UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Aspectual distinctions in Sk̲wx̲wú7mesh Bar-el, Leora Anne 2005-12-31

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
831-ubc_2006-12976X.pdf [ 17.02MB ]
Metadata
JSON: 831-1.0092760.json
JSON-LD: 831-1.0092760-ld.json
RDF/XML (Pretty): 831-1.0092760-rdf.xml
RDF/JSON: 831-1.0092760-rdf.json
Turtle: 831-1.0092760-turtle.txt
N-Triples: 831-1.0092760-rdf-ntriples.txt
Original Record: 831-1.0092760-source.json
Full Text
831-1.0092760-fulltext.txt
Citation
831-1.0092760.ris

Full Text

A S P E C T U A L DISTINCTIONS IN S K W X W U 7 M E S H  by LEORA ANNE BAR-EL B . A . , University of Western Ontario, 1996 M . A . , University of British Columbia, 1998  A THESIS S U B M I T T E D IN P A R T I A L F U L F I L M E N T OF T H E R E Q U I R E M E N T S FOR T H E D E G R E E OF DOCTOR OF PHILOSOPHY In T H E F A C U L T Y OF G R A D U A T E STUDIES  (Linguistics)  T H E UNIVERSITY OF BRITISH C O L U M B I A December 2005  © Leora Anne Bar-el, 2005  Abstract The classification of predicates according to their aspectual properties has a long history, dating back to Aristotle. Perhaps the most influential classification can be attributed to Vendler (1967). The time schemata to distinguish his four classes relies on a combination of entailment patterns and behaviours of "verbs" in different structures. Since Vendler, many researchers have revisited this classification, differing on both the proposed number of classes as well as the ways in which they are derived. Although they use different diagnostics to motivate their systems, what these approaches seem to share in common is the claim that aspectual classes are universal. This thesis addresses this claim and proposes that based on data from Skwxwu7mesh (a.k.a. Squamish), the representations of predicates vary crosslinguistically. I argue for a classification based on the presence/absence of intrinsic initial and final points in predicate representations. Chapters Two and Three are concerned with final points and initial points, respectively. I present four diagnostics which I argue test for the presence of final points and two diagnostics that test for the presence of initial points. Based on the results of these tests, I propose a modification of Rothstein's (2004) predicate templates (that in turn are a modification of Dowty's 1979 templates) to account for the classification of Skwxwu7mesh predicate classes that emerges. Chapters Four and Five are concerned with perfectivity and imperfectivity, respectively. In these chapters, I motivate the claim that Skwxwu7mesh has both a progressive marker and an imperfective marker. I propose that adopting Dowty's (1979) analysis of the progressive and Kratzer's (1998) analysis of the imperfective, along with the predicate representations introduced in chapters two and three, can derive the readings of progressive and imperfective predicates in Skwxwu7mesh. Based on a small study involving 10 native speakers of English who are not linguists, in Chapter Six I briefly revisit English aspectual classes. Using the results of some of the diagnostics from chapters two and three, I show the contrast between English and Skwxwu7mesh predicate representations, highlighting the claim that aspectual classes do indeed vary cross-linguistically.  ii  Table of Contents Abstract  ii  Table of Contents  iii  List of Abbreviations  viii  Acknowledgements  x  Chapter 1: Determining Aspectual Meaning 1. On the (non-)universality of aspectual meaning 2. What is aspect? 3. Aspectual Classes and Viewpoint Aspect in Skwxwu7mesh: overview 3.1. "Open Early. Open Late": Initial points and Final Points (Chapters 2-3) 3.2. Perfective and Imperfective (Chapters 4-5) 3.3. Cross-linguistic variation (Chapter 6) 4. Proposal: the organization of predicate meanings in Skwxwu7mesh 5. Previous aspectual classifications 5.1. "Verb" classes 5.1.1. Actualitity, Movement and Action: Aristotle 5.1.2. Activities, Accomplishments, Achievements and States:  Vendler (1967) 5.1.3. Advantages and Disadvantages 5.2. Feature-based definitions of predicate classes 5.2.1. [±POINT, ±EXTENDED, CONTINUOUS]: Carlson (1981) 5.2.2. [±STATIC, ±DURATIVE, ±TELIC]: (Smith 1997)  5.2.3. Advantages and Disadvantages 5.3. Structurally-based definitions of predicate classes  1-61 1 3 5 5 6 6 7 9 9 9  10 14 : 15 15 19  28 30  5.3.1. [±ADD TO, ±SPECIFIED QUANTITY OF A, ±TERMINATIVE]:  Verkuyl (1993) 5.3.2. STATE, PROCESS, TRANSITION: Pustejovsky (1991, 1995) 5.3.3. Advantages and Disadvantages 5.4. Operator-based definitions of predicate classes 5.4.1. BECOME, DO, CAUSE: Dowty (1979) 5.4.2. BECOME, DO: Rothstein (2004) 5.4.3. Implementation: Theoretical assumptions 6. Grammatical Aspect 7. The Study of Aspect in Salish 8. Research Methods 8.1. Elicitation 8.2. Texts 9. The Fieldwork Experience 10. Contribution of this thesis  Chapter 2: Intrinsic Final Points  30 35 37 39 40 45 49 50 53 57 57 59 59 61  62-136  iii  1. On the relevance of final points 62 1.1. The presence of intrinsic final points in Skwxwu7mesh predicates 62 1.2. Outline of this chapter 64 2. Diagnosing final points 64 2.1. Felicity conditions for culmination 65 2.1.1. Entailment vs. Implicature 65 2.1.2. Why these diagnostics require the perfective viewpoint 66 2.1.3. Test 1: Culmination Cancellation 67 2.1.4. Test 2: Event Continuation 69 2.2. Event cancellation vs. Event non-completion 70 2.1.1. Test 3: The scope of kiln 'almost' 70 2.1.1.1. Ambiguity induced by almost 70 2.1.1.2. Almost as Evidence for Lexical Decomposition 72 2.1.2. Test 4: The scope of negation 74 3. Diagnosing final points in Skwxwu7mesh: felicity conditions for culmination 75 3.1. Skwxwu7mesh activities 75 3.1.1. Skwxwu7mesh activities do not culminate 75 3.1.2. Skwxwu7mesh activities can continue 77 3.1.3. Skwxwii7mesh activities have no intrinsic final points 79 3.2. Skwxwu7mesh accomplishments 81 3.2.1 Skwxwu7mesh accomplishments need not culminate 81 3.2.2. Skwxwu7mesh accomplishments can continue 83 3.2.3. Skwxwu7mesh accomplishments have no intrinsic final points 84 3.3. Skwxwu7mesh achievements 87 3.3.1. Skwxwu7mesh achievements necessarily culminate 87 3.3.2. Skwxwu7mesh achievements cannot be continued 88 3.3.3. Skwxwu7mesh achievements have final points 92 3.4. Skwxwu7mesh inchoative states .....94 3.4.1. Skwxwu7mesh inchoative states can be continued 94 3.4.2. Skwxwu7mesh inchoative states have no final points 95 3.5. Summary and concluding remarks 96 3.5.1. Testing perfectivity vs. testing telicity 97 3.5.2. Cross-Salish comparison 98 4. Diagnosing final points in Skwxwu7mesh: event cancellation vs. event non-completion 100 4.1. Test 3: The scope of Skwxwu7mesh kiln 'almost' 101 4.1.1. Skwxwu7mesh activities: kiln induces event cancellation only 104 4.1.2. Skwxwii7mesh accomplishments: kiln induces event cancellation only 105 4.1.3. Skwxwu7mesh achievements: kiln induces event cancellation only... 109 4.1.4. Skwxwu7mesh inchoative states: kilh induces event cancellation only 112 4.1.5. Summary of Test 3 114 4.2. Test 4: The scope of Skwxwu7mesh negation 115 4.2.1. Skwxwu7mesh activites: negation induces event cancellation only... 116  iv  4.2.2. Skwxwu7mesh accomplishments: negation induces event cancellation only 4.2.3. Skwxwu7mesh achievements: negation induces event cancellation only 4.2.4. Skwxwu7mesh inchoative states: negation induces event cancellation only 4.2.5. Summary of Test 4 4.3. Summary: Event cancellation vs. Event non-completion 5. Intrinsic final points vs. pragmatically-conditioned final points 5.1. Pragmatically-conditioned final points in out-of-the-blue contexts 5.1.1. Skwxwu7mesh activities: past or present 5.1.2. Skwxwu7mesh accomplishments: past culminated 5.1.3. Skwxwu7rnesh achievements: past culminated 5.1.4. Skwxwu7mesh inchoative states: past inchoative or present stative... 5.1.5. Summary 5.2. Deriving the Skwxwu7mesh Accomplishment Culmination Implication 6. Conclusion  Chapter 3: Intrinsic Initial Points  117 119 120 120 121 122 123 123 124 125 126 127 127 135  137-201  1. On the relevance of initial points 137 1.1. The presence of intrinsic initial points in Skwxwu7mesh predicates 138 1.2. Outline of this chapter 139 2. Diagnosing Initial Points 140 2.1. Test 1: readings induced by punctual clauses/adverbials 140 2.1.1. On the reported effect of punctual clauses/adverbials 141 2.1.2. Punctual clause/adverbials in Skwxwu7mesh 149 2.2. Test 2: readings induced by mi 'come' 149 3. Test 1: The readings induced by Punctual Clauses/Adverbials in Skwxwu7mesh 151 3.1. Skwxwu7mesh activities 153 3.1.1. Punctual clauses induce inceptive readings 153 3.1.2. Skwxwu7mesh activities have initial points 157 3.1.3. Punctual clauses trigger a shift operation: activities-achievements.... 157 3.1.4. Summary 159 3.2. Skwxwu7mesh accomplishments 160 3.2.1. Punctual clauses induce inceptive, medial and culminative readings..160 3.2.2. Skwxwu7mesh accomplishments do not have intrinsic initial points.. 166 3.2.3. Summary 167 3.3. Skwxwu7mesh achievements 168 3.3.1. Skwxwu7mesh achievements: punctual clauses induce entire event reading 168 3.3.2. Skwxwu7mesh achievements have intrinsic initial points :.169 3.4. Inchoative States 171 3.4.1. Skwxwii7mesh inchoative states: punctual clauses induce inchoative readings 171 3.4.2. Skwxwu7mesh inchoative states have intrinsic initial points 173 3.5. Summary and a cross-Salish comparison 174  4. Test 2: The readings induced by the auxiliary mi 4.1. Activities 4.2. Accomplishments 4.2.1. Speaker Variation 4.3. Achievements 4.4. Inchoative States 4.5. Summary 5. Some alternative analyses 5.1. Initial points are introduced by the perfective 5.2. The sub-interval property: are activities homogenous? 5.3. Skwxwu7mesh accomplishments have initial points 6. Summary and concluding remarks  Chapter 4: Perfectivity  176 176 179 182 184 188 190 192 192 194 196 199  202-242  1. Introduction 2. The Perfective in Skwxwu7rnesh: overview 3. Skwxwu7mesh and the Typology of Perfectivity 3.1. Standard Perfective: English 3.2. Completive Perfective: Dene Syhne 3.3. Semi-perfective: Thai 3.4. Neutral perfective: Hindi 3.5. Summary 4. The Skwxwu7mesh Perfective is of the "standard" type 4.1. Perfective Activities 4.2. Perfective Accomplishments 4.3. Perfective Achievements 4.4. Perfective Inchoative states 4.5. Summary 5. The alternatives 5.1. Alternative 1: Skwxwu7mesh has a null imperfective or progressive 5.2. Alternative 2: Skwxwu7mesh has a neutral viewpoint 6. Conclusion  Chapter 5: Imperfectivity  202 203 205 207 209 214 217 223 224 226 228 229 230 231 231 231 237 242  243-304  1. Introduction 2. Imperfectivity in Skwxwu7mesh: overview 3. Skwxwu7mesh and the typology of imperfectivity 3.1. The Progressive 3.1.1. A cross-linguistic look 3.1.2. Analyses of the Progressive 3.2. The Imperfective 3.2.1. A cross-linguistic look 3.2.2. Analyses of the Imperfective 4. The Skwxwu7mesh CV- reduplicant is a progressive marker 4.1. Progressive Activities 4.2. Progressive Accomplishments  vi  243 244 246 247 248 256 260 260 263 265 266 268  5.  6. 7. 8.  4.3. Progressive Achievements 4.4. Progressive Inchoative States 4.5. Summary and concluding remarks Skwxwu7mesh wa is an imperfective marker 5.1. Imperfective Activities 5.2. Imperfective Accomplishments 5.3. Imperfective Achievements 5.4. Imperfective Inchoative States 5.5. Summary and concluding remarks 5.5.1. Potential problem 5.5.2. Generic quantification 5.5.3. A comparison to St'at'imcets Co-occurrence of wa and C V Verbal Number vs. Aspect Conclusion  '.  Chapter 6: English Activities and Accomplishments 1. Overview 2. Activities 2.1. Initial Points: Punctual Clauses 2.1.1. Skwxwu7mesh activities 2.1.2. English activities 2.1.3. Summary: Initial Points 2.2. Final Points: Culmination cancellation and event continuation 2.2.1. Skwxwu7mesh activities 2.2.2. English activities 2.2.3. Summary: Final Points 2.3. English activities have initial points 3. Accomplishments 3.1. Initial Points: Punctual Clauses 3.1.1. Skwxwii7mesh accomplishments 3.1.2. English accomplishments 3.1.3. Summary: Initial points 3.2 Final Points: Culmination cancellation and event continuation 3.2.1. Skwxwu7mesh accomplishments 3.2.2. English accomplishments 3.2.3. Summary: Final points 3.3. English accomplishments have final points 4. Conclusion  269 273 274 277 278 282 287 289 290 291 291 292 293 296 304  305-349 304 308 308 308 310 314 315 315 317 322 322 325 325 325 327 333 334 334 335 344 345 348  References  350-359  Appendix A: The Skwxwu7mesh Language  360-373  Appendix B: States  374-376  vii  List of Abbreviations Skwxwu7mesh data first person 1 second person 2 third person 3 causative CAUS = conjunction CONJ = CNJ = conjunctive DEM = demonstrative determiner DET = ergative ERG = future FUT = imperfective IMPERF = instrumental INSTR = intransitive INTR = irrealis IRR = negation NEG = nominalizer NOM = object OBJ oblique OBL particle PART = PASS = passive PAST = past plural PL = POSS = possessive question marker Q reduplicant REDUP = realis RL = relative pronoun RLP = subjunctive SBJ singular SG subject S = transitivizer TR = limited control transitivizer TR(LC) = Comox data Cnj = CTr = Erg = Indc Neg = NTr = sg lv  conjunctive subject control transitive ergative subject indicative subject negative noncontrol transitive singular link vowel  St'at'imcets data act DIR MID ooc su  = = = = =  Sanca6an ACC = AUX = CTR = D NCTR = SUB =  active intransitivizer directive middle out of control subject  accusative auxiliary control transitivizer determiner noncontrol transitivizer subject  Dene Suiine data CI CM gd O P Perf S th  = = = = = = = -  classifier conjugation marker gender object adposition perfective subject thematic  Acknowledgements Writing dissertation acknowledgements is a daunting task - to list all those who have helped me in some way during this process seems almost impossible. I hope those of you I may have missed here know who you are and know that you the omission was unintentional. I don't know if in these few pages I can adequately express my thanks for everything that everyone has done for me, but I will do my best and hope that you can fill in where words cannot. Although it is true that this thesis could not have been written without the help of many people, it is without a doubt the case that this thesis could not have been written about the Squamish language without the unbounded patience, kindness, and support from the Squamish elders Lena Jacobs, Margaret Locke, the late Chief Lawrence Baker, the late Tina Cole, the late Yvonne Joseph, the late Eva Lewis, the late Frank Miranda and the late Doris Williams. It has been an honour working with all of them, an opportunity that I know very few people get, and one which I will cherish. I have learned so much from each of them, not only about language, but about life. They are all inspiring individuals and I hope in some way this thesis is a small illustration of my appreciation, huy chap. ha71h kwis ts'its'ap'chat! Many other members of the Squamish community who work with the Squamish language have provided their support in so many ways. They have not only made the work possible and very rewarding, but a whole lot of fun. M y thanks to Roy Baker, Becky Campbell, Vanessa Campbell, Peter Jacobs, Val Moody, and Kirsten Baker Williams. I have learned a great deal from working with them and appreciate all their time and efforts. Their dedication to the Skwxwu7mesh language is inspiring, huy chap! Thanks are also due to Val Newman and Deborah Jacobs of the Squamish Nation Education Department for their help and support over the years. Other members of the community who I did not work with, but did get to know on a number of occasions, were always very supportive and enthusiastic about our work. Thank you to Layla and Lucile for the opportunity to work with their lovely mother, Eva. Thank you to Audrey Rivers for the opportunity to work with her wonderful sister, Yvonne. And thank you again to Becky for the opportunity to work with her amazing papa, Lawrence. I will remember each of them, and all the elders, very fondly. If this thesis is any good, it is because of the undying efforts of my supervisor and good friend, Lisa Matthewson. Constantly stretched beyond her limits, she always made time for her students and for that I am eternally grateful. She challenged me and supported me through not only this thesis and my many drafts and analyses, but through the ups and downs of grad student life. I am so lucky to have been able to work with her, and have learned so much from her not only about theoretical linguistics, but about doing good fieldwork. If I am even the slightest bit the linguist that she is, I will have accomplished an enormous amount. This thesis and much of my research have hugely benefited from the work and support of my two other committee members, Henry Davis and Rose-Marie Dechaine. M y sincere thanks to both of them who also managed to find the time whilst under huge stress and workloads to guide my research, read drafts of my thesis and teach me about numerous things that I probably have yet to realize. I worked as a teaching assistant for both of them and learned a great deal from each of them on the teaching front as well, for which I am very grateful. I want to also take this opportunity to thank my university examiners Hotze Rullman and Laurel Brinton for reading my thesis, for helping to make the defense process not only possible, but a smooth process and for very providing me with helpful feedback. M y sincere  x  thanks to my external thesis examiner, Angelika Krazter, who agreed to read my thesis over a year after it was meant to be finished. I cannot thank you enough for doing so and taking the time to send such detailed and helpful comments. As we all know, being a graduate student does not begin and end at the dissertation. My life at U B C involved the help and support of all the professors there, past and present. In particular, I would like to thank Strang Burton (my dancer partner), Hamida Demirdache, Laura Downing (whom I am very lucky to have reconnected with in Europe), Paivi Koskinen, Felicia Lee, Doug Pulleyblank and Martina Wiltschko. I learned a great deal about linguistics, teaching, and various other non-linguistic pursuits from all of you. A special thank you to Gunnar Hansson for not only being a great friend but for printing my thesis for me. And finally, thanks to Su Urbanczyk (now at UVic) who has been so incredibly supportive in so many ways during my graduate studies, I can't thank you enough! I had a number of opportunities to study outside of U B C , including a very rewarding and very fun month in Girona at the International Summer School in Linguistics there. Thanks to all those professors and fellow students who made that experience very memorable. After Girona, in the second year of my PhD program, I spent two semesters at the University of Southern California in Los Angeles as a visiting student. Many thanks to Hagit Borer who helped make that possible and to the other professors there who took the time to meet with me and to make my time there inspiring: Bill Rutherford, Barry Schein, Phillipe Schlenker, Rachel Walker and Maria Luisa Zubizarreta. At USC, I also met a number of students who made me feel a home. Thanks to all of them, and in particular, Teruhiko Fukaya, Shadi Ganjavi, J.-R. Hayashishita, Karine Megerdoomian, and Antonella Vecchiato (fellow L A cyclist). Thanks also to Harold Torrence and Ivano Caponigro ( U C L A graduates) who made life in L A (and a road trip to Fresno) much fun! A special thank you to Amelia Llombart, who I met in Girona and reconnected with in L A . Although I have not kept in touch these last few years, I think of you often and am so thankful for all your support as well as all the tours! And to K i m McCann, my (non-linguist) roommate in L A . What a treat it was to have met you - you definitely made life there very memorable. I was a student at U B C for longer than I care to admit, and in that time I had the privilege of knowing a lot of students. M y thanks to all of them for making grad student life fun and colourful. In particular, my thanks to Leszek Barczak, Jason Brown, Marion Caldecott, Yunhee Chung, Kristin Johannsdottir (fellow floor hockey enthusiast and squash partner), Karsten Koch, Masaru Kiyota, Yumiko Nakamura, Add Ruangjaroon, Scott Shank (who passed on many words of wisdom about thesis completion), Ian Wilson, and Rachel Wojdak. Thanks also to Kyungsook Chung (from SFU) for her support over the years. Many folks who started off as fellow students have also become very good friends and have provided a lot of support for me both in and out of school. A huge thanks to Solveiga Armoskaite, Elizabeth Currie, Suzanne Gessner, Naomi Sawai, Linda Watt and Simo von Wolff. I spent many fun days/nights with Carrie Gillon, Peter Jacobs and James Thompson. Much, though not all, of that time was taken up with episodes of A T H F , Sealab and on occasion, South Park (and then much time was spent reminiscing about episodes over sushi at the Samurai). I'm not sure how much those programs actually contributed to the thesis itself, but they sure made procrastination fun. Thanks to James for help and support through stressful times. A big thanks to Peter not only for facilitating my fieldwork, but for many hours of much needed chats, and for proofreading all the Squamish data in this thesis. Special thanks also to Carrie for venting sessions and for submitting my thesis on my behalf.  