh [ˈduːhɪɡʲ] ‘region’ (ii) dúthaí [duːˈhiː] ‘regions’ b. (i) gainimh [ˈɡanʲɪvʲ] ‘sand’ (ii) gainimhe [ɡaˈnʲiː] ‘sand (GEN.SG)’
There are various explanations of this stress shift pattern within the literature that depend on structure – from recursive syllables and feet to positing the inclusion of prosodic colon in the prosodic hierarchy (see Green 1996, 1997 and Iosad 2013 for discussion). I do not motivate a sub-ω prosodic analysis of the stress shift pattern here; rather, I provide a potential diachronic catalyst for how that pattern became phonologized, based on the previously discussed phonetic evidence. The hypothesis regarding the origins of MI stress shift that I advance here comes from Edge Repulsion Effects based on Tonal Crowding (Gordon 2014) and the alignment of prominences.
3.5.2. Edge repulsion and accent alignment Gordon (2014) identifies two types of languages in a typology of prominence-assigning languages: Those languages which assign prominence in a bottom-up manner, and those which assign prominence in a top-down manner. One of the characteristics of bottom-up prominence-assigning languages is that the same syllable will host both stress and accent. For example, a language may assign stress to the leftmost syllable within a ω, then accent is aligned to the stress-bearing syllable. A top-down language, on the other hand, may assign stress and accent independently of one another – due to their own realizational constraints. For example, a language might assign accent to the syllable aligned with the left edge of a φ, but assign stress to the heaviest syllable in a ω (which may, or may not, align with the accent). I will argue here that one difference between CI and MI is that CI is a bottom-up variety while MI is a top-down variety. It is relatively uncontroversial that CI is a bottom-up prominence-assigning language. With very few schwa-initial exceptions, stress in CI is realized on the leftmost syllable of the ω (Elfner 2015). This suggests that, with the exception of syllables headed by schwa being banned from hosting stress, there is no influence of syllable weight on stress assignment (as in 69-71 above and 75 below). Further, all tokens examined for the
Syntactically informed phonology i n I r i s h | 65 alignment of stress and accent (n = 132) showed that stress and accent were perfectly aligned on the same syllable in CI.48 (75) Alignment based rather than weight sensitive stress in CI a. cailín [ˈka.lʲiːnʲ] ‘girl’ b. casóigín [ˈka.sˠoː.ɡʲiːnʲ] ‘little jacket’ c. sagartóireachta [ˈsˠɑ.ɡəɾ.t̪ˠoː.ɾɑx.t̪ˠə] ‘performing priest duties’ d. griangrafadóraiochta [ˈɡɾi.ən.ɡɾæ.fə.doː.ɾɑx.t̪ˠə] ‘photography’
If, however, MI is a top-down prominence assigning language, as I argue here, we predict (based on Gordon 2014) that there will be misalignment between which syllable accent is realized on and which syllable stress is realized on. This prediction is borne out in our data: All Speakers/Tokens MI Speakers Only CI MI Stress-initial Shifted stress Aligned 132 90 56 32 Misaligned 0 47 14 33 χ2 54.872 14.06 p < 0.001 < 0.001 Table 5: Stress-Accent alignment patterns As can be seen in the confusion matrix in Table 5, 34% of tokens from MI speakers display misalignment between the syllable realizing stress and the syllable realizing accent; this is significantly different from the amount of misalignment found in CI, which displayed no misalignment. Looking within the MI data, we see that 70% of the tokens displaying misalignment are tokens without initial stress (shifted stress tokens); therefore, misalignment is more common in tokens with shifted stress than initial stress, making up roughly 50% of those tokens, again, a significant difference. The reason that this misalignment is expected in top-down prominence assigning varieties is due to Tonal Crowding and Edge Repulsion Effects (Gordon 2014). In his typology of prominence assignment, Gordon (2014) argues that realizing multiple tones on a single syllable is dispreferred. When multiple tones are realized on a single syllable, one of those tones is repulsed from the edge as a repair strategy for tonal crowding. I suggest
48 One token in the current study was excluded due to the anglicized pronunciation by all participants. The loan word siompóisiam ‘symposium’ was routinely produced with English stress as [tsɪm.ˈpoː.zi.ʌm] by all speakers. (Note also that [z] is not a sound in the Irish inventory.) Despite the stress being realized on the second syllable by CI speakers, accent was still aligned with the primary stressed syllable in this token when it appeared in the correct prosodic environment.
Syntactically informed phonology i n I r i s h | 66 here that tonal crowding may explain the phonetic findings reported in §3.4. Recall that stress and accent were both realized in CI using a common phonetic cue, but that stress and accent in MI were realized using separate phonetic cues (summarized in Table 3, repeated here as Table 6 for convenience): Connemara Irish Munster Irish Range Hz : Mean Hz Min Hz : Mean Hz φ-prominence (accent) [G2 (1,435) = 10.016; p = 0.001] [G2 (1,352) = 30.714; p = 0.003] Range Hz : Range dB [G2 (1,139) = 8.4895; p = 0.007] Max Hz : Range dB : Max dB ω-prominence (stress) Max Hz : Range dB [G2 (1,107) = 11.1146; p = 0.012] [G2 (1,140) = 11.8231; p < 0.001] Table 6: Summary of phonetic predictors of prominence In CI varieties, both stress and accent are realized via an interaction with the range that pitch (Hz) travels across the syllable; this suggests that both levels of prominence, which are realized on a single left-aligned syllable, are realized by a single tone. However, MI varieties examined here show no common phonetic correlates of stress and accent; this suggests that both levels of prominence are realized by separate tones. Arguably, then, there is pressure to realize the two tones on separate syllables (i.e., Tonal Crowding, to be discussed further below). Indeed, misalignment between stress and accent is frequently observed in MI, with the majority of misalignment being found on tokens displaying shifted stress than initial stress – these significant differences are summarized in Table 5, and show no effect of speaker (within dialect regions). In the following section, I suggest a possible diachronic path for the phonologization of stress shift in MI due to a tonal crowding catalyst.
3.6. A hypothesis into the origins of Munster Irish stress shift Blankenhorn (1981) and Dalton & Ní Chasaide (2003) have suggested in passing that stress shift in MI varieties may have arisen due to a temporal misalignment between stress and accent. In this chapter, we have seen that there is a significant amount of temporal misalignment between the two prominences in MI, especially in tokens with non-initial stress. I suggested in the previous section that the temporal misalignment arises as a repair strategy for the problem of tonal crowding because stress and accent are realized via separate phonetic cues in that dialect. In this section, I outline a plausible diachronic path that this change is likely to have taken.
Syntactically informed phonology i n I r i s h | 67
Stage 1: I suggest here that the catalyst for the diachronic change being described presently was a shift to phonetically realize stress and accent via separate cues. The observation that MI varieties use separate phonetic cues to realize stress and accent was a measurable result presented in §3.4 and is a dialectal difference between MI and CI, the latter of which has a common phonetic cue for both stress and accent. When stress and accent came to be realized as separate tones on the left edge of the ω, this created pressure to repulse one of the tones from the left edge due to tonal crowding (cf. Gordon 2014). This disfavourable alignment of multiple tones on a single syllable would have to be repaired in stage 2. Stage 2: Gordon (2014) describes edge repulsion effects that move an accent away from the right edge of a φ. The edge repulsion that Gordon describes is due to the tonal crowding of a terminal pitch fall (potentially due to an intonational phrase) on top of an accent on the right edge of a φ. He suggests that initial pitch accents are a potential target for edge repulsion as well, but that edge repulsion at the left edge would be due to interaction with the utterance. Gordon states: Because the factors conditioning edge repulsion of the pitch peak are relatively weak at the left edge, it is not surprising that peninitial prominence is rare cross- linguistically. It also follows that the left edge counterpart to antepenultimate stress, post-peninitial stress, appears to be unattested. (109)
This portion of Gordon’s prominence typology provides a substantive49 motivation for the fact that non-finality effects are quite common cross-linguistically, but non-initial effects are rare, if attested at all.50 Notwithstanding the rarity of edge repulsion effects at the left edge of a φ, it is well documented that φ- and ω-level prominence are both realized on the initial syllable of a ω at the left edge of a φ in Irish (see Elfner 2012, 2015). If the alignment of stress and accent created a tonal crowding problem, as suggested here, the phonological component would need to utilize plausible repair strategies to resolve this issue. In the case of MI, tonal crowding on the left edge would have been repaired by shifting stress rather than accent,
49 By ‘substantive’, I mean ‘phonetically motivated’ in the sense of Moreton & Pater’s (2012) substantive bias (cf. White 2014 and references therein). 50 Kager (1999) describes contextual markedness constraints banning a feature in non-initial syllables as awkward at best. He argues that this is because “non-initial syllables” is not a natural class. Initial syllables may be prosodically targeted, but not the class of non-initial syllables.
Syntactically informed phonology i n I r i s h | 68 and non-initial heavy syllables provided an ideal anchoring site for stress – something substantively motivated by the Stress-to-Weight principle (Myers 1987; Riad 1992). By shifting stress from the left edge of the word to a non-initial heavy syllable, MI phonology partially resolved the tonal crowding problem, particularly in lexemes that allowed stress shift. However, the default strategy would still realize stress and accent at the left edge of the ω/φ. This could have led to the phonologization of the Stress-to-Weight pattern as well as the demotion of the tonal crowding constraint in stage 3. Stage 3: In this stage, the process of Stress-to-Weight became phonologized as the motivation for stress shift since edge repulsion failed to repair tonal crowding in all instances. This stage of the diachronic process explains the current observations of prominence in MI, and how that dialect differs from the CI variety investigated in this chapter: Stress and accent are realized by separate phonetic cues in MI, and stress is attracted to non-initial heavy syllables, though left-alignment remains the default.51 The hypothesized diachronic path of phonologizing the stress shift pattern in MI is consistent with the phonetic data presented in this chapter. However, the opposite direction of change —that stress shift facilitated the realization of stress and accent by separate phonetic cues in MI— is also possible, though perhaps not quite as probable. Misalignment occurred in only 34% of analyzed tokens in the preceding study. If stress shift, rather than tonal crowding, were the diachronic catalyst, we might expect that the temporal misalignment of the two tones to be far more regular across all tokens. If, however, tonal crowding was the diachronic catalyst which resulted in stress shift, as advanced here, then one might expect to find significantly more temporal misalignment in tokens displaying shifted stress; this is precisely the result that was found in the preceding sections: 70% of the temporal misalignment of the two tones was observed in tokens with shifted stress (as reported in Table 5, above).52
51 Analysing the word-internal structure of MI stress is beyond the scope of the present research. Presumably the pattern of stressing the initial rather than the third syllable of an H́ LH word such as údarás ‘authority’ ([ˈuːdərɑːs]), as in (73g) has to do with trochaic rhythm constraints and the word-internal prosodic structure. For relevant discussion, see Green 1996, 1997; Gussmann 1997; Thrift 1999; Ó Sé 2008; and, Iosad 2013. 52 See Windsor et al. (to appear) for predictions made by this hypothesis for other Irish dialects and other languages in Gordon’s (2014) typology.
Syntactically informed phonology i n I r i s h | 69
3.7. Summary and discussion In this chapter, we examined the phonetics of two levels of prominence in two dialects of Irish – ω-level stress, and φ-level accent in Connemara Irish (CI) and Munster Irish (MI). The primary phonetic findings of this study were that accent is realized as an interaction with mean pitch (Hz) that differs by dialect, and that stress is realized as an interaction with the range that amplitude (dB) changes across the syllable, which also varies by dialect. The significant interactions in a Generalized Linear Model (GLM) showed that a single phonetic cue (the range that pitch changes across the syllable) was used to realize both stress and accent in CI, but that the cues of stress and accent were completely separate in MI. The phonetic results provided evidence for two phonological conclusions: That the prosodic structure of the ω and φ is the same in both dialects, as is the distribution of accent,53 and that there is a substantively motivated diachronic path to explain MI stress shift. While the phonological conclusions are important in and of themselves, it is the methodology that made the phonetic testing and phonological analysis possible in the first place that is integral to this thesis. Elfner (2015: 1173) suggested, based on an analysis of pitch accents in CI, that because stress and accent are always realized on the same syllable, stress is recognizable by pitch when associated with an accent. Elfner’s generalization is supported by the statistical findings in CI based on the phonetic testing conducted in the present study. However, it was unclear from Elfner’s generalization how stress would be realized when not associated with an accent – an admittedly rare circumstance given the distribution of phrasal accents in the language. A controlled test was needed to answer that question and to discover that stress is primarily realized by an interaction with amplitude and that pitch is primarily in the domain of accent. Specifically, because one of the goals was to measure pitch, which decreases across the breath group, we needed a way to elicit a single lexeme that could be realized both with and without a phrasal accent in the same position within the breath group. An understanding of the syntax that provides the underlying structure to be matched by the phonological component allowed that controlled study to be done. In Windsor (2016a), I argued that demonstratives dominate the DP containing the rest of the nominal expression,
53 See Bennett et al. (forthcoming) for a discussion of these representations in Ulster Irish, which shows that the distribution of accents is the same for that dialect as the ones explored here, but only the alignment and shapes of the tones vary.
Syntactically informed phonology i n I r i s h | 70
(see also McCloskey 2004 and Adger 2013). This previous syntactic analysis allowed the manipulation of the prosody, providing a way to force an accent onto an otherwise accentless object noun simply by adding a single syllable to the end of the utterance: seo [sʲɤ] ‘this’, sin [sʲɪn] ‘that’, siúd [sʲuːd] ‘yon’. The addition of this single syllable maintained the object noun’s position within the breath group, but allowed significant manipulation of the prosodic structure such as to provide a way to measure stress and accent in isolation. The case study presented in this chapter is one example of syntactically informed phonology: How phonological analysis is made possible by utilizing a knowledge of the syntactic structure that prosodic structure is matched to. I return to the topic of providing prosodic and syntactic evidence for the constituency of demonstratives in Chapter 5. But first, in the next chapter, I provide a case study of Blackfoot which utilizes the syntax- phonology interface as a tool for research in the other direction – where phonological structure and phonetic testing is used to decide between competing morphosyntactic analyses of structure. fin
Phonologically informed syntax in Blackfoot | 71
4. Phonologically informed syntax in Blackfoot54 In the previous chapter, I demonstrated how understanding an underlying syntactic structure and matching that structure in the phonological component allows for controlled study of the phonetic correlates of prosodic categories. In the present chapter, I detail two studies on prosodic constituency in the Algonquian language, Blackfoot. I argue that the prosodic data gained in these studies provides evidence against the analysis that agglutinated nominal expressions are the result of successive head-movement, which is achieved via the Mirror Principle (Baker 1985; Bliss 2013; Wiltschko 2014): (76) The Mirror Principle (Baker 1985: 375) Morphological derivations must directly reflect syntactic derivations (and vice versa) Specifically, I demonstrate that the morphological affixation of suffixes displays distinct phonological patterning as compared to affixation of prefixes, which I argue here to be syntactic in nature. Because the two processes have different phonological realizations, I argue that they should be treated separately – by separate morphological and syntactic components where prefixation is achieved by syntactic operations and suffixation is achieved through morphological operations.55 The analysis that I put forward in this chapter, which is further supported in Chapter 5 on demonstratives, has consequences for the mirror principle. I argue that the data presented here are best explained if suffixation is analyzed as the morphological correspondence of morphosyntactic features realized on a single X0 through agreement relationships; if this is so, the mirror principle must allow for “syntactic derivations” to include Agree operations as well as Merge operations. In order to make this argument, I first detail two studies used to identify prosodic constituency which is matched from an underlying syntactic structure.
4.1. Prosodic categories and phonetic correlates In this section, I detail two elicitation experiments that test the phonetic correlates of prosodic boundaries in Blackfoot. By identifying methods through which prosodic categories can be identified, I provide evidence for the prosodic constituency of Blackfoot
54 The studies reported in this chapter previously appeared in Windsor 2016b, 2017. 55 For similar treatment of the differences between prefixes and suffixes in polysynthetic languages, see Russell 1999 on Cree (Algonquian) and Dakota (Siouan).
Phonologically informed syntax in Blackfoot | 72 nominal elements (the focus of subsequent studies reported in this, and the following chapter). The assumption that underpins the elicitation experiments reported in this section is that different prosodic categories are matched from different morphosyntactic categories. Based on that assumption, I study prosodic effects that are observable between roots and bound suffixes, between roots and bound prefixes, and between separate (orthographic) words. The first study shows a significant correlation between vowel devoicing (what I subsequently refer to as aspiration) and what I will ultimately argue is the right edge of a phonological phrase (φ). The second study provides evidence that an obligatory process of vowel coalescence is active between adjacent vowels across syllables, but that an optional phonetic process —which mimics the obligatory coalescence one— applies across prosodic word (ω) boundaries. The elicitation experiments detailed here provide evidence for how syllables (σ), ωs, and φs can be recognized in Blackfoot acoustic data. The motivated prosodic constituency is then used to provide cross-component evidence for a hypothesis concerning syntactic constituent structure in Blackfoot, which is the subject of §4.2.
4.1.1. Study 1: Right edge aspiration The study detailed in this section examines the role of sentential position as a predictor of what is typically described as final vowel devoicing in the Blackfoot literature (Frantz & Russell 1995; Elfner 2006; Gick et al. 2011; Bliss 2013; Frantz 2017). Within this literature, vowel devoicing, what I will henceforth refer to as aspiration,56 is typically described as occurring at the end of an orthographic word, and can apply to multiple words within an utterance.57 In this chapter, I follow Russell & Reinholtz 1997: 447) who conclude for the related Cree language (specifically the Plains Cree and Swampy Cree dialects), that “the
56 See §4.1.2. for a discussion of this relevant departure from the literature based on an analysis that unifies vowel devoicing and consonantal aspiration using the phonological feature [SPREAD GLOTTIS] – the phonological feature responsible for aspiration. 57 The term “orthographic word” refers to a discernable unit in the spelling convention of the language. The orthographic word has previously been used to argue for phonological and syntactic constituency (see Russell & Reinholtz 1997 and Bliss 2013). For example, Bliss (2013: 150, fn. 76) notes that diphthongization (coalescence in the dialect studied in this dissertation) and glide deletion operate within words, but not across words. Thus, in akáímahkihkinaiksi ‘old sheep’ (in example 77), the two instances of /ai/ in this word are expected to surface as [ɛː], but this coalescence is not found between a demonstrative and the noun it introduces because they are separate orthographic (and thus, frequently considered to be separate prosodic) words. As shown in (77), however, akáímahkihkinaiksi is analyzed here as one orthographic word, but two ωs.
Phonologically informed syntax in Blackfoot | 73 units of Cree which are usually called “words” are in fact phrases at the phonological level.”58 There are several reasons to believe that the conclusions for Cree also hold true for Blackfoot, and they are detailed throughout the remainder of this section. As can be seen in the structure in (77), the domain of aspiration (indicated by rightward descending arrows) is smaller than the clause: (77) An example of Blackfoot prosodic constituency
As depicted above, there are three instances of aspiration (demarcated as voiceless vowels), all of which happen to coincide with the orthographic words in that sentence. However, as we will see in §4.1.1.2, not all orthographic words display aspiration at their right edge (a finding from the present study). Finally, an orthographic word, in Blackfoot may contain several ωs (as shown in (77), a finding from the study presented in §4.1.3). These observations provide evidence that the domain of aspiration epenthesis in Blackfoot is higher than the ω but lower than the intonational phrase (ι) – suggesting, given the prosodic hierarchy assumed here, that the domain of right edge aspiration is the φ. The first step in substantiating these claims is to show that right edge aspiration is a positional alternation based on prosodic constituency. By determining which prosodic representations are responsible for the observable phonological alternations, we can hypothesize if they are matched from X0s or XPs (or CPs if the domain is found to be the intonational phrase) and test previous hypothesized syntactic structures (Bliss 2013 and Wiltschko 2014) against the attested prosodic structure for potential mismatches.
58 Donald Frantz (p.c. 2012) clarifies that in his grammar of Blackfoot, when describing the phonological process of vowel devoicing at the end of a “word,” he did not intend that statement to be directly linked to a prosodic ω, but was rather discussing a process which takes place at the end of an orthographic word so that it could be understood by non-linguists. He agrees that these strings most often represent φs rather than ωs in the strict phonological sense.
Phonologically informed syntax in Blackfoot | 74
4.1.1.1. Methodology The elicitation experiment detailed in this section examines the amount of observable aspiration on the right edge of a target noun when that noun appears in sentence-final position versus sentence-medial position. This was done by analyzing elicited tokens in Praat (Boersma & Weenink 2016) and measuring the duration for which stable formant bands could be observed without voicing (as agreed upon by two researchers).59 A representative noun can be seen in Figure 7, where the spectrogram for the noun ninaa ‘man’ is shown, with glottalic pulses indicated. The final portion of the vowel which displays stable formant bands is measured from the end of periodic voicing (0.109 seconds in this example):
Figure 7: Final aspirated vowel In Blackfoot, nouns typically end in a vowel due to inflectional suffixes that indicate animacy, number, and reference tracking (proximate/obviative), as shown in Table 7. This study utilized fully inflected, animate singular, nouns so that the same lexical form was compared in both sentential positions (as shown in the sample carrier sentences in (78) below). Animate Inanimate60 Proximate -wa -- Singular Obviative -yi Plural -iksi -istsi Table 7: Blackfoot nominal inflection (Bliss 2013: 30)
59 From an original study reported in Windsor & Cobler 2013. 60 Inanimate nouns cannot serve as the subjects of sentences in Blackfoot, and therefore cannot take proximate morphology.
Phonologically informed syntax in Blackfoot | 75
For the experiment used in the present study, a series of proximate animate singular nouns were elicited from three native speakers of the Kainai ‘Blood’ dialect of Blackfoot61 in one of two carrier sentences which translate as: ‘this is an (animate noun)’ and ‘this (animate noun) is my pet/possession’: (78) Sample elicitation sentences for study 1 a. ámo anistápssiwa pííta (sentence-final target) amo anistápssi-wa pííta-wa DEM be.called.AI-3.SG eagle-PROX ‘This is an eagle.’ b. ámo pííta nitsináána (sentence-medial target) amo pííta-wa n(it)-itsináán-wa DEM eagle-PROX 1.SG-possession-PROX ‘This eagle is my pet.’ The hypothesis underlying these elicitations is that the amount of aspiration observable on the animate singular (proximate) suffix -wa can be measured and contrasted between the two positions; if the amount of aspiration significantly differs between the two positions, then the orthographic word cannot represent the domain of aspiration, and it can be concluded that aspiration is a positionally-dependent epenthetic feature in the language. If no difference in the amount of aspiration exists between the two sentential positions, then the orthographic word can be said to represent the domain of aspiration, possibly with that feature being part of the underlying representation. In other words, if the orthographic word coincides with the phonological unit responsible for right-edge aspiration in Blackfoot, then aspiration should be observable on the right edge of that word regardless of its position within the utterance. However, if utterance position significantly alters the amount of observable aspiration on the target noun, then the orthographic word cannot coincide with the phonological unit responsible for this alternation. Elicitations were based on a translation task with the researchers asking the subjects how to say the equivalent English sentence in Blackfoot over multiple recording sessions. Recordings were made using the same device and settings used for the Irish study reported in the previous chapter. The recordings were subsequently imported into Praat for analysis. The duration of aspiration for each of the target nouns (n = 137; 71 tokens in final position
61 A fourth speaker was excluded from this study for dialectal (generational, rather than regional) differences which made the carrier sentence ungrammatical for him due to animacy mismatches, which did not exist for the three other (younger) speakers.
Phonologically informed syntax in Blackfoot | 76 and 66 in medial position – 5 medial tokens were excluded because a different form was provided by the consultants in this position) was measured and categorized based on the sentence-medial or sentence-final position of the target.
4.1.1.2. Results Vowels, both phonemically long and short, were found to be aspirated in the final third of the overall duration in final position (ranging between 27%-44%), averaging 0.1202 seconds. Separating aspiration duration by position, a two-sample t-test with Welch correction was run on the data with the following result: [t(91.207) = 6.0408; p < 0.001]. This result shows that the amount of aspiration is significantly different in the two positions with almost no aspiration observable in the sentence-medial position and an average of 0.1202 seconds of aspiration observable at the end of target nouns in sentence-final position. This observation is depicted in Figure 8.
Figure 8: Aspiration by position
Phonologically informed syntax in Blackfoot | 77
The boxplot in Figure 8 is a visual representation of the data used in the t-test described above. This data shows that when the target appears in sentence-final position (on the right of Figure 8), there is a large increase in the amount of observable aspiration. Except for a few outliers, almost no aspiration is observed on the target when it put in medial position (depicted on the left of Figure 8).62 If the orthographic word necessarily corresponded to the phonological unit which caused aspiration at its right edge, or if aspiration were part of the underlying representation of that word, then the position that the unit occupied within the sentence would not alter the amount of observable aspiration. As this prediction is not borne out, and aspiration is dependent on sentence position, it can be concluded that aspiration is a positional alteration, and the orthographic word is not the domain of aspiration – although it frequently, but not necessarily, corresponds to the prosodic unit which does define the domain of aspiration. In sum, this first elicitation experiment shows that aspiration at the right edge of some yet-to-be-defined prosodic domain is a statistically significant indicator of a particular prosodic boundary, one which can encompass a morphosyntactically complex grammatical unit (minimally a √noun and its inflectional suffixes). This result suggests that the unit in question is likely to be larger than the ω. At present, the given evidence does not allow the conclusion that the unit in question is necessarily a φ as opposed to some other prosodic constituent. Discerning the correct prosodic constituents and their grammatical (i.e., syntactic) correlates is the subject of the second study, reported in §4.1.3. First, however, I will defend my claim from §4.1.1, that what affects the right edge boundaries, what has been commonly referred to as final devoicing, is better analyzed as aspiration.
4.1.2. Laryngeal features in Blackfoot: voicing versus aspiration The Blackfoot language is typically described as lacking any phonological contrasts based on laryngeal settings; it is typically analyzed as lacking aspiration, voiced obstruents, and the segment [h] (see Elfner 2006 and Frantz 2017). The fact that sonorants appear as voiced and obstruents as voiceless could be the result of redundancy rules (à la Stanley 1967) or phonetic implementation rather than phonological contrast (Keyser & Stevens 2006; Stevens
62 Note that aspiration inside the sentence is still possible, as seen in example (77). However, the target word in the elicitation sentences for study 1 displays little-to-no aspiration when removed from the right edge of the sentence by the word nitsináána ‘my possession’.
Phonologically informed syntax in Blackfoot | 78
& Keyser 2010). However, the fact that vowels do partially devoice at the right edge of a prosodic constituent —what I will ultimately argue is a φ— suggests that the phonological component can manipulate a laryngeal setting to some degree. In this short section, I support my claim that vowel devoicing is the result of aspiration and show that the alternation is phonological in nature, the result of demarcating the right edge of a prosodic constituent. Frantz & Russell (1995: 441) and Frantz (2017: 6) suggest that there is contrast neutralization between phonemically long and short vowels in final position due to final devoicing: (79) Vowel length contrasts (Windsor 2016b: 62) a. áakokaawa ⟦aːk퐨kaːʍḁ ⟧ yáak-okaa-wa FUT-rope.AI-3.SG ‘(S)he will rope.’ b. áakookaawa ⟦aːk퐨ːkaːʍḁ ⟧ yáak-ookaa-wa FUT-sponsor.sundance.AI-3.SG ‘(S)he will sponsor a Sundance.’63 c. nitopi ⟦nitop𝐢̥̥⟧ nit-opi 1.SG-possess.archery.equipment.AI ‘I had a bow and arrow.’ d. nitopii ⟦nitop𝐢̥ː̥⟧ nit-opii 1.SG-sit.AI ‘I sat/stayed.’ e. opiwa ⟦op𝐢̥ʍḁ ⟧ opi-wa possess.archery.equipment.AI-3.SG ‘(S)he had a bow and arrow.’ f. opiiwa ⟦op𝐢̥ːʍḁ ⟧ opii-wa sit.AI-3.SG ‘(S)he sat/stayed.’
63 This verb is translated in Frantz & Russell (1995: 168) as ‘sponsor the primary religious ceremony associated with the Sundance’. I use the abbreviated translation for simplicity (see Frantz 2017: 2, from which examples 79a-b are adapted).
Phonologically informed syntax in Blackfoot | 79
As can be seen in examples (79a-b), contrasts in vowel length signal meaningful differences in Blackfoot. Despite a lexical contrast between the forms in (79c-d), however, Frantz & Russell (1995) report that vowel devoicing in final position neutralizes the perceptual phonetic cues to this contrast. The fact that the contrast does exist at the level of the input can be observed through the addition of morphology to separate the vowel from the right edge, as is the case when the 3.SG morpheme -wa is added to the verbs in (79e-f). The data in the previous section shows that the amount of final vowel devoicing (aspiration) was, on average, 0.1202 seconds, regardless of the phonemic length of the vowel. This finding suggests that the contrast neutralization is not complete, allowing more voicing to persist in phonemically long vowels than in phonemically short vowels (approximately 0.18 seconds of voicing retained in long vowels compared to the approximately 0.08 seconds of voicing retained in short vowels). In Windsor & Cobler (2013), we reported that the average duration of vowel devoicing was roughly equivalent to the average duration of short vowels, which was approximately half the duration of the average long vowel that we observed. In Windsor & Cobler (2013), we included a small number of consonant-final verbs ending in the imperative singular suffix -t to compare final vowels and consonants in both sentence-medial and sentence-final positions (n = 14): (80) Sample elicitation sentences using imperative verbs (Windsor & Cobler 2013)64 a. matóíkskimaat (sentence-final -t) mato-íkskimaa-t go.to.do-hunt.game.AI-IMP.SG ‘Go hunt!’ b. matóíksimaat áwakaasii (sentence-medial -t) mato-íkskimaa-t áwakaasii-yi go.to.do-hunt.game.AI-IMP.SG deer.OBV ‘Go hunt deer!’ We found that sentences consisting of an imperative verb and a direct object (80b) showed little-to-no aspiration on the final [t]. By contrast, an imperative verb with no overt object (80a) showed an almost identical amount of aspiration following the final [t] consonant to the amount of devoicing recorded for final vowels. The average duration of final vowel devoicing was 0.1202 seconds, and the average duration of final consonant
64 These consonant-final forms were elicited in addition to the vowel-final forms exampled in (78) above.
Phonologically informed syntax in Blackfoot | 80 aspiration was 0.1295 seconds. These durations corresponded almost exactly to the average duration of a short vowel (0.12 seconds) and just under half the average duration of long vowels (0.27 seconds):
Average long vowel length
Average short vowel length
Duration of aspiration
Duration of devoicing
0 0.05 0.1 0.15 0.2 0.25 0.3 Seconds
Figure 9: Average durations in Blackfoot Based on the results from that preliminary study, we argued that what appeared to be three separate phonological rules (long vowel reduction, short vowel deletion, and consonant aspiration) could all be explained by the epenthesis of a [SPREAD GLOTTIS] ([SG]) feature at the right edge of a φ, which produces the phonetic realizations in (81).65 (81) Derivation of Blackfoot phrases Input /iːʔniː/ ‘buffalo’ /piːta/ ‘eagle’ /piːt/ ‘enter’ h [SG] epenthesis [iːʔniːʰ]φ [piːtaʰ]φ [piːt ]φ Phonological output [iːʔniːʰ] [piːtaʰ] [piːth] Phonetic realization ⟦iːʔnii͡ ⟧̥ ⟦piːtḁ⟧ ⟦piːtʰ⟧
The phonetic realization of the [SG] feature is the abduction of the vocal folds. To produce aspiration on a consonant, there needs to be a supra-glottal constriction in the vocal tract to build up sub-glottal pressure. When the occlusion is released (e.g., in the case of plosives) with the vocal folds abducted, air is released past the obstruction and the consonant is produced with an additional burst of turbulent noise – aspiration. If this process is instead added to a vowel, there is no additional build up of sub-glottal pressure to produce an extra burst of air, but the abduction of the vocal folds prevents phonation in this case. Oberly
65 These measurements supported the three instructions provided in the lecture material of an introductory Blackfoot language class at the University of Calgary (2010-2015) to aid students in their pronunciations: At the end of a “word,” pronounce long vowels as if short, delete short vowels, and strongly release (aspirate) consonants.