xi  As anyone knows, it is not possible to get on in life at a university without the indispensable help of administrative staff. Thanks once again to Carmen de Silva for her unending kindness and support (and curry!) during her time in the Department of Linguistics at U B C , and for her ongoing invaluable friendship. And to Edna Dharmaratne, it is impossible to express my thanks to you for everything you have done for me since you arrived in the Department of Linguistics and since I have left. You are a true friend and my life at U B C would not have been the same without you there. I have had financial support from a number of sources during my PhD: a S S H R C C Doctoral Fellowship, a British Columbia Heritage Trust Post Secondary Scholarship and various Research Assistant positions on faculty grants as well as Department of Linguistics Teaching Assistant positions. The fieldwork on which most of this work is based has been supported by S S H R C C grants to Henry Davis and to Lisa Matthewson. Many thanks for all these opportunities. My thoughts and analyses have changed many times over the years, and have benefited from questions and comments from, and conversations with, audience members, abstract reviewers and other linguistics that I met at various conferences, workshops and seminars. M y thanks to all of them. I was lucky enough to be awarded a Post-Doctoral Fellowship when I was (as I thought at the time) near completion of my thesis. The Endangered Languages Academic Program at the School of Oriental and African Studies in London (UK) kindly allowed me to take up the position before finishing my thesis. Since arriving in London my colleagues have helped and supported me through this process and my thanks to each of them. I have made a number of new friends who have helped and supported me during my thesis completion endeavors. In particular, I would like to thank Birgit Hellwig for support (and chocolate) and for proofreading a chapter of my thesis. Thanks to Anthony Jukes, my fellow thesis-writing sufferer (and weird website locator). Thanks to Zara Pybus for help and support in all things E L A P and SOAS. I would like also to thank Friederike Lupke who has not only provided support in all forms since my arrival in London (as well as proofreading a chapter), but has become a very good friend. You have all confirmed that my choice to go to London was indeed a good one. Many non-linguist friends have provided much needed non-linguistic support for years on end. M y thanks to all of you, and in particular, Wayne, Mariposa, Keinan, Martin C , Vanessa, Shana, and Greg P. Thanks to Janine, who has become a very good friend since my arrival in the U K and has gone out of her way on so many occasions to make my transition here an easy one. M y friends from my "other other" home (Toronto) have been extremely supportive through these years. Thank you to Zorana, Jay, Leslie, Dana (my partner in dissertation commiseration), Yoram, Marnie and Kevin. A special thank you to Sarit who has never wavered in her support and has helped in many steps along the way from Toronto to Vancouver to London. Thanks also to my "other" family who has stuck by me through this whole process: Ted, Joanne, Dave, Peggy, Heather, Dylan, Jordi, Terran, Alison, Maggie, and little Sara. And of course, a big thanks to my "sister-in-law" Dominique, for love and support, as well as good conversation (and for helping me prepare for my French Reading Exam!). M y sincere thanks to my parents, Shirley and Avihu (who celebrate their 50 wedding anniversary this year). Thank you for your support and enthusiasm for all that I do. I 111  xii  am sure you are as happy a I am to see me at this end of this process. I hope that I make you proud (even if my thesis is unreadable). Many many thanks to my brother Dan, who, as ever, has been so supportive through all the extremes. I cannot possibly express in words what your love and encouragement has meant over the years. I hope that I can give back half as much as you have given me. Lastly, thank you to Ross, the other (albeit "untrained") linguist of the family. Thank you for sticking by me and taking care of me during what ended up being a much longer pursuit than originally planned. I have been a grad student for as long as we have known each other, which is a scary thought but also a sign of your everlasting love and support. I am so happy that you are here in London with me. Any errors in this thesis are, of course, my own and should not reflect in any way on the speakers or linguists with whom I have worked. I took the utmost care to best to present the Squamish language as true as I understood if from my work with the elders. As I will forever be a student, mistakes are in some sense inevitable.  This thesis is dedicated to the memory of some very courageous people that I am very honoured to have known: Chief Lawrence Baker, Tina Cole, Yvonne Joseph, Eva Lewis, Frank Miranda and Doris Williams - friends, colleagues and dedicated speakers of the Skwxwu7mesh snichim Brenda Thomas, friend and fighter. Fran Watkins, a kind and caring woman who, like all these individuals, meant a great deal to a lot of people.  xiii  Chapter 1: Determining Aspectual Meaning  The study of aspect has been likened to a dark and savage forest full of "obstacles, pitfalls, and mazes which have trapped most of those who have ventured into this much explored but poorly mapped territory" (Macaulay 1978, as quoted in Binnick 1991:135)  1. On the (non-)universality of aspectual meaning The goal of this thesis is to examine predicate meanings in Skwxwu7mesh, a Central (Coast) Salish language of the Northwest coast of North America.' A t first glance, Skwxwu7mesh predicates seem to cover the same range of meanings as their English counterparts. This is illustrated with some examples below:  (1)  a.  chen ls.SG  paym rest  (i) 'I rested.' (ii) ' I ' m resting.'  b.  ACTIVITY  chen  p'ayak-an  ta  tetxwem  ls.SG  fix-TR  DET  car  T fixed the car.' c.  chen  wi7xw-em  ls.SG  fall-INTR  ACCOMPLISHMENT  'I fell.'  ACHIEVEMENT  However, on closer inspection, while the English forms are associated with a single meaning, the corresponding forms in Skwxwu7mesh may have two interpretations, as the activity example in (a) shows. This is one example of the way in which languages differ from each other in terms of how they organize predicate meaning. Another example of the way in which predicate meaning in Skwxwu7mesh differs from English has to do with whether or not the event described by the predicate must come to an end. That is, while some verbs have a final point of culmination as part of their inherent meaning, others do not. This difference may not be immediately obvious. For example, at See Appendix A for background on the Skvvxvvu7mesh language.  -I-  first glance, the meaning of the English predicate 'write a story' is identical to its Skwxwu7mesh counterpart. The sentences in (2) below are understood as describing situations in which the action of writing has come to an end and has culminated in a letter being written:  (2)  a.  I wrote a letter.  b.  chen xel'-t ls.SG wrote-TR 'I wrote a story.'  ta sxwexwiy'am' D E T story  However, i f we try to explicitly deny that the event came to an end and culminated, the two languages are observed to behave differently: in English this yields a contradiction (marked by #), but in Skwxwu7mesh it does not. This is illustrated by the sentences in (3) below:  (3)  a.  #1 wrote a letter, but I didn't finish writing it.  b.  chen  xel'-t  ta  sxwexwiy'am'  welh  haw  k-an  lS.SG  wrote-TR  DET  Story  CONJ  NEG  IRR-lCNJ  i  huy  kwi-n-s  wa  xel'-nexw  PART  finish  DET-lPOSS-NOM  IMPERF  Write-TR(LC)  'I wrote a story but I didn't finish writing it.' These are the types of subtle differences that provide us with evidence that, although all languages divide verbs into classes, the basis for these classifications is not necessarily the same in all languages. This is one of the main claims of the thesis. I will show that, in Skwxwu7mesh, predicate meanings are classified in terms of whether the event described has a distinct beginning (an "initial point"), or a distinct end (a "final point"). These findings converge with previous studies of verb meanings, where the importance of final points in the classification of verb types has long been recognized. Final points are often discussed in terms of notions such as telicity (initially by Garey 1957) or boundedness (Allen 1966), for example. However, the role of initial points has not figured prominently in analyses of aspectual meanings, so I believe the present study is novel in that respect.  In this thesis, I argue that (i) predicates can be distinguished on the basis of initial and final points and the types of (sub)events they contain, (ii) the readings induced by perfective/imperfective markers is predictable from the representations (and the meaning of the perfective/imperfective markers) and (iii) languages can differ with respect to presence/absence of initial and final points in their predicate representations. In the following section, I present an overview of how these claims are motivated.  2. What is aspect? Aktionsart refers to the aspectual properties of a given predicate; for many, the contrast that is crucial here is whether or not the predicate is intrinsically bounded with respect to its temporal properties. Consider, as an example, the sentences below (all given in the simple past, for consistency); the predicates in these sentences belong to different classes based on their inherent meanings:  (4)  a.  John was tall.  STATE  b.  John arrived.  ACHIEVEMENT  c.  John ran.  ACTIVITY  The predicate tall in (a) says nothing about the beginning or ending of John's tallness, only that John had, at some time in the past, the property of being tall; in other words, the predicate is unbounded. The predicate arrived in (b) tells us that the entire event has occurred - it is bounded. Finally, the predicate ran in (c) tells us that the running must have started at some specific identifiable time (unlike John's tallness), though it tells us nothing about when the running ended. In fact, it could be the case that John is still running (contra (b) where it could not be the case that John is still arriving), as shown by the data below:  (5)  a.  John ran and is still running.  b.  #John arrived and is still arriving.  This suggests that the predicate run is bounded with respect to its beginning, but unbounded with respect to its ending. The differences between these sentences are attributed to the  predicates themselves, and while their use may vary from sentence to sentence, they are understood as inherently coded with this information, which is why this is often referred to as lexical aspect, or in Smith's (1997) terms, situation aspect. Once a predicate is situated within a sentence, it can then be viewed in different ways; this is what grammatical aspect refers to - how the event or state of a given sentence is expressed. Consider the following sentences:  (6)  a.  John built a house last year (# but he never finished it). PERFECTIVE ACCOMPLISHMENT  b.  John was building a house last year (but he never finished it). IMPERFECTIVE ACCOMPLISHMENT  The sentences above contain the same type of event (an accomplishment),  but are expressed  in sentences with different grammatical aspects: the sentence in (a) is in the perfective and conveys that the entire event of building a house took place; this is why the sentence is incompatible with a conjoined sentence that contains an assertion of incompletion. The sentence in (b), on the other hand, is given in the imperfective aspect and conveys only that the building event was in-progress last year; the sentence does not provide any information about whether or not the event was completed. This is why the sentence, unlike the perfective sentence in (a), is compatible with a conjoined sentence that contains an assertion of incompletion. The term aspect is often used to refer to both lexical aspect and grammatical aspect; however, in more recent years, aspect is increasingly used to refer to grammatical aspect alone. Regardless of the terminology used to describe aspect or how the classification of aspect is arrived at, researchers seems to be in agreement that there are at least two different levels of aspectual classification: an inherent classification of a predicate itself and the classification of a sentence in a particular context. Thus far, most research in this area has assumed that the aspectual classifications found in English, for example, are universal; this thesis will look at both levels of aspect, and consider what a classification system that varies cross-linguistically might look like.  3. Aspectual classes and viewpoint aspect in Skwxwu7mesh: overview In this section, I provide a brief overview of the chapters to follow. Chapters T w o and Three examine initial and final points in the representations of Skwxwu7mesh predicate classes. Chapters Four and Five examine "viewpoint" aspect, and in particular, the perfective/imperfective distinction in Skwxwu7mesh. Finally, in Chapter Six, I apply the diagnostics used in the previous chapters to English activities and accomplishments. I suggest a revised account of English predicate representations.  3.1. "Open Early. Open Late": initial points and final points (Chapters 2-3) 2  In Chapter T w o , I use four diagnostics to test the presence/absence of intrinsic final points in the representations of Skwxwu7mesh predicates. : (i) culmination cancellation, (ii) event continuation (iii) the scope of kilh 'almost' and (iv) the scope of negation. I argue that these tests show that activities, accomplishments and inchoative states do not have intrinsic final points, while achievements do. I also show that the results of out of the blue judgements illustrate that accomplishments in Skwxwu7mesh, while they lack intrinsic final points, have pragmatically conditioned final points, as they have an implicature of culmination. This implicature arises with control marked transitive predicates and is derived by the addition of modality in the representation of Skwxwu7mesh accomplishments. Chapter Three is, in some sense, the flip side of Chapter T w o in that in this chapter, I examine initial points. Here I motivate the presence/absence of intrinsic initial points in the representations of Skwxwu7mesh predicates on the basis of two diagnostics: (i) the readings induced by the addition of punctual clauses/adverbials and (ii) the readings induced by the addition of the auxiliary mi 'come'. I argue that these tests show that activities, achievements and inchoative states have intrinsic initial points, while accomplishments do not. I also propose that punctual clauses trigger a shift operation in predicates with complex structures that include an initial B E C O M E sub-event (activities and inchoative states). I discuss some possible alternate analyses and show that they are not the right avenue of analysis for Skwxwu7mesh as they alone do not account for the facts.  2  Thanks to Jeff Muelbauer (p.c.) for pointing this phrase out.  -5-  3.2. Perfective and imperfective (Chapters 4-5) In Chapter Four, I examine the Skwxwu7mesh perfective. I argue that although sentences in Skwxwu7mesh not overtly marked for imperfective do share some of the same properties as other "non-standard" perfectives, the Skwxwu7mesh facts suggest that the perfective in this language is of the standard type. I show that Skwxwu7mesh does not have a completive perfective (as is claimed for Dene Suline; W i l h e l m 2003), a semi-perfective (as is claimed for Thai, Koenig and Muansuwan 2000), nor a neutral perfective (as is claimed for Hindi; Singh 1998). Instead, I argue that Skwxwu7mesh has a standard perfective, along the lines of English. However, unlike in English, Skwxwu7mesh accomplishments do not have culmination entailments or termination entailments. I propose that this is due not to the meaning of the perfective, but the meaning of the predicates. Adopting Kratzer's (1998) analysis of the perfective, I argue that the proposed representations of Skwxwu7mesh predicates predict the readings of perfective predicates that are exhibited in the language. I examine, and ultimately reject two possible alternatives to the analysis; in particular, I argued that unmarked sentences in Skwxwu7mesh cannot be claimed to have null imperfective or progressive markers, nor can they be claimed to be in the neutral viewpoint. A s with the relation between Chapters T w o and Three, Chapter Five is the flip-side of Chapter Four. In this chapter, I examine imperfectivity in Skwxwu7mesh. I propose that the Skwxwii7mesh C V reduplicant is a progressive marker and that the morpheme wa is the Skwxwu7mesh imperfective marker. Adopting Dowty's (1979) analysis of the progressive, I show that the predicate representations proposed in Chapters T w o and Three correctly predict the readings associated with sentences containing C V reduplicated/progressive predicates. I further show that adopting Kratzer's analysis of the imperfective, along with the proposed predicate representations, correctly predict the in progress readings associated with sentences containing wa/imperfective predicates.  3.3. Cross-linguistic variation (Chapter 6) In Chapter Six, I revisit aspectual classes in English. In particular, I examine initial points and final points of activities and accomplishments in the language, and compare them to the results of the present study on Skwxwu7mesh. I apply the punctual clause/adverbial diagnostic to test for initial points in English activities and accomplishments. I use the  culmination cancellation and event continuation diagnostics to test for final points. I shed some light on the discrepancies in the reported facts for English in the literature by collecting data sets from non-linguists and essentially conducting fieldwork on the language. I propose some generalizations and discuss some other factors that might affect the data. Comparing the data collected from both languages, I propose that Skwxwu7mesh and English activities have the same representation. The data suggests, however, that English accomplishments are different from Skwxwu7mesh, and thus have different representations. In the following section, I outline my proposal for Skwxwu7mesh predicates in greater detail.  4. P r o p o s a l : t h e o r g a n i z a t i o n o f p r e d i c a t e m e a n i n g s i n Skwxwu7mesh  Smith's (1997) temporal schema for aspectual classes refer to the following properties:  (7)  a.  '  b.  [+ Static]  Undifferentiated period of states  [+ Dynamic]  Successive stages of events  c.  I  F  A r b  Initial and arbitrary endpoints  d.  I  F  N a t  Initial and natural endpoints (p. 23; ex. 7)  These properties distinguish the following five temporal schemata in her system:  (8)  a.  Activities  I  F  A r b  b.  Accomplishments  I....  F  N a t R  c.  Semelfactives  E  d.  Achievements  e.  States  E (I)  R  (F) (p. 23-32)  A s Smith states, "[t]he initial endpoints of events are natural since they represent a change from a state of rest. The final endpoints are natural or arbitrary" (p. 22). In this thesis, I formalize these notions of initial and final points in the representations of Skwxwu7mesh  predicates. However, I propose that the presence/absence of initial and final points in Skwxwu7mesh are not like Smith's inventory for English given above. Initial points are distinguished from final points based on where they appear in the event, that is, whether they are an initial sub-event (e,) or a final sub-event (e ). This would 2  suggest the following possible structures, for example:  (9)  a.  Initial point:  X e . B e ^ e ^ e ^ e ^ e ^ A (BECOME(P))(e,) A (DO(P))(e )]  b.  Final point:  Xe.3e 3e [e= (e Ue ) A ( D O X P ) ) ^ ) A (BECOME(Q))(e )J  2  s  1  2  1  2  2  The representation in (a) states that there is an event that consists of two sub-events, the first of which is an initial B E C O M E event and the second of which is a final D O event. The representation in (b) states the reverse: there is an event that consists of two sub-events, the first of which is an initial D O event and the second of which is a final B E C O M E event. If the predicate contains a single B E C O M E event, this event is both the initial and final point:  (10)  Ae.(BECOME(P))(e)  Initial/Final point:  I propose the following inventory of predicate classes in Skwxwu7mesh, based on the presence/absence of initial and final points:  (11)  Skwxwu7mesh predicates: initial and final points  Activity  Initial Point  Final Point  •/  X  X  X  swim, rest, laugh  Accomplishment write a book, fix the car  Achievement  •/  win, arrive, find a rock  Inchoative State  3  X  (get) angry, (get) cloudy  3  See Appendix B for further discussion on states in Skwxwu7mesh.  -8-  A comparison between the initial and final points of Skwxwu7mesh and English predicates is given in the table below:  (12)  Predicate representations: Skwxwu7mesh vs. English  Skwxwu7mesh Predicates Activity swim, rest, laugh  Accomplishment write a book, fix the car  Achievement  English Predicates  Xe.3ei3e [t= (ei\Je ) (BECOME(P))(e,) A (DO(P))(e )]  te.(DO(P))(e)  k.[DO(P))(e) A [Vw' fw' is an inertia world w.r.t. w at the beginning of e -*• [3e' [culminates (e') in w' A e causes e'inw']]]]  Xt.3e 3e [e=Xe \Je ) A (DO(P)Xe,) A (BECOME(P))(e )]  Ae.(BECOME(P))(e)  A.e.(BECOME(P))(e)  A.e.3e 3e [e= (e Ue ) A (BECOME(P))(e,) A P(e )]  Ae.P(e)  s  A  2  2  2  l  2  l  2  2  win, arrive, find a rock  Inchoative State (get) angry, (get) cloudy  s  1  2  1  2  2  In the following section, I examine previous aspectual classifications in some detail and comment on their strengths and drawbacks.  5. Previous aspectual classifications In this section, I examine some of the earlier literature on verb classes (§5.1), some featurebased classifications (§5.2.), structurally-based aspectual classifications (§5.3) and operatorbased classifications (§5.4.). For each, I outline the conceptual claims of these classifications, and their empirical results. I discuss both the advantages and disadvantages of each of these approaches.  5.1. "Verb" classes 5.1.1. Actualitity, movement and action: Aristotle The classification of aspectual classes dates back to Aristotle. He suggests a three-way classification that distinguishes states from events, and among the events, distinguishes those which have inherent endpoints and those that do not: actuality, or the "existence of the thing" (corresponding to the more common label of "state" (Rosen 1999)), movement (an incomplete process or an event lacking an inherent terminus) and action (a process with an inherent terminus).  Kenny (1963) adopted Aristotle's three-way classification and developed tests to distinguish between events with and without natural/inherent endpoints; in his classification, the test distinguished between activities (actions with no terminus) and performances (actions with terminus). The test uses semantic entailments in the imperfective to determine whether an event can be construed as having taken place when it is still in progress. If it can, the event is an activity (a), if it cannot, the event is a performance (b)  (13)  a.  Activity 'John is pushing a cart' entails that John pushed a cart.  b.  Performance 'John is walking to school' does not entail that John walked to school.  5.1.2. Activities, accomplishments, achievements and states: Vendler (1967) Probably the most influential classification can be attributed to Vendler (1967), upon which much of current research is based. He classifies verbs into four groups: states, activities, 4  accomplishments,  and achievements. Vendler uses the contrast between verbs that can occur  in the progressive and verbs that cannot as a diagnostic to distinguish between activities and accomplishments on the one hand and states and achievements on the other. The examples below illustrate that activities and accomplishments can occur in the progressive while states and achievements cannot; the question in (a) can only be answered by the sentences in (b) and (c) but not by those in (d) or (e):  (14)  a.  Q: What are you doing?  b.  A l : I am running.  (activity)  c.  A 2 : 1 am building a house.  (accomplishment)  d.  A 3 : #1 am knowing him.  (state)  e.  A 4 : #1 am recognizing him.  (achievement) (adapted from Vendler 1967:99)  1 use the term "verbs" here on purpose (as opposed to VPs), as this is Vendler's term and became the focus on much discussion following his work. In particular, it has been discussed at length that accomplishments are not verbs but rather VPs; often they are distinguished from activities based on the presence and/or the type of object with which they occur (see Verkuyl 1993 and Rosen 1999, and references therein for discussion).  4  -10-  Vendler further points out the contrast between a state and activity where a yes/no question containing a state in the simple present and the answer given below is an appropriate sequence (15), while the same sequence with an activity is not (16) (setting aside, as Dowty emphasizes, a different meaning of running (presumably the habitual reading) that may be induced):  (15)  a. b.  Q: D o y o u k n o w . . . ? A : Y e s I do.  (16)  a. b.  Q: D o you run? A : Y e s I do.  (adapted from Vendler 1967:99)  With respect to this contrast between states and activities, Vendler suggests that  ...running, writing, and the like are processes going on in time, that is, roughly, that they consist of successive phases following one another in time.. .But although it can be true of a subject that he knows something at a given moment or for a certain period, knowing and its k i n are not processes going on in time. (p. 99-100)  Vendler distinguishes between activities and accomplishments (the class of verbs that admit, as he states "continuous tenses") in the following way. Take for example the following two sentences:  (17)  a.  John is pushing the cart,  (activity)  b.  John is running a mile.  (accomplishment)  If John stops pushing the cart, it will still be true that he did push a cart; however, if John stops running a mile, it may not be true that he did run a mile. A s Vendler states,  .. . i f someone stops running a mile, he did not run a mile.. .But the man who stops pushing the cart did push it. Thus we see that while running or pushing a cart has no set terminal point, running a mile and drawing a circle do have a  -11-  "climax", which has to be reached i f the action is to be what it is claimed to b e  "  (p. 100)  A l o n g the same lines, Vendler contrasts the two classes with the questions "For how l o n g . . . " and " H o w long did it take...". A n activity can go on for a time but does not take any definite time, while an accomplishment can also go on for some time, it takes a certain time to complete. Consider the following sentences where "for how long" is appropriate only with activities while "how long did it take" is only appropriate with accomplishments:  (18)  Activity a. For how long did he push the cart? b. #How long did it take fo-push the cart?  (19)  Accomplishment a. H o w long did it take to draw the circle? b. #For how long did he draw the circle?  (adapted from Vendler 1967: 100-1)  Activities and accomplishments can also be contrasted by their entailment patterns. M u c h like Bennett and Partee's (1978) sub-interval property, which draws on these intuitions, Vendler notes that  If it is true that someone has been running for half an hour, then it must be true that he has been running for every period within that half hour. But even if it is true that a runner has run a mile in four minutes, it cannot be true that he has run a mile in any period which is a real part of that time, although it remains true that he was running, or that he was engaged in running a mile, during any substretch of those four minutes. It appears, then that running and it kind go on in time in a homogenous way; any part of the process is of the same nature as the whole. Not so with running a mile or writing a letter; they also go on in time, but they proceed toward a terminus which is logically necessary to their being what they are. (p. 101)  With respect to the verbs that do not occur in the progressive, Vendler suggests that they divide into two types; some "can be predicated only for single moments of time.. .while  -12-  others [verbs that cannot occur in the progressive] can be predicated for shorter or longer periods of time" (p. 102). Achievements occur at definite moments, while states occur over short or long periods of time; this contrast is illustrated in the examples below:  (20)  a. b.  At what time did you reach the top? At noon sharp. At what moment did you spot the plane? At 10:53 am.  (21)  a. b.  For how long did you love her? For three years. How long did you believe in the stork? Till I was seven. (Vendler 1967: 102-3)  The fact that achievements are described as occurring at definite moments poses a problem for sentences such as the following; the following felicitous sentences containing achievements are predicted to be infelicitous since the sentence suggests that the achievement took time, rather than occurring at a moment:  (22)  a.  It took him three hours to reach the summit,  b.  He found it in five minutes.  For sentences such as these, Vendler suggests that  one does not mean that the 'reaching' of the summit went on during those hours. Obviously it took three hours of climbing to reach the top....If I write a letter in an hour, then I can say I am writing a letter at any time during that hour; but if it takes three hours to reach the top, I cannot say I am reaching the top at any moment of that period. (p. 104) Vendler concludes his discussion with the following examples that illustrate the time schemata he argues for: (23)  a.  Activities  A was running at time t means that time instant t is on a time stretch throughout which A was running.  -13-  b.  Accomplishments A was drawing a circle at t means that t is on the time stretch in which A drew that circle.  c.  Achievements A won a race between t, and t means that the time instant at which A won that 2  race is between f and t . 7  d.  2  States A loved somebody from t, to t means that at any instant between t, and t A 2  2  loved that person.  5.1.3. Advantages and disadvantages One of the major criticims of Vendler's approach is that he classifies "verbs" and as such ignores the fact that subjects and objects can have an effect on the classes to which a verb belongs. It should be noted that Vendler alludes to this problem himself (boldface has been added)  Distinctions have been made among verbs suggesting processes, states, dispositions, occurrences, tasks, achievements, and so on. Obviously these  differences cannot be explained in terms of time alone: other factors, like the presence or absence of an object, conditions, intended states of affairs, also enter the picture. Nevertheless one feels that the time element remains crucial; at least it is important enough to warrant separate treatment. Indeed, as I intend to show, if we focus our attention primarily upon the time schemata presupposed by various verbs, we are able to throw light on some of the obscurities which still remain in these matters. (P- 97)  Although the predicate classes I propose for Skwxwu7mesh do not correspond precisely to Vendler classes, I use his terminology for the following reasons: (i) The classes that Vendler refers to as accomplishment, achievement, activity and state share the same basic meanings that are associated with predicate classes in Skwxwu7mesh. (ii) I argue that the meanings associated with predicate classes can differ, which accounts for the differences between Skwxwu7mesh and English, for example. This is an alterative to claiming that the same classes are found in every language and the difference is that Skwxwu7mesh simply does not exhibit those classes, but has different ones, (iii) Skwxwu7mesh, like all Salish languages, has a rich transitivity system that encodes control as well as transitivity (see Thompson 1979,  -14-  among others). This suggests that some of the distinctions observed cross-linguistically may be attributed to these differences in transitivity and not strictly a difference in the aspectual system per se.  5.2. Feature-based definitions of predicate classes Featurally based aspectual systems use features, which are often binary, to derive aspectual classes. Some of these systems have two binary features that derive the four aspectual classes introduced by Vendler; for example, Hoeksema (1983), following Mourelatos (1978), proposes the features ±count (whether events can be counted) and ±duration "(whether event takes place over time). Other systems (e.g., Moens 1987) also propose use two binary features, but claim that they derive four event classes, leaving states aside. Smith (1997) proposes a three binary feature system that derive five aspectual classes (Vendler's four classes, and an additional class Smith labels seme If active), leaving three potential classes unattested. Carlson (1981) also proposes three binary features, but derives six aspectual classes (Vendler's four, and an additional two classes momentaneous and dynamic), leaving two classes unattested. I limit my discussion in this section to two feature-based aspectual systems: Carlson (1981) and Smith (1997). See Rosen (1999) for further detailed discussion of other featurebased systems.  5.2.1. [±POINT, ±EXTENDED, ±CONTINTJOTJS]: Carlson (1981) Using a number of tests involving grammaticality judgements, Carlson (1981) proposed the following three binary features:  (24)  a.  ±POINT  - whether or not a momentaneous adverbial applies to a given sentence - e.g., at once, at that (very) moment, at 8:30 b.  ±EXTENDED  - whether or not the progressive aspect (-ing in English) applies to a given sentence  -75-  C.  ±CONTINUOUS  - whether or not a durative adverbial applies to a given sentence - e.g., for a while, from one to ten o'clock, all day (long) These three binary features derive six classes: four of these classes parallel Vendler's (a-d) two additional classes are introduced (e-f, shaded in the chart below):  (25)  Carlson's Feature System (1981)  Point  Class a. b. c. d. >,e:\ :  K.  Activity Accomplishment Achievement State Momenlaneous D\namic  -  + + + 1  +  Extended + + + -  Continuous + -  + +  Vendler's achievements divide into two classes in Carlson's system: achievement and momentaneous predicates. They both are [+POINT] and [-CONTINUOUS], but the achievement predicates are [+extended] while momentaneous predicates are [-EXTENDED]. Momentaneous sentences are compatible with momentaneous adverbs in non-progressive sentences. This is illustrated below:  (26)  a.  A t that point. I hit him.  b.  I noticed it at once.  c.  Just when the light went on, he blinked. (p. 37, ex. 6a-c)  However, a contrast emerges when the two types of predicate are given in the progressive aspect. Achievement predicates are compatible with the progressive while momentaneous predicates are incompatible. This is shown in the two sets of data below; the underlined clauses and prepositions in each set of data are momentaneous adverbials and both predicate types are compatible with them (yielding a +POINT feature assignment). However, as these data show, only achievement predicates are A L S O compatible with the progressive (yielding a +EXTENDED feature assignment), while momentaneous predicates are marked in these cases  -16-  (yielding a - E X T E N D E D feature assignment). In the examples below, momentaneous 5  adverbials are underlined and predicates are bolded:  (27)  Achievement  [+POINT, + E X T E N D E D , - C O N T I N U O U S ]  a.  A t that point he was closing the door.  b.  H e was winning the tournament when he missed that ball.  c.  The dog was attacking me when the owner intervened.  d.  A t that point, the plane was taking off from the ground.  6  7  (p. 38, ex. 7f-i)  (28)  Momentaneous  [+POINT, - E X T E N D E D , - C O N T I N U O U S ]  a.  !At that point I was hitting him.  b.  !I was noticing it at once.  c.  !Just when the light went on. he was blinking.  8  (p. 38, ex.  7a-c)  Carlson also proposes a class of dynamic predicates which lie somewhere between Vendler's statives and activities. A l l three types share the [-(-CONTINUOUS] feature; stative and dynamic predicates share the feature [+POINT], while activity predicates do not. Thus, in the examples below, activities are incompatible with momentaneous adverbials (yielding marked sentences) while stative and dynamic predicates are compatible:  Carlson uses the "!" notation to indicate that the sentence is marked in some way; for her, marked means that some additional comment is needed. This is to say that sentences may have other interpretations that are not under consideration, or they may require artificial contexts to be judged grammatical. It is not crucial for Carlson to demonstrate that these sentences are N E V E R accepted, only that they differ from the ones that are always accepted, under any context. The marked sentences translate to a "-" notation in the chart above which suggests that they lack a given feature, or have a "-" assignment for that feature.  5  6  This has a different interpretation than the others in the set in that no part of the winning event can have taken  place, though a part of the opening of the door can. It is not clear why Carlson categorizes attack as an achievement. The sentences in (a) and (c) have a different interpretation than the sentence in (b). (a) and (c) can be interpreted as events that occur over and over (hit over and over, blink over and over), (b) lacks this meaning.  7  8  -17-  (29)  Activity [-POINT, + E X T E N D E D , +CONTINUOUS]  a.  '.The children played when I returned.  b.  !At sunrise, I walked eastward.  9  (p. 38, ex. 6n-o)  (30)  Dynamic a.  [ + P O I N T , + E X T E N D E D , -(-CONTINUOUS]  A t seven o'clock, the caravan  stood in its old place. (p. 38, ex. 6j)  (31)  Stative  [+POINT, - E X T E N D E D , +CONTINUOUS]  a.  He  b.  A t that point I  was a full-grown man when I was remembered the  born.  rule. (p. 37, ex. 6d-e)  On the other hand, activity and dynamic predicates share the feature  [+EXTENDED]  while  stative predicates do not. This is shown in the data below where activity and dynamic predicates are compatible with the progressive while stative predicates are not:  (32)  Activity [ - P O I N T , - ( - E X T E N D E D , - ( - C O N T I N U O U S ]  a.  The children were  b.  A t sunrise I was  playing when I returned.  walking  eastward. (p. 38, ex. 7n-o)  (33)  Dynamic  [+POINT, -(-EXTENDED, -(-CONTINUOUS]  At seven o'clock, the caravan was  standing in its old place. (p. 38, ex. 7j)  9  The sentences in (a) and (b) are completely grammatical not only by my o w n judgement, but by the judgement  of native English speakers who are non-linguists; the interpretation is an inceptive/ingressive event (i.e., the children began to play when the speaker returned). 1 0  Laurel Brinton (p.c.) suggests that the category dynamic is questionable and notes that Carlson provides only  one example of this type.  -18-  (34)  Stative [+POINT, - E X T E N D E D , +CONTINUOUS]  a.  !He was being a full-grown man when I was born.  b.  !At that point I was remembering the rule. (p. 38, ex. 7d-e)  Carlson claims that many of his examples are quite admissible with a secondary interpretation. For example, activity predicates yield inceptive readings with momentaneous adverbials (a) and states in the progressive can yield the sense o f an intentional activity (b):  (35)  a.  !At sunrise.  I walked eastward. (p. 38, ex 6(o))  b.  !He was being a full-grown man when I was born. (p. 38, ex 7(d))  The activity is (a) is taken to mean that the speaker began to walk eastward. A s for (b), the state is taken to mean that the subject was behaving as though he was a full-grown man (difference in volition)." This difference, according to Carlson, can be in acceptability, meaning of the main verb, or contexts of use. A s such, he does not take their possible use in different contexts or with different interpretations as counter examples, but as a means of showing that there is a fundamental difference between the different classes of predicates (i.e. those which do not require special contexts, or those that do not change meaning in different contexts...).  5.2.2. [±STATIC, ±DURATIVE, ±TELIC]: (Smith 1997) Another prominent figure i n the aspectual literature is Smith (1997). She presented probably the most ambitious and detailed attempt to account for the similarities and differences in aspectual systems across the world's languages. Smith proposes the following three features that derive five aspectual classes:  " This reading may not be available for all speakers.  -19-  (36)  a.  ±STATIC  b.  ±DURATIVE  c.  ±TELIC  L i k e Carlson's system, these three binary features derive eight classes: four of these classes parallel Vendler's (a-d) one additional class is introduced (e, shaded in the chart below):  (37)  Smith's aspectual feature system (1997:20) Situation a. b. c. d. c.  Activity Accomplishment Achievement State Semelfactive •  • .  ±Telic  ±Durative  ±Static -  + +  -  -  + +  +  -  -  +  -  •  WM,  Smith's system generates a predicate type that she labels semelf active* which she describes 1  as instantaneous non-culminating single-stage events, but which result in no change of state. Some examples are given below:  (38)  a.  Mary coughed.  b.  Mary knocked at the door.  c.  Mary hiccupped.  d.  The bird flapped its wing.  Semelfactives are distinguished from achievements in her system, which Smith classifies as instantaneous culminating events.  From the Latin semel ('once'), which as Smith notes, is used in Slavic linguistics to refer to a suffix that indicates a single event. I argue that semelfactives do not constitute a separate aspectual class but rather they are types of activities with very defined sub-events. 1 2  -20-  ±STATIC The feature S T A T I C distinguishes states ( + S T A T I C ) from events ( - S T A T I C ) . + S T A T I C predicates cannot occur in imperative sentences (a) or be complements of verbs like persuade or command (b), while - S T A T I C predicates can (where Smith does not cite the relevant examples, I have included my own, and my own judgements):  (39)  (39)  a.  Imperative sentences  i.  Push the cart!  Act. [-STATIC,  +DUR., -TELIC]  ii.  Wash your car!  Acc. [-STATIC,  +DUR., +TELIC]  iii.  Tap his shoulder!  Sem. [-STATIC,  iv.  Break the glass!  Ach. [-STATIC,  -DUR., +TELIC]  v.  #Know Greek!  Stat. [+STATIC,  +DUR., -TELIC]  (p. 40)  b.  Complements of persuade or command  i.  W e persuaded E m i l y to push the cart.  Act. [-STATIC,  +DUR., -TELIC]  (p. 44)  ii.  W e persuaded Sam to open the door.  +DUR., +TELIC]  (p. 45)  iii.  I persuaded Mary to tap his shoulder.  Sem. [-STATIC,  iv.  I persuaded Mary to break the glass.  Ach. [-STATIC,  v.  #1 persuaded Mary to know Greek.  +STATIC  13  Acc. [-STATIC,  -DUR., -TELIC]  (p. 40) (p. 46)  -DUR., -TELIC]  (p. 46)  -DUR., +TELIC]  (p. 46)  States [+STATIC,  +DUR., -TELIC]  predicates cannot occur with adverbs of manner (e.g., carefully, attentively) (c) and  adverbs of instrument (e.g., with a key) (d), while - S T A T I C predicates can:  (39)  1 3  c.  Adverbs of manner  i.  E m i l y voluntarily pushed the cart.  Act. [-STATIC,  +DUR., -TELIC]  ii.  John carefully washed his car.  Acc. [-STATIC,  +DUR., +TELIC]  iii.  She carefully tapped his shoulder.  Sem. [-STATIC,  -DUR., -TELIC]  iv.  She carefully broke the glass.  Ach. [-STATIC,  v.  *John carefully knew Greek.  States [+STATIC,  (p. 44) (p.40)  (p. 46)  -DUR., +TELIC]  +DUR., -TELIC]  (p. 40)  T h i s test seems to be more sensitive to volition rather that a static feature, as there are a number of states that  are volitional that are compatible with imperatives:  stay still, hold the baby.  -21-  (39)  d.  Adverbs of instrument  i.  E m i l y pushed the cart with a stick.  Act. [-STATIC, + D U R . , - T E L I C ]  ii.  M a r y opened the door with a key.  ACC. [-STATIC, + D U R . , + T E L I C ] (p. 40)  iii.  She tapped his shoulder with a pencil.  Sem. [-STATIC,-DUR., - T E L I C ]  iv.  She broke the glass with a ball.  Ach. [-STATIC, - D U R . , + T E L I C ]  v.  *The door was open with a key.  Stat. [+STATIC, + D U R . , - T E L I C ] (p. 40)  In English, + S T A T I C predicates cannot occur in the progressive (e) or in a do pseudo-cleft construction (f), while - S T A T I C predicates can:  (39)  (39)  e.  Progressive  i.  E m i l y was pushing the cart.  Act. [-STATIC, + D U R . , - T E L I C ] (p. 44)  ii.  John was washing the car.  Acc. [-STATIC, + D U R . , + T E L I C ] (p.40)  iii.  She was tapping his shoulder.  iv.  He was winning the race.  Ach. [-STATIC, - D U R . , + T E L I C ] (p. 46)  v.  * K i m was knowing the answer.  Stat. [+STATIC, + D U R . , - T E L I C ] (p. 40)  f.  D o pseudo-cleft  i.  What E m i l y did was push the cart.  Act. [-STATIC, + D U R . , - T E L I C ] (p. 44)  ii-  What John did was wash the car.  ACC. [-STATIC, + D U R . , + T E L I C ] (p. 40)  iii.  What M a r y did was tap his shoulder.  Sem. [-STATIC, - D U R . , - T E L I C ] (p. 46)  iv.  What he did was w i n the race.  Ach. [-STATIC, - D U R . , + T E L I C ] (p. 46)  v.  #What John did was know Greek.  Stat. [+STATIC, + D U R . , - T E L I C ] (p. 40)  Sem. [-STATIC, - D U R . , - T E L I C ]  construction  Finally, in English, + S T A T I C predicates yield a stative interpretation in the present perfective, while - S T A T I C predicates yield habitual events (g):  (39)  g.  Present  perfective  i.  E m i l y pushes the cart.  (habitual)  Act. [-STATIC, + D U R . , - T E L I C ] (p.44)  ii.  Sam opens the door.  (habitual)  Acc. [-STATIC, + D U R . , + T E L I C ] (p. 45)  iii.  M a r y taps his shoulder,  (habitual)  Sem. [-STATIC, - D U R . , - T E L I C ] (p. 46)  -22-  He wins the race.  (habitual)  Ach.  Ellen believes in ghosts, (stative)  [-STATIC, - D U R . , + T E L I C ]  Stat. [ + S T A T I C , + D U R . ,  -TELIC]  (p. 46) (p. 41)  A summary of these facts are given in the table below:  (40)  Properties associated with Smith's feature  ±STAT1C  +STATIC  -STATIC  (states)  (events)  X  •/  a. imperative sentences b. complements of persuade/command c. adverbs of manner (carefully, attentively) adverbs of instrument (with a key) d. e. Progressive f. do pseudo-cleft construction g- interpretation in the present perfective  X </  X X X X  habitual  stative  ±DURATIVE  The feature  ±DURATIVE  distinguishes states, activities and accomplishments  from semelfactives and achievements ( - D U R A T I V E ) .  +DURATIVE  (+DURATIVE),  predicates are compatible  with "direct durative adverbials" (e.g., for an hour, in an hour) (a);  -DURATIVE  predicates are  incompatible with these adverbials (as in iv), or they yield a different interpretation, which Smith suggests is a shift to a durative interpretation of the predicate. For example, (iii) below has an ingressive reading, according to Smith, which focuses on preliminary stages of the tapping event. As I show below in the discussion of the 14  ±TELIC  feature, there is a further  distinction between in and for adverbials:  a.  1 4  Direct durative adverbials [-STATIC, +DUR., -TELIC] (p.  41)  [-STATIC, +DUR., +TELIC] (p.  41)  Sem.  [-STATIC, - D U R . , -TELIC] (p.  46)  #The bomb exploded for an hour.  Ach.  [-STATIC, - D U R . , +TELIC] (p.  41)  Marv was sick for a week.  Stat. [+STATIC, +DUR., -TELIC] (p.  41)  Mary  walked in the park for an hour.  ii.  Mary  built the sandcastle in an hour.  iii.  She, tapped his shoulder in an hour.  iv. v.  This reading is not available for all speakers.  -23-  Act. Acc.  +DURATIVE  predicates are compatible with inceptive and terminative morphemes  (begin, stop, finish) (b), but - D U R A T I V E predicates are not (as in iv), or have different interpretations: (41)  The  b.  Inceptiv elterminative morphemes  .  i.  He began/stopped pushing the cart.  ii.  Sam  iii.  She began/stopped tapping his shoulder. Sem. [ - S T A T I C , -DUR., - T E L I C ] (p. 46)  iv.  #The bomb stopped exploding.  Ach. [ - S T A T I C , -DUR., + T E L I C ] (p. 42)  v.  Sam  Stat. [ + S T A T I C , +DUR., - T E L I C ] (p. 47)  began/stopped walking to school.  began to be angry.  Act. [ - S T A T I C , +DUR., - T E L I C ] (p. 45) ACC.  [-STATIC,  +DUR., + T E L I C ] (p. 45)  sentence in (iii) above is predicted to be ungrammatical or to have a different  interpretation; according to Smith, semelfactives involve a shifted interpretation as multiple events. A s for the sentence in (v) above, Smith predicts it to be grammatical with no change in interpretation. Smith suggests that sentences like those in (v) have change of state interpretation. +DURATIVE  predicates have inceptive interpretations with "momentary adverbials"  (e.g., at noon, at 5 o'clock exactly) (c), - D U R A T I V E predicates have a direct interpretation:  (41)  c.  