Phonologically informed syntax in Blackfoot | 81
(2008: 125), discussing Southern Ute, states that it is possible to detect a devoiced vowel by the aspiration which appears where the first and second formants would have appeared if the vowels were voiced; this distinction can be seen in the two sample spectrograms in Figures 10 and 11 showing a Blackfoot devoiced vowel and an aspirated consonant respectively:
Figure 10: Final vowel aspiration: Figure 11: Final consonant aspiration apasstamiinaamma ‘apple’ piit! ‘enter!’ In Figure 10, the final portion of the final /a/ vowel in apasstamiinaamma ‘apple’ is selected in the highlighted portion of the spectrogram. Here, the formants of the /a/-vowel persist even though phonation has ceased – exactly where Oberly states vowel aspiration can be found. By contrast, in Figure 11, we see the occlusion and release burst of the final /t/ segment in píít ‘enter’, which is accompanied by high frequency noise resulting from the aspiration of this segment. Although the frequency (Hz) of the aspiration differs between vowels (Figure 10) and consonants (Figure 11), both gestures are analyzed as aspiration. Even though the articulatory gesture that the [SG] feature attaches to may differ, causing slight differences in phonetic realization, both consonantal aspiration and vowel devoicing are achieved through vocal fold abduction – arguably the phonetic realization of the phonological feature, [SG]. The argument that aspiration is responsible for final devoiced vowels in Blackfoot is made even stronger by the fact that both alternations occur in the same phonological environment – at the right edge of (what I will argue in the following section is) a φ.66
66 See Windsor (2016b) for additional arguments for aspiration based on phonological contrast and the Contrastive Hierarchy Theory of phonology (Dresher 2009 et seq.).
Phonologically informed syntax in Blackfoot | 82
Because the alternations in question take place in a particular phonological environment and manipulate a phonological feature to produce the sound changes, I argue that they are truly phonological, not simply phonetic implementation. This phonological alternation, the epenthesis of an [SG] feature, is used to demarcate the right edge of a prosodic constituent. In the following section, I provide evidence to suggest that that prosodic constituent in question is the φ.
4.1.3. Study 2: Intersyllable vowel coalescence and sandhi phenomena Study 1, reported in §4.1.1, concluded that the domain of aspiration in Blackfoot did not necessarily correspond to the orthographic word. This conclusion was based on the fact that certain orthographic words could be observed in different sentential positions either with or without final aspiration. Figure 8 (repeated here as Figure 12 for convenience) showed that the amount of aspiration on a lexeme was significantly different between sentence-medial and sentence-final positions:
Figure 12: Aspiration by position
Phonologically informed syntax in Blackfoot | 83
As a result of the data presented in Study 1, no firm conclusion was reached as to the correct prosodic domain of aspiration. The conjecture in Study 1 was that the domain of aspiration needed to encompass a morphosyntactically complex unit, larger than a word, but smaller than a clause. We saw in the previous section that a sentence consisting of an imperative verb and a direct object represents a single aspiration domain, with aspiration observable only on the direct object, not on the verb: (82) Aspiration domains in imperative constructions a. matóíkskimaat [matɔɪkskimaːtʰ] mato-íkskimaa-t go.to.do-hunt.game.AI-IMP.SG ‘Go hunt!’ b. matóíkskimaat áwakaasii [matɔɪkskimaːt awakaːsii̥ ] mato-íkskimaa-t áwakaasii-yi go.to.do-hunt.game.AI-IMP.SG deer.OBV ‘Go hunt deer!’ The example in (82a) shows that when an imperative verb is in sentence-final position (with no overt direct object), aspiration is realized on the final [t] segment.67 By contrast, when an overt direct object is included in the construction, as in (82b), the imperative verb is removed from the right edge of the sentence, and presents with no aspiration. Instead, the aspiration is realized on the final vowel of the direct object, which is on the right edge of the sentence. These data show that the domain of aspiration includes a morphosyntactically complex construction, minimally consisting of a transitive verb and object DP in the case of (82b). The evidence in (82) may be taken to show that the domain of aspiration epenthesis in Blackfoot is the intonational phrase (ι – matched from the syntactic clause due to the
MATCHCLAUSE constraint provided in §2.4). However, we have also seen evidence of multiple instances of final aspiration within a clause:
67 I specify this statement with “overt” object because there is some evidence to indicate that null subjects and objects are possible in Blackfoot – see Frantz (2017) on attached pronouns, something which could possibly be analyzed as resumptive pronouns indicating a null object. There is no evidence to suggest that Blackfoot prosody is sensitive to null syntactic elements (see Nespor & Vogel 1986 [2007] and Windsor 2012 on (a lack of) prosodic sensitivity to null syntactic outputs).
Phonologically informed syntax in Blackfoot | 84
(83) An example of Blackfoot prosodic constituency (repeated from 77)
In the example above, three instances of final aspiration are indicated by rightward- descending arrows. Data such as that in (83) indicate that the domain of aspiration epenthesis is smaller than the ι, but still large enough to encompass a morphosyntactically complex unit, as in (82b). In the discussion to follow, I present evidence from a second elicitation experiment that correlates the domain of aspiration epenthesis to a φ, matched from syntactic XPs. This elicitation experiment examines both aspiration, as detailed in this section, as well as two processes of vowel coalescence, which I describe presently.
4.1.3.1. Vowel coalescence It is well established that a process of vowel coalescence occurs when a syllable (σ) with no coda consonant comes into contact with a σ lacking an onset in Blackfoot (Elfner 2006; Bliss 2013; Frantz 2017). For example, when an animate noun that ends in /a/ is pluralized with the suffix -iksi, the final /a/ of the noun plus the initial /i/ of the plural is pronounced [ɛ]:68 (84) Coalescence in Blackfoot plurals /imitaː/ + /iksi/ → [imitɛːksi̥ ] dog ANIM.PL dog-ANIM.PL ‘dogs’ A process of vowel coalescence is likewise observed between vowel-final adjective prefixes and vowel-initial nouns that they adjoin to:69
68 Some dialects of Blackfoot show a process of diphthongization in this environment rather than coalescence (Bliss 2013; Frantz 2017). There is no reason to expect that the results I present in this section would be any different for a diphthongization dialect. 69 Bliss (2013) analyzes Blackfoot adjectives as prefixes which are syntactically adjoined to a noun in a process distinct from cliticization, which would linearize them after the noun and all inflectional suffixes in her system. The syntactic assumptions about the attachment of adjectives to a √noun are discussed in §4.2, and revisited in §5.3.
Phonologically informed syntax in Blackfoot | 85
(85) Coalescence between Blackfoot adjective prefixes and nouns /áka/ + /iːmaxhiçkinaːʍḁ / → [akɛ́ːmaxkiçkinaːʍḁ ] old sheep-PROX old-sheep-PROX ‘old sheep’ Similar to the noun + plural example given in (84), the example in (85) shows that vowel coalescence can occur between nouns and their adjectival prefixes. However, the process of coalescence between vowel-final adjectives and vowel-initial nouns is not obligatory in the same way that it is between vowel-final nouns and vowel-initial plural suffixes: (86) A lack of coalescence between Blackfoot adjective prefixes and nouns /áka/ + /iːmaxhiçkinaːʍḁ / → [akaíːmaxkiçkinaːʍḁ ] old sheep-PROX old-sheep-PROX ‘old sheep’ According to my consultant70 (from whom all the data for this second study are drawn), both of the pronunciations provided in (85) and (86) are acceptable. One possible explanation for the optional application of vowel coalescence in the type of construction provided in (85- 86) would be to attribute the application of coalescence to a phonetic sandhi process applying between ωs through overlapping gestures (Quantal Theory: Keyser & Stevens 2006; Stevens & Keyser 2010; cf. Zsiga 1994, 1997, 2000). Russell (1999, 2008), examining another Algonquian language, Cree, shows that optional phonetic processes occur at the boundaries between ωs. Further, Russell argues that optional sandhi phenomena between ωs often mimic obligatory ω-internal processes. However, the phonetic sandhi phenomena do not occur with the same regularity in application as the phonological alternations that occur ω-internally. Russell (2008) cites a variety of phonetic studies that show that such optional sandhi phenomena are most frequently found at the boundaries between ωs (Zsiga 1994, 1997, 2000; Nolan et al. 1996; Ellis & Hardcastle 2002; Ladd & Scobbie 2003; Tserdanelis 2005). Although categorical phonological processes can occur between ωs, Russell (2008) and references therein argue that this environment is where optional phonetic processes are most expected (see also Sadock 1980). Coupling these previous studies with the match theoretic notion, discussed in §2.4, that syntactic heads are expected to be matched as separate ωs in the phonological component informs a hypothesis of Blackfoot prosodic constituency, which I test in the following two
70 The consultant for this study is the eldest of the three speakers used in study 1. The other two speakers were not available to participate in study 2 at the time it was conducted.
Phonologically informed syntax in Blackfoot | 86 sections. I hypothesize, for this study, that nouns (N0s) and their adjectival prefixes (A0s) are matched as separate ωs. If this hypothesis is correct, I predict significant differences in the rates of inter-word coalescence compared to intra-word coalescence, both of which are predicted to occur in different environments from aspiration. Finally, each of these three phenomena are predicted to be correlated to separate morphosyntactic categories (as shown below, aspiration is correlated with the right edge of demonstratives, optional vowel coalescence with the boundary between adjectival prefixes and noun, and obligatory coalescence between nouns and the plural morpheme), which are necessarily matched as different prosodic constituents.
4.1.3.2. Methodology In order to test the hypothesis that adjectival prefixes and nouns represent separate ωs, and to posit a prosodic representation of nominal expressions in Blackfoot, a series of sentences was elicited similar to those in (87), using the same elicitation methodology and equipment reported for Study 1 in §4.1.1. (87) Sample elicitation sentences for study 271 a. anniksi akáímahkihkinaiksi inóókawa [anːiksi̥ akɛːmaxkiçkinɛːksi̥ inoːkaʍḁ ] ~ [anːiksi̥ akaiːmaxkiçkinɛːksi̥ inoːkaʍḁ ] ann-iksi áka-íímachkihkinaa-iksi ino-oka-wa DEM-ANIM.PL old-sheep-ANIM.PL see.TA-INV-21 ‘Those old sheep see us.’ b. anna imitáawa inóokawa [anːḁ imitaːʍḁ inoːkaʍḁ ] ann-wa imitaa-wa ino-oka-wa DEM-PROX dog-PROX see.TA-INV-21 ‘That dog sees us.’ Using elicited sentences such as those in (87), the boundaries between the demonstrative and the (adjective)noun were analyzed for the presence or absence of aspiration and/or vowel coalescence to understand what prosodic boundary existed between these lexemes, and whether or not it was different from (orthographic) word-internal boundaries. The same observations were recorded for the boundaries between adjectives and nouns, and between
71 Note that the example in (87a) contains an adjectival prefix which may be realized with or without vowel coalescence, as indicated by the two possible transcriptions provided for this sentence. No optional coalescences exist for the sentence in (87b), and so only one transcription is provided.
Phonologically informed syntax in Blackfoot | 87 nouns and the plural suffix. A Generalized Linear Model (GLM) was then run on the binary (yes/no) results of this analysis to determine if any of the phonetic observations could be correlated to particular morphosyntactic boundaries. Any pauses that existed between grammatical elements were measured in Praat (Boersma & Weenink 2016), and considered as a possible predictor variable in the GLM. As with the previous uses of GLMs with binomial distribution (in Chapter 3), there are three assumptions of this statistical test: i. that the outcome (i.e., the dependent variable) is dichotomous; ii. that the outcome is a linear combination of the predictor (i.e., independent) variables; and, iii. that the various measures are independent of one another. Because only a single speaker was available for this study, as noted in fn. 70, the third assumption is partially violated in that all tokens are produced by a single speaker, and thus not independent. However, because the phonological effects under investigation are analyzed as belonging to mutually exclusive environments (as discussed below), the predictor variables are necessarily independent of one another. Still, a question may arise as to whether the conclusions of this research apply to a single speaker’s grammar, or can be generalized (at least) to the Kainai dialect of Blackfoot that this speaker belongs to. To address the generalizability issue, subsequent acceptability judgements were gained from a second Elder from that dialect region. The Elder deemed the variations found in the elicited data to be acceptable, providing qualitative support that the conclusions presented below are generalizable beyond a single speaker’s grammar.
4.1.3.3. Results The findings from study 2 are provided in the bar graph in Figure 13.
Phonologically informed syntax in Blackfoot | 88
Aspiration No aspiration Coalesence No coalesence
36 36 36 34 35
27
9 Number Number tokens of 2 0 0 0 1
Dem-(A)N A-N N-PL Syntactic boundary
Figure 13: Phonological effect by boundary Figure 13 shows that 94.4% of demonstratives display aspiration at their right edge (34 of 36 tokens), and 0% of demonstratives display vowel coalescence with an adjacent adjective or noun (0 of 36 tokens).72 By contrast, no aspiration was observed at all between adjectives and nouns, or between nouns and the plural suffix (0 of 74 tokens). At the boundary between nouns and the plural suffix, 97.2% of the tokens displayed vowel coalescence (35 of 36 tokens) with only one token not displaying this sound change. Coalescence at the boundary between adjectives and nouns was found to be probable, but not obligatory; only 75% of the tokens in this category displayed coalescence (27 of 36 tokens), including some tokens which were produced both with and without coalescence between the adjective and the noun. A GLM with binomial distribution was run on the data, considering aspiration, coalescence, and the length of pauses as possible predictor variables for the various morphosyntactic boundaries. In attempting to predict the difference between an adjective- noun boundary as opposed to a demonstrative-(adjective)noun boundary, no significant interactions between these factors were found. The length of pause between grammatical elements at these boundaries was likewise not found to be significant (p = 0.107). However, the main effects of aspiration and coalescence were both significant predictors. In this first GLM, aspiration was found to be a significant predictor of demonstrative-(adjective)noun
72 Tokens were considered to display aspiration if stable formant bands persisted after the end of periodic voicing. The 34 tokens determined to display aspiration all contained roughly equivalent amounts of aspiration to the average duration reported for Study 1.
Phonologically informed syntax in Blackfoot | 89 boundaries: [G2 (1, 70) = 84.143, p < 0.001]. Vowel coalescence, on the other hand, was found to be a significant predictor of adjective-noun boundaries: [G2 (1, 69) = 5.24, p = 0.022]. The same test was conducted to predict the boundaries between nouns and the plural suffix. This subsequent test also yielded vowel coalescence as a significant predictor of noun-plural boundaries: [G2 (1, 69) = 11.852, p < 0.001]. A two-sample test of proportions was completed on the patterns of vowel coalescence found between the adjective-noun boundaries and the noun-plural boundaries with the following result: [z = 2.7226; p = 0.007]. The test of proportions shows that the ability for each of the morphosyntactic boundaries to be predicted by vowel coalescence is significantly different between the two grammatical correlates. In other words, vowel coalescence obligatorily applies between the noun and the plural morpheme, but is optional between adjectival prefixes and the nouns that they attach to. The results of the predictive modeling in the above GLMs provide evidence for three distinct levels of phonological constituency in Blackfoot nominal expressions. These conclusions are summarized in Table 8. Syntactic Hypothesized Right-edge Coalescence acceptable, Obligatory boundary prosodic boundary aspiration but not obligatory coalescence ✓ Dem-(A)N φ (p < 0.001) ✓ A-N ω (p = 0.022) ✓ N-PL σ (p < 0.001) Table 8: Prosodic boundary cues by morphosyntactic category The data examined in this section provide evidence that the demonstrative (as well as the entire nominal expression itself in these examples) is prosodically matched as a φ; this suggests that the demonstrative element is phrasal in the underlying syntactic structure. Adjectives and nouns are each matched as ωs, suggesting that these elements are heads in the underlying syntactic structure. Finally, the plural suffix does not appear to be matched from an underlying syntactic structure, surfacing prosodically as a σ rather than as a ω. In the following section, I argue that the prosodic evidence explored here motivates a distinction between prefixes, which are argued to attach to a noun through syntactic processes (Bliss’ 2013 syntactic linearization algorithms are discussed subsequently and revisited in §5.3), and suffixes, which are argued to result from agreement relationships (see
Phonologically informed syntax in Blackfoot | 90
Russell 1999 for similar conclusions based on evidence from the related Algonquian language, Cree).
4.2. Uncovering the underlying syntactic structure In this section, I return to the overall goal of the dissertation: Using the phonology-syntax interface as a research tool. In the previous sections, I have motivated an analysis of Blackfoot prosodic constituency. In this section, I test that prosodic constituency against the predicted output of the syntax for Blackfoot nominal expressions provided in Bliss (2013) and Wiltschko (2014). I show that the prosodic constituency motivated in the previous section provides a basis to refute an analysis of Blackfoot suffixation through successive head-movement. Instead, I offer a possible alternative analysis for suffixation via agreement relationships. The alternative analysis suggested in this section is further motivated in Chapter 5 using additional evidence from the linearization of demonstratives, definite articles, and adjectives. Blackfoot is an agglutinating language, allowing multiple affixes to attach to a root: (88) Blackfoot nominal agglutination a. kitomitaamiksi kit-wimitaa-m-iksi 2.SG-dog-POSS-ANIM.PL ‘your dogs’ b. kitomahkomitaamiksi kit-omahk-wimitaa-m-iksi 2.SG-big-dog-POSS-ANIM.PL ‘your big dogs’ To analyze the underlying structure of Blackfoot affixation, Bliss (2013) and Wiltschko (2014) utilize relative syntactic height as a diagnostic tool. This tool relies on Baker’s (1985) Mirror Principle (repeated below as 89 for convenience): (89) The Mirror Principle (Baker 1985: 375) Morphological derivations must directly reflect syntactic derivations (and vice versa) This principle is adopted as part of the theoretical underpinnings of Wiltschko’s (2014) Universal Spine Hypothesis framework, which Bliss and Wiltschko each adopt for their analyses. This assumption allows Bliss (2013: 12) to develop a linearization algorithm for Blackfoot suffixation which relies on successive head-movement to order suffixes relative
Phonologically informed syntax in Blackfoot | 91 to a root. She argues that the closer an affix is to the root, the lower it must associate with the syntactic spine: (90) Blackfoot suffix linearization algorithm (Bliss 2013: 12) a.
Suffix3
Suffix2
Suffix1
Root b. [ Root – Suffix1 – Suffix2 – Suffix 3 ]
According to the algorithm above, Bliss derives the sequence √noun > possessive > plural (as in 91 below) by movement of the N0 to n0 (where she argues the possessive -m is located).73 Subsequently, the incorporated n0 then moves to the Φ0 (where she argues that the plural suffix -iksi is associated with the syntactic spine). This successive head-movement operation correctly predicts the order of morphemes as N > n > Φ in an example such as ‘your dogs’:74
73 The possessive suffix -m indicates alienable possession. This suffix co-occurs with a person prefix that indicates the possessor (e.g., kit- in example 91). 74 Bliss (2013: 117) analyzes the person prefix kit- in (91) as a DP in Spec,nP. I Merge it in the D0 for the purposes of linear order only. The analysis does not hinge on the location of this morpheme since the algorithm that Bliss develops for prefixes is different from that of suffixes (Bliss 2013: 14). See Bliss & Gruber (2015) for a full syntactic and semantic analysis of these forms.
Phonologically informed syntax in Blackfoot | 92
(91) Linearization of suffixes in possessed animate plurals a. kitomitaamiksi kit-omitaa-m-iksi 2.SG-dog-POSS-ANIM.PL ‘your dogs’ b.
In the above example, the N0 omitaa undergoes head-movement to the n0 containing the possessive suffix -m. The complex n0 then undergoes successive head-movement to the Φ0 which is argued to contain the animate plural suffix -iksi. A similar analysis can be applied to the data used in the studies reported previously in this chapter, once we understand Bliss’ treatment of adjectives. Bliss (2013: 118) suggests that adjectives attach to the NP layer in Blackfoot, as in her example (87-88) adapted as (92) here: (92) Structure of Blackfoot adjectives (Bliss 2013: 118) a. kitomahkonssta kit-omahk-w-insst-wa 2-big-3-sister-prox ‘your big sister’ (pejorative connotation, i.e., obese) b.
Bliss’ analysis of the position of adjectives stems from the relative height of the adjective compared to the long-form person prefix (kit-) that she had previously argued to be in
Phonologically informed syntax in Blackfoot | 93
Spec,nP, and the short-form person prefix (w-) that she argued to be a ΦP in Spec,NP.75 Bliss footnotes another example that shows that adjectives can also intercede between the short- form person prefix (n- in example 93) and the noun: (93) Alternate linearization of Blackfoot adjectives (Bliss 2013: 118) nissoko’siksi n-iss-o’kos-iksi 1-young-offspring-ANIM.PL ‘my grandchildren’ Because the adjective iss ‘young’ appears between the prefix n- (argued to be in Spec,NP) and the √noun o’kos, Bliss suggests that adjectives in Blackfoot can have multiple attachment sites, allowing for both NP- and N'-adjunction, a topic she leaves for future research. However, neither of the examples in (92) or (93) allow for the movement of the √noun, which she argued to be necessary to achieve suffixation in (91). For example, the √noun in (93) must minimally raise to Φ0 to gain the -iksi suffix to mark it as animate plural under that analysis. However, assuming that Bliss’ conjecture is correct and that there are multiple possible sites for adjective adjunction —and there is no reason to doubt this notion— the problem could be solved by assuming a higher adjunction site, such as nP or ΦP. Allowing for this slight modification to Bliss’ analysis, we can draw the structure for the nominal expressions used in Study 2 earlier in this chapter:76
75 Bliss (2013) (see also Ritter & Rosen 2010 and Bliss & Gruber 2015) argues that the different forms of the person prefix (and their different syntactic position) is due to differences in alienability. Short-form person prefixes indicate inalienable possession (e.g., familial relations) and do not appear with a possessive -m suffix. Long-form person prefixes indicate alienable possession (e.g., pets or material possessions) and do (normally) appear with possessive -m suffix. 76 For now, I assume Bliss’ analysis that demonstrative phrases (DemP) are in Spec,DP, and later raise to a higher functional projection. The subsequent raising of the demonstrative to Spec,KP/Spec,LinkP does not affect the present analysis. I return to this topic in Chapter 5, which specifically deals with demonstratives.
Phonologically informed syntax in Blackfoot | 94
(94) Structure of the Blackfoot nominal expression (preliminary) a. anniksi akáímahkihkinaiksi (inóókawa) ann-iksi áka-íímahkihkinaa-iksi (ino-oka-wa) DEM-ANIM.PL old-sheep-ANIM.PL (see.TA-INV-21) ‘Those old sheep see us.’ b.
Given what we know about the syntax-phonology interface, let’s consider the prosodic structure predicted by the morphosyntactic structure in (94), where XPs containing pronounceable material are matched as φs, and X0s containing pronounceable material are matched as ωs: (95) Unattested matching structure of (94) φ ← DP
φ ← DemP φ ← ΦP
anniksi ω ← Adj ω ← Φ aka- ω ← N ω ← Φ íímachkihkinaa -iksi In the above structure, I have not matched separate φs to the two ΦPs of the structure in (94) since they are part of the same extended projection. This creates a structure more in line with the prosodic facts examined previously, which suggest that the adjective and √noun are ωs in the prosodic representation, and do not have an interceding φ boundary. However, there is still a mismatch between the structure in (95) and the prosodic structure evidenced in the previous section; that is the status of the plural morpheme, -iksi.
Phonologically informed syntax in Blackfoot | 95
In the previous section, I argued that optional phonetic processes of sandhi (which may mimic ω-internal categorical processes) most frequently occur at the boundaries between ωs. Such a process was found at the boundaries between adjectives and the nouns that they attach to, which was the domain of optional vowel coalescence. The ability for a between-ω process to occur between adjective prefixes and nouns is captured by the representation in (95). However, the representation in (95) additionally predicts that the plural morpheme -iksi should also project its own ω, since it was argued to be a syntactic head by Bliss (2013). However, this prediction was not borne out in the data presented in §4.1.3.3. In the following section, I explore possible explanations for this mismatch. In the end, however, the data point to an alternate syntactic representation rather than a phonological repair strategy.
4.2.1. Phonological representation of Blackfoot affixes Because Match Theory is a theory of indirect reference, phonological constraints can refer only to phonological structure, not morphosyntactic structure. Truckenbrodt’s Wrap-XP theory (detailed in §2.3) was also a theory of indirect reference. In Truckenbrodt’s theory, representing prosodic categories for functional elements was avoided by making the labels invisible to the phonological component through the LCC (Lexical Category Condition), repeated as (96) for convenience: (96) Lexical Category Condition (Truckenbrodt 1999: 226) Constraints relating syntactic and prosodic categories apply to lexical syntactic elements and their projections, but not to functional elements and their projections, or to empty syntactic elements and their projections By contrast, Match Theory has no such mechanism to prohibit the matching of functional syntactic heads. Lee & Selkirk (2013) allow for separate constraints to match functional versus lexical material (e.g., MATCHWORDLEX and MATCHWORDFUNC). However, it is not clear how the phonological component could identify lexical versus functional heads without making reference to the morphosyntax. I follow Elfner (2012: 20, fn. 7) who assumes that “both lexical and functional projections must be matched […] in contrast to the proposal of Selkirk & Shen (1990), Selkirk (1995), Truckenbrodt (1995, 1999), and others.” Elfner provides evidence from Irish constituent structure to show that both lexical and functional elements are evaluated by constraints such as MATCHWORD and MATCHPHRASE, finding no difference between the realization of lexical projections such as VP and functional
Phonologically informed syntax in Blackfoot | 96 projections such as DP. Similarly, in Blackfoot, to argue for a distinction between functional and lexical categories, one would have to explain why demonstratives are matched in the prosodic constituent structure, but plurals (Φ0) are not. Because MATCHWORD must equally evaluate the Φ0 and any other X0, the only way to prevent -iksi from being matched as a ω —assuming the syntactic representation in (94)— is to appeal to well-formedness constraints.
In §3.3, I discussed the well-formedness constraint BIN-MIN(φ) which was a higher- ranking constraint than MATCHPHRASE in Irish. This crucial ranking caused non-binary APs in Irish to be matched as ωs, rather than as φs, as shown in examples (97) and (98), repeated from (53) and (54). (97) Unattested mapping from simultaneously maximal/minimal A to φ AP/AMax *φ | → | A ω áille áille ‘beautiful’ (Elfner 2015: 1179) (98) Attested mapping from simultaneously maximal/minimal A to ω AP/AMax | A → ω áille áille ‘beautiful’ (Elfner 2015: 1179) Likewise, ωs have similar constraints which require a defined prosodic weight in order to be well-formed. In Windsor (2012), I argued for the constraint WORDSIZE(μμ) to account for minimal words and tense/lax vowel distinctions in English as well as reduplication in Old Irish (see also McCarthy & Prince 1986 [1996] for a pre-Optimality Theory use of this constraint):
(99) WORDSIZE(μμ) (Windsor 2012: 48) A prosodic word must minimally contain two mora (μ) This constraint can also be used to evaluate the Blackfoot plural suffix -iksi to check if it meets minimal ω-weight restrictions. If it does not meet minimal weight restrictions, it cannot be realized as a ω in prosodic constituent structure. Consider the structure of -iksi provided in (100) below:
Phonologically informed syntax in Blackfoot | 97
(100) The prosodic structure of the Blackfoot plural -iksi σ σ
μ μ
i k͡ si (‘ANIM.PL’) As can be seen in the representation in (100), -iksi consists of two μs and two σs. This form contains the same prosodic weight as the adjective used in (94) and (95), which was shown to be a ω in prosodic constituent structure:77 (101) The prosodic structure of áka ‘old’78 ω
σ σ
μ μ
a ka (‘old’) By comparing the representations in (100) and (101), we can see that no prosodic weight differences exist between the two grammatical forms. Therefore, we cannot appeal to a constraint such as WORDSIZE(μμ) to cause the adjective to be matched as a ω but not so the plural. In sum, I argue that the prosodic facts observed do not support the hypothesis that the plural suffix -iksi is part of a recursive (i.e., incorporated) head structure in the syntax before spell-out (contra Bliss 2013 and Wiltschko 2014) – a similar conclusion to that reached by Russell (1999) for Cree. In the following section, I propose an alternate analysis of the structure of Blackfoot nominal expressions – one that fits with the phonological facts examined in this chapter.
4.2.2. A hypothesis on Blackfoot suffixation The Blackfoot phonological data examined thus far have shown: i. That the right edges of phonological phrases (φ) are demarcated with aspiration; ii. that the boundaries between prosodic words (ω) display optional vowel coalescence in a sandhi process that mimics ω-
77 The metrical foot (Σ) is not yet well understood in Blackfoot, and so I omit it from this representation for simplicity. This omission does not effect the ω-status of the representation as this does not change the prosodic weight of the form in question. See Weber (2016a, b) on the role of this prosodic unit in Blackfoot. 78 Note that the accent over the vowel in áka ‘old’ indicates a pitch accent, not a long vowel.
Phonologically informed syntax in Blackfoot | 98 internal phonological processes; and, iii. that adjacent vowels not separated by ω-boundaries undergo obligatory vowel coalescence. These three phenomena were correlated to three morphosyntactic boundaries such that: i. Aspiration was found at the right edge of demonstratives; ii. optional vowel coalescence was found between adjectival prefixes and the nouns they attach to; and, iii. obligatory vowel coalescence was found between nominal roots and the plural suffix. Finally, no prosodic constraints were found to be able to force the non-isomorphism between the two components such that the A0 containing the adjective was matched as a ω but the Φ0 containing the plural was not. This phonological analysis has led to the conclusion that the plural suffix cannot be attached to the nominal root by head- movement and incorporation (pace. Bliss 2013 and Wiltschko 2014). Otherwise —like adjectival prefixes— the plural would display the sound change associated with adjacent ωs, rather than the sound change associated with adjacent syllables (σ). Similar phonological conclusions were reached by Russell (1999), investigating two polysynthetic languages: Dakota, a Siouan language, and Cree. Russell (1999) concluds that the two languages show a distinct patterning for what he determines are syntactic processes versus morphological processes. He argues that in Cree, syntactic processes (i.e., head- movement) attach morphemes to the root as prefixes (as suffixes in Dakota) and morphological processes (discussed below) attach morphemes to the root as suffixes (as prefixes in Dakota). The key phonological difference between the two underlying processes (in both languages) is that affixation by syntactic processes yields separate ω structures, while affixation by morphological processes are contained within the same ω as the root. Here, I use Russell’s conclusions for Cree as a starting point to suggest a new hypothesis about the morphosyntactic structure of Blackfoot nominal expressions. I start by re- examining the syntactic structure suggested in §4.2 that incorrectly predicted suffixes would be contained within their own ωs:
Phonologically informed syntax in Blackfoot | 99
(102) Structure of the Blackfoot nominal expression (preliminary) a. anniksi akáímahkihkinaiksi (inóókawa) ann-iksi áka-íímachkihkinaa-iksi (ino-oka-wa) DEM-ANIM.PL old-sheep-ANIM.PL (see.TA-INV-21) ‘Those old sheep see us.’ b.
The above structure (repeated from 94) was matched as the phonological representation in (95), repeated as (103) here: (103) Unattested matching structure of (102) φ ← DP
φ ← DemP φ ← ΦP
anniksi ω ← Adj ω ← Φ aka- ω ← N ω ← Φ íímahkihkinaa -iksi We know from the previous discussion that the prosodic structure given in (103) does not fit the phonological data examined in this chapter. The repair I propose, following Russell (1999), is to treat suffixation in Blackfoot as a morphological process, attaching suffixes to the root through an Agree relationship rather than via successive head-movement. The resulting syntactic structure for the nominal expression in (102) would then be as in (104) below:
Phonologically informed syntax in Blackfoot | 100
(104) Structure of the Blackfoot nominal expression (second attempt)79 a. anniksi akáímahkihkinaiksi (inóókawa) ann-iksi áka-íímahkihkinaa-iksi (ino-oka-wa) DEM-ANIM.PL old-sheep-ANIM.PL (see.TA-INV-21) ‘Those old sheep see us.’ b.
In the structure provided in (104), I assume, following Bliss (2013), that N-to-n movement is basic – presumably for nominal classification (see Wiltschko 2014 for discussion). The √noun then agrees with the classifying head (n0) to value the [uANIM] (gender)80 feature as [animate], and agrees with the Φ0 to value the [uNUM] feature as [plural]. The corresponding pronunciations for the feature bundles are then selected, and spelled out on the syntactic head containing the lexeme, predicting that the root and all suffixes are matched as a single ω – which has already been shown to be the case. It should be noted that only the position of the adjective in (104) suggests that n-to-Φ movement is not basic (i.e., driven by strong uninterpretable features) in the same sense as I assume N-to-n to be. However, the position of the adjective, as discussed previously, is not a foregone conclusion. If it is accepted that adjectives may adjoin to ΦP —a repair proposed
79 The location of the DemP will be further explored in Chapter 5, it is inconsequential for the present analysis. As with the previous discussion of this structure, I continue to assume multiple possible adjunction sites for the adjective. The analysis does not hinge on whether the adjective is adjoined to nP, ΦP, or some other position. 80 Here, I adopt the standard categorization for these morphemes (cf. Ritter 2014, who challenges the majority view of animacy as gender in Blackfoot).
Phonologically informed syntax in Blackfoot | 101 here only as a way to account for the linearization patterns along with the raising analysis provided in Bliss (2013)— then the argument for n-to-Φ movement could be argued to be basic as well. At present, though, I am not aware of any analyses that conclude that ΦP- attachment is the position of the adjective in Blackfoot. If we compute the phonological structure for the nominal expression given in (104b) based on the matching constraints assumed in this dissertation, we arrive at the structure in (105), which accurately predicts the sound alternations found to demarcate the various prosodic categories earlier in this chapter: (105) Matching structure of (104) φ ← DP
φ ← DemP φ ← nP
anniksi ω ← Adj ω ← N aka- íímachkihkinaiksi The prosodic matching structure in (105) predicts that a [SPREAD GLOTTIS] feature ([SG]) will be epenthesized at the right edge of the demonstrative and the right edge of the noun. Because both of these elements end in a vowel, that [SG] feature will be phonetically realized as a devoiced vowel. It further predicts that the boundary between the adjectival prefix and the noun can be variably realized with coalescence as [ɛː] or without coalescence as [a.iː] since this is the boundary between two ωs. Finally, this structure predicts that the final /aː/ of the nominal root combined with the initial /i/ of the plural morpheme will be obligatorily realized as [ɛː]. According to the data in Study 2, all of these predictions are borne out: (106) Sample elicitation sentence for study 2 (repeated from 87) anniksi akáímahkihkinaiksi inóókawa [anːiksi̥ akɛːmaxkiçkinɛːksi̥ inoːkaʍḁ ] ~ [anːiksi̥ akaiːmaxkiçkinɛːksi̥ inoːkaʍḁ ] ann-iksi áka-íímachkihkinaa-iksi ino-oka-wa DEM-ANIM.PL old-sheep-ANIM.PL see.TA-INV-21 ‘Those old sheep see us.’ The observable phonological alternations in (106) align with the predicted matching structure of a syntactic spine that achieves suffixation via agreement and realizes the inflectional morphology under a single head, rather than by successive head-movement. The syntactic derivation that achieves suffixation through successive head-movement predicts
Phonologically informed syntax in Blackfoot | 102 that affixes (provided they meet minimal prosodic weight restrictions) will be matched as a ω, and therefore display optional vowel coalescence. In the data examined in this chapter, the phonological predictions by the structure assuming head-movement are not borne out. These empirical observations suggest that suffixation in Blackfoot is achieved through agreement, rather than head-movement. At present, though, this is conjecture, not a firm conclusion. There is a plausible alternative analysis that makes sense of inflectional morphology being realized on a single syntactic head through head-movement that I will briefly outline here. It is possible that uninterpretable features in the morphosyntactic component could be checked through head-movement but, rather than spelling out the suffixes on separate incorporated X0s, the exponence of those features is realized on the head where all the features are valued. A structure for such a hypothesis is provided in (107) using the same nominal expression as (104) above: (107) Structure of a feature valuation through head-movement analysis
As discussed previously, this analysis makes use of the adjective attaching to ΦP. In this structure, successive head-movement allows the uninterpretable features of the √noun to be valued by the features assumed to be Merged in n and Φ, but allows the observable prosodic effects to be successfully predicted by spelling out the morphology out on a single head: the N0.
Phonologically informed syntax in Blackfoot | 103
The phonological data explored in this chapter do not rule out the structure in (107) and thus far cannot be used to decide between that structure and the one provided in (104). However, in the following chapter where I analyze the organization of demonstratives (prosodic and morphosyntactic), I present additional linearization evidence that makes a head-movement analysis difficult to maintain. This evidence comes in part from the linearization of definite articles, which intercede between the demonstrative and the adjective + noun, making it unlikely that the noun raises outside of the DP to gain a proximate -wa or obviative -yi suffix (pace. Bliss’ 2013 and Wiltschko’s 2014 head- movement analysis): (108) Linearization of syntactic elements in the Blackfoot nominal expression amo na omahkómitaawa (áóhkiwa) amo-wa na omahk-wimitaaa-wa á-ohki-wa DEM-PROX DEF big-dog-PROX DUR-bark.AI-3.SG ‘That big dog (is barking).’ If the definite article in (108) can be assumed to be in D0, and the adjective omahk- ‘big’ is as high as ΦP, it is difficult to conceive how the noun, with the bound adjectival prefix, might raise outside of the DP to Link0 – where Bliss (2013) and Wiltschko (2014) argue it must move to gain the proximate -wa suffix. These additional arguments against successive head-movement in Blackfoot are elucidated in §5.3 where I analyze demonstrative constructions, which have been previously argued to undergo word-order movement around the raised nominal in Link0. Before examining demonstrative constructions, though, I first recap what can be concluded based on the evidence in this chapter and comment on the consequences of the hypothesized Blackfoot structure for the mirror principle, which was an assumption behind the head- movement analysis.
4.3. Summary and discussion In this chapter, I have detailed tests done on elicited speech data used to uncover prosodic constituency in Blackfoot. From those tests, I found: i. that the right edges of phonological phrases (φ) could be predicted by right-edge aspiration (aspirated consonants in the case of imperative verbs, or devoiced vowels in most other contexts); ii. that inter-word (ω) boundaries presented with optional vowel coalescence; and, iii. that intra-ω boundaries presented with obligatory vowel coalescence.
Phonologically informed syntax in Blackfoot | 104
The conclusions from the tests of prosodic constituent structure were then evaluated against a hypothesis of syntactic suffixation through successive head-movement. By matching the syntactic spell-out of the structure that achieves suffixation through head- movement (à la Bliss 2013 and Wiltschko 2014), I determined that the animate plural morpheme -iksi was predicted to be matched as a ω in prosodic constituent structure. However, I found no phonological evidence to suggest that the suffix -iksi is a ω in prosodic constituent structure, and there were no phonological well-formedness constraints that could motivate a syntax-prosody mismatch. This apparent mismatch provides evidence that the underlying syntactic structure, from which the prosody is matched, may have a different representation – one that does not achieve suffixation through head-movement before spell- out. I offered an alternative head-movement analysis whereby feature valuation could still occur through successive head-movement with the exponence of those features being realized on a single X0 and thus accurately predicting the attested phonological patterns in pronunciation. However, this alternative analysis is difficult to maintain in light of evidence from the linearization of definite articles. I provided one such example of determiner linearization above, and return to that line of argumentation in Chapter 5. There are two primary benefits to the hypothesis advanced in this chapter. First, it is consistent with previous analyses of Blackfoot syntax. It remains consistent by assuming that the various morphosyntactic features are still provided by the same functional heads motivated in previous syntactic studies (e.g., Bliss 2013 and Wiltschko 2014). Second, this analysis utilizes cross-component evidence from phonetic measurement, phonological constituent structure, and morphosyntactic argumentation to provide a conclusion that fits with the evidence from each component. If this hypothesis is accepted, it does have consequences for the mirror principle, which I discuss now. I provide further evidence for this proposal in Chapter 5.
4.3.1. On the Mirror Principle At the beginning of this chapter, I stated that the prosodic evidence gained herein would provide a rationale to reject an analysis of Blackfoot suffixation through successive head- movement that was achieved via the Mirror Principle, repeated here as (109). (109) The Mirror Principle (Baker 1985: 375) Morphological derivations must directly reflect syntactic derivations (and vice versa)
Phonologically informed syntax in Blackfoot | 105
The evidence presented in this chapter does not directly contradict this principle in its strict interpretation. The mirror principle rests on two morphological assumptions. First, it assumes no morphological distinction between inflection and derivation. Second, it assumes that morphological operations are ordered and cyclic, i.e., that affixation occurs in a specific order, attaching a single affix at a time. Baker (1985: 377) provides the following example using the English word derivationally to illustrate this assumption, stating that the word has the layered structure in (110a), not the flat structure of (110b), nor an arbitrary binary branching structure as in (110c). (110) Composition of derivationally (Baker 1985: 377) a. [[[[derive] ation] al] ly] b. [derive + ation + al + ly] c. [[derive [[ation] al]] ly]
Baker adds that, given those two assumptions, we are naturally led to a simple conclusion: Affixes that appear closer to the root are attached before affixes further from the root. The mirror principle, then, contends that the syntactic process associated with each of the affixes must occur in the same order as morphological affixation. The mirror principle also makes several syntactic assumptions, which I update here to reflect current generative syntactic theory where I am able to do so. The mirror principle assumes (at least) two levels of syntactic structure – an initial Merge location and a final Merge location as the result of movement/re-Merge. The principle also assumes an interface with semantic-conceptual structure that we will not be concerned with here. Importantly for the present discussion, the syntactic assumptions of the mirror principle do not rely on movement operations, but could also allow for agreement relations, the latter of which I argue is responsible for Blackfoot suffixation. Provided that an additional assumption is permitted, the Blackfoot data explored in this chapter are not contradictory to the mirror principle. The necessary assumption is that syntactic derivation may be interpreted as applying to the Agree operation, not only movement (or re-Merge) operation. This assumption allows for the relative height diagnostic, as given in Wiltschko (2014), to remain a useful tool for syntactic analysis, but it requires that movement versus agreement be diagnosed by other means (such as through the prosodic evidence presented in this chapter).
Phonologically informed syntax in Blackfoot | 106
If the mirror principle allows for this additional assumption, it continues to be a useful tool for diagnosing which syntactic heads provide which morphosyntactic features. The cross-component data from Blackfoot presented in this chapter, however, show that we cannot assume movement/re-Merge operations are necessary via this principle, and that we need additional tests to motivate Move versus Agree. In the case of Blackfoot, I have shown how these tests may be conducted by looking at the language from a phonological perspective, and then relating those structures back to the underlying syntactic constituent structure.
4.4. Interim conclusion The goal of this dissertation is to show how cross-component evidence can be utilized as a tool for conducting research in one grammatical component or to inform analysis in another. This cross-component evidence can allow for more controlled testing of a particular linguistic element that is the topic of study, and it can also provide independent evidence in support of an analysis supported by component-internal arguments. In Chapter 3, I showed how knowledge of an underlying syntactic structure from previous syntactic analyses allowed me to isolate a lexeme both with and without phrasal accenting in the same position within the breath group. By utilizing the knowledge of syntactic constituent structure, I gained an understanding of how phrasal accents and word stress were realized in different dialects of Irish. Further, understanding the differences in prominence realization in the dialects motivated a diachronic analysis of the catalyst behind a stress-shift phenomenon in one of those dialects. These phonological analyses would not have been possible without an understanding of the syntactic constituent structure that allowed me to control for prosodic constituent structure. In the current chapter, I began with an observation of “word”-final sound alternations in Blackfoot, which led to an analysis of the prosodic constituent structure in the language. Once I better understood the prosodic constituent structure of Blackfoot, I was able to motivate an underlying syntactic structure that the prosody matched. The evidence gained in this study suggested that suffixation in Blackfoot may not be the result of successive head- movement operations as previously suggested in the literature. I argued that both the phonological and morphosyntactic evidence available could be understood if the syntactic
Phonologically informed syntax in Blackfoot | 107 operation responsible for suffixation in Blackfoot were agreement relations, rather than movement operations, and will provide further evidence in support of this hypothesis in the following chapter. The evidence presented in the last two chapters was not meant to provide a definitive conclusion regarding the constituent structures of Irish and Blackfoot. Rather, the discussions in these chapters were meant to illustrate how the interface could be used as additional evidence towards conclusions about constituent structures. I have, instead, provided converging evidence from the morphosyntactic component as well as the prosodic component to support a hypothesis that can only be achieved via cross-component data. The studies thus far have highlighted how evidence from one component can be used to inform study in the other. In an ideal scenario, though, the evidence from each component can be used in tandem to reinforce one another to yield an empirically-supported conclusion. In the following chapter, I present two ideal test cases, using data already introduced, to analyze the structure of demonstrative elements in both Irish and Blackfoot. The two languages are from separate language families and have different structures and diagnostics, but each can be researched using both prosodic and morphosyntactic evidence in unison. fin
The demonstrative in Irish and Blackfoot | 108
5. The demonstrative in Irish and Blackfoot In the preceding chapters, we have explored the hierarchical structures of nominal expressions at the interface between syntax and phonology. By utilizing a knowledge of syntactic structure at spell-out, we have gained insight into how prominence is instantiated in two Irish dialects, and motivated a possible origin of a stress-shift phenomenon in Munster Irish. Additionally, investigating the phonetic correlates of prosodic boundaries in Blackfoot has provided evidence for agglutinating patterns through agreement relationships rather than successive head-movement in that language (additional evidence to be presented in this chapter). Each of the preceding discussions of Irish and Blackfoot nominal expressions has utilized demonstrative elements in the analysis —either to elicit a particular prosodic prominence structure in Irish, or to motivate the phonetic correlates of prosodic phrases (φ) in Blackfoot— but these syntactic elements have not been the primary focus of discussion until now. The first problem I highlighted in this dissertation, that the syntax-phonology interface could provide insight into, was deciding between two opposing analyses on the structural position of demonstratives in Celtic. Demonstrative elements have become a topic of much inquiry within the syntactic literature over the past few decades with many authors contributing to questions such as: Do demonstratives represent a homogenous cross- linguistic category; should all demonstratives receive the same analysis; do all demonstratives serve the same syntactic functions; and, where do demonstratives connect to the syntactic spine? (Giusti 1992, 1996, 2015; Cinque 1994, 2005; Szabolcsi 1994; Bernstein 1997, 2008; Diessel 1999, 2006; Lyons 1999; Rosen 2003; McCloskey 2004; Leu 2008; Wiltschko 2009, 2014; Adger 2013; Bliss 2013; Roehrs 2013; Lewis 2014; Windsor 2014a, 2016b, 2017, to appear; Windsor & Lewis to appear; Roberts to appear, among many others). In this chapter, I explore what the evidence from the interface between phonology and syntax can teach us about the constituency and composition of demonstratives (prosodic and syntactic) in the unrelated languages of Irish (Celtic/Indo-European) and Blackfoot (Algonquian). I begin by first discussing what a demonstrative is, and what it has in common between the two languages of under discussion.
The demonstrative in Irish and Blackfoot | 109
5.1. Demonstratives Diessel (1999, 2006) provides four facts about the category of demonstratives cross- linguistically: i. Demonstratives are among the first words acquired during L1 acquisition; ii. they are universally present in all languages; iii. they appear to be very ancient roots that do not derive from other words, but rather are frequently the roots from which other lexemes (e.g., articles and adjectives) are derived; and, iv. unlike other syntactic categories, they can be linked to a particular gesture. One of these generalizations, however, is controversial in the syntactic literature, namely, what is universal about demonstratives (i.e., do they belong to the same category, associate with the same position of the syntactic spine, or perform the same syntactic function cross-linguistically)? 81 Through the study of two languages, I cannot reach any conclusions about universal structures. However, since the discovery of typological universals and the range and nature of cross-linguistic variation are goals of syntactic research (Chomsky 1965 et seq.; Chomsky & Lasnik 1993; Wiltschko 2014), the evidence explored herein will suggest a hypothesis about a possible universal structure and make predictions for future cross-linguistic study to support or reject that hypothesis. In discussing the universality of demonstratives, some authors such as Giusti (2015) argue that demonstratives universally occupy the specifier position of DP where they realize the referential index of a nominal expression. By contrast, Sybesma & Sio (2008) argue for differences in nominal expressions between Chinese languages and the Tai language, Zhuang, based on differences in the syntactic organization of demonstratives. They argue that the demonstrative represents a phrase raised to pre-nominal position in Chinese languages, but a post-nominal head that does not undergo movement in Zhuang. Irish and Blackfoot show similar differences in realization: The Blackfoot demonstrative is obligatorily pre-nominal and the Irish demonstrative is obligatorily post-nominal. To suggest any commonalities between these two languages, I provide a brief background on demonstrative constructions in both, and highlight what traits they have in common, and where the differences lay.
81 The composition of demonstratives, whether they are an individual root or composed of multiple syntactic roots (e.g., a D0 plus an A0) is also not agreed upon within the literature. I do not focus on this issue in this chapter, but do provide some evidence against the compositional hypothesis based on Irish. See Windsor (2014a) for additional arguments against the analysis that demonstratives are compositional roots consisting of a determiner plus an adjective based on further evidence from Irish.
The demonstrative in Irish and Blackfoot | 110
5.1.1. Demonstratives in Irish and Blackfoot – background information Irish has a three-demonstrative system that provides a contrast between proximal and distal demonstratives, and within the distal system, between visible and invisible referents: (111) Irish demonstrative interpretations (Windsor 2016a: 4) DEM
PROX DIST seo VIS INVIS sin siúd/úd82 Demonstratives in Irish obligatorily select for a definite DP and may not appear with an indefinite:83 The definite DP that allows a demonstrative to be merged can take the form of a DP with a definite article (112b), a pronoun (112e), a vocative (112f), or a proper noun (112g). 84
82 Ó Siadhail (1995: 36) notes that úd is used with nouns, and siúd is used with pronouns. McCloskey (2004:2) includes the form udaí among the most distal demonstratives; it is a variant form found in some dialects, but not in Connemara Irish, from which most of the data in this chapter are drawn. 83 Demonstratives in Irish may also be used independently as a pronoun (as in i and ii) or as what den Dikken (2006) calls a defective copula as in (iii and iv): i. seo dhuit DEM to.2.SG ‘This to you.’ (Said when handing something to someone else i.e., the colloquial ‘here you go.’)
ii. tabhair dom sin give .IMP to.1.SG dem ‘Give me that!’
iii. sin é! DEM.COP 3.SG.M ‘That’s it!’
iv. sin é an chaoi is diabhal neart air DEM.COP 3.SG.M DEF way & devil strength on.3.SG.M ‘That is the way, and there’s devil all can be done about it.’
Demonstrative pronouns and defective copulas will not be considered in this chapter. This dissertation deals only with post-nominal demonstratives as part of a nominal expression. 84 Unlike pronouns and proper nouns, which I assume, following Giusti (2015), are in Spec,DP, Wiltschko (2014) argues that vocatives are in a higher structure, K0. In Windsor (to appear), I support Wiltschko’s vocative-in-K0 analysis based on evidence from Irish and Attic Greek. See Appendix I for details. McCloskey (2004) does not offer an analysis of Irish vocatives, but contends that they fit the general pattern of definite DPs in Irish. Given Diessel’s (2006) generalization that demonstratives can be linked to a particular gesture, and that presumably vocative addresses could also be linked to the same gesture, it is likely the case that vocatives must also select a definite DP in Irish, but that the vocative particle prevents the realization of an overt definite article with the noun that they dominate.
The demonstrative in Irish and Blackfoot | 111
(112) Demonstratives and definite DPs in Irish85 a. an tráchtas beag seo DEF thesis small PROX ‘this short thesis’ b. an tráchtas beag DEF thesis small ‘the short thesis’ c. * tráchtas beag seo thesis small PROX d. chuaigh [sé seo] ar seachrán go.PST 3.SG.M PROX on stray ‘This person went astray.’ (Lit: ‘This he went astray.’) (adapted from McCloskey 2004: 2) e. a bhean udaí VOC woman INVIS ‘Hey you over there (addressed to a woman)!’ (Lit: ‘Hey yon woman!’) (adapted from McCloskey 2004: 3) f. bhí urradh as miosúr i nGoll seo be.PST strength out.of measure in Goll PROX ‘this (guy) Goll had astonishing strength’ (Lit: ‘(There) was strength beyond measure in this Goll.) (adapted from McCloskey 2004: 2) Demonstratives are generally the final element in an Irish nominal expression (with the exception of some adnominal modifiers and PPs), which is a fact that has led to some researchers positing that they are very low in the nominal structure (Brugè 2002; Roberts to appear). However, scope effects provide evidence that demonstratives are actually organized quite high in the nominal expression: (113) Irish Coordinated DP structure [[na fir]DP agus [na mná]DP ]DP sin… DEF.PL man.PL and DEF.PL woman.PL DIST ‘those men and women…’ (Adapted from McCloskey 2004, brackets my own) McCloskey (2004) uses the example provided in (113) to show that a single demonstrative scopes over multiple coordinated DPs. If demonstratives were not able to scope over both DPs in (113), to specify all referents as distal, we would expect the demonstrative sin to be Merged after fir ‘men’ and again after mná ‘women’, but that is not required.
85 Translations in (112d-f), are adapted from McCloskey 2004 to include literal translations. McCloskey notes that, since these constructions are not easily available in English, translations can only be approximate.
The demonstrative in Irish and Blackfoot | 112
Blackfoot is typically analysed as having a five-root demonstrative system (e.g., Frantz 2017) that operates on proximity or familiarity of a referent relative to speech participants.86 The characterization that Frantz provides for each of the five demonstrative roots in Blackfoot is given in Table 9 below:87 Label Blackfoot form Interpretation DEM1 am Proximal and familiar to the speaker DEM2 amo Proximal to the speaker but not to the addressee DEM3 anno Proximal to the speaker and proximal or familiar to the addressee DEM4 ann Proximal or familiar to the addressee, but not proximal to the speaker DEM5 om Proximal to neither the speaker nor the addressee Table 9: Blackfoot demonstrative root interpretations (adapted from Frantz 2017) Unlike their Irish counterparts, Blackfoot demonstratives obligatorily precede nominal expressions. Bliss (2013) provides the data in (114) to show that, although multiple linearizations are possible, the demonstrative cannot surface in post-nominal position. As can be seen in this data, the verb can precede the nominal expression (114a), follow it (114c), or even intercede between the demonstrative and the noun (114b).88 (114) Linear order of demonstratives and nouns in Blackfoot a. áohkiwa oma imitááw [V > Dem > N] a-ohki-wa om-wa imitaa-wa IMPF-bark.AI-PROX DEM-PROX dog-PROX ‘That dog is barking.’ b. óóma áohkiwa imitááw [Dem > V > N] (Bliss 2013: 150) c. omi ponokáómitaayi iksíkkaayiyináyi [Dem > N > V] om-yi ponokaomitaa-yi ik-ikkaayi-yini-ayi DEM-OBV horse-OBV INTNS-canter.AI-OBV-3.SG.PRN ‘That horse is naturally swift.’ (Lit: it canters) d. * ponokáómitaayi omi iksíkkaayiyináyi *[N > Dem > V] (Bliss 2013: 157)
86 Bliss (2013) and Wiltschko (2014) argue for a three-demonstrative system in Blackfoot, suggesting that the
The demonstrative in Irish and Blackfoot | 113
The data in (114a-c) show that, while Blackfoot allows multiple possible orderings of the verb, demonstrative, and noun within a clause, it does not permit a noun to precede the demonstrative that specifies it (114d). Another striking difference between Irish and Blackfoot is whether demonstrative roots can inflect: Irish demonstratives do not show inflection for any of the grammatical features in that language (Table 10), whereas, under Bliss’ (2013 et seq.) analysis, demonstrative inflection in Blackfoot can produce over 900 unique combinations (Figure 14). Article89 Noun Adjective Demonstrative Case Number Gender ‘the’ ‘man/woman’ ‘big’ PROX DIST INVIS MASC an fear mór SG NOM/ FEM an bhean mhór ACC MASC na fir móra PL FEM na mná móra MASC an fhir mhóir SG FEM na mná móire GEN MASC na bhfear móra PL FEM na mban móra seo sin úd MASC an bhfear mór SG FEM an mbean mhór DAT MASC na fir móra PL FEM na mná móra MASC a fhear mhóir SG FEM a bhean mhór VOC MASC a fhir móra PL FEM a mná móra Table 10: Agreement patterns in Irish nominal expressions
Figure 14: Demonstrative inflection template (Bliss 2013: 13)
89 The vocative particle, a is not an article and so it does not agree with the noun for other morphosyntactic features, unlike the definite article, but does denote an argument as definite (see fn. 84).
The demonstrative in Irish and Blackfoot | 114
Some examples of how these inflections are realized on demonstrative roots in Blackfoot are provided in (115), including two examples of verbalized demonstratives in (115c, d). (115) Inflected demonstratives in Blackfoot (Frantz 2017: 71-75) a. annaahka Sámahka áako’toowa ann-wa-hka Sam-wa-hka yáak-o’too-wa DEM-PROX-INVIS Sam-PROX-INVIS FUT-arrive.AI-3.SG ‘Sam will arrive.’ b. áámoksika isttsiíksinai’kokaiksika amo-iksi-ka isttsííksina-i’kokaa-iksi-ka DEM-ANIM.PL-O.T. snake-paint^lodge-3.PL-O.T. ‘the ones who (used) snake-painted lodges here’ c. áannayao’ka naapi anna-ya-o’ka náápi DEM-MOVG-VBZR Naapi ‘That one is Naapi.’ d. ki áámoksimao’kiaawa ki amo-iksi-ma-o’ka-yi-aawa & DEM-ANIM.PL-STAT-VBZR-3.PL-PRN ‘...and here they were’ On the surface, Irish and Blackfoot demonstratives could not be more different. However, in Windsor (to appear), I argue that there is one property that is common to demonstratives in both languages: Expressing the function of referentiality. I follow Cinque (1990, see also de Cuba & Ürögdi 2009; de Cuba & MacDonald 2012, 2013) for a definition of referentiality: (116) Referentiality (Cinque 1990:16) The ability to refer to specific members of a set in the mind of the speaker, or pre- established in the discourse. In Irish, when a referent is pre-established in the discourse (D-linked), it becomes more proximal to the conversation, selecting a distal visible demonstrative rather than an invisible one. In the spontaneous speech data below, taken from Ó Conghaile (1999: 27), the speaker is discussing a location approximately 50km away which should be introduced by the invisible demonstrative úd, but instead, the visible demonstrative sin is selected:
The demonstrative in Irish and Blackfoot | 115
(117) D-linking invisible referents (sentence from Ó Conghaile 1999: 27) … amach faoi na cnic, amach faoi na Beanna Beola out about DEF.PL hill.PL out about DEF.PL ‘Twelve Bens’
[na cnic mhóra sin]DemP taobh ó thuaidh de Chonamara DEF.PL hill.PL big.PL DIST.VIS side from north of Connemara ‘…near the hills, near the Twelve Bens, those large hills on the North side of Connemara’ In the example in (117), the referent Beanna Beola ‘Twelve Bens’ is established in the discourse. When it is subsequently referred back to, a distal demonstrative is used, but the visible, rather than the invisible one – an alternation based on referentiality and D-linking. Blackfoot displays a similar alternation where referentiality (i.e., familiarity in Frantz’ 2017 description) can substitute for proximity. Like in Irish, the choice of which Blackfoot demonstrative root is selected can be influenced by different referentiality readings. The data below in (118) illustrate the same sentence under different referentiality readings: discourse-new (118a), D-linked with an invisible referent (118b), and D-linked with a proximal referent (118c).
The demonstrative in Irish and Blackfoot | 116
(118) Blackfoot demonstratives and D-linking90 a. nitohpómmaa náápioyii nit-ohpómma-wa náápioyis-i 1.SG-buy.s.t.AI-PROX house-NON^REF.SG ‘I bought a house.’ [Context: The speaker is calling a friend to inform them of a new purchase, the friend had no previous knowledge of the house in question.] b. nitohpómmatoohpa omi náápioyisi nit-ohpómmatoohp-wa om-yi náápioyis-yi 1.SG-buy.s.t.TI-PROX DEM5-OBV house-OBV ‘I bought that house.’ [Context: The speaker and a friend are sitting down for coffee, and the speaker reminds the friend that s/he bought the house they had been speaking about previously, but the house in question is in a different town.] c. nitohpómmatoohpa anno náápioyisi nit-ohpómmatoohpa-wa anno-yi náápioyis-yi 1.SG-buy.s.t.TI-PROX DEM3-OBV house-OBV ‘I bought this house.’ [Context: The speaker, located at the new house, calls the friend, who is in another town, to discuss the newly purchased house now that s/he is moving in.] The data in (118a) show that when a referent is discourse-new in Blackfoot, it is introduced without a demonstrative and with the non-referential suffix -i (discussed below). In (118b), the house in question is not proximal to either the speaker or the addressee, but is referential because there is a specific house under discussion, and known to both the speaker and the addressee. Therefore, in (118b), the distal demonstrative root om is selected. However, when the speaker is proximal to the referent but the addressee is not, the familiarity of the referent to the addressee can substitute for physical proximity, allowing the speaker to use the anno root rather than the ann root. The anno root denotes a deictic relationship where both the speaker and the addressee are proximal to the referent while the ann root denotes a deictic relationship where the speaker is physically proximal to the referent, but the addressee is not. That the demonstrative is concerned with the syntactic function of referentiality is further supported by the complementary distribution between the demonstrative and the non-referential nominal suffix in Blackfoot, as shown in (118a) and (119) below:
90 Note: These contexts do not build on one another, they are separate situations. For example, the sentence in (118c) could be used if the addressee knew that the speaker was contemplating buying the house, but it would be pragmatically odd if the addressee was already aware of the purchase.