Momentary  adverbials  i.  H e pushed the cart at noon.  ii.  They ate dinner at noon.  iii.  She tapped his shoulder at noon.  iv.  The clock struck at noon.  v.  John was dumbfounded when Harry threw the glass.  Act. [ - S T A T I C , +DUR., - T E L I C ] (p. 45) Acc. [ - S T A T I C , +DUR., + T E L I C ] (p. 42) Semel. [ - S T A T I C , -DUR., - T E L I C ] (p. 46) Ach. [-STATIC, -DUR., +TELIC] (p. 42)  States [ + S T A T I C , +DUR., - T E L I C ] (p. 49)  While the semelfactive in (iii) can have the predicted direct interpretation of predicates with a [-durative] feature, it also has an inceptive interpretation, which is not accounted for. Furthermore, while the state in (v) has an inceptive interpretation, it also has different reading which is unaccounted for here. In fact, with many stative sentences, the inceptive reading is not necessarily even the most natural, as illustrated by the data below:  -24-  (41)  v'.  The baby was asleep at noon.  (p. 47)  The data in (cii) above is not representative of the entire class of accomplishments, which in general seem to require specific contexts in order to be judged felicitous with an inceptive interpretation. The following sentences illustrate that many accomplishments are not judged felicitous with momentary adverbials (see Chapter Six for further discussion):  (41')  a.  #Mary fixed the car at 5 o'clock.  b.  #Mary built a sandcastle at 5 o'clock.  c.  #Mary read a book at 5 o'clock.  +DURATIVE  predicates are compatible with what Smith calles "indirect durative  adverbials" (e.g., slowly, quickly) (d), while - D U R A T I V E predicates are not:  (41)  d.  Indirect durative  adverbials  i.  He slowly pushed the cart.  ii.  Sam slowly walked to school.  iii.  She slowly tapped his shoulder.  iv.  #John slowly broke the glass.  v.  '"Mary was slowly sick.  Act. [ - S T A T I C , +DUR., - T E L I C ] (p. 45) ACC.  [-STATIC,  +DUR., + T E L I C ] (p. 45)  Seme I. [ - S T A T I C , -DUR., - T E L I C ] (p. 46) Ach. [ - S T A T I C , -DUR., + T E L I C ] (p. 47) States [ + S T A T I C , +DUR., - T E L I C ] (p. 47)  The sentence in (iii) is predicted to be ungrammatical as semelfactives have -durative features; however, Smith suggests that these predicates get ingressive interpretations. A s for the stative in (v), Smith predicts the sentence to be grammatical as statives have a +durative feature; however, as shown above, Smith judges the sentence ungrammatical and suggests that these adverbs imply activity. Finally, imperfective viewpoints (e) focus on internal states of + D U R A T I V E situations, but preliminary stages of - D U R A T I V E events:  -25-  (41)  (42)  e.  Imperfective  i.  E m i l y was pushing the cart.  (inter)  ii.  The door was opening.  (inter)  iii.  Mary was knocking.  (?)  Sem. [ - S T A T I C , -DUR., - T E L I C ]  iv.  Mary was reaching the top.  (prelim)  Ach. [ - S T A T I C , -DUR., + T E L I C ] (p.42)  v.  # K i m was knowing the answer. (?)  Act. [ - S T A T I C , +DUR., - T E L I C ] (p.44) ACC.  [-STATIC,  Stat. [ + S T A T I C , +DUR., - T E L I C ] (p. 40)  Properties associated with Smith's feature  ±DURAT1VE  +DURATIVE (states, act., acc.)  a.  Direct durative adverbials (for an hour, in an hour)  b.  Inceptive, terminative morphemes (begin, stop, finish)  c.  Momentary adverbials  Inceptive interpretation  (at noon, at 5 o'clock exactly)  -DURATIVE (semel., ach.) X  1 5  X  1 6  Direct interpretation X  d. Indirect durative adverbials (slowly, quickly)  e.  +DUR., + T E L I C ] (p.42)  internal stages  Imperfective viewpoint  preliminary stages  ±TELIC Smith's feature ± T E L I C distinguishes accomplishments and achievements (events with an inherent endpoint, +TELIC), from semelfactives and activities (events without an inherent endpoint, - T E L I C ) . + T E L I C predicates are compatible with verbs and adverbials of completion (e.g., finish, in an hour) (a), while - T E L I C predicates are not:  (43)  Verbs and adverbials of completion i. .  ?Mary finished walking in the park.  Act. [ - S T A T I C , + D U R A T I V E , -TELIC]  ii.  Mary finished walking to school.  Acc. [ - S T A T I C , + D U R A T I V E , +TELIC] (p. 43)  iii.  ?She finished tapping his shoulder.  Sem. [ - S T A T I C , - D U R A T I V E , -TELIC]  iv.  #John finished breaking the glass.  Ach. [ - S T A T I C , - D U R A T I V E , +TELIC] (p. 47)  v.  ?John finished being angry.  17  (p. 43)  (p. 46)  States [ + S T A T I C , + D U R A T I V E , -TELIC]  Or have different interpretations. Or have different interpretations. This sentence is predicted to be felicitous. Smith says elsewhere that achievements are "instantaneous and Telic..They disallow terminative and simple duratives" (p. 46). 15  16  17  -26-  +TELIC predicates are incompatible with adverbs of simple duration (e.g., stop, for an hour) (b), while -TELIC predicates are compatible. However, Smith states that the verb stop is fine with telic predicates, but does not indicate completion (p. 43). This explains why the sentence in (ii) is fine, even though it is predicted to be bad. Smith claims that the sentence in (iii) is a shifted multiple event reading.  (44)  b.  Adverbs of simple  duration  1.  Mary stopped walking in'the park.  Act.  [-STATIC, +DURATIVE, - T E L I C ]  (p.  43)  ii.  Mary stopped walking to school.  Acc  [-STATIC, +DURATIVE, + T E L I C ] (p.  43)  iii.  She stopped tapping his shoulder.  Sem.  [-STATIC, - D U R A T I V E , - T E L I C ]  (p.  46)  iv.  #John broke the glass for an hour.  Ach.  [-STATIC, - D U R A T I V E , + T E L I C ]  (p.  47)  v.  Mary was sick for three days.  Stat. [+STATIC, +DURATIVE, - T E L I C ] (p. 47)  +TELIC predicates are compatible with main verbs of time such as take an hour, but -TELIC predicates are compatible with main verbs of time such as spend an hour (c).  (44)  c.  Main verbs of time  l.  I spent an hour listening to music.  i'  ?It took me an hour to listen to music.  ii.  ?I spent an hour writing the letter.  ii.'  It took me an hour to write the letter.  iii.  I spent an hour knocking.  iii'  It took me an hour to knock.  iv.  ?John spent an hour leaving.  iv.  It took John an hour to leave.  v.  I spent an hour being angry.  v'  ?It took me an hour to be angry.  Act. [-STATIC, +DURATIVE, -TELIC]  18  (p. 43) Acc. [-STATIC, +DURATIVE, + T E L I C ] (p. 43) (p. 43) Sem.  [-STATIC, - D U R A T I V E , - T E L I C ]  Ach.  [-STATIC, - D U R A T I V E , + T E L I C ]  (p. 47)  19  Stat. [+STATIC, + D U R A T I V E , - T E L I C ]  This sentence is fine for some speakers. Smith suggests that this sentence suggests that the letter was not finished. It is unclear to me how this bears on the claims. Smith claims this sentences has an ingressive/preliminary stage reading. 18  19  -27-  (p. 43)  (44)  a.  Properties associated with Smith's feature ±TEL1C +TELIC  -TELIC  (acc, ach.)  (states, act., semel.) X  Verbs and adverbials of completion {finish, in an hour)  b.  X  Verbs and adverbials of simple duration (stop, for an hour)  c.  spend  take  Main verbs of time  5.2.3. Advantages and disadvantages Both Carlson's and Smith's systems have the advantage of attempting to account for a wide range of data that is comparable across languages. Both approaches use a variety of diagnostics and make interesting predictions that are testable. This is very appealing in both systems. However, there are a few problems with both the features and the data. One problem is that both features systems derive more classes than are attested, and their explanations for why this is so are somewhat unsatisfactory. For Carlson, there are two classes unattested:  (45)  Carlson's Feature System (1981) Point -  Class a. b. c. d. e. f.  Activity Accomplishment Achievement State Momentaneous Dynamic  JI.  •>  + + + + -  h.  Extended  Continuous  + + +  +  -  +  +  +  -  +  The types of predicates classified as (g) are those which would be incompatible with momentaneous adverbials, the progressive and durative adverbials. The predicates in (h) would be incompatible with momentaneous adverbials and the progressive, but compatible with durative adverbials. Carlson states that these classes are not relevant; that is, in cases where both features point and extended have negative values can never be true or false as "all periods are either moments or extended" (p. 41), thus there cannot be a predicate that does  -28-  not allow either of them. The problem with this argument is that her feature [ + E X T E N D E D ] is defined as whether or not it allows the progressive. It is possible, for example, for predicates to not allow the progressive, but still go on for an extended period. This crucially relies on the assumption that there are only those two periods, moments and extended, and no others. Smith's system only includes five classes, which leaves three classes are unattested (f-g shaded in the chart below):  (46)  Smith's aspectual feature system (1997 :20) r  Situation a. b. c. d. e. r. g. h.  Activity Accomplishment Achievement State Semelfactive ?  ,  ±Static -  -  -  +.  +  -  -  ±Telic -  + + -  + •  •>  ?.  ±Durative + +  +  -  + . +  With respect to these classes, Smith suggests that the feature [ ± T E L I C ] is irrelevant to situations with the property [+STATIC]. It is unclear to me at this point why this would be the case. It may be that telicity is a feature of events only, but this may be a stipulation of the system. A second problem with both systems is that the reported data are inconsistent with judgements reported by other native speakers, as I highlighted in the footnotes throughout each section. Neither Carlson nor Smith suggest how robust the judgements are, and thus it is difficult to evaluate in this respect. Many of the examples they both use are not necessarily representative of an entire class of predicates they refer to. A problem quite relevant to this thesis is that Smith does not explain the relationship between the features she proposes for the different aspectual classes and the diagnostics she uses; for example, it is not clear why predicates with a [+DURATIVE] feature would get inceptive readings with momentaneous adverbs. For Carlson, many of the "secondary interpretations" that she refers to are not in fact, secondary at all. I would argue that they do not involve, as she suggests, coercion; the fact that many of her sentences are easily admissible, I argue is evidence that these sentences have these interpretations to begin with. I  -29-  further propose that the interpretations that Carlson would like to set aside tell us something about these predicates and that these readings are crucial in establishing the meaning of these predicate types.  5.3. Structurally-based definitions of predicate classes 5.3.1. [±ADD T O ,  ±SPECIFIED QUANTITY OF A, ±TERMINATIVE] :  Verkuyl 1993  Verkuyl's compositional system has been influential in the re-examination of the traditional classes introduced by Vendler (and the ones used by Carlson and Smith in the previous section). He argued that the event classes themselves are not primitive, but that the classification is based on primitive features of the event. Furthermore, he proposed that the classification is not as useful as an understanding of the binary features that generate the classes. Verkuyl argued that the N P subject and object of a given sentence affect whether to 20  consider a verb a member of one of Vendler's classes or another and thus suggests that the classification of predicate classes cannot be simply lexical, but must be structural as well, and he proposes a system that incorporates both. Verbs are assigned lexical features, while the features of NPs, V P s and S are composed structurally. V e r k u y l ' s system assigns binary features to heads, phrases and sentences. The following is an inventory of features in his system:  (47)  a.  [±ADD TO]  b.  [±SPECIFIED Q U A N T I T Y OF A (SQA)]  (where A = a set of individuals in the  domain o f interpretation) C.  [ ± T E R M I N A T I V E (T)]  The binary feature [ ± A D D T O ] is assigned to the verb. This is the only lexical feature of the system as it is a property of the verb alone. [ ± A D D T O ] refers to the dynamic semantic information that distinguishes non-stative verbs such as eat, walk, drink, knit stative verbs such as want, hate  2 0  [+ADDTO]  from  [-ADDTO]:  Note that V e r k u y l ' s system could have been included in the previous section on feature-based systems, but  since (contrary to what he claims) his system is also structural, I have included it here.  -30-  (48)  a.  V[+ADDTO]  b.  eat  V [-ADD  TO]  want  The binary feature [±SQA] is attributed to the NP subject and the NP object of the clause and is a result of the combination of a determiner and the head N of the NP. A [+SQA| NP pertains to a specified quantity of things or mass denoted by its head noun whereas [SQA] pertains to an unspecified quantity:  (49)  a.  b.  a sandwich  three sandwiches NP [+SQA]  NP [+SQA]  c.  Det  N  Det  a  sandwich  three  d.  sandwiches  sandwiches  no sandwiches NP [-SQA]  NP [-SQA]  Det  N  N  Det  sandwich  no  N  sandwiches  Finally, a binary feature [±T] is attributed to the VP and the sentence itself. A terminative [+T] feature is assigned to events (accomplishments, in Vendler's terms), which Verkuyl calls "temporal entities that can be counted and quantified over" (1993:19). A [-TJ feature indicates durativity, and is assigned to processes and states (in the sense of Bach 1986). Durativity refers to "something going on in time unboundedly, or about something not going on in time" (19-20). As with [±SQA], this feature is structurally determined.  -31-  T o calculate the aspectual value of a sentence, first the value of the V P is calculated based on the composition of the verb and its object (a); then the value of S is calculated based on the composition of the subject and the V P (b): (50)  Calculating aspect (Verkuyl 1993) a.  V [ ± A D D T O ] + object [±SQA] = [±T ]  b.  subject[±SQA] + [ ± T J = [±T]  VP  T o calculate the aspectual value of V P , the lexical information encoded in V is combined with the structural information of the N P object (if present).  21  This is illustrated for  [+ADD  TO] verbs in (a-b) and [-ADD TO] verbs in (c-d):  (51)  a.  ate a sandwich  b.  ate  sandwiches V P [-T]  V P [+T]  V  eat  2 1  NP  [ + A D D TO]  V  [+SQA]  N P [-SQA]  [ + A D D TO]  Det  Det  N  a  sandwich  eat  See V e r k u y l (1993) for details of the assignment of [ ± S Q A ] features.  -32-  N  sandwiches  c.  d.  want a sandwich VP  V  [-ADD TO]  VP  [-T]  NP  want  want sandwiches  V  [+SQA]  Det  N  a  sandwich  [-T]  NP  [-ADD T O ]  [-SQA]  Det  want  sandwiches  A n y [-] feature in the calculation results in the feature [-T]; thus, only if all the features are positive can you get a terminative feature [+T]. Regardless of the value of the N P object in (c-d) above, the V P is [-T] since the V is [-ADD T O ] . This is what Verkuyl refers to as the plus-principle. T o calculate the aspectual value of S, the structural information encoded in the subject N P is composed with the structural information encoded in the V P :  (52)  The boy ate a sandwich S  NP  Det  The  [+SQA]  Structural  [+T]<-  VP[+T]  Structural Structural  Lexical Structural  The predictions of the system are summarized in the chart below:  (53)  Verkuyl's feature system  State Process Event  NP feature [± SQA] ±  V feature [±ADD TO]  VP/S feature [±T]  -  -  -  + +  +  +  Verkuyl suggests that these three features result in an ontological tripartition of state, process and event, which parallel Bach's (1986) classification of predicates into these three classes. Example sentences of each type are given below:  (54)  States (durative) a.  b.  c.  d.  a sandwich  [+SQA]  [-ADD TO]  [+SQA]  Children  wanted  a sandwich  [-SQA]  [-ADD TO]  [+SQA]  Judith  wanted  sandwiches  [+SQA]  [-ADD TO]  [-SQA]  Children  wanted  sandwiches  [-ADD TO]  [-SQA]  Judith  ate  sandwiches  [+SQA]  [+ADD TO]  [-SQA]  =durative [-T]  Children  ate  sandwiches  =durative  [-SQA]  [+ADD TO]  [-SQA]  [-T]  Children  ate  a sandwich  =durative  [-SQA]  [+ADD TO]  [+SQA]  [-T]  Judith  ate  a sandwich  =terminative  [+SQA]  [+ADDTO]  [+SQA]  [+SQA]  (55)  =durative [-T] =durative [-T] =durative [-T]  Processes (durative) a.  b.  c.  (56)  =durative [-T]  wanted  Judith  Events (terminative) a.  -34-  [+T]  The system requires an additional mechanism to account for intransitive process predicates such as the following:  (57)  a.  John walked  b.  Mary sang  c.  Bill rehearsed  These are problematic for the system presented above in that they are predicted to be [+T] since they combine  [+SQA]  subject NPs with  [+ADDTO]  verbs. However, these predicates are  processes [-T], not events. In earlier work, Verkuyl (1972) suggested that unergative predicates such as these were two-place predicates with  [-SQA]  complements, which would  make the correct [-T] predication. In more recent work, Verkuyl (1993) proposes that some verb stems, like walk, are neutral with respect to [±ADDTO| and thus can choose whether to be inserted in a transitive frame  [+ADD TO]  and an intransitive frame  [-ADD T O ] .  22  For Verkuyl, activities that take object complements are a central problem in that the system predicts that a verb with the feature  |+ADD TO]  with an object with the feature  [+SQA]  will be terminative, rather than durative. This is shown below:  (58)  John/The man  pushed/stroked  a cart/three carts  [+SQA]  [+ADD TO]  [+SQA]  = [+T]  Verkuyl suggests that these types of verbs are different from typical [+add to] verbs; they are in some sense 'incomplete' and that these actually mean 'give one or more pushes to' or 'to exert force to', which would result in terminative aspect, as predicted.  5.3.2.  STATE, PROCESS, TRANSITION:  Pustejovsky (1991,1995)  Following the proposal that verbs have event variables in their logical semantics (Davidson 1967) and the proposal that [e] is an argument of verbs (Higginbotham 1985, Kratzer 1995), some research has attempted to address the relation between verb meaning and the syntactic realization of its arguments by representing events in the lexicon/syntax mapping. Tenny It seems to be that this amounts to saying that verbs like walk can be states since states have a [-ADD TO] specification; it is not clear to me that this is a desired result.  2 2  -35-  (1994) (among others) suggests that the verb lexically determines a set of event roles that determine how and where the arguments of the verb are mapped into the syntax. Pustejovsky (1991, 1995) proposes a representation of the event in the lexical semantics of the verb. Pustejovsky proposes that events have internal structure that can be decomposed into smaller parts. He defines three event types: state (a single event which is evaluated relative to no other event), process (a sequence of events identifying the same semantic expression) and transition (an event identifying a semantic expression which is evaluated relative to its opposition). Their structures are given below: (59)  Pustejovsky's event structures T  P  s  y\  e State  / ei  / \  \ e  -iE E Transition  n  Process  Crucially, unlike any previous analysis, Pustejovsky assumes that events can have subeventual structures. A n accomplishment, for example, can be represented as a transition consisting of a process as a first subevent with a final resulting state, as shown below:  (60)  23  Mary built a house T  ei  e  n  The advantage of this representation is that it allows us to refer to the sub-structure of events in order to account for the behaviour of adverbial modification and its scope effects. 2 3  Although this structure does not appear to contain an opposition per se, at the level of L e x i c a l Conceptual  Structure, the final state that results is negated at the process subevent; that is, the resulting state of the house being built is negated at the process level since the house is not yet built.  -36-  A n adverb can yield different readings in combination with transition predicates. Pustejovsky argues that these different readings can be attributed to scope effects and that his representation of transition predicates can derive the two readings. Consider the manner and stative interpretations of the sentence below:  (61)  Mary rudely departed. a.  = M a r y departed in a rude manner (manner interpretation).  b.  = It was rude of M a r y to depart (stative interpretation).  The sub-event structural theory allows us to distinguish the two readings with respect to scope: the manner reading in (a) is derived by modifying the process sub-event whereas the stative reading in (b) is derived by modifying the entire transition itself.  5.3.3. Advantages and disadvantages A n advantage of both V e r k u y l and Pustejovsky's approaches is that they both attempt to minimize the components of the system that derives the Vendler classification. Verkuyl's system aims to improve earlier classifications by accounting for the effect of arguments on the aspectual nature of a given sentence. This approach is advantageous in accounting for a large range of data (but see the discussion on transitive activities above) and focusing on primitive features. However, V e r k u y l ' s system cannot account for the Skwxwu7mesh data. In V e r k u y l ' s system, the N P and its determiner are relevant in determining the aspectual value of the V P as well as the aspectual value of the sentence. In particular, bare plurals are assigned a [-SQA] which when combined with a |+ADDTO] verb w i l l result in a | T] feature of the V P or S. Bare plurals can have generic interpretations which have been claimed to be absent from Skwxwu7mesh (Gillon in prep; see also Matthewson 1996 on St'at'imcets). Determiners are obligatory in Skwxwu7mesh (Gillon in prep) and most likely Salish in general (Matthewson 1996 on determiners in St'at'imcets and other Salish languages). When asked to translate an English sentence containing a bare plural 'John saw eagles yesterday', in some cases, speakers w i l l offer a sentence that is translated with a quantified D P . This is illustrated below:  -37-  (62)  John  na  kw'ach-nexw-as  kwi  kex  John  RL  look.at-TR(LC)-3  D E T many  sp'akw'us  kwi  chel'aklh  eagle  D E T yesterday  'John saw a lot of eagles yesterday.'  Gillon (p.c.) suggests that it may be the case that all D P s in Skwxwu7mesh are [+SQA]. DPs composed of the determiner kwi (which are the closest to bare nouns) appear to be  [+SQA].  For example, kwi seplin [DET bread] is translated as 'a bread'. If it is the case that all D P s in Skwxwu7mesh are [+SQA], then it cannot be the status of a D P that determines to which class a predicate belongs. The only feature left to distinguish among classes in V e r k u y l ' s system is the [± A D D T O ] feature, which would predict only the following two possible sentence types:  (63)  a.  [+SQA] + [ + A D D T O ] + [+SQA] =  [+T]  b.  [+SQA] + [-ADD T O ] + [+SQA] =  [-T]  This incorrectly predicts two classes of predicates: a class of states and a class made up of accomplishments and achievements, leaving no class of processes:  (64)  a.  [ + S Q A ] + [ + A D D T O ] + [ + S Q A ] = [+T]  ACHIEVEMENTS, ACCOMPLISHMENTS  b.  [+SQA] + [-ADD T O ] +  STATES  [ + S Q A ] = [-T]  This is not a desired result since (i) there is evidence for a class of process-like predicates in Skwxwu7mesh (namely, activity predicates that test as durative, not terminative):  (65)  a.  na  t'ich-im  lha  RL  swim-lNTR  DET  Mary welh haw k-as i Mary CONJ NEG IRR-3CNJ P A R T ' M a r y swam but she didn't finish swimming.'  huy finish  (ii) accomplishments and achievements do not behave as though they are members of the same class of predicates in Skwxwu7mesh. A m o n g other things, accomplishments are easily compatible with C V reduplication, while achievements are not always:  T h i s is not the full picture. See Chapter Five for detailed discussion.  -38-  24  (66)  a.  na  p'a-p'ayak-ant-as  ta  snexwilh-s  RL  REDUP-f i X - T R - 3 ERG  DET  canoe-3POSS  'He's i n the process of fixing his canoe.' b.  *chen xe-xewtl'-an  ACCOMPLISHMENT  xel'ten pencil . attempted: 'I am breaking the pencil.' lS.SG  REDUP-break-TR  ta  DET  ACHIEVEMENT  Thus, using V e r k u y l ' s system to try to account for the aspectual classification of Skwxwu7mesh predicates leads us along the wrong path. Not only would the application of this system for Skwxwu7mesh rely on a feature that seems to be irrelevant for the language as a whole  [+SQA],  but it would not derive the correct classification of predicates that are  observed. This thesis w i l l show that there is evidence for more than two classes of predicates in Skwxwu7mesh which result not from the feature specifications of DPs, but from a more articulated classification of predicates, which cannot be derived by a binary feature. T o preserve the V e r k u y l system for Skwxwu7mesh, other factors that would derive this classification would have to be introduced, which again, is not a desired result. A n advantage of Pustejovsky's system is that it is constructed in such a way that it can derive a number of possible predicates. In earlier work on Skwxwu7mesh, I adopted a Pustejovsky-type system to attempt to account for the Skwxwu7mesh facts (Bar-el 1998, 2003). However, it was not possible to map the Skwxwu7mesh data directly to the representations suggested in the event structure approach, but instead, this system needed to be adapted to account for the facts observed in Skwxwu7mesh. Although my analysis here has changed, the facts remain the same; thus I do not think Pustejovsky's approach is the right one for Skwxwu7mesh.  5.4. Operator-based definitions of predicate classes To conclude this overview of the different types of aspectual classifications, I end with a presentation of Dowty's (1979) aspect calculus and a brief discussion of Rothstein's extension of his sytem. It is these classifications on which I base my proposals for Skwxwu7mesh in the following chapters. Many of the diagnostics that Dowty proposes to distinguish among Vendler's four verb classes are not applicable to Skwxwii7mesh, and possibly Salish in general, Dowty and Rothstein do provide an account of the meaning of the  -39-  predicate classes that are observed in English. Given that the focus here is on the initial and final points of predicates, the system presented in Dowty provides a straightforward way to derive these facts.  5.4.1. BECOME, DO, CAUSE: D o w t y (1979) Drawing on Vendler's verb classes, Dowty (1979) observes a number of diagnostics, in addition to those proposed by Vendler, that distinguish among four classes of verb phrases. These diagnostics are summarized in the chart below. For expository purposes, I have divided them into two sets (Dowty's notations are as follows: • / / x = does or does not exhibit the features of that test, " O K " = grammatical, semantically normal, "bad" = ungrammatical, semantically anomalous, " N / A " = test does not apply to verbs of this class): (67) 1.  Criteria that distinguishes among the four classes (adapted from Dowty 1979:60) Ach. Acc. Act. States Criterion y y X meets non-stative tests (see below)  2.  has habitual interpretation in simple present  X  y  y  y  3.  4> for an hour entails <j) at all times in the hour  •/  y  X  N/A  4.  x is tb-ing entails x has cb-ed  NA  y  X  N/A  5.  ambiguity with almost  X  X  y  X  6.  x ())-ed in an hour entails that x was cb-ing during that hour  N/A  N/A  y  X  OK  OK -  OK  Bad  bad  bail  OK  OK  OK  OK  :,OK  Bad  bad.  bad  OK  Bad  bad  OK  OK  H;id  7. . (() for an hour; spend an hour (j)-ing 8.  ' ..  ,<j) in an hour; take ah hour to (b  9. , complement of stop 10. complement of finish 1L  •  .  occurs with studiously, attentively, careful 1\..  The non-stative tests that Dowty refers to in the above chart are as follows:  2 5  Dowty suggests that achievements are like statives according to some stativity tests, but not others: *John  persuaded Bill to notice the stranger in the room vs. John is noticing a stranger in the room. 2 6  While Dowty suggests that these diagnostics test for agentivity, the fact is they seem to test for volition, or  possibly control (or lack of).  -40-  (68)  a.  Only non-statives occur in the progressive i. *John is knowing the answer.  (* state)  i i . John is running.  (activity)  i i i . John is building a house.  (accomplishment)  Only non-statives occur as complements of force and persuade  c.  d.  e.  i. *John forced Harry to know the answer.  (*state)  i i . John persuaded Harry to run.  (activity)  i i i . John forced Harry to build a house.  (accomplishment)  Only non-statives occur as imperatives i. * K n o w the answer!  (*state)  ii. Run!  (activity)  i i i . B u i l d a house!  (accomplishment)  Only non-statives co-occur with deliberately,  carefully  i. * John deliberately knew the answer.  (*state)  i i . John ran carefully.  (activity)  i i i . John carefully built a house.  (accomplishment)  Only non-statives appear in psuedo-cleft  constructions  i. *What John did was know the answer.  (*state)  i i . What John did was run.  (activity)  i i i . What John did was build a house.  (accomplishment)  Dowty proposes that these classes are distinguished from one another in which operators are present in their Logical Structure. He uses M c C a w l e y ' s (1968) B E C O M E and C A U S E operators, as well as introducing a D O operator, and proposes that these operators are part of the basic meanings of verb phrases. The result is the aspect calculus shown below, with example sentences for each (c^ and (3, stand for arbitrary individual terms, jt and p stand n  -41-  m  for arbitrary stative (n-place) predicates, and 4> and i|> are arbitrary formulas, either complex or atomic)  (69)  Simple statives 'John knows the answer.'  jt  (70)  Simple activities  D O (a  'John is w a l k i n g . '  n  a„)  (a,,  u  [7i (a n  ...,a )])  u  n  27  28  (71)  Achievements a.  Simple achievements 'John discovered the solution.'  B E C O M E [jt ( a „ . . . , a ) ]  b.  Inchoation of activity 'John began to walk.'  B E C O M E [ D O (a,, K  n  n  ( a „  a„)])]  Inchoation of accomplishment B E C O M E (|) (where can have the structure of any of the accomplishments below) ' John began to build a house'  (72)  Accomplishments a.  Non-agentive accomplishments  [ [ B E C O M E <(>] C A U S E [ B E C O M E ip]] (where (j) and are stative sentences)  ' The door's opening causes the lamp to fall down.' b.  (Non-intentional) agentive accomplishments [[DO ( a „ [Jt ( a , , . . . , « „ ) ] ) ] C A U S E [ B E C O M E [p 'John broke the window.' n  c.  Agentive accomplishments with secondary agent [[DO ( a „ [7t ( a , , a „ ) ] ) ] C A U S E [DO n  (ft,  [p  m  m  ([3,,..., | 3 J ] ] ]  ((3,,..., | 3 J ] ) ] |  'John forced B i l l to speak/build a house.'  2 7  It should be noted that the example that Dowty provides is a present progressive, which cannot really be  described as "simple". 2 8  Dowty points out that the inchoation of an activity or accomplishment does not lexicalize as s single verb in  English, or it does so marginally; he suggests that germinate might be an example of the inchoation of an activity.  -42-  d.  Intentional agentive accomplishments (?) [DO ( a „ D O ( a „ n (a,, ...,a )) C A U S E <> |] (where (j) may be any non-stative sentence) 'John murdered B i l l . ' n  n  (Dowty 1979: 123-125) Dowty points out that there is a possible structure that his system allows but does not fit into Vendler's categories; this is shown below:  (73)  D O - ( a , , B E C O M E [JT„ ( a „  a )]) n  He suggests that "these would be basic actions, events under the unmediated control of an agent that are not brought about by any subsidiary activity. Plausible candidates would be John opened his eyes and John raised his arm - no conscious causal activity is apparent" (p. 125) Using Bennett and Partee's (1978) notion of intervals, but introducing the notion of initial boundary interval and final boundary interval, Dowty defines the truth conditions for [ B E C O M E <[)] in the following way:  [BECOME  <f>) is true at / iff there is an interval J containing the initial bound of / such that -• (p  is true at J and there is an interval K containing the final bound of / such that (j) is true at K (Dowty 1979: 140)  Thus,  [BECOME  <p] w i l l be true at the interval / in the following situation:  (74) I  E  ^  is true  -i 0 is true  K  J  29T h e  question mark here is given in D o w t y (1979).  -43-  •  Dowty notes that the truth conditions of B E C O M E as stated above do not put any requirements on the truth value of (p at /, or any times within /. This predicts  [BECOME <p]  to  be true in counterintuitive situations, where, for example, there is a large interval where —> cp) is true followed by a large interval where (|) is true. The example he gives is a situation where the door is closed for a long period and then it suddenly comes to be open and stays open for some time. It would be strange in this case to say the door opens is true of any interval in that whole period as long as the interval contains the first moment that the door opened. The preference would be to limit the truth of the door opened to the smallest interval over which the change of state has taken place. Dowty suggests that pragmatics will rule these situations out and force the relevant reading; he argues that adding another clause to the truth conditions can be too restrictive, raising its own problems. Dowty's notion of D O dates back to Ross (1972) who proposed that "every verb of action in embedded in the object complement of a two-place predicate whose phonological realization in English is do" (p. 70). Dowty notes that " D O does not necessarily connote action in the usual sense" (p. 117) as thiswould not explain examples such as the following where there are no actions:  (75)  a.  John is being quiet  b.  John is ignoring Mary  These sentences involve activities that have been derived from states by the addition of D O ; note, that D O seems to add control, rather than simply agentivity, to the sentence. Furthermore, the notion of intentionality or volition is a close approximation to the meaning of D O . In the following examples, Dowty suggests that there is no necessary intention:  (76)  a.  John is being obnoxious  b.  John is being a fool  However, he suggests that some property under the control of John qualifies him as obnoxious, for example.  -44-  Thus, Dowty concludes that D O should satisfy the following condition:  (77)  [ D O ( a , <> | ) ** <|> A u.tu.c.o.a. ((f))] = "is under the unmediated control of the agent (individual denoted by a ) "  Many of the diagnostics that Dowty uses to generate the Vendler classes cannot be used to distinguish among predicate classes in Skwxwu7mesh. For example, tl'iyi7 'stop' and huy 'finish' do not distinguish between activities and accomplishments in Skwxwu7mesh. Furthermore, there does not seem to be a contrast between in X time and for X time. A number of the tests are language-specific; for example, there is no simple present in Skwxwu7mesh and thus testing for whether it induces a habitual reading is not relevant. Finally, some of the entailment tests are not easily tested in fieldwork settings.  30  Mattina (1996) takes another approach to classifying the aspectual classes of Okanagan (an Interior Salish language), namely a morphological one. She argues for an inventory of aspectual classes based on morphological evidence where certain "bases" are compatible with certain affixes. I follow Mattina's intuition that we must rely on languageinternal evidence for classification; however, I extend this idea beyond morphology and into semantics by relying on not only on speakers' grammaticality judgements of given sentences, but on the meanings they give to sentences and the various contexts that they are permitted in. I argue that predicate classes are distinguished based on the presence/absence of initial and final points; the diagnostics I use precisely pick out these points and thus relate the data directly to the representations.  5.4.2. BECOME, DO: R o t h s t e i n (2004) Based on Dowty (1979), Rothstein proposes the following verb class templates in a neoDavidsonian theory of verb representations where verbs are predicates of events (the relevant thematic roles are set aside here). States are bare event predicates, activities are bare event predicates under the scope of a D O operator and achievements are bare event predicates under the scope of a B E C O M E operator. Accomplishments  3 0  are more complex in that they are  See Matthewson (2004) and Kiyota (2005) for claims that standard tests do not apply in St'at'imcets or  SsncaOan, respectively.  -45-  created by summing an activity and a culmination point (accomplishments are discussed in further detail below):  (78)  a.  States  Ae.P(e)  b.  Activities  Xe.(DO(P))(e)  c.  Achievements  Ae.(BECOME(P))(e)  d.  Accomplishments  A.e.3e,3e [e= (e ue ) A (DO(P))(e,) A (BECOME(P))(e )] s  2  1  2  2  In each of the templates, P is a variable over the idiosyncratic content of particular lexical items. Rothstein uses these templates to account for a variety of behaviours of predicate classes in different environments. Culminations: after suggesting three alternatives to the question of what are culminations, Rothstein finally settles on a fourth. She argues that the culmination of an accomplishment is the final minimal event in an incremental process. Thus, the accomplishment template is as follows C ^ a n d . ^ give the content of the activity and B E C O M E events):  (79)  A.yX,e.3e 3e [e= (e Ue ) s  1  2  A ACTIVITY ^ A BECOME  < Y >  1  2  (e,) A Ag(e,)=x A Th(e,)=y (e ) AArg(e )=Th(e!) 2  2  AlNCR( ,e ,C(e ))] e ]  2  2  Relevant to the analysis in this thesis is Rothstein's shift operation which she proposes for progressive accomplishments. She shows that although achievements are predicted to be incompatible with the progressive (since they are near instantaneous events), they are sometimes compatible:  (80)  a.  The old man is dying,  b.  The plane is landing. (Rothstein 2004: 36; ex. 2d-e)  -46-  Rothstein argues that previous analyses that attempt to account for progressive achievements (Mittwoch 1991, Verkuyl 1989) incorrectly predict that accomplishment and achievement verbs should show the same behaviour in the progressive. Based on a number of diagnostics, Rothstein shows that this is not the case: (i) achievements that are felicitous in the progressive are limited (some are never accepted), as shown below:  (81)  a.  # B i l l is noticing that Mary has dyed her hair.  (p. 42)  b.  # Mary is spotting her friend at the party.  (p. 42)  (ii) progressive achievements have a "slow-motion" reading where accomplishments do not; (iii) a temporal modifier (in futurate progressive) can modify the activity part of an accomplishment, but never the telic point; with achievements, this is the most natural reading. This is shown below:  (82)  a.  W e are eating dinner in half an hour.  ACCOMPLISHMENT  (p. 43)  b.  The plane is landing in half an hour.  ACHIEVEMENT  (p. 43)  In (a) the modifier in half an hour suggests that either the entire event of eating dinner will occur in that time, or that the eating (activity portion) will begin in half an hour. What it cannot mean is that dinner w i l l be over in half an hour. This contrasts with the achievement in (b) where the most natural meaning is that the culmination (the landing) will take place after half an hour. (iv) no "stops along the way" with progressive achievements: they cannot be part of a bigger eventuality. This following example is an illustration:  (83)  a.  Accomplishment Mary is running to the Netherlands. In fact she is running to Amsterdam.  b.  Achievement M a r y is arriving in the Netherlands. In fact she is arriving in Amsterdam. (p. 43)  -47-  In the accomplishment sentence in (a), Mary first arrives at the Dutch border, then she runs to Amsterdam (the running to the Netherlands is a stage of the running to Amsterdam). However, in the achievement sentence in (b), the arrival at the Dutch border is identical to arrival in Amsterdam (the first cannot be seen as a stage of the second). (v) achievements cannot be modified by "halfway through". This is shown below:  (84)  a.  She is halfway through walking to the station.  ACCOMPLISHMENT  b.  #She is halfway through arriving at the station.  ACHIEVEMENT  31  (p. 44)  Finally, (vi) only progressive achievements can be paraphrased as "about to"; take the following sentences as an example:  (85)  a.  Jane is building a house.  ACCOMPLISHMENT  b.  The vase is falling.  ACHIEVEMENT  (p. 44)  The accomplishment in (a) cannot be paraphrased as 'Jane is about to build a house', although the achievement in (b) can be paraphrased as 'The vase is about to fall'. Given these data, Rothstein argues that "while the progressive does not treat the achievement as a special kind of lexical accomplishment verb, it does trigger a type-shifting operation which results in an accomplishment being derived from the achievement" (p. 37). The proposed shift rule is given below:  (86)  SHIFT(VP ) : Ae.(BECOME)(e) Ae.3e 3e [e='(eiUe ) A (DO(a))(e,) A ( B E C O M E ( P ) ) ( e ) A Cul(e)=e ] punctual  1  2  2  2  2  The achievement V P is thus shifted into the structure of an accomplishment V P , the culmination of which is given by the achievement V P . A n example is given below:  H . Davis (p.c.) states that 'the plane is landing' can be modified by halfway through. I do not share this judgement and unfortunately have not tested this with other English speakers. I thus set it aside here. 3 1  -48-  (87)  S H I F T (X.e.(ARRIVE A T T H E S T A T I O N ( e ) A Th(e)=x = Xe.3e 3e [e= (e ue ) A (DO(a))(e,) A A R R I V E A T T H E S T A T I O N ( e ) A T h (e )=x A (Cul(e)=e ] s  1  2  1  2  2  2  2  The denotation of the shifted V P is a set of derived accomplishment events which are sums of an activity and a culmination, where the culmination is in the denotation of arrive at the station. The S H I F T operation thus preserves the meaning of the original lexical predicate intact, and uses it as a piece in the construction of a new derived predicate meaning. This is a formal analysis of Smith's (1991) suggestion that progressive achievements focus on "detachable" preliminary stages of the achievement (see also K a m p and Reyle 1993).  5.4.3. Implementation: theoretical assumptions The proposed representations of Skwxwu7mesh aspectual classes rely heavily on Rothstein's (and thus, Dowty's, as well); however, as the inventory of aspectual classes in Skwxwu7mesh is not identical to that of English, the templates need to be adapted in some way. I follow Rothstein in using a system that provides the meanings of various types of predicates in Skwxwu7mesh using the operators D O and B E C O M E . I propose that predicates can be distinguished from each other based on the presence of intrinsic initial and final points. In the system used here, initial and final points are represented as B E C O M E events in the predicate representations. With respect to D O , this thesis will not present a more detailed discussion than that given by Dowty. I w i l l assume D o w t y ' s condition on D O , namely, that the event in question is under the unmediated control of the agent. This is an appropriate way to think about Salish predicates in that the two classes that invoke a D O operator are the two classes of predicates that have overt (in)transitivizers that also denote the control of the agent. There is one problem with this assumption as well as one prediction that is yet to be understood. Beginning with the prediction, it is crucial here that in Skwxwu7mesh (in)transitivizers encode not only control over the event, but limited control as well as noncontrol. This would suggest that in cases where an accomplishment predicate, for example, 32  is marked by a limited control or non-control transitivizer, we might predict that D O would  3 2  T h e Skwxvvu7mesh transitivity system is quite complex; a complete discussion is beyond the scope of this  thesis. See Kuipers (1967) and Jacobs (1999) for discussion.  -49-  not be present in the denotation of the predicate. This may very well be the case, and preliminary data do suggest minimally that there is a difference in culmination requirements for control and limited control marked accomplishments. The problem with the correlation between D O and the (in)transitivizers is that there are cases where Skwxwu7mesh achievement predicates seem to be marked by an intransitivizer:  (88)  chen  mekw'-em  kwetsi  smant  ls.SG  find-lNTR  DEM  rock  'I found a rock.' Achievements make up the class of predicates that denote B E C O M E events only. Thus, under this analysis of D O , further explanation is required to explain why there are control intransitivizers in cases where there is no D O in the denotation of the predicate.  6. Grammatical aspect In this section, I outline the assumptions I make with respect to the perfective/imperfective distinction, beginning with the perfective. Perfective sentences view the situation or event as a whole, while imperfective sentences look inside of a situation or event (see Comrie 1976, among others; the behaviour of the Skwxwu7mesh perfective is discussed in greater detail in Chapter Four). The contrast can be seen in the following data:  (89)  a.  John baked a cake.  PERFECTIVE  b.  John is baking a cake.  IMPERFECTIVE  The perfective sentence in (a) entails that John did some baking and that baking led to a cake being baked. The imperfective sentence in (b) has no such entailment; in fact, it may be the case that the cake never gets baked (e.g., John could drop the batter on the floor). Accounting for the contrast between the perfective and imperfective, Kratzer (1998) proposes that aspect heads denote operators that map properties of events into properties of times. They can then impose conditions on the relation between event time and reference  -50-  time ("topic time", in K l e i n ' s (1994) terms). Following Kratzer, I assume the semantics for the two grammatical aspectual oppositions given below, beginning with the perfective:  Perfective (90)  a.  event time included in reference time  [REFERENCE TIME [EVENT TIME]  b.  XQktSe [Q(e) A x(e) C t]  c.  A property of events Q is mapped into a property of times and it is true of a time t (the reference time) just in case t includes the running time (x; the event time) of a Q-event.  