The demonstrative in Irish and Blackfoot | 117
(119) Referential and non-referential expressions a. nitsíín (*oma) aakííkoani nit-íín (om-wa) aakíí-koan-i 1.SG-see.AI (DEM5-PROX) woman-DIM-NON^REF.SG ‘I saw some girl.’ b. nitsínoawa oma aakííkoana nit-íno-aa-wa om-wa aakíí-koan-wa 1.SG-see.TA-DIR-3.SG DEM5-PROX woman-DIM-PROX ‘I saw that girl.’ As the data in (119) show, it is ungrammatical to Merge a demonstrative when a nominal is marked as non-referential, indicating the mutual exclusivity between the demonstrative, a referential element, and the non-referential suffix, -i. Taken with the previous evidence from D-linking in both Blackfoot and Irish, the fact that demonstratives in Blackfoot are in complementary distribution with the non-referential suffix provides strong evidence that demonstratives in these languages encode referentiality. If we additionally consider Giusti’s (2015) strong conclusion that the cross- linguistic function of demonstrative elements is to realize the referential index of a nominal expression, we can conclude that the syntactic property that demonstratives in these languages share is (minimally) that they are deictic elements that encode referentiality. Therefore, I provide the following definition of a demonstrative for future cross-linguistic comparison and testing of the hypotheses advanced in this chapter: (120) Demonstrative (definition) A deictic element in the nominal syntactic domain which encodes referentiality – frequently, but not necessarily, associated with spatial proximity91 Having established what is meant by “demonstrative” as a grammatical category for the present discussion, I now proceed to investigations of Irish and Blackfoot demonstratives in turn to highlight what can be learned about this category through the use of cross-component evidence.
91 Giusti (2015: 78), citing Matthewson (1998) argues that demonstratives are not necessarily associated with spatial proximity, but they do all encode the referential index of a nominal expression. She provides the following list as a set of additional properties that demonstratives have cross-linguistically: i. Definiteness English ii. Specificity Turkish (Enç 1991); Polynesian (Chung 1978) iii. Visibility Bella Coola (Davis & Saunders 1975) iv. Proximity St’at’imcets (Van Eijik 1997)
The demonstrative in Irish and Blackfoot | 118
5.2. Irish demonstratives at the interface At the beginning of this dissertation, I provided a brief synopsis of two syntactic analyses of Celtic demonstratives, and suggested that we could reject the low demonstrative analysis of Roberts (to appear) by utilizing phonological data at the interface. The data I briefly outlined in Chapter 1 involved initial consonant mutation (lenition) in Irish. I repeat that evidence here as it will again bear on the present discussion. I made the prediction —based on the analysis of lenition patterns in Windsor (2012) and a knowledge of interface matching constraints provided by Elfner (2012, 2015) and Bennett et al. (2016, forthcoming)— that if demonstratives are in the same structural position as adjectives (adjoined to nP, as argued by Roberts to appear), they would display lenition just as adjectives do. But, if they dominate the DP, as argued by McCloskey (2004) and Adger (2013), they would display no lenition because they are separated from the noun and its adjectives by a phonological phrase (φ) boundary. The lenition hypothesis is based on the observable differences in initial consonant weakening between adjectives that are part of a DP (121a), and those that are external to the DP (121b): (121) Adjective lenition in Irish (repeated from 5) a. tá [an chloch ghorm mhór]DP ag rolladh /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ ə ɾol̘a/ [t̪ɑː ən xl̪ ɑx ɣɔɾm woːɾ ə ɾol̘a] be DEF stone blue big PRT roll.VN ‘The big blue stone is rolling.’
b. tá [an chloch ghorm]DP mór /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ/ [t̪ɑː ən xl̪ ɑx ɣɔɾm moːɾ] be DEF stone blue big ‘The blue stone is big.’ In (121a), the adjectives g(h)orm ‘blue’ and m(h)ór ‘big’ begin with the radical consonants /ɡ/ and /m/ respectively. In the prosodic structure of this construction, each of these adjectives is matched as a word (ω) without being separated by a φ-boundary, and so they are lenited to [ɣ] and [w], respectively. However, in (121b) where the adjective mór ‘big’ is outside of the DP, the initial radical /m/ surfaces faithfully. The difference between these two realizations of the adjective is dependent on structure:
The demonstrative in Irish and Blackfoot | 119
(122) Structure of sentences in (120) a. b.
As can be seen in (122a), the adjective m(h)ór is inside the subject DP (which is matched as a φ in prosodic constituent structure), but is outside of the DP (and thus outside of the
φDP) in (122b). The internal structure of the DP (not shown in (122a)) adjoins the AP to the nP, which is the same position that Roberts (to appear) argues to be the attachment site of the DemP, as in (123), repeated from (1). Conversely, McCloskey (2004) and Adger (2013) argue that the demonstrative is outside of the DP (and thus outside of the φDP), as in the structure in (124). (123) Structure of the low demonstrative (cf. Brugè 2002; Roberts to appear) a. an leabhar nua seo (faoi teangeolaíocht) DEF book new PROX under linguistics ‘this new book (on linguistics)’ b.
The demonstrative in Irish and Blackfoot | 120
(124) Structure of the high demonstrative (cf. McCloskey 2004; Adger 2013)92
According to the lenition data presented above, the structure in (123) predicts that we will observe lenition on the demonstrative, which is in the same structural position as adjectives in that analysis (since the prosody is not sensitive to the distinction between specifiers and adjuncts – see §5.2.2.2 for full discussion). Contrastively, the structure in (124) predicts that we will not observe lenition on the demonstrative, which is outside of the DP (and thus outside of the φDP). As shown in (125), the prediction of the high demonstrative in (124) is borne out: (125) Lenition and demonstratives in Irish adjective strings (repeated from 6) tá an chloch ghorm mhór sin/*shin ag rolladh /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ sʲɪn ə ɾol̘a/ [t̪ɑː ən xl̪ ɑx ɣɔɾm woːɾ sʲɪn/*hɪn ə ɾol̘a] be DEF stone blue big DIST PRT roll.VN ‘That big blue stone is rolling.’ The data in (124) show that in a string where adjectives such as g(h)orm ‘blue’ and m(h)ór ‘big’ are lenited (as indicated by the weakening of the initial /ɡ/ and /m/ to [ɣ] and [w], respectively), the initial /sj/ of the demonstrative sin is not weakened to [h]. This result indicated that the lenition of demonstratives in such a construction was not grammatical, and therefore, the demonstrative must occupy a position different from adjectives.93 The lenition data support the analyses of Celtic demonstratives by McCloskey (2004) and Adger (2013). However, there is additional evidence (prosodic and syntactic) that can be used to differentiate the subtle differences in the analyses that each of these authors proposes as well. This additional evidence comes from the pitch accent data as described in Chapter 3 and additional lenition data on the prosodic side, and linearization of nominal elements on the syntactic side. Before providing evidence for one syntactic analysis over
92 As discussed below, McCloskey and Adger do not provide a definite conclusion on the label that dominates the DP and the Dem. This structure is meant to capture the similarities between the two analyses. The differences are discussed below. 93 See also Windsor (2012) for a full discussion of Irish lenition as a positional prosodic alternation.
The demonstrative in Irish and Blackfoot | 121 the other, I first describe the analyses of Celtic demonstratives by McCloskey (2004) on Irish and Adger (2013) on Scottish Gaelic, and then highlight the differences between them.
5.2.1. Previous syntactic analyses 5.2.1.1. Dem as the right branching sister of DP In Chapter 1, I described how Brugè (2002) and Roberts (to appear) analyzed the DemP as the external argument of nP due to the height of the demonstrative relative to a PP complement of the noun such as [faoi teangeolaíoct]PP in (126), repeated from (2). (126) Linear relationship of demonstratives and PPs in Irish an leabhar (*seo) nua seo [faoi teangeolaíocht] (*seo) DEF book PROX new PROX under linguistics PROX ‘this new book on linguistics’ (adapted from Brugè 2002: 41, corrections my own) However, Adger (2013) did not take it as a foregone conclusion that the PP in the above construction was necessarily a complement. Instead, Adger argued that such PPs are adjuncts that attach very high in the nominal spine – above the rest of the nominal expression, including demonstratives and quantifiers. Adger argues that elements like the Scottish Gaelic universal quantifier uile and demonstratives scope over nominal expressions, citing McCloskey’s (2004) coordination test (repeated from 113) as evidence of this scope relationship: (127) Irish Coordinated DP structure [na fir]DP agus [na mná]DP sin… DEF.PL man.PL & DEF.PL woman.PL DIST ‘those men and women…’ (Adapted from McCloskey 2004, brackets my own) In the example in (127), the demonstrative sin specifies both na fir ‘the men’ and na mná ‘the women’ as spatially distal. This evidence shows that the demonstrative does indeed scope over a nominal expression. Adger extends this analysis of Irish to his own analysis of quantifiers in Scottish Gaelic. Combining the scope data with the relative height data from PPs, as in the construction in (126) above, Adger concludes that the PPs must be high- adjoining adjuncts rather than complements to the noun. The analysis of the type of structure in (126) that Adger (2013: 110) provides as his examples (83) and (85) is given as (128-129), repeated from (3-4).
The demonstrative in Irish and Blackfoot | 122
(128) Linear relationship of demonstratives, quantifiers, and PPs in Scottish Gaelic na dealbhan mòra ud uile de Mhàiri the pictures big that all of Màiri ‘all those big pictures of Màiri’ (129) Structure of the high demonstrative in Scottish Gaelic
PP uile de Mhàiri def ud
na dealbhan mòra Part of Adger’s rationale behind the structure in (129) stems from the fact that Celtic demonstratives must take a definite DP (indicated by a definite article, a proper name, or a pronoun) – as shown in (130), repeated from (112) for convenience: (130) Demonstratives and definite DPs in Irish a. an tráchtas beag seo DEF thesis small PROX ‘this short thesis’ b. an tráchtas beag DEF thesis small ‘the short thesis’ c. * tráchtas beag seo thesis small PROX d. chuaigh [sé seo] ar seachrán go.PST 3.SG.M PROX on stray ‘This person went astray.’ (Lit: ‘This he went astray.’) (adapted from McCloskey 2004: 2) e. a bhean udaí VOC woman INVIS ‘Hey you over there (addressed to a woman)!’ (Lit: ‘Hey yon woman!’) (adapted from McCloskey 2004: 3) f. bhí urradh as miosúr i nGoll seo be.PST strength out.of measure in Goll PROX ‘This (guy) Goll had astonishing strength.’ (Lit: ‘(There) was strength beyond measure in this Goll.) (adapted from McCloskey 2004: 2)
The demonstrative in Irish and Blackfoot | 123
The data in (130) show that while a nominal expression can be definite without merging a demonstrative (130b), a nominal expression cannot be Merged with a demonstrative without being marked as definite (130c). Definiteness is indicated by a definite article in (130a, b), by a pronoun in (130d), by a vocative in (130e), and by a proper name in (130f).94 Adger argues that if demonstratives are low in the structure (as argued by Brugè 2002 and Roberts to appear), we would be required to assume some sort of definiteness feature on the Dem0 which would be checked by merging a definite D0 above the N, representing a look-ahead problem. In the case of proper names and pronouns (possibly themselves in Spec,DP according to Giusti 2015), we would further need a way to stipulate that the definite D0 be null, but still satisfy the definiteness feature of the Dem0. The other possibility, according to Adger, is a right-branching node above DP, as depicted in (129). However, this analysis, too, has its problems. McCloskey (2004: 3) notes that “many theoreticians would be unwilling to cede to a given head the ability to determine whether its complement appeared on the right or on the left.” McCloskey’s statement stems in part from the fact that all other complements in the language are on the right, so it seems improbable that demonstratives (and quantifiers in Adger’s system) would have their complements on the left. As a result of being unwilling to cede such power to a given head, McCloskey instead suggests that the demonstrative realizes a syntactic head, possibly labelled D itself, which takes a definite DP as its specifier (raised to that position from within its own domain).
5.2.1.2. Recursive DPs The alternative DP-dominating analysis of Celtic demonstratives to Adger’s right- branching structure is McCloskey’s (2004) recursive DP analysis. Recognizing that Irish is (otherwise) a rigidly head-initial language, McCloskey proposes a movement analysis that moves a definite DP from the complement of the demonstrative D to the specifier of that phrase:
94 See fn. 84 for additional details.
The demonstrative in Irish and Blackfoot | 124
(131) A movement analysis of Irish demonstrative structure (McCloskey 2004: 3) a. DP b. DP
Dem DP DP [DEF] [DEF]j Dem tj seo seo According to McCloskey, in order to account for the movement shown in (131), we need to assume some property (or feature) on the demonstrative which drives movement of the DP around the demonstrative (what I have previously claimed in Windsor (2014a) is a strong uDEF feature, see also Doyle 2002 for another possible motivation for this type of movement). Leaving the explanation of how movement is theoretically achieved in (131) aside, McCloskey provides evidence from modifier stranding to suggest that movement of a DP within a nominal expression does occur. He suggests that when a DP moves to the higher specifier position, it may be extracted from the phrase containing the adnominal modifier eile ‘other’, stranding the modifier in a low position (as in 132a), or the entire phrase containing the DP and the modifier may be moved as a single unit (as in 132b). (132) Modifier stranding in Irish nominal expressions (adapted from McCloskey 2004: 5) a. an fear seo
The demonstrative in Irish and Blackfoot | 125 grammatical attachment sites for the modifier, similar to the Irish universal quantifier, uile, which may appear pre- or post-nominally: (133) Multiple positions for quantifier attachment95 a. na daoine uile DEF.PL person.PL every ‘everyone/all men/anyone’ b. an uile dhuine DEF every person ‘everyone/all men/anyone’ c. gach/an uile dhuine eile each/DEF every person other ‘everyone else’ d. na dhuine sin uile DEF people DEM every ‘all those people’ The data in (133) show that, at least in Irish, the universal quantifier may appear post- nominally (133a) or appear between the definite article and the noun (133b) without changing the meaning of the phrase. The quantifier may also appear in conjunction with the adnominal modifier eile (133c) or following a demonstrative (133d) (the standard linearization of this element in Scottish Gaelic according to Adger 2013). However, the order of the quantifier relative to the modifier eile or the demonstrative seems to be less variable.96 The evidence presented in (133) is not necessarily contradictory to McCloskey’s interpretation of modifier stranding, but does give us reason to question how strong that evidence is for movement. Because we cannot use the data in (132) to decisively conclude that there is DP-movement within the nominal expression, we are left with the problem of choosing an analysis based on which theoretical problem we take the least issue with: If we opt for a DP-raising analysis, we must allow for comp-to-spec movement, a problem according to Abels’ (2003) anti-locality constraint; if we opt for a right-headed demonstrative projection, we must allow this (and only this) head to require a different
95 An explanation of the interaction of quantifier position and number is beyond the scope of the present discussion. See fn. 98 for an example showing that this quantifier precedes overt numerals in Irish. 96 My consultants judge an uile dhuine sin (DEF every person DEM) as opposed to na daoine sin uile in (133d) for ‘all those people’ as odd, but not ungrammatical.
The demonstrative in Irish and Blackfoot | 126 branching direction relative to the complement. Thus, what is needed is additional evidence to decide between Adger’s (2013) right-branching analysis of demonstrative structure and McCloskey’s (2004) recursive DP analysis. The evidence I provide in the following section for the constituency of the demonstrative in Irish once again comes from pitch accent distribution (as discussed in Chapter 3) and lenition patterns (as discussed in Chapter 1). This evidence supports a left-branching, movement analysis (à la McCloskey), but suggests that the specifier position to which the DP moves has a different label – which I argue is DemP.
5.2.2. Phonological and syntactic evidence for Irish DemP structure Recall that in Chapter 3, I provided the structure in (134, repeated from 55) to diagram how HL pitch accents are associated with the stressed syllable of the rightmost word in every phonological phrase (φ) (as indicated by rightward descending arrows), and LH accents are associated with the stressed syllable of the leftmost word in every φNon-min (as indicated by leftward descending dotted-line arrows): (134) The distribution of phrasal LH and HL pitch accents (Elfner 2015: 1179) díolfaidh leabharlannaí dathúil bláthanna áille sell.FUT librarian handsome flower.PL beautiful.PL ‘The handsome librarian will sell beautiful flowers.’ a. Morphosyntactic structure b. Prosodic structure ΣPa φa ← ΣP
ω φb ← TP Vi TPb díolfaidh DP → φc φd,e← vP/VP/DP DPc …d ω ω ω ω N A leabharlannaí dathúil bláthanna áille
ti DPe
N A In the discussion of the above structures in Chapter 3, I provided additional commentary on the internal structure of the DP. It is commonly held that the noun in Irish obligatorily raises to a higher functional projection to achieve the Noun > Adjective ordering. The prosodic facts of pitch accent distribution could be accounted for equally well
The demonstrative in Irish and Blackfoot | 127 by an analysis that assumed N-raising to Num/NumP (e.g., Roberts to appear) or N-to-D raising (e.g., Windsor 2012). Crucially, however, because the nP contains only a non-binary
AP in terms of pronounceable material, the BIN-MIN constraint prevented the nP/AP from being matched as a φ. Thus, only the DP (containing the ωN and the ωA) is matched as a φ in constructions like that provided in (134). In Chapter 3, I also assumed a structure of the demonstrative —syntactic and prosodic— in line with McCloskey’s (2004) and Adger’s (2013) analysis that the demonstrative was organized outside of the DP. That structural assumption provided the testable hypothesis that, if accurate, a demonstrative could force a pitch accent onto an object noun by making it the leftmost daughter of a φNon-min. This hypothesis is contrasted in (135) with the prediction that a pitch accent cannot be forced onto the object noun if the demonstrative is in Spec,nP – the analysis of the low demonstrative by Roberts (to appear, see also Brugè 2002): (135) Prosodic predictions of accents based on demonstrative location (attested structure) phlucáil an buchaill dóighiúil [na bláthanna corcra úd] PST-pick DEF boy handsome DEF.PL flower.PL purple.PL DEM ‘The handsome boy picked yon purple flowers.’ a. Dem as sister to DP prediction φ ← DemP
φ ← DP ω ← Dem úd ω ← D/N ω ← A HL corcra na bláthanna HL LH (adapted from Windsor 2016: 9) b. Dem as Spec,nP prediction (unattested structure)
φ ← DP
ω ← D/N ω ← A ω ← Dem corcra úd na bláthanna HL (adapted from Windsor 2016: 8) In Chapter 3, I showed that pitch accents could be reliably forced onto an object noun when a demonstrative was Merged with that nominal expression (χ2 = 50.6192; p < 0.001). This empirical evidence independently supported the analysis presented in Chapter 1 based on lenition patterns.
The demonstrative in Irish and Blackfoot | 128
In Chapter 1, I argued that lenition patterns showed that the demonstrative was outside the φ containing the noun and any adjectives. This was based on the difference in lenition realization between DP-internal adjectives (136a) and DP-external adjectives (136b), repeated from (5) and (121). (136) Adjective lenition in Irish a. tá [an chloch ghorm mhór]DP ag rolladh /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ ə ɾol̘a/ [t̪ɑː ən xl̪ ɑx ɣɔɾm woːɾ ə ɾol̘a] be DEF stone blue big PRT roll.VN ‘The big blue stone is rolling.’
b. tá [an chloch ghorm]DP mór /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ/ [t̪ɑː ən xl̪ ɑx ɣɔɾm moːɾ] be DEF stone blue big ‘The blue stone is big.’ As shown in the data above, (feminine) adjectives inside a DP display initial consonant weakening (lenition) where the initial /ɡ/ and /m/ of g(h)orm ‘blue’ and m(h)ór ‘big’ become lenited to [ɣ] and [w] respectively. However, in (136b), the adjective mór ‘big’ is a predicate of the DP subject, and surfaces with the radical consonant [m], displaying no lenition. The prediction based on this pattern was that if the demonstrative was in the same position as adjectives (Spec,nP), it would display the same lenition pattern. If, however, the demonstrative were external to the φDP, no lenition would be present on the demonstrative: (137) Lenition and demonstratives in Irish adjective strings tá an chloch ghorm mhór sin/*shin ag rolladh be DEF stone blue big DIST PRT roll.VN /t̪ɑː ən kl̪ ɑx ɡɔɾm moːɾ sʲɪn ə ɾol̘a/ [t̪ɑː ən xl̪ ɑx ɣɔɾm woːɾ sʲɪn/*hɪn ə ɾol̘a] ‘That big blue stone is rolling.’ As can be seen in (137, repeated from 6 and 125) no lenition is found on demonstratives in similar constructions, as indicated by the fact that the radical /sj/ surfaces faithfully rather than being lenited to [h]. Independently, lenition patterns and pitch accent distribution each provide evidence to reject the low demonstrative analysis, supporting an analysis that attaches the demonstrative high in the nominal structure, above the DP. However, in conjunction with evidence from linearization, these two sources of evidence can also help us to reject the high demonstrative analysis of Adger (2013), in favour of the movement analysis of
The demonstrative in Irish and Blackfoot | 129
McCloskey (2004) except for one minor necessary change to the category label, which I address subsequently.
5.2.2.1. Adnominal modifiers and lenition Recall that in §5.2.1.2, I showed that the adnominal modifier uile had multiple possible linearizations – able to occur post-nominally after the demonstrative, as well as pre- nominally between the article and the noun. These two linearizations are provided in (138) using the Irish translation of Adger’s (2013: 110) Scottish Gaelic example: (138) Multiple linearizations of uile ‘all’ in Irish (Windsor 2016: 12) a. na pictiúra mhóra úd uile de Mháire DEF.PL picture.PL big.PL DEM all of Mary ‘all yon big pictures of Mary’ b. na uile phictiúra mhóra úd de Mháire DEF.PL all picture.PL big.PL DEM of Mary ‘all yon big pictures of Mary’ The important part of the constructions in (138) is not the position of uile ‘all,’ but the effect it has (or does not have) on its neighbouring lexemes. In (138a), the initial consonant of the preposition de ‘of’ is un-lenited, surfacing faithfully as the radical [d]. This can be explained if the PP de Mháire is matched as a φ in prosodic constituent structure, preventing uile from causing lenition because it is separated from the lexeme by a φ-boundary. We know that uile is capable of causing lenition due to the construction given in (138b) where the initial /p/ of p(h)ictiúra is realized in the lenited form [f]. We also know, from the previous discussion, that a ω not separated from the ω containing the √noun by a φ- boundary can cause lenition of the noun.97 Based on the fact that uile causes lenition to be realized on phictiúra, we can conclude that uile represents a ω in prosodic constituent structure, indicating it is a syntactic head inside the DP, which may be linearized either pre- nominally (after the definite article) or post-nominally, after the demonstrative.98
97 In (138a), pictiúra is not lenited because the article na [nə] is monomoraic and therefore not prosodically heavy enough to be a well-formed ω. 98 Although rare, I was able to find one instance of uile preceding a numeral in an nominal expression, suggesting that it is Merged above number. This example comes from an 1817 printing of the New Testament (see An Biobla Naomhthata on Google Books for a searchable version): i. … uile dhá threabh dhéag Isreal all two tribe ten Israel ‘all the twelve tribes of Israel’ (Genesis 49:28)
The demonstrative in Irish and Blackfoot | 130
The ability of adnominal modifiers like uile ‘all’ and eile ‘other’ (discussed in §5.2.1.2) to be linearized in multiple positions is similar to the multiple linearization patterns of adverbial modifiers, such as uaireanta ‘sometimes’ as shown in (139) with examples taken from a short story by Máire Mhac an tSaoi (2002: 1139). (139) Multiple linearizations of adverbial modifiers (Windsor 2016: 12) a. uaireanta thugadh sí fuáil léi… sometimes PST.bring 3.SG.F sewing with.3.SG.F ‘sometimes she brought sewing with her…’ b. thuightí di uaireanta… PST.understand.PASS for.3.SG.F sometimes ‘sometimes it occurred to her…’ Data such as those in (138) show that adverbials may either be left- or right-linearized in Irish. If adnominal modifiers such as uile ‘all’ and eile ‘other’ can be analyzed in the same way as adverbials —that they are adjunct phrases which may be either right or left adjoined— then we can account for the various possible realizations of their linear orders as well. The hypothesis I suggested in Windsor (2016a) for these patterns of linearization and lenition was that adverbial/adnominal modifiers are not restricted (i.e., specified) in the directionality of their syntactic adjoinment – allowing right or left linearization relative to the constituent that they modify (subject to speaker preference). This does not contradict the fact that Irish is rigidly head-initial, nor does it make demonstratives syntactically unique in Irish in that they would be the only syntactic element that is head-final. Instead, this is a property (unspecified directionality) of a natural class of syntactic elements – adnominal/adverbial modifiers. This hypothesis is also supported by the lenition facts as well – showing that the adnominal can be Merged as part of the syntactic constituent containing the determiner and noun, and is not necessarily DP-external at the spell-out to PF (contra Adger 2013). The evidence from adnominal linearization and lenition support the DP-raising analysis of McCloskey (2004) over the right-branching analysis of Adger (2013). However, as I show in the following section with evidence from pitch accent distribution, there is one modification required to the structure that McCloskey suggested for this movement-based analysis.
The demonstrative in Irish and Blackfoot | 131
5.2.2.2. Recursive syntactic labels and φ-matching In §4.2.2, looking at matching Blackfoot nominal structure, I stated that Match Theory would not match recursive ΦP labels as separate φs because they represent the same projection. In other words, because the phonological component seeks to match φs to maximal syntactic projections, not intermediate projections, only the highest projection of a single label will be matched as a φ. Elfner (2015: 1177-8) clarifies this component of the
MATCHPHRASE constraint by providing the simple matching principle: XMax → φ. This principle prevents non-maximal recursive XPs (as in the case of adjunction, for example) as well as intermediate projections (i.e., bar-levels) from being realized as φs in prosodic constituent structure. I formulate Elfner’s XMax → φ matching principle as (140): (140) Condition on φ matching A node XPj is matched as a φ iff: i. XPj is not dominated, or ii. XPj is dominated by a node XPk, where j ≠ k The condition on φ matching expressed in (140) formalizes the generalization that, because the phonological component cannot reference morphosyntactic indices, it cannot tell the difference between a recursive XP structure (as in adjunction structures) and an XP in the specifier position of an XP with the same label. Therefore, in the recursive DP structure suggested by McCloskey (2004) for demonstrative constructions, recursive DPs would be realized as a single φ: (141) Matching a recursive DP in Irish (adapted from Windsor 2016a: 13) a. Syntactic spell-out b. Unattested prosodic matching structure DPk φ ← DPk
DPj[DEF] D'[uDEF*] ω ω ω N A D(em) N A D(em)
The demonstrative in Irish and Blackfoot | 132
By contrast, if we assume that the category label of the phrase whose specifier the definite DP moves to is not D itself, but Dem, we can predict the attested prosodic structure for this construction: (142) Matching a DemP in Irish (adapted from Windsor 2016a: 13) a. Syntactic spell-out b. Attested prosodic matching structure DemP φ ← DemP
DPj[DEF] Dem'[uDEF*] φ ← DP ω Dem N A Dem
DP[DEF] Dem φ ← DP ω Dem N A ω ω N A The structure Adger argues for can predict the attested distribution of pitch accents. However, it cannot account for the multiple linearization patterns of adnominal modifiers found in Irish since it dominates the DemP in that analysis and thus would not be able to appear between the article and noun, as in (138b) above. Further, it cannot account for the fact that the lenition evidence indicates that the adnominal can be spelled out to PF inside the DP. Adopting McCloskey’s (2004) analysis for (142a) allows us to maintain the strict left- headed structure attested in Irish, and avoids the need to treat demonstratives as exceptions to this left-head characteristic of Irish syntactic phrases. It does assume comp-to-spec movement, but arguably does so to satisfy a strong uninterpretable definiteness feature on the demonstrative that requires its specifier be a definite DP (see McCloskey 2004 and references therein for further argumentation in favour of this movement).
The demonstrative in Irish and Blackfoot | 133
The structure in (142b) correctly predicts that demonstratives will not show lenition effects following feminine adjective strings (as shown in 137), and that the addition of a demonstrative will cause an object noun to become the leftmost daughter of a φNon-min, forcing a pitch accent to be realized on the stressed syllable (as shown in 135a).
5.2.2.3. Summary In the preceding section, we re-examined two types of phonological evidence to help us come to a conclusion about the syntactic position and category of demonstratives in Irish. However, we may have come to the incorrect conclusion based on the prosody alone without additional syntactic data in the form of multiple linearization possibilities for adnominal modifiers. The data from lenition patterns and pitch accent distribution supported an analysis that organized the demonstrative outside of the DP headed by the definite determiner, in line with the analyses of McCloskey (2004) and Adger (2013). Further exploring the prosodic structures predicted by the matching of each of those analyses revealed that Adger’s right- branching structure correctly predicted the attested prosodic structure in Irish but that McCloskey’s left-branching structure required a different syntactic label to do so. As stated early on, both syntactic analyses had their own theoretical problems: McCloskey contended that it was problematic to allow a single syntactic category to dictate an alternate head directionality, in contrast to the general headedness principle of the rest of the language (the notable counter-example to McCloskey’s statement being head-final VP in German while all other categories are head-initial – see Dopke 1998, and references therein). However, many theoreticians would also argue that comp-to-spec movement is problematic due to an anti-locality constraint that prohibits too-local movement (Abels 2003). The anti-locality constraint contends that a head-complement relation is the closest syntactic relationship possible, and thus comp-to-spec movement could not possibly be driven by feature checking. Scottish Gaelic, the primary focus of Adger’s investigation of Celtic demonstratives, does not allow multiple linearizations of adnominal modifiers. But examining these modifiers in Irish shows that both adnominal and adverbial modifiers have multiple linearization patterns – adjoining either to the left or to the right of the syntactic constituent
The demonstrative in Irish and Blackfoot | 134 that they modify. This evidence suggested that a natural class of elements in Irish, adnominal and adverbial modifiers, is unspecified for adjunction site and may be either left- or right-linearized. Demonstratives, however, do not share this variability, and thus we are still left with the problem that they would be the only syntactic category in the language with an obligatory right-headed structure under Adger’s analysis. Further, the fact that demonstratives obligatorily select for a definite DP makes an uninterpretable definiteness feature on the demonstrative more plausible. Because McCloskey only suggests that the label of the demonstrative may be D itself, there is nothing riding on that label for his analysis. By altering the label that the demonstrative projects to DemP (necessarily distinct from a DP based on the attested prosodic matching structure examined above), we are able to successfully predict the attested prosodic structure without stipulating that demonstratives violate the general headedness principle of the language. Using both the syntactic and the phonological evidence from Irish, we have motivated a DemP projection that is high in the nominal domain, dominating DP.99 In the following section, I use the DemP projection in Irish as a cross-linguistic possibility to base an alternative hypothesis of demonstrative constituency in Blackfoot on. I suggest that Blackfoot, too, can be analyzed as having a demonstrative projection that dominates DP, rather than in Spec,DP, where it is frequently assumed to be (e.g., Bliss 2013; Wiltschko 2014).
5.3. Blackfoot demonstratives at the interface In Windsor (2017, to appear), I present an alternative analysis of the constituency of Blackfoot demonstratives to that provided by Bliss (2013) and Wiltschko (2014).100 The structure I proposed in those articles was in line with the analysis given in the previous
99 This conclusion suggests that when a demonstrative is present, the nominal constituent is (minimally) a DemP in Irish. What the structure of indefinites (which a demonstrative cannot appear in) is, whether a null DemP projection is Merged, or whether a null DP projection is Merged (note, there is no indefinite article in Irish), or whether indefinites are maximally ΦPs or nPs and what the consequences are for feature checking or case (for example) are not addressed here. The interested reader is referred to Kane (2015) for relevant discussion. Kane (2015) follows Adger (2013) in assuming that the demonstrative is merged in a right-headed projection above DP, in what she labels LocP. 100 See also Windsor & Lewis (to appear) for additional cross-component support of this analysis from the syntax-semantics interface.
The demonstrative in Irish and Blackfoot | 135 section for Irish demonstrative constituency that concluded demonstratives head a DemP projection that dominates DP in the nominal structure. The structure I propose for Blackfoot nominal expressions is given in (144). (144) Hypothesized structure of the Blackfoot DemP a. amo na sikomitaa (i’nitsi ni poosi) amo-wa na sik-omitaa-wa i’nit-ii-wa ni poos-yi DEM2-PROX DEF black-dog-PROX kill.TA-DIR-PROX DEF cat-OBV ‘This black dog (killed the cat).’ b.
The evidence provided in this section comes from the previously-established phonetic correlates of prosodic boundaries, the interface that allows the established prosodic structure to match the morphosyntactic structure, and predictions about aspiration that a head-movement analysis versus an agreement analysis each make. In the following section, I outline the previously proposed structure of Blackfoot demonstratives by Bliss (2013) and Wiltschko (2014). I then use phonological evidence from vowel coalescence and aspiration together with syntactic evidence from linearization and relative height to give further evidence in support of the agreement-based proposal of suffixation from Chapter 4. Finally, I offer a proposal of a common structure for demonstrative projections in Irish and Blackfoot, supported by phonological well-formedness and a goal of working towards a universal spine – as motivated by Bliss (2013) and Wiltschko (2014).
The demonstrative in Irish and Blackfoot | 136
5.3.1. Previous syntactic analysis: DemP in Spec,DP Bliss (2013) and Wiltschko (2014) use relative syntactic height via the mirror principle (described in Chapter 4) to diagnose the positions of various syntactic heads. As described in §4.2, Bliss argues that the position of possessive morphemes is in n0 and the plural morphemes are in Φ0. The positions that Bliss argues for these morphosyntactic features are assumed in the analysis I present here. She contends that suffixation in Blackfoot occurs by successive head-movement to the various heads that contain those features/morphemes. This linearization analysis correctly predicts the order of morphemes in (145), repeated from (91). (145) Linearization of suffixes in possessed animate plurals101 a. kitomitaamiksi kit-omitaa-m-iksi 2.SG-dog-POSS-ANIM.PL ‘your dogs’ b.
A consequence of the assumption of suffixation via head-movement, though, is an unmotivated stipulation that the demonstrative raise to a higher specifier to satisfy word order. Bliss (2013) and Wiltschko (2014) argue that proximate -wa and obviative -yi are suffixes gained in a Link0 or K0, respectively – these are different labels for a projection that immediately dominates DP in their system. Thus, for a nominal expression to be valued as proximate (for example), the √noun must move successively from N0 to n0, then from n0 to Φ0, from Φ0 to D0, and finally from D0 to Link0 (or K0 for obviative nouns), as shown in (147), below. As shown in §5.1.1, demonstratives obligatorily precede nouns in Blackfoot (repeated below from 114 as 146).
101 See fn. 74 on the analysis of person prefixes in Bliss’ (2013) analysis, which I alter in this structure for convenience only.
The demonstrative in Irish and Blackfoot | 137
(146) Linear order of demonstratives and nouns in Blackfoot a. áohkiwa oma imitááw [V > Dem > N] a-ohki-wa om-wa imitaa-wa IMPF-bark.AI-PROX DEM-PROX dog-PROX ‘That dog is barking.’ b. óóma áohkiwa imitááw [Dem > V > N] (Bliss 2013: 150) c. omi ponokáómitaayi iksíkkaayiyináyi [Dem > N > V] om-yi ponokaomitaa-yi ik-ikkaayi-yini-ayi DEM-OBV horse-OBV INTNS-canter.AI-OBV-3.SG.PRN ‘That horse is naturally swift.’ (Lit: it canters) d. * ponokáómitaayi omi iksíkkaayiyináyi *[N > Dem > V] (Bliss 2013: 157) Thus, under Bliss’ analysis, the demonstrative must raise to Spec,LinkP to achieve the correct linearization. Bliss states (2013: 157) that the motivation for demonstrative raising is not clear, something she leaves for future research, but is forced to assume that it does so because of the word order facts and the head-movement analysis she adopts: (147) A head-movement structure of Blackfoot proximate nominal expressions102 oma nitákkaawa (inóókawa) om-wa n-itákkaa-wa ino-oka-wa DEM-PROX 1-friend-PROX see.TA-INV-21 ‘That friend (sees us).’
102 I assume Bliss’ analysis of short-form person prefixes as belonging to a ΦP in Spec,NP for this tree. The N0 presumably moves to the n0, the Φ0, and D0 in turn, but I omit that additional structure for simplicity since those heads are morphologically null in this example.
The demonstrative in Irish and Blackfoot | 138
I contended, in Chapter 4, that because the plural morpheme -iksi was not matched as a ω in prosodic constituent structure, it could not represent a head in syntactic constituent structure. I suggested there that this pointed to an analysis where suffixation in Blackfoot is the result of agreement relationships rather than head-movement. In theory, it is possible that the two analyses are not incompatible. One possible analysis that unifies the feature- checking analysis and the head-movement analysis would be to suggest that features are checked via head-movement, but that the morphological correspondence of those features are spelled out to PF on a single head. In the following sections, I draw on linearization of adjectival prefixes, a definite article emergent among younger speakers of the language, and prosodic constituency, to provide further evidence against a head-movement analysis of suffixation in Blackfoot and to ultimately provide evidence in support of the constituent structure provided in (144) above.
5.3.2. Linearization of Blackfoot adjectives and articles Bliss’ (2013) analysis of the structure of Blackfoot provides two linearization algorithms, one that explains the linearization of suffixes, and one that explains the linearization of prefixes and other words.103 The suffixation algorithm was given in §4.2 as example (90), repeated here as (148) for convenience: (148) Blackfoot suffix linearization algorithm (Bliss 2013: 12) a.
Suffix3
Suffix2
Suffix1
Root b. [ Root – Suffix1 – Suffix2 – Suffix 3 ]
The prefix (and word order) linearization algorithm that Bliss provides is based on scope relations, and thus relative height within the syntactic spine:
103 Bliss (2013: 14) also provides the clitic linearization algorithm below. I do not deal with clitics in this dissertation. i. [[Prefix(es) – Root – Suffix(es)] – Clitic(s)]
The demonstrative in Irish and Blackfoot | 139
(149) Blackfoot prefix linearization algorithm (Bliss 2013: 14) a. XP
Prefix1 YP
Prefix2 ZP
Prefix3 Root b. [ Prefix1 – Prefix2 – Prefix3 – Root ]
Bliss (2013: 14) describes the algorithm in (149) by stating that each of the prefixes are in specifier/adjunct positions, and their relative hierarchical positions determine their linear position. The same is true of separate words in the syntactic spine, though they do not need to be in specifier/adjunct positions – we are able, to a certain extent, to suggest hierarchical position from the linear order of elements. In her discussion of nominal expressions, Bliss (2013: 118) suggests that there are multiple possible attachment sites for adjectives in the nominal spine, providing evidence for what she determines to be NP and N' attachment. This analysis provides the structure in (150), repeated from (92). (150) Structure of Blackfoot adjectives (Bliss 2013: 118) a. kitomahkonssta kit-omahk-w-insst-wa 2-big-3-sister-PROX ‘your big sister’ (pejorative connotation, i.e., obese) b.
However, in order to deal with the N-raising analysis of N-to-n and subsequent n-to-Φ (and because the adjective has scope over the fully-inflected noun, which determines its
The demonstrative in Irish and Blackfoot | 140 syntactic position according to the linearization algorithm in 149), I suggested that adjectives may also adjoin to ΦP in a construction such as that provided in (151), repeated from (94). (151) Structure of the Blackfoot nominal expression (preliminary) a. anniksi akáímahkihkinaiksi (inóókawa) ann-iksi áka-íímachkihkinaa-iksi (ino-oka-wa) DEM-ANIM.PL old-sheep-ANIM.PL (see.TA-INV-21) ‘Those old sheep see us.’ b.
However, if we consider that a singular noun must obligatorily raise to Link0 under this analysis to get the proximate singular suffix -wa, this linearization becomes difficult to maintain. In a construction such as that in (152), it is difficult to decide where an adjective must be Merged in order to intercede between a demonstrative, which is analyzed as being in Spec,LinkP, and a noun in the Link0. Given that Bliss (2013: 118) suggests that N' is one possible attachment site for adjectives, it stands to reason that perhaps Link' is as well. This speculation is represented in (152).
The demonstrative in Irish and Blackfoot | 141
(152) Adjectives modifying singular nouns in a head-movement analysis a. amo omahkómitaawa (áóhkiwa) amo-wa omahk-wimitaa-wa á-ohki-wa DEM-PROX big-dog-PROX DUR-bark-AI-3.SG ‘That big dog (is barking).’ b.
The repair strategy in (152) that places the Merge location of adjectives even higher in the tree, adjoining to Link', can be made to work in order to fit the data. However, a recent development amongst younger speakers of Kainai Blackfoot shows the emergence of a definite article that co-occurs with a demonstrative in constructions such as in (152). The emergence of a definite article in Kainai Blackfoot follows the expected path of diachronic development whereby the article has developed from an existing demonstrative form (Manolessou 2001; Diessel 2006; Giusti 2015). These new articles that have emerged in Blackfoot are na (definite proximate) and ni (definite obviative), and are undoubtedly (diachronically) derived from the demonstrative root ann with a proximate -wa or obviative -yi suffix. The articles are linearized between the demonstrative and the fully inflected noun, which may contain an adjectival prefix:104
104 It is likely that the articles na and ni are bi-morphemic, consisting of a definite morpheme plus a proximate/obviative marker. However, the relevant part of the current examples is to show the linear order of these elements, so that morphological breakdown is not shown in the examples that follow.
The demonstrative in Irish and Blackfoot | 142
(153) Definite articles in Blackfoot (Windsor & Lewis to appear) a. amo na ninaa aahkaakomimmii ni
amo-wa na ninaa-wa aahk-waakomimm-ii-wa ni DEM-PROX DEF man-PROX might-love-DIR-3.SG DEF imitaa imitaa-yi dog-OBV ‘That man might love the dog.’ b. amo na omahkómitaa áóhkiwa amo-wa na omahk-wimitaa-wa á-ohki-wa DEM-PROX DEF big-dog-PROX DUR-bark.AI-3.SG ‘That big dog is barking.’ If we assume that the articles in (153) are Merged in D0, the linear order does not support an analysis that moves the noun above DP to Link0 to gain the -wa suffix. However, if we utilize the structure provided in §4.2.2 as (104), where the distinction between singular and plural is made by the agreement with a different feature in the Φ0 rather than successive head-movement, the linear order is explained without additional word-order movements: (154) Structure of the Blackfoot nominal expression (second attempt)
Like the structure proposed in Chapter 4, the prosodic structure matched from the agreement and feature-checking structure in (154) above also correctly predicts the
The demonstrative in Irish and Blackfoot | 143 phonological realization discussed in that chapter.105 However, there is one more piece of co-occurrence data to consider before concluding that this structure is correct. In the subsequent section, I present additional evidence from Irish and Blackfoot to provide a hypothesis on the cross-linguistic generalizability of the structures under investigation here.
5.4. Cross-linguistic generalizability In the preceding section, I presented data from the relative height of adjectival prefixes and definite articles in Blackfoot to further support my (Chapter 4) analysis of Blackfoot suffixation where feature checking is the result of Agree rather than Move. The structure I motivated in that section was the same as the structure that was proposed at the end of Chapter 4 where the Blackfoot demonstrative is in Spec,DP. That structure, which I labelled as the second attempt, successfully predicts the attested phonological structure based on the matching system assumed in this dissertation. But it is a substantially different structure from the demonstrative constituency argued to exist in Irish in §5.2. The difference between an Irish DemP dominating DP as part of the nominal spine but the Blackfoot DemP being Merged in Spec,DP could be an accurate representation of cross- linguistic differences. However, in this section, I pursue an alternate hypothesis that allows cross-linguistic generalization. The notion that nominal elements belong to a universal spine is part of Wiltschko’s (2014) Universal Spine Hypothesis (preliminary versions of which appear in Ritter & Wiltschko 2009, 2014) and is adopted by Bliss (2013) in her analysis of Blackfoot.106 Therefore, I pursue this hypothesis in the spirit of providing further cross-linguistic evidence towards future typological studies of what universal categories the nominal spine includes. I conclude with testable hypotheses that could allow us to support or reject this analysis with future cross-linguistic study.
105 One of the speakers who provided the data containing definite articles used in the syntax-semantics interface study of Blackfoot demonstratives reported in Windsor & Lewis (to appear) was the same speaker consulted for elicitation data in Study 2 reported in Chapter 4. However, that speaker did not provide any definite articles for the prosodic analysis in that study. The discourse factors that influence article usage is not yet understood. 106 See Appendix I for a description of the Universal Spine Hypothesis (Wiltschko 2014) and a brief summary of how the DemP constituency argued for here fits with that framework.
The demonstrative in Irish and Blackfoot | 144
To begin exploring this hypothesis, I provide data from the co-occurrence of demonstratives and proper nouns, which in turn provides a reason to suspect demonstratives may be higher than Spec,DP in both Irish and Blackfoot.
5.4.1. Co-existence with proper nouns In her book on nominal syntax that provides a typology of languages with articles, Giusti (2015) argues that demonstratives and proper nouns both occupy the position of Spec,DP where they can realize a referential ι-Op(erator) – the referential index of a nominal expression. Because both demonstratives and proper nouns are argued to universally occupy the same specifier position, Giusti predicts that they will never co-occur. Giusti’s prediction is not borne out in either Irish or Blackfoot. In Irish, proper nouns may provide the definite DP that the demonstrative selects for, and in Blackfoot, proper nouns are obligatorily introduced by a demonstrative:
The demonstrative in Irish and Blackfoot | 145
(155) Demonstrative – proper noun co-occurrence in Irish and Blackfoot a. Muiris Bhidí seo (Irish) Muiris Bhidí DEM ‘this (person) Muiris Bhidí’ (adapted from McCloskey 2004) b. bhí urradh as miosúr i nGoll seo (Irish) be.PST strength out.of measure in Goll DEM ‘This (guy) Goll had astonishing strength.’ (Lit: ‘(There) was strength beyond measure in this Goll.) (adapted from McCloskey 2004) c. anna Tsóó (Blackfoot) ann-wa Tsóó-wa DEM-PROX Joey-PROX ‘Joey’ d. anna Tsím áyimmiwa (Blackfoot) ann-wa Tsím-wa á-yimmi-wa DEM-PROX Jim-PROX DUR-laugh.AI-3.SG ‘Jim is laughing.’ e. annááhka Sámahka áako’toowa (Blackfoot) ann-wa-hka Sam-wa-hka yáak-o’too-wa DEM-PROX-INVIS Sam-PROX-INVIS FUT-arrive-3.SG ‘Sam will arrive.’ (Frantz 2017: 71) f. * Sámahka áako’toowa (Blackfoot) Sam-wa-hka yáak-o’too-wa Sam-PROX-INVIS FUT-arrive-3.SG ‘Sam will arrive.’ If Giusti (2015) is correct that proper nouns occupy Spec,DP, then co-occurring demonstratives cannot also occupy that position. In languages such as Blackfoot where proper nouns are obligatorily introduced by a demonstrative (as in 155c-f), either the proper noun or the demonstrative must not be in Spec,DP. In this section, I suggest a hypothesis that preserves Giusti’s analysis of proper nouns being in Spec,DP, but organizes the demonstrative projection as part of the nominal spine – as I showed to be the case in Irish. Assuming that demonstrative projections can be part of the hierarchy of projections/universal syntactic spine, dominating the DP, we can model the matching prosodic constituency of a DP-dominating DemP in Blackfoot and check it against the attested phonological output.
The demonstrative in Irish and Blackfoot | 146
5.4.2. Modeling prosodic constituency of a DP-dominating DemP in Blackfoot Entertaining the possibility that the constituency of the DemP in Blackfoot is the same as in the genetically unrelated language, Irish (but without the strong uninterpretable definiteness feature that drives comp-to-spec movement in Irish), we arrive at the structure provided in (156). This structure crucially differs from that in (154) in that, similar to the conclusion for Irish demonstratives, the DemP is suggested to dominate DP, rather than being a specifier to DP. (156) Hypothesized structure of the Blackfoot DemP a. anna akaomitaa (inóókawa) ann-wa aka-omitaa-wa ino-oka-wa DEM4-PROX old-dog-PROX see.TA-INV-21 ‘That old dog (sees us).’ b.
By matching the structure in (156b), we incorrectly predict the prosodic structure in (157). (157) Unattested prosodic matching structure of (156) φ ← DemP
ω ← Dem φ ← DP anna ω ← Adj ω ← N/n aka- omitaawa The prosodic structure in (157) correctly predicts the possibility of vowel coalescence between the final /a/ vowel of the adjective and the initial /o/ vowel of the noun, which tend
The demonstrative in Irish and Blackfoot | 147 to surface (according to the data elicited for study 2 in §4.1.3) as [ɔ]. However, the right φ- boundary associated with the DemP is not aligned with the right edge of the demonstrative, but rather, coincides with the right edge of the φDP boundary. Therefore, this structure does not predict that aspiration would be observable on the final /a/ vowel of the demonstrative – a sound alternation we know to be present on 94.4% of the tokens analyzed in §4.1.3. One principled mechanism exists in the literature that could cause the structure in (156b) to make the correct predictions for phonological constituency: The spell-out to PF through cyclic phases. Phase Theory (Chomsky 2000, 2001, 2008; Gallego 2010) contends that syntactic material is spelled out to LF (Logical Form) and PF (Phonological Form) in cyclic spell-outs determined by phase heads in the syntactic component. Phase Theory preserves what has become known as the Generative inverted Y model of the grammar (interfaces), which is also assumed in Match Theory and adapted here: (158) Y-model of grammar (Elfner 2012: 4) Abstract Syntactic Structure
Spell-out
LF PF Under Phase theory, when the derivation of the phase head is complete (i.e., when the XMax label is generated), the sister to the phase head is spelled-out to the other grammatical components for interpretation and pronunciation, while the rest of the syntactic derivation continues. For example, if v0 is analyzed as a phase head, Merge combines the v0 with a VP and projects a v' (intermediate projection). Merge then combines the intermediate v' projection with a subject DP and projects a vP. Once the maximal projection vP has been generated, the sister to the v0, the VP and any VP complements, is spelled-out to LF and PF. Once a phase has been spelled-out to the other grammatical components, material within that phase is no longer accessible to the rest of the derivation (i.e., for feature valuation etc.). Phase Theory has been used to motivate prosodic phrasing by disallowing a single φ to contain material from separate phases (Kahnemuyipour 2004, 2009; Kratzer & Selkirk 2007; Ishihara 2007). Further, recent research has made used of Phase Theory and Match Theory in tandem, showing that the two are not incompatible (Selkirk 2011; Elfner 2012; Clemens 2016; Clemens & Coon to appear and references therein). The phase theoretic
The demonstrative in Irish and Blackfoot | 148 principle that prevents a single φ from containing material spelled out in separate phases does not contradict the MATCHPHRASE constraint in any way, but rather, it provides an additional condition that provides a more restrictive theory of prosodic phrasing. Clemens (2016) and Clemens & Coon (to appear) use Phase Theory to explain why their constraint
(ARGUMENT-φ) appears to impact prosodic constituency in some cases, but not others, while MATCHPHRASE always evaluates candidates in the formation of φs. Clemens (2016) and Clemens & Coon (to appear) adopt Chomsky’s (2001, 2008) position that the nominal parallel of C0 and v0 are phase heads, based on the parallelism that exists between the two spines – a parallelism that is a critical assumption of Bliss’ (2013) and Wiltschko’s (2014) analyses of Blackfoot. Chomsky assumes that the nominal parallel of C0 is (minimally) D0. For Blackfoot (and several other languages), though, Bliss and Wiltschko argue that D0 is the nominal parallel of I0, not C0 (see also Abney 1987). Therefore, in Blackfoot, the nominal parallel of C0 is the projection above D0 – the Link0 or K0 in Bliss or Wiltschko’s analyses, or the Dem0 in the hypothesis I put forward here (see also Windsor 2017, to appear; Windsor & Lewis to appear):107 (159) Nominal-verbal parallels a. [DemP [DP [ΦP [nP [NP ]]]]] b. [CP [IP [AspP [vP [VP ]]]]] c. [KP [DP [ΦP [nP [NP ]]]]]
The parallels given in (159) provide two comparisons with the verbal spine in (159b). The parallel in (159a) is the one argued for in Windsor (2017, to appear), Windsor & Lewis (to appear), and what I suggest here. The parallel in (159c) is the one argued for in Bliss (2013) and Wiltschko (2014). The major difference between the nominal spines in (159a and c) are that Bliss and Wiltschko do not consider the possibility of a DemP as a projection in the universal spine, but rather, as a specifier to DP.108 If we accept Chomsky’s (2001, 2008) conjecture that the nominal parallel of C0 is a phase head, then under the present hypothesis, we expect Dem0 to be a phase head (K0 in
107 Since I assume N-to-n movement, the noun obligatorily escapes the lower phase, headed by the nominal parallel of v0, the n0. Therefore, no pronounceable material is in the lower phase and it does not bear on the present discussion. 108 I suggest (Windsor to appear), based on evidence from Irish, Attic Greek, and Cree, that the KP is higher than the DemP, the nominal parallel of GroundP (Thoma 2014; Wiltschko 2014). See Appendix I for further details.
The demonstrative in Irish and Blackfoot | 149
Bliss and Wiltschko’s system). It follows from this prediction that, given that the evidence from adjective and article linearization presented in the previous section shows that the nominal cannot be moved outside of the DP, that the demonstrative and the DP are spelled out in separate phases, and must therefore be realized in separate φs: (160) Prediction of the Dem phase-head hypothesis (Windsor 2017: 13) a. anni sikomitaayi (áóhkiwa) ann-yi sik-omitaa-yi á-ohki-wa DEM-OBV black-dog-OBV DUR-bark.AI-3.SG ‘That black dog (is barking).’ b. DemP Predicted phase boundary Dem DP anni sikomitaayi If we match a prosodic structure to the spell-out of (160), we correctly predict aspiration at the right edge of the demonstrative once again: (161) Prosodic matching structure of (160) φ ← DemP φ ← DP
ω ← Dem sikomitaayi
anni The prosodic structure in (161) correctly predicts the phonological alternations that were reported in Chapter 4 and is in line with the originally proposed prosodic structure of Blackfoot presented in that chapter (repeated as 162 from 77). (162) An example of Blackfoot prosodic constituency
The structure proposed in (162) differs from the matched structure of a Dem in Spec,DP structure in one subtle way. The structure in (162) suggests that the demonstrative
The demonstrative in Irish and Blackfoot | 150 and the adjective + noun are each parsed into a φ (which we know to be the case through the aspiration evidence), and each of those φs are parsed into an intonational phrase (ι). The predicted structure from the Dem in Spec,DP analysis (which spells the demonstrative out in the same phase as the rest of the nominal expression) has the DemP and the adjective + noun both matched as φs in the same way, but then each of those φs are parsed into a recursive φ, matched from the DP: (163) Matching structure of DemP in Spec,DP φ ← DP
φ ← DemP φ ← nP
anni ω ← Adj ω ← N sik- omitaayi Based on all of the evidence found for prosodic categories in Chapter 4, there is no reliable way to discern between a φ and a recursive φ. Further, recursive prosodic categories are expected in Match Theory, so there is no theoretical reason to oppose the structure in (163) either. The benefit of assuming the structures in (160-162) lies in their cross-linguistic generalizability and the fact that the same spine that is motivated in Irish would also exist in Blackfoot (if we entertain the theoretical possibility of a universal spine hypothesis à la Wiltschko 2014). Therefore, the DP-dominating placement of the DemP is theoretically desirable, but not empirically necessitated. There is, however, one piece of phonological evidence that lends some support to this hypothesis. Like the distinction between prefixes and suffixes in Blackfoot, this evidence is from minimal weight restrictions and well-formedness constraints.
5.4.2.1. Well-formedness and the Blackfoot φ In §3.3 and §4.2.1, I discussed the necessity of the constraint BIN-MIN(φ) as a prosodic well-formedness constraint, which demands that a φ contain at least two ωs. This constraint prevented non-binary APs from being matched as φs in Irish. The following examples are repeated from (53, 97) and (54, 98) respectively:
The demonstrative in Irish and Blackfoot | 151
(164) Unattested mapping from simultaneously maximal/minimal A to φ AP/AMax *φ | → | A ω áille áille ‘beautiful’ (Elfner 2015: 1179) (165) Attested mapping from simultaneously maximal/minimal A to ω AP/AMax | A → ω áille áille ‘beautiful’ (Elfner 2015: 1179) There is reason to suspect that this constraint is also active in Blackfoot phonology since adjectival prefixes in Blackfoot nominal expressions, being non-binary, are realized as ωs, not φs – just as in Irish. We know from the aspiration evidence presented in Chapter 4 that the right edge of a φ is aligned with the right edge of a demonstrative in Blackfoot, but demonstratives are also non-binary: The φDemP does not contain two ωs. To make sense of the fact that the non-binary AP is matched as a ω but the non-binary DemP is matched as a φ is difficult, if not impossible, without appealing to phases. The observable evidence suggests that an adjectival prefix is matched as a ω and, together with the ω encompassing the noun and its inflections, is parsed into a well-formed, binary φ. But the non-binary φ encompassing only the demonstrative is not well-formed – it is prosodically too light. However, if the constraint demanding that all pronounceable material be parsed into a φ (e.g., the PARSE constraint of Prince & Smolensky 2004) dominates BIN-MIN(φ), then the DemP, as the sole head in its phase, can be realized as a φ despite violating well-formedness. Considering this well-formedness constraint on φs, we can examine the constituencies predicted by the Dem in Spec,DP analysis as compared to the DP-dominating DemP hypothesis: (166) Well-formedness and unattested matching of a Dem in Spec,DP structure a. anna akáómitaawa (inóókawa) ann-wa aka-wimitaa-wa ino-oka-wa DEM-PROX old-dog-PROX see.TA-INV-21 ‘That old dog (sees us).’ φ ← DP
ω ← DemP φ ← nP anna ω ← AP ω ← N/n aka- wimitaawa
The demonstrative in Irish and Blackfoot | 152
By considering well-formedness constraints and applying the same restrictions to demonstratives as was applied to other heads in Blackfoot nominal expressions, we once again fail to predict a φ boundary at the right edge of the demonstrative. However, the structure matched by an analysis that puts a phase boundary between the demonstrative and the rest of the nominal expression does make the correct phonological predictions: (167) Prosodic matching structure of Dem and DP in separate phases φ ← DemP φ ← DP
ω ← Dem ω ← AP ω ← N/n anna aka- wimitaawa The theoretical evidence from well-formedness supports the DP-dominating hypothesis of demonstrative constituency in Blackfoot. This hypothesis is based on cross- linguistic generalizability and allowing the same syntactic nominal spine attested in Irish to be applied to Blackfoot as well. However, what is needed to support or reject this hypothesis is further cross-linguistic testing in other languages outside of the Algonquian or Celtic families.
5.4.3. Summary In this chapter, I have provided both theoretical and empirical evidence for the syntactic and prosodic constituency of demonstrative elements in the two unrelated languages of Irish and Blackfoot. Further, I have offered a hypothesis that allows cross-linguistic generalizability of the constituency between the two languages based on a universal hierarchy of projections/universal spine (Wiltschko 2014). The evidence explored in §5.2 builds on the evidence provided in Chapter 3 and supports the conclusion that demonstrative elements are part of the Irish (Celtic) nominal spine, dominating the DP. Further linearization evidence suggested that the DP complement of the DemP undergoes raising to Spec,DemP for a strong uninterpretable definiteness feature on the Dem, which is checked by movement, in line with McCloskey’s (2004) syntactic analysis. The evidence explored in §5.3 builds on the evidence provided in Chapter 4 and rules out a head-movement approach to Blackfoot suffixation based on prosodic constituency and the linearization of demonstratives, definite articles, adjectives, and nouns. This evidence runs counter to the analysis of Blackfoot linearization algorithms for suffixation
The demonstrative in Irish and Blackfoot | 153 presented in Bliss (2013), but retains the position of morphosyntactic features motivated in that analysis. Instead of a head-movement analysis, I advance a proposal of agreement relations rather than movement for feature checking. This analysis was proposed in Chapter 4, but required additional evidence from demonstratives and determiners presented in the present chapter. Finally, I hypothesized a structure that allows cross-linguistic generalizability between Blackfoot (Algonquian) and Irish (Celtic), which gained some support through well- formedness restrictions. This hypothesis allows for a unified analysis of all nominal elements in Blackfoot. The hypothesis rests on the phase theoretic assumption that the Dem0 is a phase head, causing the demonstrative and the rest of the nominal expression to be spelled out to PF in two separate cycles. The cyclic spell-out therefore prevents the demonstrative and the rest of the nominal expression from being phonologically phrased together. The proposed structure allows for an analysis of Blackfoot that follows the same hierarchical structure as that motivated for Irish earlier in this chapter:
The demonstrative in Irish and Blackfoot | 154
(168) Dem0 as a phase head in Blackfoot anna akáómitaawa (inóókawa) ann-wa aka-wimitaa-wa ino-oka-wa DEM-PROX old-dog-PROX see.TA-INV-21 ‘That old dog (sees us).’ a. Morphosyntactic structure b. Prosodic matching structure φ ← DemP
ω ← Dem φ ← DP anna ω ← AP ω ← N/n aka- omitaawa
The structures in (168) accurately predict aspiration at the right edge of the demonstrative and at the right edge of the noun in Blackfoot. They provide a rationale for why the DemP should be matched as a φ, but not so the AP. They accurately predict the optional vowel coalescence of the final /a/ of the adjective with the initial /o/ of the noun, which is variably realized as [a.i] or [ɔ]. Moreover, they make use of the cross-linguistic generalizability (i.e., universality) of the nominal spine that underpins Bliss’ (2013) and Wiltschko’s (2014) analyses by being in line with the attested constituency of demonstratives in Irish: (169) Dem0 as a phase head in Irish a. Syntactic spell-out b. Attested prosodic matching structure DemP φ ← DemP
DPj[DEF] Dem'[uDEF*] φ ← DP ω Dem N A Dem
The demonstrative in Irish and Blackfoot | 155 demonstratives will not be lenited in this construction. They account for coordination and scope data (e.g., that a single demonstrative can scope over coordinated DPs). They do not force us to assume that a single syntactic category in Irish violates the headedness parameter-setting of the language. Furthermore, they do not cast doubt on the notion of phases contributing to the prosodic constituency of nominal expression since the movement of the DP to Spec,DemP allows the noun and adjective to escape the lower phase and be pronounced in the same φNon-Min as the demonstrative (a φNon-Min that is empirically motivated by the distribution of pitch accents). The structures that I hypothesize in this chapter for both Irish and Blackfoot are empirically motivated and theoretically desirable for cross-linguistic generalizability. However, further support from other, unrelated, languages is needed in order to understand whether or not the DP-dominating DemP position does indeed instantiate a universal function of the nominal syntactic spine.109 In §6.2 of the concluding chapter, I outline a typology that this analysis predicts, and I also outline evidence that could allow us to support or reject the hypothesis for demonstrative constituency presented here. All of the analyses presented in this dissertation rely on evidence from syntax and phonology in order to be successful. Syntactic structures informed phonological analysis in the case of Munster Irish stress shift, and worked in tandem with the phonological evidence as a deciding factor between the competing syntactic analyses of a right-headed demonstrative by Adger (2013) and a recursive DP by McCloskey (2004). Phonological data has allowed us to reject an analysis of Blackfoot suffixation through head-movement, and propose an alternative that utilizes agreement relations. This analysis was contingent on converging evidence from the syntax and phonology, and has the benefit of cross- linguistic generalizability. In the concluding chapter to follow, I summarize the various arguments from the syntax and phonology that have led to the conclusions drawn here, and provide testable predictions for future cross-linguistic study. Fin
109 See Appendix I (and Windsor 2017, to appear) for a Universal Spine Hypothesis (Wiltschko 2014) analysis of what that universal function may be, and further discussion of the role of KP in that spine.