The representation of the aspectual classes given earlier replace Q in Kratzer's definition of the perfective. I illustrate with the following French sentence (Smith 1997 claims that the passe compose in French is a perfective):  (91)  L' ete passe ils ont DET.M summer PAST 3PL.M have 'Last summer, they built a cabin.'  construit built  une INDEF.F  cabine cabin  In this perfective sentence, the reference time is I'ete passe 'last summer, and the event time is the running time of the event ils construire une cabine 'they build a cabin'. That the event time must be included inside the reference time is emphasized in examples such as the following where it is not possible for the event to be still on-going:  (92)  # L' ete DET summer peut-etre maybe  passe past  ils ont 3PL.M have  construit built  qu' that  ils la contruisent 3PL.M 3SG.F build  une DET  cabine; cabin  encore still  'Last summer they built(PC) a cabin; perhaps they are still building(PRESENT) it.' (Smith 1997: 194; ex. (2a))  The perfective is schematically illustrated below:  -57-  (93)  Perfective: event time included in reference time L'ete passe Us out construit une cabine  Event Time 4— ils construire une cabine  —•  —h,  L, eie passe  ^  Reference Time  The following are Kratzer's semantics for the imperfective:  Imperfective (94)  a.  reference time included in event time  b.  XQktBe [Q(e) A t C ( ) ]  c.  A property of events Q is mapped into a property of times and it is true of a time t just in case t is properly included in the running time (x) o f a Q-event.  T  [EVENT T I M E [REFERENCE TIME]]  e  The following French sentence containing the Imparfait is an illustration:  (95)  L' ete passe ils construisaient une DET summer past 3PL.M built DET 'Last summer, they were building a cabin.'  cabine cabin  For this sentence, the event time is again the time during which they built a cabin, and the reference time is last summer. Since the sentence is in the imperfective, the beginning and end of the event are outside of the reference time; that is, the imperfective makes no reference to the beginning or end of the event, and it could be the case that they began building the cabin before last summer and did not finish by the end of the year; as the sentence below shows, they could still be building the cabin:  -52-  (96)  #L' DET  ete summer  passe past  ils construisaient une 3 P L . M built DET  cabine; cabin  peut-etre maybe  qu' that  ils la contruisent 3 P L . M 3SG.F build  encore still  'Last summer they were building(iMPF) a cabin; perhaps they are still b u i l d i n g ( P R E S E N T ) it.'  (Smith 1997: 198; ex. (11a))  The imperfective is illustrated schematically below:  (97)  Imperfective: reference time included in event time L'ete passe ils construisaient une cabine  Event Time ils construire une cabine  L'ete passe Reference Time  This analysis of the perfective and imperfective, along with the proposed representations for predicates account for the readings of Skwxwu7mesh perfective and imperfective sentences. This is discussed in details in Chapters Four and Five.  7. The study of aspect in Salish  33  The Salish language family is made up of 23 languages that are or were spoken  "in an area encompassing coastal British Columbia and Washington State (including parts of Vancouver Island, the G u l f Islands, and the San Juan Islands), the interior of British Columbia as far north as the southern Cariboo  While Salish aspectual morphemes, their shape, phonology and morphology have been discussed in a number of works, including the grammars of a various Salish languages, I limit the discussion and references cited here to work that attempts to address the semantics of some of these aspectual morphemes, as well as the classification of predicates in terms of their aspectual properties.  3 3  -53-  region, most of the northern interior of Washington, Idaho, and into Montana, with a pocket on the Oregon Coast" (Czaykowska-Higgins and Kinkade 1998:1)  The Salish languages fall into five groups. T w o of these consist of a single language: Bella Coola and Tillimook. These languages are geographically isolated from the remaining Salish languages and are divergent from the other languages. The list below comprises the non-indigenous appellation most commonly used in the linguistics literature:  (98)  34  Salish Language Family (based on Czaykowska-Higgins and Kinkade 1998: 3) I. Bella Coola II. Central/Coast Salish Comox Pentlatch Sechelt  Skwxwu7mesh (Squamish) Halkomelem Northern Straits Klallam Nooksack Lushootseed Twana III. Tsamosan Quinault Lower Chehalis Upper Chehalis Cowlitz IV. T i l l i m o o k V . Interior Salish Lillooet Thompson Shuswap Colville-Okanagan Columbian Spokane-Kalispel-Flathead Coeur d'Alene Aspect has been a topic of research in the Salish literature for some time. There has been much descriptive work, many discussions of which have been the focus of sections of See C z a y k o w s k a - H i g g i n s and K i n k a d e (1998) and Poser (2003) for discussions on names of Salish languages.  3 4  -54-  grammars for various languages across the family. Czykowska-Higgins and Kinkade (1998) provide a brief overview of some central issues in the study of aspect in Salish and the related references. They address four main areas of research on aspect across Salish: (i) stative aspect (the stative marker), (ii) a distinction which they label imperfective/continuative/actual/habitual aspect on the one hand and perfective/noncontinuative/nonactual aspect on the other, (iii) unrealized, repetitive, inchoative and (iv) the use of aspectual morphemes and their interactions with transitivity, control and mood. More recently, there have been more works that have focused on the semantic properties of expressions of aspect across Salish. In particular, there has been some research that has focused on understanding of the meaning of events and the classification of aspectual classes in Salish languages. The range of aspectual markers across the Salish language 35  family is outlined in detail in Kinkade (1996) where he provides a reconstruction of aspectual markers in Salish. In particular, he examines the imperfective, stative, unrealized, inchoative, repetitive, affective, state (distinguished from stative) and mutative. Martina's (1996) thesis explores aspect in Okanagan (Southern Interior); her focus is on the role of aspect and category in the word formation processes in the language from a morphological perspective. Martina uses a number of diagnostics that show the compatibility of certain affixes with certain types of "bases" to motivate an inventory of Okanagan predicate classes. There have been a number of papers on the aspectual classification of predicates in St'at'imcets. Davis and Demirdache (1995) was the first attempt to analyze St'at'imcets lexical aspect adopting a Pustejovsky approach. Davis and Demirdache (2000) (building on the work of Davis 1997 and Demirdache 1997) argue that all roots in St'at'imcets (Interior Salish) are derived from unaccusative bases, but that these bases have underlying causative lexical semantics. Transitive and intransitive predicates are derived via suffixation of (in)transitivizers and these suffixes also serve to foreground the relevant sub-events of the causative structure (a la Pustejovsky). Burton and Davis (1996) explore states in St'at'imcets (Northern Interior). Their work is the first to formalize the relationship between stative aspect  35  1 refer the reader to C z y k o w s k a - H i g g i n s  and Kinkade (1998) and van Eijk's (2001) Salish bibliography for  overviews and references o f research in this area.  -55-  in the verbal domain and possessive aspect in the nominal domain. They also draw on the Pustejovky model. Bar-el (1998) was the first look at aspectual classes in Skwxwu7mesh. In this work, I also proposed an inventory of Skwxwii7mesh aspectual classes using a Pustejovsky model. In particular, I showed that the effect of adverbial quantification provides a clue in this regard. More recently, Kinkade and Kiyota (2004) examine different morphemes in Salish languages which they argue indicate beginning, middle and end of change in states across the language family. K i y o t a (2005) examines the aspectual classification of verbs in SsncaOan (Northern Straits, Central). He argues that most English based classification diagnostics do not apply to Saanich and argues for cross-linguistic variation in the aspectual classification of predicates . Kiyota uses the following three language-internal diagnostics:  (99)  a.  interpretation of tenseless forms  b.  interpretation with the particle k l  c.  stative prefix s-  w  Based on the results of these tests, Kiyota argues for the following three primitive aspectual verbal classes in SancaGan:  (100)  a.  States  b.  Activities  c.  Achievements  Building on data first introduced in J. Davis (1978), Watanabe (2003) discusses the contrast of control marked transitives and non-control marked transitives with respect to culmination in Comox. He suggests that "encoding completion" may be a more important function of the non-control transitivizer than "control" since non-control marked transitives cannot be cancelled although control marked transitives can. Originally stemming from data discussed in Davis and Matthewson (2001), Matthewson (2004) take a close look at the nature of accomplishments in St'at'imcets and  -56-  argues that unlike in English where accomplishments have a culmination entailment, in St'at'imcets, accomplishments have a culmination implicature. Bar-el, Davis and Matthewson (2004) extended this analysis to account for data from Skwxwu7mesh as well. I build on this emerging body of literature on aspect in Salish in providing a semantic analysis of aspectual classes in Skwxwu7mesh based on a larger number of diagnostics than previously proposed for Skwxwu7mesh. Bar-el (2004a) was the first attempt at an account that takes into consideration the initial point of Skwxwu7mesh predicates and the present study extends this analysis even further. It is my hope that the diagnostics used here will continue to be used for other Salish languages, which w i l l result in an even larger inventory of data that is easily comparable.  8. Research methods The Skwxwii7mesh data on which this thesis is based stem from original fieldwork conducted with eight Skwxwu7mesh elders. A l l the elders spoke Skwxwu7mesh as a first language and as an only language for a number of years. Their ages vary from mid-60s to early 90s. Previously, I worked with the late Doris White for two years, the late E v a Lewis for seven years, the late Tina Cole for five years, the late Yvonne Joseph for eight years, the late Chief Lawrence Baker for seven years. More recently, I have worked with Margaret Locke and Lena Jacobs for approximately two years and Frank Miranda for about half a year.  8.1. Elicitation Elicitation is the primary method of data collection used in my fieldwork. This involves any of the following tasks:  (101)  a.  Providing an English sentence to the speaker and asking for a Skwxwii7mesh translation.  b.  Providing a Skwxwu7mesh sentence to the speaker and asking for an English translation.  c.  Providing a context, then presenting a Skwxwu7mesh sentence to the speaker and asking for a felicity judgement.  -57-  d.  A s k i n g speakers to compare two Skwxwu7mesh sentences and provide a judgement on which sentence is more appropriate and/or whether there is a difference in meaning.  e.  Presenting a visual context (e.g., picture) to the speaker as well as a Skwxwu7mesh sentence and asking whether the sentence is appropriate given the context(s).  f.  Presenting a variety of pictures to the speaker, as well as a Skwxwu7mesh sentence and asking which picture provides the most appropriate context.  36  With the exception of two speakers, the speakers with whom I work do not read or write the Skwxwu7mesh language. A s a result, each session involves oral discussion alone; speakers are not asked to read and translate Skwxwu7mesh sentences, nor are they asked to write Skwxwu7mesh sentences down. With respect to one speaker who did read and write the Skwxwu7mesh language, the same methods were used (as with the other speakers), but he chose to write the sentences down for himself. Elicitation sessions are attended by elders, U B C linguists and members of the Squamish Nation Education Department. In some cases, sessions were attended by more than one speaker, and data was collected from both elders. A s well, often more than one U B C linguist attended a given elicitation; sessions were open to all members o f the Squamish Nation Education Department, so often administrators and teachers would attend, asking their own questions as well. Recorded along with the Skwxwu7mesh and English sentences, are all volunteered sentences and glosses as well as all comments made by speakers. The importance of recording speaker comments has been noted in the literature (e.g., Mithun 2001); furthermore, Matthewson (in press) cautions the researcher to record all comments made by speakers as their translations alone may not be sufficient to base a conclusion; as she notes, "consultants' comments should be given the same weight as translations - they are a clue to meaning". Matthewson takes comments to include "statements the consultant makes about context, alternative ways of saying things, fine-grained grammaticality judgments, meanings of parts of words, comments about formality, alternative word order possibilities, and so on".  In many cases, a variety o f methods from the list above were used. A s for pictures, this was not attempted until later in my research and thus data elicited v i a pictures are limited in scope.  3 6  -58-  Wherever possible, the same sentence has been elicited with different speakers as well as the same speaker on a different day/different elicitation session.  8.2.  Texts  There are two sets of textual materials in Skwxwu7mesh available for reference: Kuipers (1967) and Kuipers (1969), the first and second parts of the only Skwxwu7mesh grammar that has been published. Texts are an important resource to be used in conjunction with elicitation. These texts offer extended discourses, as well as a picture of the Skwxwu7mesh language in the middle of the last century. Texts are limited resource, however, since they provide only possible structures in the language, and do not illustrate what is impossible, which is of great importance to this research as ungrammatical/infelicitous sentences show just as much as grammatical/felicitous sentences. Furthermore, the lack of certain structures in texts cannot be taken as an indication that those structures are not possible, nor can we assume that the meaning ascribed to a given sentence in a text is the only possible meaning. Thus, while textual materials are very relevant, they are used in this study as supplementary materials. Where possible, I have included data taken from the aforementioned texts in the relevant places throughout this thesis.  9. T h e f i e l d w o r k  experience  3 7  While this thesis is a formal linguistic study, the collection of data and relationships that developed through the fieldwork on which this work is based was an experience that stretched beyond theoretical linguistics. Since this is a major part of the involvement in this type of work, it is important and necessary to discuss the issues that lay outside the scope of formal linguistics that were, and continue to be, part of this experience. Conducting fieldwork with Skwxwu7mesh elders in conjunction with the Skwxwu7mesh Nation Education Department is an experience that involves more that just data collection. Researchers have been given the opportunity to hear firsthand accounts of the lives of elders, the history of Vancouver, and the culture of the Squamish Nation, including traditions and songs. W e developed strong relationships not only with the elders but with Skwxwu7mesh language teachers.  3 7  T h i s section is taken from Bar-el, G i l l o n and Jacobs (2004).  -59-  An important result of our research was an understanding of the relevance of our work for the Skwxwu7mesh community, that fieldwork can benefit the Skwxwu7mesh community and not just the academic community. In particular, we came to understand that (i) the participation of language teachers in these sessions has a direct impact on their teaching of the Skwxwii7mesh language, as well as their development of linguistic awareness of their language, (ii) the data collected can be used for curriculum development as part of the language revitalization efforts undertaken by the Skwxwu7mesh community, and (iii) our data directly contribute to the documentation of the Skwxwu7mesh language for future generations. Through our research we discovered that speakers appreciate our work and our interest in their language and culture. Furthermore, we became aware of the fact that our fieldwork sessions provided an opportunity for speakers to use the Skwxwu7mesh language. A major result of this work is the uncovering of variations among speakers' grammars. Our research has shown that these differences should not be ignored. In my own research, this has become particularly relevant in my work on the Skwxwu7mesh auxiliary mi, which is often translated as 'come' and can be used as a main verb or an auxiliary (Kuipers 1967). These facts have been replicated in my own fieldwork, along with an 38  additional reading 'start/begin to' that occurs with some predicates. This reading, however, is not available for all speakers; crucially, the 'start to' reading was available to speakers that are no longer alive. The fact that there are only a handful of fluent native speakers of Skwxwu7mesh remaining has driven us to record these differences as they may hot be visible to future researchers. Perhaps most significantly, we have learned how important the Skwxwu7mesh language is to Skwxwu7mesh speakers. In particular, (i) they want the Skwxwu7mesh language and culture to survive, (ii) they see the connection between language and culture as inseparable, (iii) they want to pass the language down to the next generation, and (iv) they want to hear the children in the community speaking it.  This is discussed in greater detail in Chapter Three.  -60-  10. Contribution of this thesis This thesis makes four major contributions. First, it provides a formal semantic examination of the Skwxwu7mesh aspectual system. Aspect in Salish languages is a relatively new area of study in Salish linguistics and has become the focus of much research in recent years (see section 3.2). Thus far, work in this area has challenged many of the universal claims about aspect, this thesis w i l l contribute to this body of work, hopefully inspiring new avenues of research on aspect in Salish languages. Second, this thesis contributes to cross-linguistic theoretical research on aspect. M u c h of the work on aspect, though it spans a number of different languages and language families, has assumed that aspectual systems are all structured in the same way. This thesis challenges this claim by proposing that aspectual classification systems are a locus of cross-linguistic variation. Furthermore, this thesis forces us to reconsider some of the claims made for a language such as English, on which much work has been done, but has left many issues unexplained. Moreover, although the issue of aspect dates back to Aristotle, and has been the focus of a huge body of research since, a complete understanding of the range of aspectual systems cannot possibly be understood until as many languages as possible have been studied. A fourth contribution that this thesis makes relates specifically to cross-linguistic data on aspect. Skwxwu7mesh, like all Salish languages, is an endangered language, with very few native speakers remaining. This work, based on data that may not be available in the next generation, contributes to the collection of cross-linguistic data on aspectual issues. The final contribution pertains to the documentation of the Skwxwii7mesh language; Skwxwu7mesh is greatly endangered, with less than twenty fluent native speakers remaining whose ages range from their seventies to nineties. There is a great deal of work being done within the community to document the language, and to say that time is running out is a vast understatement. I hope that this thesis will not only contribute to the field of linguistics, but to the Skwxwu7mesh community as well, and their efforts to document as much of the language as possible so that it can be taught to younger generations and passed on to future generations.  -61-  Chapter 2: Intrinsic Final Points  "The difference in representations with and without endpoints ought to show up in the possibilities...for  the interpretation of sentences which follow them.  That is, one would.. .expect that when trying to place the ..constituent introduced by the next sentence within the narrative time structure we have already constructed, we would be able to make use of.. .endpoints"  (Kamp  and Rohrer 1989, Chapter One, p. 15-16, as quoted in Smith 1997, p. 66).  1. On the relevance of final points A s shown in Chapter One, there are a number of different features that have been used to distinguish among predicate classes; telicity is a common feature used to distinguishes among predicate classes on the basis of final points. The terms telic and atelic were first coined by Garey (1957) to refer to situations that include or do not include a goal, respectively. Since then, the distinction has surfaced with a number of different terms (see Dahl 1981 and Brinton 1988 for a review). Definitions of telicity do vary among researchers; Smith (1997) suggests that a telic event has "a change of state which constitutes the outcome, or goal, of the event", while an atelic event is simply a process that can stop at any time and "has an arbitrary endpoint" (p. 19). In a Vendler-type four class system, telicity distinguishes accomplishments and achievements (telic) from activities and states (atelic). I use the term "final points" to refer to final B E C O M E events in the representation of a predicate. Moreover, I use the term "intrinsic final points" to refer to the claim that final points are represented in the templates of aspectual classes; the predicates in some cases are morphologically complex, thus the final points are not necessarily part of the root meaning. This is crucial later in this chapter where I show that final points can be pragmatically conditioned; predicates with these types of final points differ from predicates with intrinsic final points in that the former can be cancelled while the latter cannot.  1.1. The presence of intrinsic final points in Skwxwu7mesh predicates Final points parallel the notion of telicity; I suggest that a final point corresponds to a final change of state, represented by a final B E C O M E event in the denotation of a given predicate.  -62-  However, in this chapter I show that the notion of "final point" (and perhaps telicity as well) does not distinguish the aspectual classes in Skwxwu7mesh along the same lines as they do for English, for example. O n the basis of four independent tests, I propose the following classification of final points in the meaning of Skwxwu7mesh predicates:  (1)  Skwxwu7mesh predicates:  Final  points  Final Point  Representation Activity swim, rest,  X e J e ^ t e ^ U e ^ A (BECOME(P))(ej) A laugh  X  (DO(P))(e )] 2  Accomplishment  X  write a book, fix the car  A,e.[(DO(P))(e) A [ V W ' [W' is an inertia world w.r.t. w at the beginning of e -> [3e' [culminates (e') in w ' A e causes e' in w']]]]  Achievement  X.e.(BECOME(P))(e)  V  A.e.3e 3e [e= (e,ue ) A ( B E C O M E ^ ) ) ^ ) A  X  win, arrive, find a rock  Inchoative State (get) angry, (get)  s  1  cloudy  2  2  P(e )] 2  Comparing this to the presence of final points in Rothstein's proposals for English predicates, there is a significant difference, namely that accomplishments in Skwxwu7mesh do not have final points in the actual world, but rather an implicature of culmination. The contrast in the representations of accomplishments in the two languages is shown below:  (2)  Accomplishments: a.  English  vs. Skwxwu7mesh  English  Xe.3e 3e [e= (e ue ) A ((DO(P))(e,) A B E C O M E ( Q ) ) ( e ) ] s  1  b.  2  1  2  1  2  Skwxwu7mesh X.e.[(DO(P))(e) A [ V W ' [W' is an inertia world w.r.t. w at the beginning of e - » [3e' [culminates (e') in w ' A e causes e' in w']]]]  1  T h e incremental theme is not included in this representation (see Rothstein 2004 for details).  -63-  Although there are other differences between the proposed representation of Skwxwu7mesh aspectual classes and those that have been assumed for English, they relate to intrinsic initial points, and thus are discussed in Chapter Three.  1.2. Outline of this chapter  This chapter is organized as follows. I begin with an overview of and motivation for the four diagnostics used in this chapter to test for the presence of intrinsic final points in Skwxwu7mesh (§2).  I then examine the first two diagnostics (culmination cancellation and  event continuation) by providing the relevant data for each predicate class under examination and show that the results of these tests provide evidence that achievement predicates are the only class of predicates in Skwxwu7mesh that have final points in their representations (§3). Turning to the second two diagnostics (the scope of kiln 'almost' and the scope of negation), I again provide the relevant data for each predicate class and show that Skwxwu7mesh predicates all have only one reading with 'almost' and negation, event cancellation; I argue that this is evidence that, crucially, accomplishment predicates in Skwxwu7mesh lack final points (§4). I end the chapter by examining the issue of pragmatically-conditioned final points by looking at out of the blue judgements of Skwxwu7mesh predicates, and in particular, how culmination implicatures of Skwxwu7mesh accomplishments are derived (§5).  2. Diagnosing final points  In this chapter, I use the following diagnostics to argue for the above claims regarding the presence of final points in the representations of Skwxwu7mesh predicates:  (3)  Diagnostics for final points  a. b.  Culmination cancellation: conjunctions and questions  c.  The scope of kilh 'almost'  d.  The scope of negation  Event continuation: conjunctions and questions  -64-  these can be subdivided into two types of diagnostics: those that test whether a given sentence or discourse is felicitous (a-b) and those that test for ambiguity (c-d). In this section 2  1 examine these diagnostics in detail, motivating their use as diagnostics for final points, and showing how they are replicated in Skwxwu7mesh.  2.1. Felicity conditions for culmination In this section, I outline the two tests used to diagnose final points: culmination cancellation and event continuation. These two tests rely on the contrast between culmination entailments and culmination implicatures, discussed in §2.1.1. Furthermore, the sentences used for these tests must be in the perfective viewpoint, a claim which is motivated in §2.1.2. The tests are then discussed in detail in §2.1.3-4.  2.1.1. Entailment vs. implicature The first two tests rely on the notion of entailments. A n entailment is a semantic relation that holds in all possible worlds and thus cannot be cancelled. Entailment is often referred to as the relationship between two propositions where the truth of one proposition requires the truth of the other. The following is a common example; the sentence in (a) entails the sentence in (b):  (4)  a.  The president was assassinated,  b.  The president is dead.  The idea is that the sentence in (a) could only be true i f the sentence in (b) is true; that is, if it is true that the president was assassinated, it must be the case that the president is dead. If this were not the case, the event in (a) would not be considered a successful assassination. This contrasts with an implicature, which can be cancelled, as it is not a requirement. Using a parallel description as above, the relationship between the two sentences is such that the truth of one merely suggests the truth of the other. Take the following sentence as an example:  2  Unless otherwise noted throughout this thesis, the judgements provided are robust in Skwxwu7mesh among  -65-  (5)  Mary had a baby and got married.  This sentence suggests that the two events occurred in the order they are uttered; that is, Mary first had a baby and then she got married). However, the sentence would be true even i f the events occurred in the reverse order. This can be tested with the following sentence that adds a clause which overtly states that the events happened in reverse (bolded below):  (6)  Mary had a baby and got married.. .but not in that order.  The second clause here cancels any entailment that the baby came first and then the wedding; but since the judgement in absence of the second clause is that it did happen in that order suggests that this is an implicature. This contrasts directly with the entailment example, which is infelicitous when continued with a clause that overtly states that the culmination was not reached (bolded below):  (7)  # The president was assassinated.. .but he isn't dead.  I propose that one way to determine whether a predicate has a final point is to test whether or not it has a culmination entailment. If it does, I argue that this is evidence that the final point is part of the meaning of the predicate. I show in this chapter that only one predicate class has a culmination entailment (achievements); thus, I argue achievements are the only class of predicates that have a final point.  2.1.2. Why these diagnostics require the perfective viewpoint Given my assumption that the perfective places the event time inside the reference time, it should be the case that each point of the event should be included inside the reference time. Thus with respect to entailments, perfective sentences are necessary to test whether or not a predicate has a final point; this is to say that i f an event contains a final point, it will be  the speakers with whom I work.  -66-  reached in the perfective viewpoint. In the imperfective viewpoint, the event is on-going; if while testing for a culmination entailment it seemed as though there were none, it would not be possible to tell whether that is due to the imperfective or the meaning of the predicate. If a predicate has no final point, the perfective sentence will not entail that a culmination has been reached. Consider the following example of a stative predicate in English:  (8)  John was sick yesterday.  This sentence is perfective, but there is no requirement for the state to be over within the reference time (that is, there is no requirement that John is well now); this is clearly shown by the fact that the sentence can be continued with .. .and he is still sick today. This is due to the fact that there is no final sub-event of culmination in the meaning of an English state, thus the state may have come to an end within the reference time (it may be the case that when John woke up this morning he was not sick, so his being sick ended at the end of the reference time), but there is no requirement to do so (as shown by the continuation of the sentence above).  3  2.1.3 Test 1: culmination cancellation I use this test in two ways; the first involves eliciting a simple sentence containing a perfective predicate and then conjoining it to a clause that overtly cancels the culmination by suggesting that the agent has not yet completed the event. Davis and Matthewson (2001) use this diagnostic to show similar facts in St'at'imcets (Lillooet; Interior Salish). A n example of the diagnostic frame in Skwxwu7mesh is illustrated below:  (9)  chen  X...  welh  haw  k-an  i  huy-nexw  lS.SG  X...  CONJ  NEG  IRR-lCNJ  PART  finish-TR(LC)  (waX) (IMPERFX)  'I X-ed...but I didn't finish (X-ing).'  3  T h e fact that the state C A N continue beyond the reference time is due to the fact that E n g l i s h states have the  sub-interval property; that is, since every sub-event of being sick is the same as being sick. Since an English  -67-  Smith (1997 p. 65) suggests that another test that can be used to "delimit the semantic meaning of a sentence" is questions. A s she notes, " i f the sentence presents an open 4  situation, questions about its continuation are reasonable; i f the situation is closed, such 5  questions are not reasonable". Smith defines open/closed situations with respect to perfective and imperfective viewpoints; that is, "perfective viewpoints are closed informationally, in the sense that they present situations as complete with both endpoints. Imperfectives are open". In the English examples below, the first statement in the narrative sequence in (a) presents a perfective sentence and the first statement in the narrative sequence in (b) presents an imperfective sentence:  (10)  a.  i. ii.  Martin walked to school, #Did he get there?  b.  i. ii.  Martin was walking to school, D i d he get there? (Smith 1997; 65, adapted from ex. 8)  Smith argues that the question is reasonable for the imperfective sentence (b), but strange with the perfective (a) because the perfective sentence provides the answer to the question; that is, the perfective predicate in (a) entails that Martin did get to school, thus the question is in some sense redundant.  6  I replicate this test in Skwxwu7mesh in the following way. I present the speakers with a perfective sentence and then present a question that asks whether the event finished. In some cases, I have presented the two sentences as a dialogue where I will give the sentence and then another researcher w i l l ask the question. I then ask the speaker if these are appropriate questions to ask given the previous statement, or i f the dialogue makes sense to  state has no final point, there is no need for the state to end at any point within the reference time. If it does, this is an implicature, as it is easily shown to not have the requirement to end. It is not clear what Smith means by "semantic", and whether it is in opposition to "non-semantic" and if so, what that would entail. 1 take Smith's "reasonable" to mean felicitous in the given context and "strange" to mean infelicitous in the given context. For Smith there is an additional factor of telicity here, since accomplishments and activities differ with respect to their entailments in the perfective (see Chapters Four and Six for further discussion about perfective activities and accomplishments).  4  5  6  -68-  them. The type of question used for this diagnostic is given below where the question in asks whether the event is complete:  (11)  na  u  huy-nexw-as  RL  Q  finish-TR(LC)-3ERG  ' D i d she finish (it)?' Although both huy 'finish' and tl'iyij  'stop' are compatible with activities and  accomplishments i n Skwxwu7mesh, speakers have clear judgements regarding their meaning. This is illustrated by the volunteered sentences and translations given below:  (12)  Speaker's comments (volunteered):  "there are two [different]  a.  chen tl'iyi7 l s . S G stop 'I stopped.' (volunteered gloss)  b.  chen ls.SG  words"  huy finish  'I finished.' (volunteered gloss) If the dialogue consisting o f the perfective sentence and following question is judged felicitous, I take this to mean that the perfective sentence does not entail culmination and thus the predicate does not have a final point. O n the other hand, i f the dialogue is infelicitous, I take this to mean that the perfective sentence does entail culmination and thus the predicate does have a final point.  2.1.4. Test 2: event continuation Another test that I use to test the presence of a final point in the meaning of a predicate is event continuation. This involves conjoining a clause containing a perfective predicate with a clause that asserting that the event (may have) continued. The diagnostic frame used for Skwxwii7mesh is given below:  (13)  na  X . . . (iw'ayti) na7-xw waX RL X . . . (maybe) RL-STILL IMPERFX 'He/She X e d . . . a n d (maybe) he's/she's still X - i n g . '  -69-  A s with culmination cancellation, questions can be used for this diagnostic as well. The questions in (a) and (b) below are two variations of asking whether the event is still in progress.  (14)  a.  na  u  RL  Q  men wa just IMPERF  we7u continue  wa... IMPERF  Ts he s t i l l . . . ? ' b.  na7-xw  u  RL-Still  Q  wa... IMPERF  Ts he s t i l l . . . ? ' The second clause in (14) and the questions in (15) are given in the imperfective in order to ensure that the sentences are not judged infelicitous for the wrong readings. Data reveal that with the exception of individual level states (which are not covered in this thesis), Skwxwu7mesh predicates that are marked by the morpheme -xw only appear in imperfective clauses (clauses marked with wa).  2.2. Event cancellation vs. event non-completion Dowty (1979) uses English almost, and the different effects it has, as a diagnostic that distinguishes activities from accomplishments, which are prototypical atelic and telic predicates, respectively. A s telicity is taken as a feature that parallels the final points of events, I use kiln 'almost' in Skwxwu7mesh as a means of testing for the presence of final points in a given predicate. Negation behaves in much the same way: negation can have different scope-taking possibilities, depending on the structure of the predicate. A s with 'almost', I use the results of negation in Skwxwu7mesh as a diagnostic for determining the structure of predicates, and in particular, the presence of final points.  2.1.1. Test 3: the scope of kiln 'almost' 2.1.1.1. Ambiguity induced by almost A s Morzycki (2004) discusses, almost (and other modifiers of the same class) can modify phrases of different categories, including D P , P P and V P . A s I am concerned with predicates  -70-  here, I focus on the way i n which almost modifies V P s ; I use this as another diagnostic to determine whether or not a predicate has an inherent final point. A s noticed initially by M c C a w l e y (1971) and discussed i n Dowty (1979), different readings arise when almost modifies a V P , depending on the class of the predicate. Dowty suggests that almost contrasts activity predicates and accomplishment predicates such that accomplishments have two possible readings when modified by almost whereas activities have only one reading. This is illustrated in the data below:  (15)  a. John almost fixed the car.  (ACCOMPLISHMENT)  (i) = John had the intention of fixing the car, but changed his mind SEVENT  CANCELLATION  (ii) = John began to work on fixing the car, but didn't quite finish SEVENT  b.  NON-COMPLETION  John almost drove the car  (ACTIVITY)  (i) = John had the intention of driving the car, but changed his mind v^EVENT C A N C E L L A T I O N  (ii) * John started to drive, but didn't finish * EVENT COMPLETION  The event cancellation reading indicates that no event of the type denoted by the predicate takes place. The event non-completion  reading indicates that the event denoted by the  predicate has begun, but has not been completed (= has not culminated). A s (a) shows, accomplishments are ambiguous between the two readings. This is a result of the fact that accomplishments are made up of both activity portions and final culminations. Almost can modify the entire event (event cancellation), or the final culmination (event non-culmination) alone. Activities on the other hand, are not ambiguous, and have only the event cancellation reading (bi). There are other available readings for the data above, which I will address later. Pustejovsky (1991) suggests that almost distinguishes between accomplishments and non-accomplishments, that is, not just activities, but other predicates as well. For example, achievements modified by almost have event cancellation readings only, as shown below:  -71-  (16)  a.  John almost reached the summit. (i) = John did not reach the summit (yet) (ii) John started to reach the summit but stopped halfway through  b.  John almost finished. (i) = John did not finish (yet) (ii) * John started to finish but stopped halfway through  A l o n g the same lines, W i l h e l m (2003) notes that "'almost' seems more sensitive to event complexity than to telicity per se" (p. 82). I agree with her comment and w i l l demonstrate in the following sections that this seems to be true for Skwxwu7mesh as well.  2.1.1.2. Almost as evidence for lexical decomposition The fact that almost gives rise to a variety of readings has been taken as evidence for the lexical decomposition of verbs; the different interpretations observed relate to differing scopes for almost (Rapp and von Stechow 1999, McCawley 1971). This is illustrated with the verb kill below that consists of the operators A C T , C A U S E and B E C O M E , as well as the resultant state dead. In the sentence John almost killed Harry, almost can take the widest scope, over the entire event (a), medial scope over C A U S E (b), or the lowest scope over the resultant state dead (c). The following are the representations given by M c C a w l e y (as discussed in Rapp and von Stechow):  (17)  7  John almost killed Harry. a. almost(ACT(John) C A U S E B E C O M E dead(Harry)) John almost did something = John resisted acting on his rage b.  ACT(John) almost(CAUSE B E C O M E (dead(Harry)) John did something, and it almost caused Harry's death = John shot at Harry, but missed  counter/actual  scalar  Rapp and von Stechow do not discuss another possible scope for almost, namely one where almost takes scope above B E C O M E but below C A U S E . This reading would correspond to a reading where John did something which caused Harry to almost become dead. It is not quite clear whether this is a different reading (b) or (c), though it is different from (a). If it is a possible reading but it not available, this will need to be explained.  7  -72-  resultative  A C T ( J o h n ) C A U S E B E C O M E almost(dead(Harry)) John did something, and it caused Harry to be almost dead = John wounded him seriously  c.  The reading in (a) above is such that John almost did something that would have had the effect of Harry's dying. Rapp and von Stechow call this "the outer or counterfactual reading of almost" (p. 157). The readings in (b) and (c) above constitute Rapp and von Stechow's "inner readings" (p. 157). In both these cases, almost modifies only a part of the verb meaning. The scalar reading in (b) suggests that John did something which almost had the effect of Harry's dying. The resultative reading in (c) suggests that John did something which had the effect of Harry's becoming almost dead. A s Rapp and von Stechow suggest, there is a clear contrast between the inner and outer readings; in the case of the outer reading, the event does not take place at all. O n the other hand, the inner readings for the sentence above both describe events in which the killing was initiated but not completed. Based on fast 'almost' i n German, Rapp and von Stechow give the following meaning rule for the core meaning of almost:  (18)  8  fast is of type <s, « s , t > , t » . Let w be any world: F(fast)(w)(p)•= 1 iff (a) and (b) hold. (a) There is a world w ' which is almost not different from w and p(w') = 1 (b)  P(w)-=0  Rapp and von Stechow's representations of M c C a w l e y ' s examples above are shown below:  (19)  John almost killed Harry a.  a]most(w)(X.w3e[AGENT (Jdhn) ew  A BECOME  c w  (A.wA.s.DEAD (Harry))]) sw  counterfactual b.  3e[AGENT (John) A almost(w)(Aw. BECOME (A,wA,s. DEAD (Harry)))J ew  ew  sw  scalar  8  T h e y note that this rule closely follows Sadock (1981) but with an additional modification: Sadock proposes  that (b) is an implicature. A s I do not believe that the exact meaning of almost affects the analysis, I adopt Rapp and von Stechow's proposal that almost yields different scopal readings and that these readings tell us something about the representations of predicates. I suggest that other similar analyses (such as Sadock's) would be compatible with my  argumentation.  -73-  c.  3 e [ A G E N T ( J o h n ) A B E C O M E ( X W A S . almost(w)(Xw. D E A D ( H a r r y ) ) ) | ew  6W  sw  resultative  Dowty suggests that it is not clear whether the different readings for sentences with almost result from structural ambiguities or vagueness; in other words, it may be the case that almost has only one meaning, like that suggested by Sadock (1981): "there is a possible world very similar to the actual world in which ' x V E R B s is true". The result would be that this meaning could be appropriate for all of the readings available. M o r c z y c k i (2004) suggests that a virtue of the scope explanation is that there is a single "almost" with a single denotation and the variation of interpretation results from the position. I adopt M o r c z y k i ' s point of view and use kiln 'almost' as a test for the representation of Skwxwu7mesh predicates, and in particular, for the presence of final points.  2.1.2. Test 4: the scope o f negation Negation provides further evidence of the presence of final points. L i k e almost, negation can take scope over different parts of a predicate's structure. For example, the English sentence below that contains an accomplishment predicate, has two possible interpretations:  (20)  9  John didn't fix his car. (John hasn't fixed his car yet) = John didn't event start fixing his car (John didn't fix his car, he decided to buy a new one instead) = John has started fixing his car, but he didn't finish yet (John didn't fix his car yet he started yesterday and is hoping to be finished today)  In English, there are two ways to reverse the truth value of a clause containing an accomplishment predicate such as fix his car. he could not start or he could not finish. Every predicate has an event cancellation reading; the question is which predicates, i f any, have an event non-completion reading in Skwxwu7mesh. If a predicate is complex and has an intrinsic final point, negation should take scope over that final point, inducing an event non-  9  For some speakers, "yet" might be needed for the second reading; I set that aside here as the analysis of this is  beyond the scope of this discussion.  -74-  completion reading. If it does not induce an event non-completion reading, I take this as evidence that the predicate has no intrinsic final point.  