Conclusions | 156
6. Conclusion In this dissertation, I have provided hypotheses about the phonological and syntactic hierarchical structure of nominal expressions in the unrelated languages of Irish and Blackfoot. Evidence was presented in the form of case studies detailing how the syntax- phonology interface could be used as a research tool by providing additional cross- component evidence to support component-internal evidence; by enabling experimentation through which a researcher can control for additional linguistic variables; or, by providing additional evidence to bear on a decision between competing analyses. In the following section, I briefly review the evidence presented to show: i. How utilizing a knowledge of the way phonological structure matches syntactic structure allowed for the control of prosodic variables and enabled research into the diachronic origins of a stress-shift phenomenon in Munster Irish; ii. how finding ways to recognize prosodic boundaries in Blackfoot was used to differentiate the syntactic operations of Merge versus Agree; and, iii. how cross-component evidence was used to motivate a hypothesis of syntactic structure that can be generalized across (at least) two unrelated languages. Following the summary presented in this section, I outline directions for future research in the area of demonstrative projections, and the expected typology of the structure hypothesized in Chapter 5 with predictions that could support or refute this proposal based on further cross-linguistic study.
6.1. Summary In Chapter 3, I used the matching of prosodic structure to morphosyntactic structure in order to control for prosodic variables in an elicitation experiment designed to learn the differences in phonetic instantiation of two levels of prosodic prominence: word (ω) stress and phrase (φ) accent. The study that I described in that chapter was contingent on being able to measure prominence on the same lexeme both with and without a φ-accent. Importantly, to avoid differences in the measures of the phonetic correlates based on other factors, the lexeme had to appear in the same position within the breath group. The hypothesized syntactic structure of Irish demonstrative phrases (DemP) —further motivated in Chapter 5— provided a way to isolate a lexeme in such a way that all of the conditions for the study were met. Based on the prosodic matching algorithms provided for
Conclusions | 157
Irish in Elfner (2012, 2015), we knew that an object noun, modified by a single adjective, is not in the correct prosodic position to receive a pitch accent. However, if the hypothesized syntactic structure of demonstratives motivated in Chapter 5 (see also Windsor 2016a) were correct, adding a demonstrative to the structure of the object nominal expression would create the correct prosodic environment to assign a pitch accent to that noun while maintaining its position within the breath group. The evidence gathered in that study suggested that the assumed structural representations were correct, and allowed the testing of the two levels of prominence. Additionally, the evidence gained in that study allowed for a hypothesis into the diachronic origins of a stress shift phenomenon in Munster Irish. Since the diachronic hypothesis was contingent upon the understanding of the differences and similarities of the two levels of prominence, it would not have been possible without starting with a hypothesized syntactic structure that allowed for the control of prosodic variables. In Chapter 4, I investigated Blackfoot, a language genetically unrelated to Irish. Rather than beginning with a syntactic assumption to allow for prosodic testing, in this chapter, I began by searching for observable demarcations of prosodic boundaries. I found: i. That the right boundary of φs could be reliably be recognized by aspiration (frequently described as devoicing when applied to a final vowel); ii. that the boundary between ωs could reliably be recognized by an optional process of vowel coalescence (applying with approximately 75% regularity); and, iii. that ω-internal adjacent vowels (the boundary between syllables (σ)) displayed an obligatory process of vowel coalescence. The evidence gained in two studies of Blackfoot prosody was then used to motivate a syntactic structure that the prosody is matched from. Specifically, the syntactic structure was examined to see if the observable prosodic alternations could be predicted by the syntactic constituent structure of Blackfoot nominal expressions provided by Bliss (2013). One mismatch between the assumed syntactic structure and observable prosodic structure emerged: The assumed syntactic structure provided by Bliss (2013) —owing to a linearization algorithm that achieved suffixation through successive head-movement— predicted that the plural suffix -iksi should be realized as a ω in prosodic constituent structure. However, I found no evidence to suggest that that was the case. Using prosodic well-formedness, I searched for a reason for the observable non-isomorphism between the assumed syntactic structure and
Conclusions | 158 observable prosodic structure, but was unable to motivate a prosodic rationale for the mismatch. Since I could not motivate the mismatch between components, I instead provided a hypothesis that Blackfoot suffixation might be achieved through agreement relationships with the heads that contain the various morphosyntactic features argued for by Bliss. On the basis of the data examined in that chapter, a firm conclusion could not be reached on Move versus Agree operations and Blackfoot suffixation. I continued the discussion of nominal expressions in both Irish and Blackfoot in Chapter 5, which explicitly investigated the constituency of demonstratives in the nominal spine of both languages. I argued that, even if the linear strings and ability to be inflected for various morphosyntactic features are very different in the two languages, the property that demonstratives in both languages have in common is their ability to encode referentiality. In examining Irish demonstratives, I utilized the evidence from pitch accent distribution gained in the study reported in Chapter 3, as well as evidence from initial consonant mutation (lenition), to reject a previous analysis of the demonstrative as low within the nominal spine (Roberts to appear). Two analyses of a high demonstrative position that dominates the DP were then contrasted. They were contrasted using the previously presented prosodic evidence in addition to evidence from the linearization of adnominal modifiers. Taken together, the prosodic and syntactic evidence for the constituency of demonstratives in Irish supported the DP-raising analysis of McCloskey (2004) over a right-headed demonstrative analysis of Adger (2013). However, to achieve the correct prosodic matching in order to explain the observable phonological phenomena, a modification to McCloskey’s analysis was suggested: That the label of the phrase projected by the demonstrative must be DemP rather than DP. Taking as a starting point the structure motivated for demonstrative constituency in Irish where the DemP dominated a DP, I then re-examined the evidence for the structure of Blackfoot nominal expressions. I showed that —provided suffixation is the result of agreement relationships providing feature checking rather than successive head- movement— the structure provided by Bliss (2013), with DemP in Spec,DP rather than dominating DP, could predict the observable phonological facts when matched. Assuming the goal of syntactic research to discover structural universals and the range of cross-
Conclusions | 159 linguistic variability, I explored the prosodic predictions made by an analysis of Blackfoot demonstratives that assumed the structure previously motivated for Irish. Utilizing the cyclic spell-out to the phonological component provided by Phase Theory (Chomsky 2000 et seq.), I showed that a structure where the DemP dominates DP could also predict the observable prosodic alternations. One piece of prosodic evidence favouring the DP-dominating DemP analysis over the DemP in Spec,DP analysis in Blackfoot was provided, once again in the form of a prosodic well-formedness constraint. I suggested that because the DemP is matched as a φ whereas AP is matched as a ω (even though neither meets minimal weight restrictions of dominating a minimum of two ωs), this distinction cannot be motivated using the DemP in Spec,DP analysis. However, this distinction can be motivated if the DemP is spelled out in a separate phase from the DP – as assumed in the DP-dominating DemP analysis. While the prosodic evidence supported the DP-dominating DemP analysis, and was theoretically desirable based on the cross-linguistic generalizability between (at least) Irish and Blackfoot, I did not provide a firm conclusion based on that evidence alone. Instead, I suggested that further cross-linguistic research is needed to support or reject that possibility for Blackfoot. Like the two preceding chapters, Chapter 5 on the prosodic and syntactic constituency of demonstratives utilized the interface as a crucial tool in conducting the analyses. In Chapter 3, the syntax was used to gain control over variables in prosodic study. In Chapter 4, observable prosodic alternations were used to challenge a syntactic conclusion that suffixation in Blackfoot is accomplished via head-movement. And, in Chapter 5, syntactic and prosodic evidence were used in concert to provide cross-component support for a hypothesis on the hierarchical structures (prosodic and syntactic) in two genetically unrelated languages. The goal of all of the studies undertaken in this dissertation was to show how the interface can be a useful tool in linguistic research, opening up additional cross-component evidence which can be brought to bear on a number of linguistic problems. That goal has been met using the interface in three ways: i. syntactically informed prosodic study; ii. prosodically informed syntactic study; and, iii. converging evidence from cross-component study. Based on the goals of this dissertation, determining whether the DP-dominating DemP hypothesis proposed in Chapter 5 represents a true syntactic universal or not is
Conclusions | 160 beyond the scope of the present study. To reach a conclusion on that issue would require an entire dissertation of its own, with a far greater range of cross-linguistic sources. Because the necessary cross-linguistic evidence for or against that hypothesized structure cannot be explored here, I provide a predicted typology in the following section, based on that hypothesis, which can be tested in the future, as well as other unresolved questions from these studies.
6.2. Predictions and limitations 6.2.1. Predicted typology The evidence presented in §5.4.2.1 from well-formedness supported the hypothesis of a DemP that dominates a DP in Blackfoot, but was not conclusive. This hypothesis could gain support, or be refuted, based on further cross linguistic study. Here, I provide a typology that the hypothesis for demonstrative constituency advanced in Chapter 5 predicts to inform future study and make testable predictions: (170) Demonstrative ordering typology (theoretical) a. Dem > D > N b. Dem > N > D c. D > Dem > N d. D > N > Dem e. N > D > Dem f. N > Dem > D
Considering the three nominal elements, demonstrative (Dem), determiner (D), and noun (N), (170) provides the logically possible orders that could exist. However, not all of these orderings are predicted based on the constituency hypothesized. Below, I provide structures for each of these typological possibilities accompanied by a language that may provide data for such a structure, and indicate which structures are not predicted by this hypothesis.
Conclusions | 161
(171) Dem > D > N (English) a. that Ø day DEM DEF day ‘that day’ b.
(172) Dem > N > D (Swedish) a. den dag-en DEM day-DEF ‘that day’ (Holmes & Hinchliffe 1994) b.
(173) D > Dem > N Not predicted: It is not clear what would drive D-to-Dem head-movement that would not contain successive N-movement. On the other hand, it should be noted that there is no specific principle that might prevent this movement (e.g., parallel T- to-C movement to form a yes/no question in English).
Conclusions | 162
(174) D > N > Dem (Irish) a. an lá sin DEF day DEM ‘that day’ b.
(175) N > D > Dem (Romanian) a. băiatul acesta (frumos) boy-DEF DEM (nice) ‘this boy’ (adapted from Brugè 2002:16)110 b.
(176) N > Dem > D Not predicted: It is not clear what would drive N-to-Dem head-movement that would skip over D0
In Chapter 5, I suggested that the difference between the Irish DemP as D > N > Dem and the Blackfoot DemP as Dem > D > N is attributed to a [uDEF] feature, which may be either strong (driving movement) or weak (not driving overt movement). This variation
110 Given the position of the adjective in this language, it is also possible that there is incorporation of the N+D0 inside the Dem0 showing only head movement and incorporation, rather than phrasal movement of any kind. Without further research into the phonology of words and clitics in Romanian, this is purely speculation given the available data. Brugè herself notes that there must be at minimum N-to-D movement.
Conclusions | 163 leads to the expectation of a language that shows both possible orderings. Michif, a contact language spoken in the Canadian Prairies, is a language that provides such evidence: (177) Multiple ordering in Michif nominal expressions (Rosen 2003) a. awa la fij DEM DEF girl ‘this girl’ (i.e., a specific girl) b. la fij awa DEF girl DEM ‘this girl’ (i.e., not the other one – contrastive reading) Given the hypothesis advanced here, we can compute the differences between the two realizations of Michif nominal expressions based on a feature in Dem:111 (178) Hypothesized structure of Michif DemP a.
b.
Based on the available data, the above structures are speculative. Additional elicitation evidence and prosodic testing are needed to support or refute this conjecture.
111 Note that the differences in interpretation between the two possible orders of nominal expression in Michif suggest a possible alternative analysis. It is possible that in (177b) and (178b), the DP raises to a higher functional projection, possibly a focus projection in the nominal domain, as discussed by Giusti (1996) and Aboh (2004). In that case, the feature that drives movement is not likely to be [uDEF*].
Conclusions | 164
All of the structures provided in (170-178) are provided as typological predictions based on the hypothesis that a DP-dominating DemP may be a linguistic universal.112 To support or reject this hypothesis, prosodic and syntactic evidence should be sought in these and other languages to see if this structure is universal, or merely a typological possibility that is evidenced in Irish. In addition to the need for future research from many more cross-linguistic sources, there are also open questions that some of the preceding discussion has raised for the languages under discussion in this dissertation. Some of these open questions are addressed below with special attention to questions that may present a problem for the hypothesis presented in Chapter 5 on DemP constituency.
6.2.2. Limitations Beyond typological predictions, there are also a few other unanswered questions that the data presented in this dissertation raise, which also require further research. These questions include: Exploring the consequences for case assignment and feature checking under the assumption that DemP dominates DP; locating the possible adjunction sites for Irish adnominal modifiers and investigating the possibility of syntactic or semantic restrictions on that attachment; determining the role of adjectives in Blackfoot (as distinct from stative verbs) and determining if there are semantic or lexical properties which limit the linearization of these elements; and, motivating the fact that verbs can appear between demonstratives and nouns in Blackfoot. This final question is of particular interest as it represents a possible problem for the DP-dominating DemP analysis, so I provide additional discussion on it now. In §5.1.1, I provided the following data as (114) to show the multiple linearization patterns that are allowable in Blackfoot:
112 Note that if this hypothesis is pursed under a Universal Spine Hypothesis (Wiltschko 2014), it is the function κ:referential that is predicted to dominate the κ:anchoring head (with which D normally associates), not necessarily a demonstrative category. See Appendix I for further details.
Conclusions | 165
(179) Linear order of demonstratives and nouns in Blackfoot a. áohkiwa oma imitááw [V > Dem > N] a-ohki-wa om-wa imitaa-wa IMPF-bark.AI-PROX DEM-PROX dog-PROX ‘That dog is barking.’ b. óóma áohkiwa imitááw [Dem > V > N] (Bliss 2013: 150) c. omi ponokáómitaayi iksíkkaayiyináyi [Dem > N > V] om-yi ponokaomitaa-yi ik-ikkaayi-yini-ayi DEM-OBV horse-OBV INTNS-canter.AI-OBV-3.SG.PRN ‘That horse is naturally swift.’ (Lit: it canters) d. * ponokáómitaayi omi iksíkkaayiyináyi *[N > Dem > V] (Bliss 2013: 157) If the DemP is a projection that dominates DP rather than being Merged in Spec,DP, the data in (179b) is unexpected. The construction in (179b) could perhaps be more easily explained if the DemP is a specifier that can undergo movement to a higher projection, above the verb, separate from the noun. It is not clear how a DP-dominating DemP would achieve this surface order. One possible way that the construction in (179b) could be accounted for in the DP- dominating DemP analysis would be to argue that the verb in that construction is nominalized and Merged as part of the nominal spine. The reason to posit this possible analysis is because the inflectional morphology on nouns and nominalized verbs is indistinguishable from one another, at least in the singular:
Conclusions | 166
(180) Morphology on nouns, nominalized verbs, and action verbs a. omiksi ííkaayaokstakiiksi sináákia’tsii om-iksi ííkaa-ya-okstaki-iksi sináákia’tsiS-i DEM-ANIM.PL PST:PERF-DUR-read.AI-ANIM.PL book-NON^REF ‘those who have read books’ (Frantz 2017: 129) b. oma áíssaakso’kaawa om-wa á-ssáak-Io’kaa-wa DEM-PROX DUR-try-sleep.AI-3.SG/PROX ‘that one who is trying to sleep’ (Frantz 2017: 129) c. áohkiyi omiksi imitáíksi á-ohki-yi om-iksi imitaa-iksi DUR-bark.AI-3PL DEM-ANIM.PL dog-ANIM.PL ‘Those dogs are barking.’ d. áohkiwa oma imitááwa á-ohki-wa om-wa imitaa-wa DUR-bark.AI-3.SG DEM-PROX dog-PROX ‘That dog is barking.’
If we compare the final -wa morpheme in (179a, b) (which Bliss 2013 glosses as PROX) with the final -wa morpheme in (180b, d), we see that they are indistinguishable. The final -wa on the verb ‘sleep’ in (180b) attaches to a nominalized verb, and the final -wa on the verb ‘bark’ in (180d) attaches to a verb in order to provide subject agreement with the 3.SG referent, ‘that dog’. However, in the plural (180a, c), there is a difference in final morphology. The nominalized verb in (180a) takes the animate plural ending -iksi while the verb with 3.PL subject agreement in (180c) takes a -yi suffix.113 Thus, if a construction such as that in (179b) where the verb intercedes between the demonstrative and the noun could be elicited with a plural referent, we would expect to be able to tell the difference between an action verb and nominalized verb based on the final morphology.114 If an analysis that locates the verbal element within the nominal spine (i.e., as a nominalization) cannot be motivated to achieve the linearization in (179b) (see also Bliss 2013: §7.2.3 on discontinuous expressions and non-configurationally), then this data is problematic for the DP-dominating DemP hypothesis advanced in Chapter 5. In such a case, we would need to explore other possibilities in order to motivate the difference between the
113 It is worth noting that if the verb was linearized at the end of the string in (180c), it would also have an attached pronoun -aawa that followed the -yi agreement morpheme. See Frantz (2017, Chapter 9) on attached pronouns for discussion. 114 For further diagnostics of nominalization in Blackfoot, see Bliss et al. (2016).
Conclusions | 167
ω-status of adjectival prefixes and the φ-status of demonstratives in Blackfoot. One possible solution to investigate, in order to resolve this difference (if the ordering in (179b) cannot be explained by the present hypothesis), would be to look closer at the prefixation of the adjective onto the noun. It is possible —since suffixation is not achieved via head-movement— that the A+N construction in Blackfoot is due to movement of N0 to A0 for compounding. If that were the case, we might expect the prosodic representation of the A+N construction to be a recursive ω, matched from a structure like: [NP (N (A adj)ω (N noun)ω)ω ]φ. Further tests to recognise recursive prosodic structure in Blackfoot would be required if this proved to be a necessary repair based on the rejection of a DP-dominating DemP structure for Blackfoot.
6.2.3. Consequences of theoretical assumptions This dissertation provides case studies that exemplify the benefit of utilizing the phonology- syntax interface as a source of additional evidence and what could be gained by doing so. In order to undertake the case studies presented in this dissertation, certain theoretical assumptions were needed, which may not necessarily align with a given reader’s theoretical commitment or interpretation. The fact that a particular theory was chosen through which to analyze the data will necessarily affect the conclusions presented herein, but does not affect the ability of the syntax-phonology interface to provide additional evidence to bear on a given problem. Take, for example, my assumption of Match Theory (Selkirk 2009, 2011). Match Theory is a phonological theory that guides the creation of prosodic categories through indirect reference to the morphosyntactic component. If a researcher makes the choice to pursue research under a theory of direct reference (e.g., Chomsky & Halle 1968), prosodic correlates of syntactic constituency are expected even more so – though the burden of explaining prosody-syntax mismatches under such an assumption might prove more challenging given the additional power of the theory. The present work also utilizes Phase Theory (Chomsky 2001, 2008) to analyze the Blackfoot DemP and assumes a common lenition environment (Windsor 2012, et seq.) to provide a hypothesized Irish nominal structure (later supported by pitch accent data). I briefly address the consequences of each of these assumptions now.
Conclusions | 168
In Chapter 5, Phase Theory was pursued as a possible motivation to explain the prosodic difference between Blackfoot demonstratives, which were evidenced to be φs, and adjectives, which were evidenced to be maximally ωs. From the analysis that promotes a DP-dominating DemP in Blackfoot, the Dem0 was predicted to be a phase head due to its parallel with the phase head C0 in the verbal domain (i.e., Chomsky’s 2001 parallelism conjecture). Following the analysis of the Dem0 as a phase head, it was predicted that the Blackfoot DemP would be phased, and thus phonologically phrased, separate from the rest of the nominal expression. The aspiration data analyzed in that chapter suggested that these predictions were borne out. Thus, Phase Theory was not contradicted by any of the data explored, and allowed further explanation of why non-branching demonstratives could be matched as φs in Blackfoot, but non-branching adjectives could not be. The assumption of Phase Theory was not problematic for any of the data explored herein. However, the danger of relying on Phase Theory to explain the attested phonological patterns in Blackfoot is that it becomes a necessary part of the DP-dominating DemP analysis; thus, if the hypothesis that the DemP is part of the nominal hierarchy of projections/universal nominal spine (see Appendix I) is extended to further unrelated languages, those languages must also be analyzed using Phase Theory. If evidence against the Dem0 as a phase head were discovered in those languages, it would necessarily cast doubt on the Dem0 as a phase head in Blackfoot. Turning now to the assumed analysis of a common prosodic environment for lenition in Irish (Windsor 2012 et seq.; cf. Green 2007; Gorrie 2011) in Chapters 1 and 3, this too may represent a point of contention for some readers. In particular, some may question my assumption that the definite article an is realized as a ω that can cause lenition on an adjacent noun. More broadly, I assume that definite articles can be matched as ωs from the D0 position in the syntactic component. This follows from the assumption provided by Elfner (2015) that, due to the fact that Match Theory utilizes indirect reference, the phonological component is not sensitive to the syntactic difference between functional and lexical categories (contra Selkirk & Shen (1990), Selkirk (1995), Truckenbrodt (1995, 1999), and others). This assumption allows Bennett et al. (2016) to match functional elements such as demonstratives and strong pronouns —D-elements themselves— as ωs in prosodic representation. The assumption that definite articles can be matched as ωs also
Conclusions | 169 correctly predicts the application of another phonological phenomenon in Irish: Coronal Blocking (Ní Chiosáin 1991; Green 2008; Windsor 2011, 2012). Coronal Blocking occurs when the contact edges of two prosodic words (or a proclitic and host, in Green’s (2008: 200) terms) are both coronal consonants, and parsed together as part of a recursive ω. In such constructions, lenition is prevented from being realized on the initial consonant of the target word. In the following examples, which are all lenition environments, the application of lenition can be recognized by the presence of
b. ard + sagart ω( ω(ard) ω(sagart)) ‘high priest’ ‘high’ ‘priest’ Coronal Blocking
c. tonn + báite ω( ω(tonn) ω(bháite)) ‘stormy ocean wave’ ‘wave’ ‘drown.VA’ Lenition
d. tonn + tuile ω( ω(tonn) ω(tuile)) ‘tidal wave’ ‘wave’ ‘flood.GEN’ Coronal Blocking
e. an + geal ω( ω(an) ω(gheal)) ‘very bright’ ‘very’ ‘bright’ Lenition
f. an + deas ω( ω(an) ω(deas)) ‘very nice’ ‘very’ ‘nice’ Coronal Blocking
g. an + bosca ω(an ω(bhosca)) ‘(of) the box’ ‘DEF’ ‘box.GEN’ Lenition
h. an + tairbh ω(an ω(tairbh)) ‘(of) the bull’ ‘DEF’ ‘bull.GEN’ Coronal Blocking
115 Examples (181b, d, f, h) are from Green (2008: 200). Examples (181a, c, e, g) are generated from de Bhaldraithe (1977 [1992/2013]).
Conclusions | 170
The representations in (181) are those assumed by Green (2008). Note that under the proposal of lenition in Windsor (2011, 2012; also assumed here) the representation of the definite article an plus noun in (181g, h) would be the same as for the (phonologically identical) degree word an ‘very’ in (181e, f): ω( ω(an) ω(bhosca)) ‘(of) the box’ and
ω( ω(an) ω(tairbh)) ‘(of) the bull’ respectively. There are several advantages to assuming that the definite article is matched as a ω, as in examples (181e, f): i. It provides a common environment for lenition/coronal blocking with the other prosodic environments provided in this list; ii. it allows the article to be treated the same as other D0-elements such as strong pronouns, which are matched as ωs; and, iii. the recursive ω structure that allows the definite article + noun combination to display coronal blocking is predicted based on the N-to-D syntactic structure provided in Chapter 3. In this case, the matching structures are:
[D [D an] [N tairbh]] = ω( ω(an) ω(tairbh)). If there were not N-to-D movement in Irish, resulting in a complex incorporated X0, it is not clear how the recursive ω structure necessary for the application of coronal blocking (as in 181h) would be created. Although there are several reasons to assume an N-to-D syntax for Irish nominal expressions, and to analyze Irish definite articles as being able to be evaluated by the
MATCHWORD constraint like all other X0s, this is not a crucial component of the analyses presented in this dissertation. These assumptions can be treated as generalizations that happen to capture the majority of lenition cases in Irish, and these generalizations were used (in Chapter 1) to form a hypothesis about the DemP structure in Irish only. That hypothesized structure provided the testable prediction that, if correct, would allow a pitch accent to be forced onto otherwise accentless object nouns. That testable prediction was investigated in Chapter 3, and was not falsified. Thus, the assumption of the lenition environment as posited in Windsor (2011, 2012) is not critical for the analyses presented in this dissertation, but it does provide supporting evidence to the analysis of the DemP presented in Chapter 3, based on data from the distribution of pitch accents. In sum, none of the assumptions above were essential to the central goal of the work: That attention to converging prosodic and syntactic evidence can deepen our understanding of phenomena in one grammatical component or the other. These assumptions merely make testable predictions, which provide a way to falsify the analyses presented herein, if not borne out upon further investigation.
Conclusions | 171
6.3. Conclusion Through the use of multiple case studies on the nominal expressions of two unrelated languages, I have illustrated how the syntax-phonology interface is a useful tool for research in either grammatical component. The interface can be used to control for linguistic variables during experimentation; it can be used to motivate hypothesized structures, which can then be supported with component-internal evidence; and, it can be used to check hypotheses against cross- component evidence to support or reject predicted structures. In sum, the cross-component evidence that the interface provides is an important instrument in the researcher’s toolbox that should not be overlooked. By understanding the interface between grammatical components, we open up a wealth of resources to inform, support, or reject an analysis. Fin
R e f e r e n c e s | 172
7. References Abels, Klaus. 2003 Successive cyclicity, anti-locality, and adposition stranding. Ph.D. dissertation, University of Connecticut. Abney, Stephen. 1987. The English noun phrase in its sentential aspect. Ph.D. dissertation, MIT. Aboh, Enoch. 2004. Topic and focus within D. Linguistics in the Netherlands 2004: 1-12. Adger, David. 1997. VSO and weak pronouns in Goidelic Celtic. Canadian Journal of Linguistics 42: 9-31. Adger, David. 2007. Pronouns postpose at PF. Linguistic Inquiry 38: 343-349. Adger, David. 2013. A syntax of substance. Cambridge, MA: The MIT Press. Acquaviva, Paolo. 2014. The categories of Modern Irish verbal inflection. Journal of Linguistics 50: 537-586. Applegate, J. R. 1958. An outline of the structure of Shilha. New York: ACLS. Arnhold, Anja, Richard Compton, & Emily Elfner. To appear. In M. Keough (ed.), Proceedings of the Workshop on Structure and Constituency in Languages of the Americas 21. Baker, Mark. 1985. The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16: 373-415. Beckman, M. E. & Jannet Pierrehumbert. 1986. Intonational structure in Japanese and English. In C. Ewen & J. Anderson (eds.), Phonology Yearbook 3: 255-309. Bennett, Ryan T. 2012. Foot-conditioned phonotactics and prosodic constituency. Ph.D. dissertation, University of California, Santa Cruz. Bennett, Ryan T. 2013. The uniqueness of metrical structure: Rhythmic phonotactics in Huariapano. Phonology 30: 355-398. Bennett, Ryan, Emily Elfner, & James McCloskey. 2016. Lightest to the right: An apparently anomalous displacement in Irish. Linguistic Inquiry 47: 169-234. Bennett, Ryan, Emily Elfner, & James McCloskey. Forthcoming. Prosody, focus and ellipsis in Irish. Bernstein, Judy. 1997. Demonstratives and reinforcers in Romance and Germanic language. Lingua 102: 87-113.
R e f e r e n c e s | 173
Bernstein, Judy. 2008. English th-forms. In H. H. Müller & A. Klinge (eds.), Essays on nominal determination: From morphology to discourse management, 213-332. Amsterdam: John Benjamins. Blankenhorn, Virginia S. 1981. Pitch, quantity and stress in Munster Irish. Éigse 18: 225- 250. Bliss, Heather Anne. 2013. The Blackfoot configurationality conspiracy: Parallels and differences in clausal and nominal structure. Ph.D. dissertation, UBC. Bliss, Heather, & Bettina Gruber. 2015. Temporal restrictions on personal pronouns: The composition of Blackfoot proclitics. Lingua 156: 175-199. Bliss, Heather, Elizabeth Ritter, & Martina Wiltschko. 2016. Blackfoot nominalization patterns. In M. Macaulay, M. Noodin, & R. Valentine (eds.), Papers of the 44th Algonquian Conference, 1-21. Boersma, Paul, & David Weenink. 2016. Praat: Doing phonetics by computer. Version 5.3.56. http://www.praat.org/. Brugè, Laura. 2002. The positions of demonstratives in the extended nominal projections. In G. Cinque (ed.) Functional structure in DP and IP, 15-53. New York: Oxford University Press. Carnie, Andrew. 1991. A domain based phonology: The evidence from Modern Irish. Proceedings of the Leiden Conference for Junior Linguists 3: 59-76. Carnie, Andrew. 1995. Non-verbal predication and head-movement. Ph.D. dissertation, MIT. Cinque, Guglielmo. 1990. Types of A’-dependencies. Cambridge, MA: MIT Press. Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance DP. In G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi, & R. Sanuttini (eds.), Paths towards universal grammar: Studies in honor of Richard S. Kayne, 85-110. Washington D.C.: Georgetown University Press. Cinque, Guglielmo. 2005. Deriving Greenberg’s universal 20 and its exceptions. Linguistic Inquiry 36: 315-332. Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.
R e f e r e n c e s | 174
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In R. Martin, D. Michaels, & J. Uriagereka (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89-155. Cambridge, MA: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In M. Kenstowicz (ed.), Ken Hale: A life in language, 1-52. Cambridge, MA: MIT Press. Chomsky, Noam. 2008. On phases. Current Studies in Linguistics Series 45: 133-166. Chomsky, Noam, & Howard Lasnik. 1993. The theory of principles and parameters. In J. Jacobs, A. von Stechow, W. Sternefeld, & T. Vennemann (eds.), Syntax: An international handbook of contemporary research, 506-569. New York: de Gruyter. Chomsky, Noam, & Morris Halle. 1968. The sound pattern of English. New York: Harper & Row. Christian Brothers. 2004. New Irish grammar. Dublin: C.J. Fallon. Chung, Sandra. 1978. Case marking and grammatical relations in Polynesian. Austin, TX: University of Texas Press. Chung, Sandra, & James McCloskey. 1987. Government, barriers, and small clauses in Modern Irish. Linguistic Inquiry 18: 173-237.
Clemens, Lauren. 2016. ARGUMENT-Φ: A prosodic account of pseudo noun incorporation. In B. Prickett & C. Hammerly (eds.), Proceedings of the 46th Meeting of the North East Linguistic Society, 217-226. Clemens, Lauren, & Jessica Coon. To appear. Prosodic constituency of verb-initial clauses in Ch’ol. Proceedings of the 21st Workshop on Structure and Constituency in Languages of the Americas. Dalton, Martha, & Ailbhe Ní Chasaide. 2003. Modeling intonation in three Irish dialects. Proceedings of the 15th International Congress of Phonetic Sciences, 1073-1076. Barcelona. Davis, Philip W., & Ross Saunders. 1975. Bella Coola deictic usage. Rice University Studies 61: 13-35. de Bhaldraithe, Tomás (ed.). 1977 [1992/2013]. Foclóir Gaeilge-Béarla. An Roinn Oideachais: An Gúm. Available online at: http://www.teanglann.ie/en/. de Cuba, Carlos. 2007. On (non)factivity: Clausal complementation and the CP-field. Ph.D. dissertation, Stony Brooke University.
R e f e r e n c e s | 175 de Cuba, Carlos, & Barbara Ürögdi. 2009. Eliminating factivity from syntax: Sentential complements in Hungarian. In M. den Dikken & R. Vago (eds.), Approaches to Hungarian vol. 11, 29-63. Amsterdam: John Benjamins. de Cuba, Carlos, & Jonathan MacDonald. 2012. Referentiality in Spanish CPs. In M. V. C. Toboada, Á. J. Fernández, J. M. González, & M. R. Tejedor (eds.), Information structure, agreement, and CP. Amsterdam: John Benjamins. de Cuba, Carlos, & Jonathan MacDonald. 2013. On the referential status of embedded polarity answers in Spanish. In J. C. Amara, G. Lord, A. de Prada Pérez, & J. E. Aaron (eds.), Selected Proceedings of the 16th Hispanic Linguistics Symposium, 312- 323. Massachusetts: Cascadilla Press. Déchaine, Rose-Marie, & Martina Wiltschko. 2002. Decomposing pronouns. Linguistic Inquiry 33: 409-442. den Dikken, Marcel. 2006. Relators and linkers: The syntax of predication, predicate inversion, and copulas. Cambridge, MA: MIT Press. Diessel, Holger. 1999. Demonstratives: Form, function and grammaticalization. Amsterdam: John Benjamins. Diessel, Holger. 2006. Demonstratives, joint attention, and the emergence of grammar. Cognitive Linguistics 17: 463-489. Doherty, Cathal. 1991. Munster Irish stress. In A. Mester & S. Robbins (eds.), Phonology at Santa Cruz 2: 19-32. Doherty, Cathal. 1996. Clausal structure and the Modern Irish copula. Natural Language and Linguistic Theory 14: 1-46. Dopke, Susanne. 1998. Competing language structures: The acquisition of verb placement by bilingual German-English children. Journal of Child Language 22: 555-584. Doyle, Aidan. 1998. Obligatory and optional movement: The evidence from Irish. In P. Stalmaszcsyk (ed.), Projections and mapping: studies in syntax, 43-56. Lublin: Wydawnictwo. Doyle, Aidan. 2002. Covert and overt pronominals in Irish. Lublin: Wydawnictwo Folium. Dresher, B. Elan. 2009. The contrastive hierarchy in phonology. New York: Cambridge University Press. Duffield, Nigel. 1995. Particles and projections in Irish syntax. Dordrecht: Kluwer.
R e f e r e n c e s | 176
Duffield, Nigel. 1996. On structural invariance and lexical diversity in VSO languages: Arguments from Irish noun phrases. In R. D. Borsley & I. Roberts (eds.), The syntax of the Celtic languages: A comparative perspective, 314-340. Cambridge: Cambridge University Press. Elfner, Emily. 2006. The mora in Blackfoot. MA Thesis: University of Calgary. Elfner, Emily. 2011. The interaction of linearization and prosody: Evidence from pronoun postposing in Irish. In A. Carnie (ed.), Formal approaches to Celtic linguistics, 17- 40. Newcastle upon Tyne: Cambridge Scholars Publishing. Elfner, Emily. 2012. Syntax-prosody interactions in Irish. Ph.D. dissertation, University of Massachusetts at Amherst. Elfner, Emily. 2014. Syntax-prosody mismatches in Irish and English verb-initial structures. Paper presented at Exploring the Interfaces 3: Prosody and Constituent Structure. McGill University, May 8. Elfner, Emily. 2015. Recursion in prosodic phrasing: Evidence from Connemara Irish. Natural Language and Linguistic Theory 33: 1169-1208. Ellis, Lucy, & William J. Hardcastle. 2002. Categorical and gradient properties of assimilation in alveolar to velar sequences: Evidence from EPG and EMA data. Journal of Phonetics 30: 373-396. Enç, Mürvet. 1991. The semantics of specificity. Linguistic Inquiry 22: 1-25. Flynn, Darin, Joseph W. Windsor, Vincent Carveth, Lisa Kennedy, & Awil Hashi. 2012. Unpacking the phonology of Somali consonant alternations. Unpublished ms., University of Calgary. Frantz, Donald G. 2017. Blackfoot grammar. 3rd edition. Toronto: University of Toronto Press. Frantz, Donald G. & Norma Jean Russell. 1995. Blackfoot dictionary of stems, roots, and affixes. 2nd edition. Toronto: University of Toronto Press. Gallego, Ángel J. 2010. Phase theory. Philadelphia: John Benjamins. Gick, Bryan, Heather Bliss, Karin Michelson, & Bosko Radanov. 2011. Articulation without acoustics: “Soundless” vowels in Oneida and Blackfoot. Journal of Phonetics 40: 46-53.
R e f e r e n c e s | 177
Giusti, Giuliana. 1992. Heads and modifiers among determiners: Evidence from Rumanian. In G. Cinque & G. Giusti (eds.), Advances in Rumanian Linguistics. Amsterdam: John Benjamins. Giusti, Giuliana. 1996. Is there a FocusP and a TopP in the noun phrase structure? University of Venice Working Papers in Linguistics 6: 105-128. Giusti, Giuliana. 2015. Nominal syntax at the interfaces: A comparative analysis of languages with articles. Newcastle upon Tyne: Cambridge Scholars Publishing. Gordon, Matthew. 2014. Disentangling stress and pitch accent: A typology of prominence at different prosodic levels. In H. Van der Hulst (ed.), Word stress: Theoretical and typological issues, 83-119. New York: Cambridge University Press. Gorrie, Collin. 2011. Adjective agreement in Gaelic: A case for mophophonological features. In A. Carnie (ed.), Formal approaches to Celtic linguistics, 321-336. Newcastle upon Tyne: Cambridge Scholars Publishing. Green, Antony D. 1996. Stress placement in Munster Irish. CLS 32: 77-91. Green, Antony D. 1997. The prosodic structure of Irish, Scots Gaelic, and Manx. Ph.D. dissertation, Cornell University. Green, Antony D. 2007. Phonology limited. Potsdam: Universität für Linguistik. Green, Antony D. 2008. Coronals and compounding in Irish. Linguistics 46: 193-213. Guilfoyle, Eithne. 1988. Parameters and functional projections. NELS 18: 193-207. Guilfoyle, Eithne. 1990. Functional categories and phase-structure parameters. Ph.D. dissertation, McGill University. Guilfoyle, Eithne. 2000. Tense and N-features in Irish. In A. Carnie & E. Guilfoyle (eds.), The syntax of verb-initial languages, 61-73. Oxford: Oxford University Press. Gussenhoven, Carlos. 1988. Intonational phrasing and the prosodic hierarchy. Phonologica 1988, 89-99. Gussmann, Edmund. 1997. Putting your best foot forward: Stress in Munster Irish. In J. Folke (ed.), Celts and Vikings: Proceedings of the Fourth Symposium of Societas Celtologica Nordica, 103-133. Goteborg: Goteborgs Universitet. Haegeman, Liliane. 2006. Conditionals, factives and the left periphery. Lingua 116: 1651- 1669.
R e f e r e n c e s | 178
Haegeman, Liliane, & Barbara Ürögdi. 2010. Referential CPs and DPs: An operator movement account. Theoretical Linguistics 36: 111-152. Hale, Kenneth, & Elisabeth O. Selkirk. 1987. Government and tonal phrasing in Papago. Phonology Yearbook 4: 151-183. Halle, Morris, & J. R. Vergnaud. 1987. An essay on stress. Cambridge, MA: MIT Press. Holmes, Philip, & Ian Hinchliffe. 1994. Swedish: A comprehensive grammar. New York: Routledge. Hyman, Larry. 1977. On the nature of linguistic stress. In L. Hyman (ed.), Studies in stress and accent, 37-82. Los Angeles: University of Southern California. Inkelas, Sharon. 1989. Prosodic constituency in the lexicon. Ph.D. dissertation. Stanford University. Iosad, Pavel. 2013. Head-dependant asymmetries in Munster Irish prosody. In S. Blaho, M. Krämer, & B. Morén-Duolljá (eds.), Nordlyd 40. Special XL Issue: A Fetschrift on the occasion of X years of CASTL phonology and Cur Rice’s Lth birthday, 66-107. Ishihara, Shinichiro. 2007. Major phrase, focus intonation, multiple spell-out. The Linguistic Review 24: 137-167. Ito, Junko, & Armin Mester. 2010. The onset of the prosodic word. In S. Parker (ed.) Phonological argumentation: Essays on evidence and motivation, 227-260. London: Equinox. Ito, Junko, & Armin Mester. 2013. Prosodic subcategories in Japanese. Lingua 124: 20-40. Kager, René. 1999. Optimality theory. New York: Cambridge University Press. Kane, Frances. 2015. The fine structure of the Irish NP. Ph.D. dissertation, Ulster University. Kahnemuyipour, Arsalan. 2004. The syntax of sentential stress. Ph.D. dissertation, University of Toronto. Kahnemuyipour, Arsalan. 2009. The syntax of sentential stress. Oxford: Oxford University Press. Keyser, Samuel J., & Kenneth N Stevens. 2006. Enhancement and overlap in the speech chain. Language 82: 81-109. Kratzer, Angelika, & Elisabeth Selkirk. 2007. Phase theory and prosodic spellout: The case of Verbs. The Linguistic Review 24: 93-135.
R e f e r e n c e s | 179
Ladd, Robert. 1986. Intonational phrasing: The case for recursive prosodic structure. In C. Ewen & J. Anderson (eds.), Phonology Yearbook 3:. 311-340. Ladd, Robert D., & James M. Scobbie. 2003. External sandhi as gestural overlap? Counterevidence from Sardinian. In J. Local, R. Ogden, & R. Temple (eds.), Phonetic Interpretation: Papers in Laboratory Phonology VI: 164-182. Laka, Itziar. 1990. Negation in syntax—on the nature of functional categories and projections. Ph.D. dissertation, MIT. Larson, Richard K. 1988. On the double object construction. Linguistic Inquiry 19: 355- 391. Lee, Seunghun, & Elisabeth O. Selkirk. 2013. Structural restrictions on H tone spread in Xitsonga. Presented at ETI3, McGill University. May 9. Leu, Thomas. 2008. The internal syntax of determiners. Ph.D. dissertation, New York University. Lewis, Blake. 2014. The syntax and semantics of demonstratives: A DP-external approach. Proceedings of the 2014 Annual Conference of the Canadian Linguistic Association, 1-15. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Mac Congáil, Nollig. 2005. Irish grammar book. Indreabhán: Cló Iar-Chonnachta. Manolessou, Io. 2001. The evolution of the demonstrative system in Greek. Journal of Greek Linguistics 2: 119-148. Matthewson, Lisa. 1998. Determiner systems and quantificational strategies: Evidence from Salish.The Hague: Holland Academic Graphics. McCarthy, John J., & Alan Prince. 1993. Generalized alignment. In G. Booij & J. van Marle (eds.), Yearbook of morphology 1993, 79-153. Dordrecht: Kluwer. McCarthy, John J., & Alan Prince. 1986 [1996]. Prosodic morphology. Report no. RuCCS- TR-32. New Brunswick, NJ: Rutgers University Centre for Cognitive Science. Reprinted in Selected Works of John J. McCarthy. Available at: http://works.bepress.com/john_j_mccarthy/54. McCloskey, James. 1991. Clause structure, ellipsis and proper government in Irish. Lingua 85: 259-302.
R e f e r e n c e s | 180
McCloskey, James. 1996. Subjects and subject positions in Irish. In R. D. Boresley & I. Roberts (eds.) The syntax of the Celtic languages: A comparative perspective, 241- 283. Cambridge: Cambridge University Press. McCloskey, James. 1999. On the right edge in Irish. Syntax 2: 189-209. McCloskey, James. 2004. Irish nominal syntax 1: Demonstratives. Unpublished ms. Available at http://ohlone.ucsc.edu/~jim/PDF/demon.pdf. Last accessed: July 2016. McCloskey, James. 2009. The syntax of clauses in Irish. Paper presented at Formal Approaches to Celtic Linguistics. March 23-27. McCloskey, James. 2011. The shape of Irish clauses. In Andrew Carnie (ed.) Formal approaches to Celtic linguistics, 143-178. Newcastle upon Tyne: Cambridge Scholars Publishing. Mhac an tSaoi, Máire. 2002. An bhean óg ‘the little woman’. In A. Bourke, S. Kilfeather, M. Luddy, M. Mac Curtain, G. Meany, M. Ní Dhonnchadha, M. O’Dowd, & C. Wills (eds.), The field day anthology of Irish writing: Irish women’s writing and traditions Vol. V. Morton, Elliott, & Joe Pater. 2012. Structure and substance in artificial-phonology learning (part I and II). Language and Linguistics Compass 6: 686-718. Mulkern, Ann E. 2003. Cognitive status, discourse salience, and information structure: Evidence from Irish and Oromo. Ph.D. dissertation, University of Minnesota. Mulkern, Ann E. 2011. Left right behind: Irish pronoun postposing and information structure. In A. Carnie (ed.), Formal Approaches to Celtic Linguistics, 179-202. Newcastle upon Tyne: Cambridge Scholars Publishing. Myers, Scott. 1987. Vowel shortening in English. Natural Language and Linguistic Theory 5: 485-518. Nespor, Marina & Irene Vogel. 1986 [2007]. Prosodic phonology. New York: Mouton de Gruyter. Ní Chiosáin, Máire. 1991. Topics in the phonology of Irish. Ph.D. dissertation. University of Massachusetts, Amherst. Ní Chiosáin, Máire. 2000. Prosodic well-formedness and sonority constraints: Epenthesis in Irish. Unpublished ms. Available at http://roa.rutgers.edu/files/89-0000/89-0000- CHIOSAIN-0-0.PDF. Last accessed: September 2017.
R e f e r e n c e s | 181
Nolan, Francis, Tara Holst, & Barbara Kühnert. 1996. Modelling [s] to [ʃ] accommodation in English. Journal of Phonetics 24: 113-137. Ó Conghaile, Eddie Bheairtle. 1999. Seoda ár sinsear. Connemara: An Chéad Chló. Ó Sé, Diarmuid. 2000. Gaeilge Chorca Dhuibhne. Baile Átha Cliath: Institiúid Teangeolaíochta Éireann. Ó Sé, Diarmuid. 2008. Word stress in Munster Irish. Éigse 36: 87-112. Ó Siadhail, Mícheál. 1989. Modern Irish: Grammatical structure and dialectal variation. Cambridge: Cambridge University Press. Ó Siadhail, Mícheál. 1995. Learning Irish. London: Yale University Press. Oberly, Stacey Inez. 2008. A phonetic analysis of Southern Ute with a discussion of Southern Ute language policies and revitalization. Ph.D. dissertation, University of Arizona. Prince, Alan & Paul Smolensky. 1993 [2004]. Optimality theory: Constraint interactions in generative grammar. Cambridge, MA: MIT Press. R core team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. http://www.R-project.org/. Riad, Tomas. 1992. Structures in Germanic prosody: A diachronic study with special reference to Nordic languages. Ph.D. dissertation, Stockholm University. Ritter, Elizabeth. 1991. Two functional categories in noun phrases: Evidence from Modern Hebrew. Syntax and Semantics 25: Perspectives on Phrase Structure, 38-62. New York: Academic Press. Ritter, Elizabeth. 2014. Featuring animacy. Nordlyd 14: 103-124. Ritter, Elizabeth, & Martina Wiltschko. 2009. Varieties of INFL: TENSE, LOCATION, and PERSON. In J. van Craenenbroeck (ed.), Alternatives to cartography, 153-201. Berlin: Mouton de Gruyter. Ritter, Elizabeth, & Martina Wiltschko. 2014. The composition of INFL: An exploration of tense, tenseless languages, and tenseless constructions. Natural Language and Linguistic Theory 32: 1331-1386. Ritter, Elizabeth, & Sara Thomas Rosen. 2010. Animacy in Blackfoot: Implications for event structure and clause structure. In M. Rappaport-Hovav, E. Doron, & I. Sichel
R e f e r e n c e s | 182
(eds.), Lexical semantics, syntax, and event structure, 124-152. New York: Oxford University Press. Rizzi, Luigi. The fine structure of the left periphery. In L. Haegeman (ed.), Elements of grammar: Handbook in generative syntax, 75-116. Dordrecht: Kluwer. Roberts, Ian. To appear. FOFC in DP: Universal 20 and the nature of demonstratives. In T. Biberauer, A. Holmberg, I. Roberts, & M. Sheehan (eds.) The final-over-final constraint: A word-order universal and its implications for linguistic theory. Cambridge, MA: The MIT Press. Roehrs, Dorian. 2013. The inner makeup of definite determiners: The case of Germanic. Journal of Germanic Linguistics 25: 295-411. Rosen, Nicole. 2003. Demonstrative position in Michif. Canadian Journal of Linguistics 48: 39-69. Russell, Kevin. 1999. The “word” in two polysynthetic languages. In T. A. Hall & U. Klenhenz (eds.), Studies on the phonological word, 203-222. Amsterdam: John Benjamins. Russell, Kevin. 2008. Sandhi in Plains Cree. Journal of Phonetics 36: 450-464. Russell, Kevin, & Charlotte Reinholtz. 1997. Nonconfigurationality and the syntax- phonology interface. In B. Agabayani & S.-W. Tang (eds.), Proceedings of the Fifteenth West Coast Conference on Formal Linguistics, 441-456. Stanford: CSLI Publications. Sadock, Jerrold M. 1980. Noun incorporation in Greenlandic: A case of syntactic word formation. Language 46: 300-319. Selkirk, Elisabeth O. 1981. The nature of phonological representation. In J. M. Anderson, J. Laver, & T. Myers (eds.), The cognitive representation of speech, 379-388. Dordrecht: North Holland Publishing Company. Selkirk, Elisabeth O. 1986. On derived domains in sentence phonology. Phonology Yearbook 3: 371-405. Selkirk, Elisabeth O. 1995. Sentence prosody: Intonation, stress and phrasing. In J. A. Goldsmith (ed.), The handbook of phonological theory, 123-171. Oxford: Blackwell. Selkirk, Elisabeth O. 2009. On clause and intonational phrase in Japanese: The syntactic grounding of prosodic constituent structure. Gengo Kenkyu 136. 35-73.
R e f e r e n c e s | 183
Selkirk, Elisabeth O. 2011. The syntax-phonology interface. In J. Goldsmith, J. Riggle, & A. Yu (eds.), The handbook of phonological theory 2nd edition, 435-484. Malden: Wiley-Blackwell. Selkirk, Elizabeth O., & Tong Shen. 1990. Prosodic domains in Shanghai Chinese. In S. Inkelas & D. Zec (eds.), The phonology-syntax connection, 313-337. Chicago: University of Chicago Press. Shattuck-Hufnagel, Stefanie, & Alice E. Turk. 1996. A prosody tutorial for investigators of auditory sentence processing. Journal of Psycholinguistic Research 25: 193-247. Shlonsky, Ur. 2004. The form of Semitic nominals. Lingua 114: 1465-1526. Stanley, Richard. 1967. Redundancy rules in phonology. Language 43: 393-436. Stevens, Kenneth N., & Samuel J. Keyser. 2012. Quantal theory, enhancement, and overlap. Journal of Phonetics 38: 10-19. Sybesma, Rint, & Joanna U. Sio. 2008. D is for demonstrative – Investigating the position of the demonstrative in Chinese and Zhuang. The Linguistic Review 25: 453-478. Szabolcsi, Anna. 1994. The noun phrase. In F. Kiefer & K. É. Kiss (eds.), The syntactic structure of Hungarian, 179-275. Bradford: Emerald Publishing Group Ltd. Szabolcsi, Anna. 2006. Strong and weak islands. In M. Everaert & H. van Riemsdijk (eds.), The Blackwell companion to syntax Vol. 4, 479-571. Malden: Blackwell. Thoma, Sonja. 2014. Bavarian discourse particles- at the syntax pragmatics interface. In The Proceedings of the Northwest Linguistics Conference 29.: 41-58. Thrift, Erica. 1997. A reanalysis of Munster Irish stress. Calgary Working Papers in Linguistics 19: 9-22. Truckenbrodt, Hubert. 1995. Phonological phrases: Their relation to syntax, focus, and prominence. Ph.D. dissertation, MIT. Truckenbrodt, Hubert. 1999. On the relations between syntactic phrases and phonological phrase. Linguistic Inquiry 30: 219-255. Tserdanelis, Georgios. 2005. The role of segmental sandhi in the parsing of speech: Evidence from Greek. Ph.D. dissertation, Ohio State University. Van Eijik, Jan. 1997. Lillooeet language: Phonology, morphology, syntax. Vancouver: UBC Press.
R e f e r e n c e s | 184
Weber, Natalie. 2016a. Initial extrametricality and cyclicity in Blackfoot accent. In A. Anghelescu, J. Dunham, & N. Weber (eds.), UBC Working Papers in Linguistics 44 (UBC Qualifying Papers 3.): 1-16. Weber, Natalie. 2016b. Accent and prosody in Blackfoot verbs. In M. Macaulay, M. Noodin, & J. R. Valentine (eds.), Papers of the 44th Algonquian Conference: Actes du Congrès des Algonquinistes, 348-369.SUNY Press. White, James. 2014. Evidence for a learning bias against saltatory phonological alternations. Cognition 130: 96-115. Wiltschko, Martina. 2009. What’s in D and how did it get there? In J. Ghomeshi, I. Paul, & M. Wiltschko (eds.), Determiners: Universals and variation, 25-66. Amsterdam: John Benjamins. Wiltschko, Martina. 2014. The universal structure of categories: Towards a formal typology. Cambridge: Cambridge University Press. Windsor, Joseph W. 2011. On the boundaries of Irish prosodic words. Calgary Working Papers in Linguistics 27: 1-15. Windsor, Joseph W. 2012. When nothing exists: The role of zero in the prosodic hierarchy. Saarbrücken, Germany: Lambert Academic Publishing. Windsor, Joseph W. 2014a. The distribution of Irish locatives (seo, sin, siúd): DP, AP, or other? Calgary Working Papers in Linguistics 28: 97-116. Windsor, Joseph W. 2014b. Recursive words and stress shift in Irish. Unpublished ms., University of Calgary. Windsor, Joseph W. 2015. Review of: Martina Wiltschko 2014. The universal structure of categories: Towards a formal typology. Linguistlist.org: October 30. Available at: https://linguistlist.org/pubs/reviews/get-review.cfm?SubID=36061317 Windsor, Joseph W. 2016a. Prosodic evidence for the syntactic constituency of demonstratives in Irish. Proceedings of the 2016 annual conference of the Canadian Linguistic Association. Windsor, Joseph W. 2016b. Contrast, phonological features, and phonetic implementation: Aspiration in Blackfoot. CWPL 29: 61-80. Windsor, Joseph W. 2017. Predicting prosodic structure by morphosyntactic category: A case study of Blackfoot. In L. Clemens & E. Elfner (eds.), Glossa: A Journal of
R e f e r e n c e s | 185
General Linguistics 2. Special issue on Prosody and Constituent Structure, 1-17. DOI: https://doi.org/10.5334/gjgl.229 Windsor, Joseph W. To appear. Blackfoot demonstratives, referentiality, and association with the syntactic spine. In M. Keough (ed.), Proceedings of the Workshop on Structure and Constituency in Languages of the Americas 21. UBCWPL. Windsor, Joseph W. & Blake Lewis. To appear. Constituency of demonstratives in Blackfoot: Evidence from the phonology, syntax, and semantics. In M. Keough (ed.), Proceedings of the Workshop on Structure and Constituency in Languages of the Americas 21. UBCWPL. Windsor, Joseph W. & J. L. Cobler. 2013. A unified explanation of three phrase-final phenomena in Blackfoot. Presented at the annual conference of the Canadian Linguistics Association. Victoria, BC. July 3. Windsor, Joseph W., Stephanie Coward, & Darin Flynn. To appear. Disentangling stress and pitch accent in Munster Irish. Proceedings of the West Coast Conference on Formal Linguistics 35. Zsiga, Elizabeth C. 1994. Gestural overlap in consonant sequences. Journal of Phonetics 22: 121-140. Zsiga, Elizabeth C. 1997. Features, gestures and Igbo vowels: An approach to the phonetics-phonology interface. Language 73: 227-274. Zsiga, Elizabeth C. 2000. Phonetic alignment constraints: Consonant overlap and palatalization in English and Russian. Journal of Phonetics 28: 69-102. fin
A p p e n d i c e s | 186
8. Appendix I: The Universal Spine Hypothesis The syntactic discussion in this dissertation is couched in a minimalist, generative theory (Chomsky 1995 et seq.). The previous research I draw on for Blackfoot syntax (Bliss 2013 and Wiltschko 2014) is couched in another syntactic theory, the Universal Spine Hypothesis (USH) (Wiltschko 2014, with earlier instantiations appearing in Ritter & Wiltschko 2009, 2014). I briefly outline the USH here to show how the hypothesis of Blackfoot demonstrative constituency advanced in the preceding chapters can be pursued under the USH.116 The USH is a framework for syntactic investigation that assumes a universal hierarchy of syntactic functions, but predicts cross-linguistic variation in what categories associate with those universal functions. This framework is based partly on research in Blackfoot, and is assumed by Bliss (2013) in her analysis of Blackfoot syntax. The USH has two core assumptions: (182) Core assumptions of the USH (adapted from Wiltschko 2014: 24) a. Language-specific categories (c) are constructed from a small set of universal categories (κ) and language-specific UoLs (Units of Language) b. The set of universal categories κ is hierarchically organized where each layer of κ is defined by a unique function (κ:linking, κ:anchoring, κ:point-of-view, κ:classification)
The core assumption of the USH given in (182a) is that syntactic categories are not homogeneous across languages. What is universal, is the function that categories associate with. For example, if (assuming the definition of the demonstrative given in (120)) the function of demonstratives in Irish and Blackfoot is referentiality (κ:referential), then it is possible that some other language could associate determiners (or other categories) with this function instead. In other words, it is the function, and that function’s position within the spine, that is universal, not the category that associates with that function. Categories in this framework are computed through the combination of language- specific UoLs and the function that they associate with: (183) Category computation (Wiltschko 2014: 24) c = κ + UoL
116 For a USH analysis of the function of Blackfoot demonstratives in line with the hypothesis presented in this chapter, see Windsor (to appear). See also Windsor (2015) for a review of Wiltschko (2014).
A p p e n d i c e s | 187
The algorithm in (183) entails the prediction that cross-linguistic variation is expected. Wiltschko (2014: §4.5) uses this heuristic to argue that, while a language like English might encode tense information under an I(nflection) category label (associated with κ:anchoring in her system), Halkomelem (a Salish language spoken in British Columbia) encodes person features in that position. An example of the algorithm in (183) is provided in (184) using Wiltschko’s (2014: 216) analysis of Blackfoot personal pronouns. A discussion of how to achieve the UoL in this system follows. (184) Computing Blackfoot personal pronouns c:1.SG.PRO = κ:anchoring + UoL To compute the language-specific UoL in the above example, we must combine a phonological form (π) and a logical form (Σ), which may also be compositional through either feature valuation (f-val) or morphological valuation (m-val) of a syntactic [ident/coin] feature.117,118 Using the personal pronoun category provided in (184), we can then compute the UoL of that category by combining the phonological and semantic forms, and associating them with a syntactic function: (185) Computing a first-person pronoun in Blackfoot (Wiltschko 2014: 216) niistó n-iistó 1.sg-self ‘myself’ c:1.SG.PRO: ⟨{{π:n-,Σ:1} {π:iistó ,Σ:identity}}, [κ:anchoring, m-val:+ident]⟩ Using the above example, Wiltschko argues that by combining the first-person n- with the reflexive identity marker iistó, and morphologically valuing the feature of the κ0 associated with the anchoring function as [+ident], we spell-out a first person, singular pronoun with the phonological form, niistó [niːstu]. This can be captured hierarchically as in the following tree diagram, adapted for Blackfoot from Wiltschko’s (2014: 212) syntax of reflexives example in German:
117 The semantic function (Σ) is distinct from the ΣP (polarity phrase) assumed in the earlier presentations of Irish clausal syntax (and from the stress foot (Σ) in prosodic representation). 118 The [±ident] feature is used in the nominal domain, and the [±coin] (coincidence) feature is used in the verbal domain.
A p p e n d i c e s | 188
(186) Derivation of niistó κ:anchoring
Proind κ
κ argind [+ident]
n- -iistó 1.SG identity The decision of which function a particular category associates with in either the nominal or verbal domain is based on the primary syntactic function that the category encodes (e.g., anchoring in the above examples using niistó). Associating a category with a particular function brings us to the second core assumption of the USH: The syntactic spine universally contains a series of hierarchical functions which are parallel between the nominal and verbal domain. Syntactic categories and features are language-specific, but the ordering of functions in the spine is universal (e.g., tense in English but person in Halkomelem both associating with the κ:anchoring function in the verbal spine). Wiltschko (2014) posits a set of common association with the functions she assumes to be part of the spine: (187) Common associations with the syntactic spine (Wiltschko 2014: 253)
To determine which part of the spine a given category associates with, Wiltschko (2014: 93) provides a diagnostic:
A p p e n d i c e s | 189
(188) The function diagnostic The absolute position of a given UoL can be diagnosed by identifying its function, which is independent of its content In Windsor (to appear), I argue that a corollary of the diagnostic given in (188) is that if a category displays evidence for a κ not contained in the original set of functions provided in (187) (and if that function is paralleled in the nominal and verbal spines), that is evidence for another level of the spine. Importantly, Wiltschko (2014) notes that the spine provided in (187) is preliminary, and encourages future research in determining what additional functions are needed within the universal spine. Thoma (2014), for example, argues for an additional layer that dominates the κ:linking function based on discourse particles in the left periphery of the clause. Thoma argues that the GroundP associates with a κ:grounding function, which dominates κ:linking. Grounding is defined as the moment-to-moment establishment of speaker and addressee conversational common ground. Importantly, I know of no research which currently argues for a nominal parallel of the κ:grounding function. However, in Windsor (to appear), I suggested that the nominal parallel of GroundP is KP (either associated with the function κ:grounding or κ:linking – deciding which function provided the best definition for the role of these syntactic categories was left to future research) and that the nominal parallel of CP is DemP, associating with the κ:referential function: (189) Revised universal spine (Windsor to appear: ex. 22)
A p p e n d i c e s | 190
The analysis that led to the revision of the syntactic spine provided in (189) was influenced by a number of factors: i. Previous research detailing the referential function of CP (Haegeman 2006; Szabolcsi 2006; de Cuba 2007; de Cuba & Ürögdi 2009; Haegeman & Ürögdi 2010; de Cuba & MacDonald 2012, 2013); ii. arguments concerning the referential function of demonstratives (see a brief synopsis of these arguments in §5.1.1); and, iii. the position of the demonstrative relative to the vocative particle in Irish and Attic Greek. Wiltschko (2014: §6.4.4.3) argues that the vocative particle is associated with the K0 (κ:linking function) in languages like Upper Austrian German where the K0 f-values a [+ident] feature on a lower head (D0), which in turn spells out the vocative morphology. In (190) I adapt Wiltschko’s (2014: 244) example of the vocative in Upper Austrian German for South East Cree: (190) Syntax of the vocative (Windsor to appear: ex. 19)
If we assume that vocative particles are in K0, as motivated by Wiltschko (2014: 244), and that the functional category below KP in the nominal spine is DemP (the hypothesis advanced in Chapter 5), then we can understand the co-occurrence of vocative particles and demonstratives in Attic Greek and Irish:
A p p e n d i c e s | 191
(191) Proposed structure of the Attic Greek vocative (Windsor to appear: ex. 20) a. Ὦ οὗτος, Αἶαν, δεύτερόν σε προσκαλῶ O hout-os Aian, deuteron se proskal-o VOC DEM-VOC Ajax.VOC second 2.SG call-1.SG.PRES.IND 'Oh Ajax, I call you again.' [Sophocles Trag., Ajax, Line 89]119 b.
(192) Proposed structure of the Irish vocative (Windsor to appear: ex. 21) a. a bhean udaí VOC woman.VOC DEM ‘Hey you over there (addressed to a woman)!’ (McCloskey 2004: 3) b.
The evidence presented in this dissertation allows me to conclude that the DemP in Irish is as provided in (192b) above. In §5.1.1, I argued that the primary syntactic function of demonstratives in Irish (and in Blackfoot) is to encode referentiality. If each of the phrases in the nominal spine must associate with a κ, then this suggests that DemP is commonly associated with κ:referential. The DemP was shown in Chapter 5 to dominate DP (at least in Irish and possibly in Blackfoot as well), and therefore, κ:referential must dominate κ:anchoring, a function that DP normally associates with. Given the relative
119 I am thankful to Blake Lewis for providing this example.
A p p e n d i c e s | 192 height diagnostic (Wiltschko 2014), since the DemP follows the vocative, which has previously been argued to associate with κ:linking (Wiltschko 2014: 244), the DemP (and thus κ:referential) must be a projection between DP (κ:anchoring) and KP (κ:linking). Finally, if such associations are common cross-linguistically, this allows us to also suggest a structure of the vocative in Attic Greek as in (191b), and provides a rationale for the hypothesized location of the DemP in Blackfoot, as suggested in Chapter 5. Fin
A p p e n d i c e s | 193
9. Appendix II: Elicitation sentences 9.1. Elicitation sentences used in Irish prominence study (Chapter 3) The data contained in this appendix are the sentences that were presented to participants in the Irish prominence study detailed in Chapter 3. The forms presented in this appendix are organized by target word. The three sentences used for elicitation of that word are then presented with the target in subject position (a), object position (b), and object position modified by a demonstrative (c). (193) amhráin ‘song.pl’ a. tógann amhráin ghra croíthe óga lift-PRES song.PL love.PL heart.PL young.PL ‘Love songs lift young hearts.’ b. canfaidh an ceoltóir doighiúil amhráin bheoga sing-FUT DEF musician handsome song.PL lively.PL ‘The handsome musician sang lively songs.’ c. canfaidh an ceoltóir doighiúil na amhráin sing-FUT DEF musician handsome DEF.PL song.PL bheoga sin lively.PL DEM ‘The handsome musician sang those lively songs.’ (194) bláthanna ‘flower.pl’ a. d’fhás na bláthanna áille faoin solas geal PST-grow DEF.PL flower.PL beautiful.PL under.DEF light bright ‘The beautiful flowers grew in the bright light.’ b. phlucáil an buachaill dóighiúil bláthanna corcra PST-pick DEF boy handsome flower.PL purple.PL ‘The handsome boy picked purple flowers.’ c. phlucáil an buachaill dóighiúil na bláthanna PST-pick DEF boy handsome DEF.PL flower.PL corcra úd purple.PL DEM ‘The handsome boy picked yon purple flowers.’
A p p e n d i c e s | 194
(195) bliain ‘year’ a. dá mbeadh bliain eile ag an seanduine cineálta PRT be.COND year other at DEF old-person kind ‘If (only) the old man had another year.’ b. imeoidh an mac-léinn óg i gceann bliain (amháin) go.away-FUT DEF student young in one year only ‘The young student will go away for one year.’ c. chonaic an mac-léinn óg san bhliain chrua sin PST-see DEF student young in.DEF year hard DEM ‘The young student saw that hard year.’ (196) bóithrín ‘country road’ a. casann an bóithrín cam go dtí twist-PRES DEF road-DIM crooked PRT come.PRES.SUBJ an trá álainn DEF beach beautiful ‘The crooked country road twists to the beautiful beach.’ b. shiúil an lánúin óg síos an bóithrín fada PST-walk DEF couple young down DEF road-DIM long ‘The young couple walked down the long country road.’ c. shiúil an lánúin óg síos an bóithrín fada sin PST-walk DEF couple young down DEF road-DIM long DEM ‘The young couple walked down that long country road.’ (197) cailín ‘girl’ a. rith an cailín láidir an camrian fada PST-run DEF girl strong DEF winding.track long ‘The strong girl ran the long winding track.’ b. ní fhaca an grianghrafadóir nua an cailín smaiseach NEG PST-see DEF photographer new DEF girl gorgeous ‘The new photographer didn’t see the gorgeous girl.’ c. ní fhaca an grianghrafadóir nua an cailín smaiseach NEG PST-see DEF photographer new DEF girl gorgeous úd dem ‘The new photographer didn’t see yon gorgeous girl.’
A p p e n d i c e s | 195
(198) caora ‘sheep’ a. ní fhaca an chaora dhubh sin an mac-tíre mór NEG PST-see DEF sheep black DEM DEF wolf big (sa pháirc) in.DEF field ‘The black sheep didn’t see the big wolf (in the field).’ b. dhíol an feirmeoir bocht a chaora dhubh PST-sell DEF farmer poor 3.SG.M.POSS sheep black ‘The poor farmer sold his black sheep.’ c. dhíol an feirmeoir bocht an chaora dhubh sin PST-sell DEF farmer poor DEF sheep black DEM ‘The poor farmer sold his black sheep.’ (199) cara ‘friend’ a. d’inis mo chara rúin dom faoin dlí aisteach PST-tell 1.SG.POSS friend bosom to.1.SG under.DEF law strange sin DEM ‘My bosom friend told me about that strange law.’ b. cuir glaoch gearr ar do chara saibhir put.IMP call short on 2.SG.POSS friend rich ‘Give a short call to your rich friend.’ c. cuir glaoch gearr ar an chara athmhalta sin put.IMP call short on DEF friend annoying DEM ‘Give a short call to that rich friend.’ (200) carraig ‘rock’ a. beidh an charraig bhan úd mar líne chríche be-FUT DEF rock white DEM as line finish ‘Yonder white rock will be the finish line.’ b. dreapfaidh na buachaillí láimhtheacha an charraig climb-FUT DEF.PL boy.PL daring DEF rock fhiontrach risky ‘The daring boys will climb the risky rock.’ c. dreapfaidh na buachaillí láimhtheacha an charraig climb-FUT DEF.PL boy.PL daring DEF rock fhiontrach úd risky DEM ‘The daring boys will climb yon risky rock.’
A p p e n d i c e s | 196
(201) casóigín ‘little jacket’ a. thit an casóigín liath san abhainn leathan PST-fall DEF jacket-DIM grey in.DEF river wide ‘The grey little jacket fell in the wide river.’ b. chniotáil an bhean Éireannach casóigín te PST-knit DEF woman Irish jacket-DIM warm ‘The Irish woman knitted a warm little jacket.’ c. chniotáil an bhean Éireannach an casóigín te seo PST-knit DEF woman Irish DEF jacket-DIM warm DEM ‘The Irish woman knitted this warm little jacket.’ (202) ceannaitheoir ‘buyer’ a. dhiultaigh an ceannaitheoir scrábach an praghas PST-refuse DEF buyer miserly DEF price réasúnta reasonable ‘The miserly buyer refused the reasonable price.’ b. scioll an siopadóir feargach an ceannaitheoir PST-scold DEF shopkeeper angry DEF buyer coigilteach frugal ‘The angry shopkeeper scolded the frugal buyer.’ c. scioll an siopadóir feargach an ceannaitheoir PST-scold DEF shopkeeper angry DEF buyer coigilteach seo frugal DEM ‘The angry shopkeeper scolded this frugal buyer.’ (203) ciorcal ‘circle’ a. (nuair a d’éirigh mé inniu) bhí ciorcal dearg ar when PRT PST-rise 1.SG today PST-be circle red on mo theilifís nua 1.SG.POSS television new ‘(When I got up today) a red circle was on my new television.’ b. rinne an tealaíontóir leiscúil ciorcal cam PST-make DEF artist lazy circle crooked ‘The lazy artist made a crooken circle.’ c. rinne an tealaíontóir leiscúil ciorcal cam sin PST-make DEF artist lazy circle crooked DEM ‘The lazy artist made that crooked circle.’
A p p e n d i c e s | 197
(204) cistin ‘kitchen’ a. tá cistin an-mhór i mo theach nua be kitchen very-big in 1.SG.POSS house new ‘A very big kitchen is in my new house.’ b. dheisigh an siúinéir oilte an chistin bháite PST-fix DEF carpenter experienced DEF kitchen flooded ‘The experienced carpenter fixed the flooded kitchen.’ c. dheisigh an siúinéir oilte an chistin PST-fix DEF carpenter experienced DEF kitchen bháite sin flooded DEM ‘The experienced carpenter fixed the flooded kitchen.’ (205) cóicín ‘cocaine’ a. ní raibh aon chóicín rúnda ag an bpríosúnach NEG PST-be any cocaine secret at DEF prisoner baolach dangerous ‘The dangerous prisoner didn’t have any secret cocaine.’ b. ghabh na gárdaí diograiseacha an cóicín rúnda PST-seize DEF.PL guard.PL diligent.PL DEF cocaine secret ‘The diligent guards seized the secret cocaine.’ c. ghabh na gárdaí diograiseacha an cóicín rúnda PST-seize DEF.PL guard.PL diligent.PL DEF cocaine secret sin DEM ‘The diligent guards seized that secret cocaine.’
A p p e n d i c e s | 198
(206) coinín ‘rabbit’ a. theith an coinín beag ón sionnach ocrach PST-run.away DEF rabbit small from.DEF fox hungry ‘The little rabbit ran away from the hungry fox.’ b. ní bhfuair an sealgaire mífhoighneach an coinín NEG PST-find DEF hunter not-patient DEF rabbit éalaitheach elusive ‘The impatient hunter didn’t find the elusive rabbit.’ c. ní bhfuair an sealgaire mífhoighneach an coinín NEG PST-find DEF hunter not-patient DEF rabbit éalaitheach úd elusive dem ‘The impatient hunter didn’t find yon elusive rabbit.’ (207) coinneal ‘candle’ a. dhóigh an choinneal gheal i rith na hoíche PST-burn DEF candle bright in run DEF night stoirmiúla stormy ‘The bright candle burned throughout the stormy night.’ b. leáigh an ghrian the na coinnle nua PST-melt DEF sun hot DEF.PL candle.PL new.PL ‘The hot sun melted the new candles.’ c. leáigh an ghrian the na coinnle nua seo PST-melt DEF sun hot DEF.PL candle.PL new.PL DEM ‘The hot sun melted these new candles.’ (208) coinnlín ‘candle-dim’ a. thug an coinnlín céarach solas geal PST-give DEF candle-DIM wax light bright ‘The little wax candle gave bright light.’ b. lasfaidh an córbhuchaill óg an coinnlín naofa light-FUT DEF choir-boy young DEF candle-DIM holy ‘The young choirboy will light the holy little candle.’ c. lasfaidh an córbhuchaill óg an coinnlín naofa sin light-FUT DEF choir-boy young DEF candle-DIM holy DEM ‘The young choirboy will light that holy little candle.’
A p p e n d i c e s | 199
(209) craidhstín ‘swear-word’ a. d’fhág an craidhstín anbhal an béal salach PST-leave DEF swear.word rude DEF mouth dirty ‘The rude swear-word left the dirty mouth.’ b. dúirt an buachaill dána craidhstín anbhal PST-say DEF boy bold swear.word rude ‘The bold boy said the rude swear-word.’ c. dúirt an buachaill dána craidhstín an anbhal seo PST-say DEF boy bold swear.word DEF rude DEM ‘The bold boy said this rude swear-word.’ (210) farraige ‘sea’ a. shuncáil an fharraige gharbh go leor bádaí iascaigh PST-sink DEF sea rough PRT ample boat.PL fishing.PL ‘The choppy sea sunk many fishing boats.’ b. tháinig an bád beag thar an fharraige gharbh PST-come DEF boat small over DEF sea rough ‘The little boat came over the rough sea.’ c. tháinig an bád beag thar an fharraige gharbh sin PST-come DEF boat small over DEF sea rough DEM ‘The little boat came over that rough sea.’ (211) máilín ‘bag-dim’ a. bhí máilín deas ag an gcailín óg PST-be bag-DIM nice at DEF girl young ‘The young girl had a nice little bag.’ b. d’fholaigh an fear mímhacánta an máilín bradach PST-hide DEF man not-honest DEF bag-DIM stolen ‘The dishonest man hid the stolen little bag.’ c. d’fholaigh an fear mímhacánta an máilín bradach PST-hide DEF man not-honest DEF bag-DIM stolen sin DEM ‘The dishonest man hid that stolen little bag.’
A p p e n d i c e s | 200
(212) mála ‘bag’ a. thit an mála daor ar an urlár salach PST-fall DEF bag expensive on DEF floor dirty ‘The expensive bag fell on the dirty floor.’ b. ceannóidh an bhean mhóiréiseach an mála daor buy-FUT DEF woman pretentious DEF bag expensive ‘The pretentious woman will by the expensive bag.’ c. ceannóidh an bhean mhóiréiseach an mála daor seo buy-FUT DEF woman pretentious DEF bag expensive DEM ‘The pretentious woman will by this expensive bag.’ (213) ógánach ‘youth’ a. dúirt an t-ógánach dána an fhírinne mhithaitneamhach PST-say DEF young.man bold DEF truth not-pleasant ‘The bold youth told the unpleasant truth.’ b. dhamhsaigh na cailíní uaibhreacha leis na hógánaigh PST-dance DEF.PL girl.PL spirited.PL with DEF.PL young.man.PL leisciúla shy.PL ‘The spirited girls danced with the shy youths.’ c. dhamhsaigh na cailíní uaibhreacha leis na hógánaigh PST-dance DEF.PL girl.PL spirited.PL with DEF.PL young.man.PL leisciúla úd shy.PL DEM ‘The spirited girls danced with yon shy youths.’ (214) oíche ‘night’ a. thit an oíche fhada thar an thír bhocht PST-fall DEF night long over DEF country poor ‘The long night fell over the poor country.’ b. longraigh na réaltaí geala san oíche dhorcha PST-shine DEF.PL star.PL bright.PL in.DEF night dark ‘The bright starts shone in the dark night.’ c. longraigh na réaltaí geala san oíche dhorcha sin PST-shine DEF.PL star.PL bright.PL in.DEF night dark DEM ‘The bright starts shone in that dark night.’
A p p e n d i c e s | 201
(215) páipéar ‘paper’ a. d’inis an páipéar nuachta an scéal brónach PST-tell DEF paper news DEF story sad ‘The news paper related the sad story.’ b. scríobhfaidh an file cairdiúil ar an bpáipéar write-FUT DEF poet friendly on DEF paper athchúrsáilte recycled ‘The friendly poet will write on the recycled paper.’ c. scríobhfaidh an file cairdiúil ar an bpáipéar write-FUT DEF poet friendly on DEF paper athchúrsáilte seo recycled DEM ‘The friendly poet will write on this recycled paper.’ (216) páipéarachas ‘stationary’ a. thug an páipéarachas oifigiúil an nuacht mhaith PST-give DEF stationary official DEF news good ‘The official stationary delivered the good news.’ b. d’athchúrsáil cailliúnaí scrábach an páipearachas saor PST-recycle spendthrift miserly DEF stationary cheap ‘The miserly spendthrift recycled the cheap stationary.’ c. d’athchúrsáil cailliúnaí scrábach an páipearachas saor seo PST-recycle spendthrift miserly DEF stationary cheap DEM ‘The miserly spendthrift recycled this cheap stationary.’
A p p e n d i c e s | 202
(217) parailéalachas ‘parallelism’ a. déanfaidh parailéalachas díreach argóint mhaith make-FUT parallelism direct argument good ‘Direct parallelism will make a good argument.’ b. úsáideann údair mhóra parailéalachas folaithe use-PRES author.PL big.PL parallelism hidden ‘Great authors use hidden parallelism.’ c. úsáideann údair mhóra an parailéalachas folaithe use-PRES author.PL big.PL DEF parallelism hiden sin DEM ‘Great authors use that hidden parallelism.’ (218) peacúlacht ‘sinfulness’ a. thaibhsigh peacúlacht nua thar an tsean-drochchathair PST-loom sinfulness new over DEF old-evil-city ‘A new sinfulness loomed over the evil old city.’ b. d’éirigh an fear mallachtach as an bpeacúlacht PST-rise DEF man accursed out.of DEF sinfulness domhain deep ‘The accursed man rose from the deep sinfulness.’ c. d’éirigh an fear mallachtach as an bpeacúlacht PST-rise DEF man accursed out.of DEF sinfulness domhain sin deep DEM ‘The accursed man rose from that deep sinfulness.’ (219) peannaireacht ‘penmanship’ a. beidh peannaireacht dheas ag gach duine anseo be-FUT penmanship nice at every person here ‘Every person here will have nice penmanship.’ b. d’fhoghlaim an buachaill óg peannaireacht dheas PST-learn DEF boy young penmanship nice ‘The young boy learned nice penmanship.’ c. scríobh an buachaill óg san pheannaireacht dheas PST-write DEF boy young in.DEF penmanship nice seo DEM ‘The young boy wrote in this nice penmanship.’
A p p e n d i c e s | 203
(220) peintiméadar ‘pentameter’ a. léigh an peintiméadar iambach don chailín cliste read.IMP DEF pentameter iambic to.DEF girl clever ‘Read the iambic pentameter to the clever girl.’ b. scríobh an Bard cóir i bpeintiméadar iambach PST-write DEF bard true in pentameter iambic ‘The true Bard wrote in iambic pentameter.’ c. scríobh an Bard cóir san bpeintiméadar iambach PST-write DEF bard true in.DEF pentameter iambic sin DEM ‘The true Bard wrote in that iambic pentameter.’ (221) pingin ‘penny’ a. ní fhaca an phingin nua dorn dúnta NEG PST-see DEF penny new fist closed ‘A new penny never saw a closed fist.’ b. níor bhronn an seanfhear scrábach ach pingin NEG.PST PST-donate DEF old-man miserly but penny rua red ‘The miserly old man didn’t donate but a read penny.’ c. níor bhronn an seanfhear scrábach ach an NEG.PST PST-donate DEF old-man miserly but DEF pingin rua seo penny red DEM ‘The miserly old man didn’t donate but this read penny.’
A p p e n d i c e s | 204
(222) píobaire ‘piper’ a. íoc an píobaire tallannach dhá phunt airgid pay.IMP DEF piper talented two pound Stirling ‘Pay the talented piper two Stirling pounds.’ b. d’íoc an bhean dathúil an píobaire PST-pay DEF woman generous DEF piper Albanach Scottish ‘The generous woman paid the Scottish piper.’ c. d’íoc an bhean dathúil an píobaire PST-pay DEF woman generous DEF piper Albanach sin Scottish DEM ‘The generous woman paid that Scottish piper.’ (223) sámhán ‘nap’ a. thit an sámhán gear ar an seanduine cam PST-fall DEF nap short on DEF old-man crooked ‘The short nap fell on the crooked old man.’ b. thóg an naíonán cipeánta sámhán fada PST-take DEF infant cranky nap long ‘The cranky infant took a long nap.’ c. thóg an naíonán cipeánta an sámhán fada PST-take DEF infant cranky DEF nap long sin DEM ‘The cranky infant took that long nap.’ (224) saoirse ‘freedom’ a. chuir an tsaoirse luachmhar fuarchrith ionainn PST-put DEF freedom costly cold-shiver in.1.PL ‘The costly freedom gave us cold shivers.’ b. shaothraigh an gaiscíoch láidir saoirse luachmhar PST-earn DEF hero strong freedom costly ‘The strong hero earned a costly freedom.’ c. shaothraigh an gaiscíoch láidir an saoirse luachmhar PST-earn DEF hero strong DEF freedom costly seo DEM ‘The strong hero earned this costly freedom.’
A p p e n d i c e s | 205
(225) sealbhán ‘group’ a. bhí sealbhán bheag san uisce fuar PST-be group small in.DEF water cold ‘A small group was in the cold water.’ b. thosaigh an t-óganach imníoch an sealbhán foscúil PST-start DEF youth anxious DEF group shady ‘The anxious youth started the shady group.’ c. thosaigh an t-óganach imníoch an sealbhán foscúil PST-start DEF youth anxious DEF group shady sin DEM ‘The anxious youth started that shady group.’ (226) searcóg ‘sweetheart’ a. fuair na searcóga deasa úlla úra PST-find DEF.PL sweetheart.PL nice.PL apple.PL fresh.PL ‘The nice sweethearts found fresh apples.’ b. phóg an buachaill dána a shearcóg PST-kiss DEF boy bold 3.SG.POSS sweetheart álainn beautiful ‘The bold boy kissed his beautiful sweetheart.’ c. d’fhluich an buachaill dána na searcóga PST-wet DEF boy bold DEF.PL sweetheart.PL armacacha seo loving DEM ‘The bold boy soaked these loving sweethearts.’
A p p e n d i c e s | 206
(227) siompóisiam ‘symposium’120 a. lean an siompóisiam leadránach thar an tseachtain PST-persist DEF symposium boring over DEF weekend fhada long ‘The boring symposium persisted over the long weekend.’ b. rachaidh an gnóthadóir rathúil go dtí go-FUT DEF business.man successful PRT come.PRES.SUBJ an siompóisiam suimiúil DEF symposium interesting ‘The successful businessman will go to the interesting symposium.’ c. rachaidh an gnóthadóir rathúil go dtí go-FUT DEF business.man successful PRT come.PRES.SUBJ an siompóisiam suimiúil sin DEF symposium interesting DEM ‘The successful businessman will go to that interesting symposium.’ (228) tacas ‘easel’ a. tá an tacas beag ag an péintéir óg sin be DEF easel small at DEF painter young DEM ‘The young painter has a small easel.’ b. dheisigh an péintéir cleachtach an tacas briste PST-fix DEF painter experienced DEF easle broken ‘The experienced painter fixed the broken easel.’ c. dheisigh an péintéir cleachtach an tacas briste seo PST-fix DEF painter experienced DEF easle broken DEM ‘The experienced painter fixed this broken easel.’
120 Excluded from study. See fn. 48.
A p p e n d i c e s | 207
(229) tosaitheoir ‘beginner’ a. bhris an tosaitheoir dearmadach an téad tanaí PST-break DEF beginner absent.minded DEF string thin (ar a ghiotár) on 3.SG.M.POSS guitar ‘The absent-minded beginner broke the thin string (on his guitar).’ b. ní raibh san iomaitheoir áitheasach ach tosaitheoir NEG PST-be in.DEF competitor successful but beginner éiritheach lucky ‘The successful competitor was not but a lucky beginner.’ c. ní raibh san iomaitheoir áitheasach ach an tosaitheoir NEG PST-be in.DEF competitor successful but DEF beginner éiritheach úd lucky DEM ‘The successful competitor was not but yon lucky beginner.’ (230) tóstalachán ‘conceited person’ a. bhéic an tóstalachán scrábach ar an bhean PST-yell DEF conceited.person miserly on DEF woman chineálta kind ‘The miserly conceited person yelled at the kind woman.’ b. ná déan plamas mór le tóstalachán NEG.IMP do.IMP compliment big with conceited.person míthréitheach not-talented ‘Don’t pay a big compliment to an untalented conceited person.’ c. ná déan plamas mór leis an tóstalachán NEG.IMP do.IMP compliment big with DEF conceited.person míthréitheach sin not-talented DEM ‘Don’t pay a big compliment to that untalented conceited person.’
A p p e n d i c e s | 208
(231) údarás ‘authority’ a. bhí údarás intuigthe ag an dlíodóir cliste PST-be authority implied at DEF lawyer clever ‘The clever lawyer had implied authority.’ b. fuair an dlíodóir sliceanta an t-údarás intuigthe PST-get DEF lawyer sly DEF authority implied (óna chliant) from.3.SG.M.POSS client ‘The sly lawyer got the implied authority (from his client).’ c. fuair an dlíodóir sliceanta an t-údarás intuigthe PST-get DEF lawyer sly DEF authority implied seo (óna chliant) DEM from.3.SG.M.POSS client ‘The sly lawyer got this implied authority (from his client).’ (232) foclóirín ‘wordlist’121 a. chríochnaigh an foclóirín iomadúil ar an gceann PST-finish DEF wordlist excessive on DEF one socair easy ‘The excessive wordlist ended on the easy one.’ (233) Filler tokens a. bhí go leor deiseanna griangrafadóraiochta agam nuair a bhí mé ann ‘I had a lot of opportunities for photography when I was there.’ b. sin é an chaoi is diabhal neart air ‘That is the way and there’s devil all you can do about it.’ c. éist le ceol na habhann agus gheobhaidh tú breac ‘Listen to the sound of the river and you’ll get a trout.’ d. is minic a bhris béal duine a shorn ‘It’s many a time (that) a person’s mouth broke his nose.’ e. cé leis pictiúr an fhir seo de chuid Picasso? ‘Who owns this picture of the man by Picasso?’ f. ní bhíonn suas gan síos ‘There’s no up without a down.’ g. tagann ciall le haois ‘Wisdom comes with age.’
121 Excluded due to the fact that it was not elicited in all contexts.
A p p e n d i c e s | 209 h. tagann ciall le haois ‘Good(ness) comes with friendship.’ i. tosach maith is leath na hoibre ‘A good beginning is half the work.’ j. más fada an lá, tiocfaidh an oíche ‘As long as the day may be, night will come.’ k. ní dhéanann eilifint dearmad ar rud ar bith ‘Elephants never forget.’ l. tá sé ráite ná fuil sé oiliúnach a bheith ag inseachtit scéalta insa lá ‘They say that it’s not right to be telling stories during the day.’ m. tá súile ina chúl aige ‘He has eyes in the back of his head.’ n. tháinig sé ar ais ar nós na drochaimsire ‘He returned like bad weather.’ o. d'fhoghlaimeofá an cineál sin ruda i gcoláiste sagartóireachta ‘You would learn that kind of think in seminary school.’ p. is olc an ghaoth nach séideann do dhuine eicínt ‘It’s an ill wind that blows nobody any good.’ q. go dtuga Dia slán go ceann cúrsa thú ‘May God carry you safe to your journey’s end.’ r. cén fáth a bhfuil mé tuirseach i gcónaí? ‘Why am I always tired?’ s. an bhfeiceann tú an fear sin a bhfuil an hata aisteach dubh air? ‘Did you see that man with the strange black hat?’ t. An fear a dhéanann é, ní chaitheann sé é, agus an fear a chaitheann é, ní fheiceann sé é ‘The man who makes it, doesn’t use it, and the man who uses it, doesn’t see it.’
A p p e n d i c e s | 210
9.2. Pitch accent distribution
A p p e n d i c e s | 211
9.3. Elicitation targets used in Blackfoot study 1 (Chapter 4) (234) Elicitation carrier sentences for study 1 (see §4.1.1.) a. ámo anistápssiwa ______(sentence-final target) amo anistápssi-wa (animate noun) DEM be.called.AI-3.SG ‘This is a(n) ______.’ b. ámo ______nitsináána (sentence-medial target) amo (animate noun) n(it)-itsináán-wa DEM 1.SG-possession-PROX ‘This ______is my pet.’ (235) Target nouns (uninflected) a. áíksini ‘pig’ b. áísaayoohkomi ‘bull’ c. apánii ‘butterfly’ d. áápotskina ‘cow’ e. áwákaasii ‘deer’ f. iiníí ‘bison’ g. imitáá ‘dog’ h. ksissohksísii ‘mosquito’ i. ksiwáwákaasii ‘spider’ j. mai’stóó ‘crow’ k. mamíí ‘fish’ l. maká’pipiitaa ‘bat’ m. naamóó ‘bee’ n. ni’tawáakii ‘chicken’ o. omahkatayo ‘mountain lion’ p. omahksstooki ‘donkey’ q. píítaa ‘eagle’ r. ponoká ‘elk’ s. ponokáómitaa ‘horse’ t. sai’áí ‘duck’ u. siipisttoo ‘owl’
A p p e n d i c e s | 212
9.4. Elicitation targets used in Blackfoot study 2 (Chapter 4) (236) Elicitation frame for study 2 (see §4.1.3.) a. anna (___)______wa inóokawa DEM4-PROX adj-animate_noun-PROX see.TA-INV-21 ‘That ((adj)animate noun) sees us.’ b. anniksi (___)______iksi inóókawa DEM4-ANIM.PL adj-animate_noun-ANIM.PL see.TA-INV-21 ‘Those ((adj)animate nouns) see us.’ (237) Adjectives used for study 2 a. aká ‘many’ b. áka ‘old’ c. akoksska ‘of paramount rank’ d. apoyi ‘earth-coloured’ e. ihta ‘lucky’ f. ikkihkini ‘sad/boring’ g. ikkina ‘slow/soft’ h. iima ‘pitiful’ i. immoyi ‘hairy’ j. inssta ‘neat’ k. i’táám ‘happy’
(238) Animate nouns used for study 2 (uninflected) a. ááattsistaa ‘rabbit’ b. áápotskina ‘cow’ c. aitsikiiskaa ‘cobbler’ d. áókatsitsimaa ‘maker of laws’ e. apánii ‘butterfly’ f. asóótokiaa ‘fool’ g. áyinnamaa ‘hawk’ h. áwákaasii ‘deer’ i. íímachkihkinaa ‘mountain goat’ j. iihtáíkahksiststakio’p ‘saw’ k. imiihkayi ‘swan’ l. imtáá ‘dog’ m. iiníí ‘bison’ n. itákka ‘friend’ o. óko’siipokaa ‘colt’ p. omahkapi’si ‘timber wolf’ q. omahkatayo ‘mountain lion’ r. omahkinaa ‘old man’ s. onista ‘calf’ t. otáá ‘weasel’ fin