3. Diagnosing final points in Skwxwu7mesh: felicity conditions for culmination In this section, I explore how the predicate classes under examination behave with respect to the first two tests: (i) culmination cancellation and (ii) event continuation. Here I rely on two methods: conjunction and questions. I examine these two tests in parallel throughout the section. Based on the data in the remaining sections, I show the following results of these first two diagnostics for final points. In particular, I illustrate which predicates have an entailment of culmination and which do not. It is these results that motivate the following claims with respect to presence of final points in the representation of Skwxwu7mesh predicates:  (21)  Skwxwu7mesh Final points: culmination  entailments  Culmination  Final Point  no entailment  X  Accomplishment  implicature  X  Achievement  entailment  •/  no entailment  X  Activity  Inchoative state 3.1. Skwxwu7mesh activities  Given that activity predicates are cross-linguistically defined as atelic, thus having no final point in its meaning, it may seem futile to even consider whether or not activities in Skwxwu7mesh have a culmination implicature. However, since the representation of activities presented in this thesis is very different from English, for example, (Skwxwu7mesh activities have no final points), I am making as few assumptions as possible regarding activities. That is, although it seems to be straightforwardly expected, I w i l l show explicitly that Skwxwu7mesh activities do not have final points either.  3.1.1. Skwxwu7mesh activities do not culminate The data below show that the termination of a Skwxwu7mesh activity can be cancelled without inducing any contradiction. This is illustrated by the following data, which are all  judged felicitous by speakers:  (22)  a.  b.  na  k-as  i  RL  lulum ta John welh haw sing DET John GONJ N E G 'John sang, but he didn't finish the song.'  IRR-3CNJ  PART  na RL  t'ich-im swim-iNTR  lha DET  huy finish  10  Mary Mary  welh  haw  k-as  i  huy  CONJ  NEG  IRR-3CNJ  PART  finish  ' M a r y swam but she didn't finish swimming.' Data involving short conversations consisting of a statement and a following question show the same results: the culmination of the activity can be cancelled without inducing any contradiction:  (23)  A:  na  t'ich-im  lha  Mary  RL  swim-iNTR  DET  Mary  ' M a r y swam.'  (24)  B:  na u huy RL Q finish 'Is she finished?'  A:  na lulum lha RL sing DET ' M a r y sang.'  B:  na u huy RL Q finish 'Is she finished?'  Mary Mary  In English, sentences such as these might be judged infelicitous as 'finish' in English is claimed to be incompatible with activities (there is not culmination, so nothing can be "finished"). The data above show that even though in Skwxwu7mesh, huy 'finish' IS  1 0  A l t h o u g h the predicate is an activity and has no overt object in the sentence, the speaker seems to coerce the  predicate into an E n g l i s h accomplishment ('sing a song'), presumably since in E n g l i s h it might sound more  -76-  compatible with English activities, sentences such as these are felicitous. This is evidence that Skwxwu7mesh activities, like English activities, lack final points as they have no entailment of culmination.  3.1.2. S k w x w u 7 m e s h activities can continue Skwxwu7mesh activities also parallel English activities in that perfective activities in Skwxwu7mesh can be continued without inducing infelicity; they differ from Smith's characterization of English activities in that they are compatible with a simple continuation, and do not need the assertion of a new unit of activity (see Chapter S i x for further discussion). This is illustrated in the examples below where the first clause contains a perfective predicate (indicated by the lack of any overt imperfective marking), and the conjoined clause contains an imperfective predicate (indicated by the addition of the imperfective morpheme wa). In most cases, the speaker translates the sentence in the English simple past (a-b), and in others, the sentences are translated in the English present progressive (c-d):  (25)  a.  na RL  11  shupn whistle  lha Carrie D E T Carrie  iw'ayti maybe  na7-xw  wa  (shu-)shupn  RL-still  IMPERF  (REDUP-)whistle  'Carrie whistled, maybe she's still whistling.' b.  na RL  kw'eyilsh dance  lha Mary D E T Mary  i  na7-xw  wa  CONJ  RL-still  IMPERF  kw'eyilsh dance  ' M a r y danced and she's still dancing.'  plausible to finish a song rather than just finish singing. I expect that the translation 'John sang, but he didn't finish singing' is also possible. "The  present progressive is an artifact of the English translation; these are not, I argue, progressive sentences.  English lacks a simple present tense that does not have a habitual reading, and thus the progressive is used. I expect simple past translations to also be available for these sentences (see Chapters F o u r and F i v e for further discussion o n grammatical aspect).  -77-  c.  na RL  imesh kwa walk DET i CONJ  John John  iw'ayti maybe  na-xw RL-still  wa IMPERF  imesh walk  'John is walking and maybe he's still walking.' d.  na RL  paym rest  kwa DET  iw'ayti maybe  John John na7-xw RL-still  wa IMPERF  paym rest  'John is resting and he's still resting.' The behaviour of Skwxwu7mesh activity predicates in sentence/question dialogues provide evidence of the same effect: Skwxwu7mesh activities do not have inherent final points. The examples below illustrate that the sequence of a perfective activity predicate followed by an interrogative that questions whether the activity has reached a culmination point is completely grammatical; speakers do not find sequences such as these unusual in any way:  (26)  (27)  a.  na paym kwa RL rest DET 'John is resting.'  b.  na7-xw u wa RL-still Q IMPERF Ts John still resting?'  a.  na t'ichim RL swim ' M a r y swam.'  b.  John John  lha DET  paym rest  kwa DET  Mary Mary  na7-xw  u  wa  (t'i-)t'ichim  RL-Still  Q  IMPERF  (REDUP)-Swim  Ts she still swimming'?  -78-  John John  The translations given for the sequence of (26a-b) is strange i n English as it induces a tautology. Recall that in Skwxwu7mesh, perfective activities can be translated in the 12  present or past. The fact that the sentence in (a) is translated in the present progressive but does not induce an infelicitous sequence with the following sentence provides evidence that while the progressive translation is given in (a) and (b), the predicate cannot actually be itself a progressive form of the predicate; otherwise the tautology that is observed in English would be expected.  3.1.3. Conclusions: Skwxwu7mesh activities have no intrinsic final points A summary of the generalizations for Skwxwu7mesh activity predicates shown in this section are given the table below:  (28)  Activities: Culmination cancellation and Event  Continuation  TEST 1  TESI 2  CULMINATION CANCELLATION  E V E N T C O N ITNUATION  Conjunctions  Questions  Conjunctions  Questions  Activities Skwxwu7mesh activity predicates parallel English activity predicates (so far) in that they have no culmination point as part of their meaning. Rothstein's template for activities straightforwardly accounts for this fact since the denotation of an activity for her is simply a bare predicate under a D O operator; in other words, there is no final B E C O M E event as part of the meaning of activity predicates:  (29)  English activities  (Rothstein)  Ae.(DO(P))(e)  Skwxwu7mesh activities are, however, different from English activities - not with respect to final points, but with respect to initial points. I argue that Skwxwu7mesh activities, in Hotze Rullman (p.c.) suggests that this need not be the case; some time may have elapsed between the event in (a) and the question in (b), or it may be the case that the sentence in (b) is an expression of surprise (most 1 2  -79-  addition to D O sub-events, have initial points in their representations (the presence of the initial point is motivated in Chapter Three):  (30)  X e . B e ^ l e ^ L J e , ) A ( B E C O M E ( P ) ) ( e , ) A (DO(P))(e )] 2  This template, although different from the representation containing D O alone, w i l l still account for the absence of culmination points in activity predicates. Combining this representation with Kratzer's (1998) semantics for the perfective results in the following representation:  (31)  3 e , 3 e [ e = ( U e ) A ( B E C O M E ( P ) ) ( e , ) A (DO(P))(e ) A x(e) C t] s  2  ei  2  2  There is no culmination entailment in the perfective here since there is no culmination included in the representation of the predicate. The fact that a perfective activity can still be continued without inducing a contradiction results from the fact that there is an activity event whose event time is included inside the reference time; i f the event is continued after the reference time it is another event of the same type. Note, however, that it is not the entire event (initial point plus D O sub-event) that is continued, but only the D O sub-event of the activity. T o continue the event (for example, when -xw ' s t i l l ' is added), an imperfective morpheme is required which, unlike the perfective, places the reference time within the event time. In the imperfective, only the D O sub-event of the activity is then visible, and thus only the D O portion continues in the second clause. The following is a summary of the representation of Skwxwu7mesh activity predicates, their absence of a final point, and the results of the final point tests that show they have no entailment of culmination:  likely with stress on  still). I  d i d not control for these other readings in  Six), thus I leave this issue for further research.  -80-  Skwxwu7mesh or  English (see Chapter  (32)  Activities: Culmination Representation  Activity  Final Point  Culmination  X  N o entailment  Xe3e 3e [e= (ei\-ie ) A (BECOME(P)Xe,) (DO(P))(e )] s  1  2  2  A  2  3.2. Skwxwu7mesh accomplishments I propose that Skwxwu7mesh accomplishments do not have final points. Given the fact that perfective accomplishments in Skwxwu7mesh yield past culminated readings in out of the blue contexts, this might appear to be a problem. However, I w i l l show in this section that the culmination observed in out of the blue contexts is only an implicature, not an entailment.  3.2.1 Skwxwu7mesh accomplishments need not culminate The culmination of a perfective accomplishment can be cancelled without inducing a contradiction. This can be seen in examples where the culmination is explicitly denied, as shown below:  (33)  a.  na  p'ayak-ant-as  RL  heal-TR-3ERG D E T  ta  John  ta  snexwilh-s  John  DET  canoe-3POSS,  welh  haw  k-as  i  CONJ  NEG  IRR-3CNJ  PART  huy-nexw-as finish-TR(LC)-3ERG  'He fixed his canoe but he didn't finish (fixing) i t ' b.  chen ls.SG  p'ats'-an sew-TR  ta DET  hem'ten blanket  welh  haw  k-an  i  huy-nexw  CONJ  NEG  IRR-CNJ  PART  finish-TR(LC)  T sewed a/the blanket but did not finish i t ' c.  na  mikw'-int-as  ta  lhxenptn  lha  Mary  RL  wash-TR-3ERG  DET  floor  DET  Mary  welh  haw  k-as  i  huy-nexw-as  CONJ  NEG  IRR-3CNJ  PART  finish-TR(LC)-3ERG  ' M a r y washed the floor but she didn't finish.'  -57-  d.  chen  ilhen  kwi  skawts  ls.SG  eat  DET  potato  welh  haw  k-an  i  CONJ  NEG  IRR-lCNJ  PART  huy-nexw finish-TR(LC)  T ate a potato but never finished it.' e.  kwa DET  John John  na  kw'el-nt-as  RL  COOk-TR-3ERG D E T  ta  skawts potato  welh  haw  k-as  i  huy-nexw-as  CONJ  NEG  IRR-3CNJ  PART  finish-TR(LC)-3ERG  'John boiled a potato but never finished it.' f.  chen ls.SG  lhen'-t weave-TR  ta hem'ten D E T blanket  welh  haw  k-an  i  CONJ  NEG  IRR-lCNJ  PART  huy-nexw finish-TR(LC)  ' I ' m making the blanket, but I didn't finish i t . '  13  Further evidence that perfective accomplishment predicates do not entail culmination comes from data involving questions. Declarative sentences that include perfective accomplishment predicates can be followed by questions that do not presuppose the completion of the event; in this respect, they behave like activities. This is illustrated by the short dialogues in the data set below; each of the sequences of sentences followed by questions is accepted by speakers and not judged unusual in any way:  (34)  a.  na  cha7-st-as  kwi  RL  do-CAUS-3ERG DET  kw'axwa7 box  lha DET  Mary Mary  ' M a r y made a box.' b.  na  u  RL  Q  huy-nexw-as finish-TR(LC)-3ERG  ' D i d she finish it?'  1 3  A g a i n , the progressive here is an artifact o f the English translation and should not, I argue, be understood as a  source o f confusion; the addition o f adverbials (such as kwi chel'aklh 'yesterday'), or clauses that indicates some past reference time (such as na.7 tkwi 7an'usk 'two o'clock' or kwi ses tl'ik kwa John 'when John arrived') consistently yield simple past English translations.  -82-  (35)  a.  na RL  kw'el-t-as cook-TR-3ERG  kwi DET  smeyts meat  kwa DET  John John  'John cooked the meat.' b.  na  u  huy-nexw-as  RL  Q  finish-TR-3ERG  ' D i d he finish?' (36)  a.  na RL  ilhen-t-as eat-TR-3ERG  ta DET  skawts kwa potato D E T  John John  'John ate a potato.' b.  na u huy-nexw-as RL Q fi ni sh - T R ( L C ) - 3 ERG ' D i d he finish it?'  ta DET  skawts potato  The culmination of accomplishment predicates in out of the blue sentences is clearly cancelable, suggesting that the culmination observed in, for example, out of the blue sentences is an implicature, rather than an entailment (see Section 5). I take this as evidence that these predicates have no final points.  3.2.2. Skwxwu7mesh accomplishments can continue The lack of culmination entailment is also illustrated in examples where the perfective accomplishment event is continued without inducing a contradiction, as the examples in (23) are accepted by speakers and are not deemed unusual in any way:  (37)  a.  na RL  xel'-t-as ta write-TR-3ERG D E T iw'ayti maybe  sxwexwiy'am' story  na7-xw RL-still  wa  lha DET  Mary Mary  xel'-t-as  I M P E R F write-TR-3ERG  ' M a r y wrote a story. Maybe she's still writing it.' b.  chen ls.SG  kw'el-t cook-TR  ta DET  smeyts meat  iw'ayti maybe  wa7-xw iMPERF-still  wa IMPERF  ti DET  natlh morning  kw'el cook  ta DET  'I cooked the meat this morning and it's still cooking.'  -83-  smeyts meat  na RL  lhen'-t-as  ta  weave-TR-3ERG  DET  na7-xw RL-still  iw'ayti maybe  Linda Linda  hem'ten blanket  lha DET  wa IMPERF  lhen'-t-as weave-TR-3ERG  ' L i n d a ' s making a blanket, maybe she's still making it.' 14 Mary Mary  na  mikw'-int-as ta  RL  wash-TR-3 ERG DET  lhxenptn floor  i CONJ  wa mikw'-int-as IMPERF wash-TR-3ERG  na7-xw RL-Still  lha DET  ' M a r y is washing the floor and she's still washing it.' Furthermore, questioning whether a perfective accomplishment in Skwxwu7mesh continues is also judged felicitous, as the example below shows:  (38)  a.  na kw'el-t-as RL COOk-TR-3ERG 'John cooked m e a t ' na  kwi DET  kwa DET  smeyts meat  u  men wa we7u RL Q just I M P E R F continue Ts he still cooking it?'  John John  wa kw'el-t-as IMPERF COOk-TR-3ERG  3.2.3. Skwxwu7mesh accomplishments have no intrinsic final points A summary of the generalizations for Skwxwu7mesh accomplishment predicates shown in this section are given the table below:  (39)  Accomplishments:  Culmination  cancellation  and Event  TESTI  CULMINATION CANCELLATION Conjunctions  Questions  Continuation TEST 2 E V E N T CONTINUATION Conjunctions  Questions  Accomplishments  The table above shows that unlike English accomplishments, Skwxwu7mesh accomplishments have no final point. This accounts for why, unlike in English, canceling the  -84-  culmination of a perfective Skwxwu7mesh accomplishment does not induce contradiction and continuing the event denoted by a perfective Skwxwu7mesh accomplishment is felicitous. Rothstein's template for English accomplishments does not make the correct predictions for Skwxwu7mesh since culmination is built into the basic meaning of English accomplishments; Rothstein's template below shows that the culmination point, here represented by the B E C O M E event, is part of the meaning of the accomplishment:  (40)  English  15  Accomplishments  Xe.3e 3e [e= (e ue ) A DO(P)(e,) A B E C O M E ( Q ) ( e ) ] s  1  2  1  2  2  Removing the B E C O M E event from the denotation of accomplishments correctly predicts that Skwxwu7mesh perfective accomplishments do not necessarily culminate. This is illustrated below (since there is no longer a conjoined event, I have removed e, and e from 2  the representation):  (41)  Skwxwu7mesh Accomplishments  (first pass)  Xe.(DO(P))(e)  However, as I show in §5, Skwxwu7mesh accomplishments are interpreted as culminated in out of the blue contexts, t h i s is shown by the data below:  (42)  na xel'-t-as ta sxwexwiy'am' RL wr ite-TR- 3ERG D E T story ' M a r y wrote a story.' Speaker's comments: "She wrote it...she finished."  lha DET  Mary Mary  A s the speaker's comments in the above data indicate, the sentence is interpreted as having culminated. However, as I have shown in this section, this culmination can be cancelled without inducing any contradiction. I conclude, then, that the culmination is an implicature.  1 4  15  See footnote 1 1 . 1 have left the thematic arguments out of this representation.  -85-  The template given above does not alone predict the culmination implicature that is observed with perfective accomplishments in out of the blue contexts. I propose the following template that accounts for the fact that accomplishments culminate in some contexts, but need not. The culmination is reached only in an inertia world:  (43)  Skwxwu7mesh Accomplishments  (revised)  Xe.[DO(P))(e) A [ V W ' [W' is an inertia world w.r.t. w at the beginning of e —* [Be' [culminates (e') in w ' A e causes e' in w']]]] A s I did above for activities, I now combine the representation proposed for accomplishments with Kratzer's (1998) semantics for the perfective. This results in the representation below:  (44)  3e[(DO(P))(e) A V W ' [W' is an inertia world w.r.t. w at the beginning of e - * [3e' [culminates (e') in w ' A e causes e' in w']]] A x(e) C t]  Since there is no final B E C O M E event, there is no entailment of culmination in perfective sentences containing this predicate. The culmination is reached only in inertia worlds where there is no evidence to the contrary. The following table summarizes the representation of Skwxwu7mesh accomplishments their lack of a final point, and the result of the final point tests that show there is no entailment of culmination, but rather an implicature:  (45)  Accomplishments:  Accomplishment  Final points  Representation  Final Point  Culmination  Xe.[DO(P))(e) A [Vw' [w' is an inertia world w.r.t. w at the beginning of e -» [Be' [culminates (e') in w' A e causes e' in w']]]]  X  Implicature  I will discuss how this implicature is derived in greater detail in §5. For the remainder of §3 I examine the final two predicate classes: achievements and inchoative states.  -86-  3.3. S k w x w u 7 m e s h achievements Thus far I have shown that Skwxwu7mesh is like English with respect to activities in that in neither language do they have final points. Skwxwu7mesh and English differ with respect to accomplishments in that I propose that they lack final points whereas English accomplishments have them. In this section, I show that Skwxwu7mesh achievements are like English in both languages, they have final points. Unlike accomplishments in Skwxwu7mesh that have culmination implications, I show in this section that Skwxwu7mesh achievements have culmination entailments.  3.3.1. S k w x w u 7 m e s h achievements necessarily culminate Contrary to activities and accomplishments, canceling the culmination of an achievement results in a contradiction, and as such, is judged infelicitous:  (46)  #chen mekw'-em ls.SG f i n d - i N T R welh CONJ  haw NEG  ta DET  ayalhkw beach  k-an SBJ-ICNJ  lhk'i7-st-an know-caus-1  k-as IRR-3CNJ  attempted gloss: 'I found the beach but I don't know where it is.' Speaker's comments: (response to "am 1 contradicting myself?") (47)  #na RL  xelk'-em fall-iNTR  ta DET  skakl baby  tina7 from  ta DET  yay'wes bed  welh  na7-xw  wa  na7  ta  CONJ  RL-Still  IMPERF  LOC  DET  encha where "yes"  yay'wes bed  attempted gloss: 'The baby fell/was falling from the bed but he is still on the bed.' 16 (48)  1 6  #na tl'exwenk kwa John welh na tl'exw-et-em RL win DET John CONJ RL get.beat-TR-PASS attempted gloss: 'John won/was winning, but he lost.' Speaker's comments: "[laughs] How could he win and lose?...you don't say those words together"  H . D a v i s (p.c.) suggests that this is not pragmatically plausible (cf., ' T h e baby fell from the airplane...'), and  as a result, the sentence might be infelicitous for the wrong reasons. Smith (1997) uses 'fall from a bed'  -87-  These data show that indeed achievements have culmination entailments as they cannot be cancelled without inducing a contradiction.  3.3.2. Skwxwu7mesh achievements cannot be continued The addition of a clause continuing the event also results in a contradiction, and thus is judged infelicitous as well, as illustrated below:  (49)  a.  #natl'ik R L arrive  kwa John D E T John  iw'ayti maybe  na7 t-kwi LOCOBL-DET  na7-xw RL-still  an'us-k two-o'clock  watl'ik IMPERF  arrive  attempted gloss: 'John arrived at two, maybe he is still arriving.' b.  #na RL  xelk'-em fall-iNTR i and  na7-xw RL-still  ta DET  skakl baby wa IMPERF  tina7 from  ta DET  yay'wes bed  xelk'-em fall-iNTR  attempted gloss: 'The baby fell from the bed, and he's still falling.' Speaker's comments: "Do you mean 'stillfalling'? No...notproper." c.  #na  tl'exwenk kwa John i na7-xw RL win DET John CONJ RL-still attempted gloss: 'John won and he's still winning.'  wa  tl'exwenk I M P E R F win  Regarding (c) above, the speaker adds (in my impression, in an attempt to make some sense of it) that if the intended meaning is that 'John won again' the following sentence would be the way to express that:  (50)  17  na kiy'at tl'exwenk RL again w i n 'He won again.'  examples to illustrate progressive achievements in Navajo. Further data is needed to show whether or not pragmatics are interfering here; I leave this for future research. 1 7  C f . French  encore,  which is translated as 'still' or 'again', depending on the context.  -88-  This data illustrates that perfective achievement predicates entail culmination in the perfective since the addition of any information to the contrary results in a contradiction. This, I argue, provides clear evidence that achievement predicates have final points. In the following case, an achievement predicate C A N be continued, but notice that the subject is plural and crucially, the speaker's comments suggest that the meaning of the sentence is such that some people had arrived and others had not yet arrived. However, it does not seem to be the case that none of them had completely arrived:  (51)  na RL  tl'ik ta-n arrive DET-IPOSS na7-xw RL-still  siiyay' friends  wa IMPERF  ti DET  natlh morning  tl'i-tl'ik REDUP-arrive  ' M y friends arrived this morning...they're still arriving.' Speaker's comments: "there were still more coming" In the following sentence, an achievement predicate seems to be able to be continued as well. However, the speaker's gloss suggests that it is not that the event has not yet culminated (as with accomplishments, for example), but that the entire event happened again (see Chapter Five on imperfective achievements):  (52)  18  chen tl'exw-et-em ls.SG get.beat-TR-PASS i CONJ  na7-xw RL-Still  chen lS.SG  wa IMPERF  tl'exw-et-em get.beat-TR-PASS  'I got beat and I got beat again.' Although the translation provided for the sentence above is not a direct translation, it is nonetheless very telling. In Chapter Five, I w i l l argue that wa (which appears in the above sentence with -xw) is an imperfective morpheme and show that it induces on-going and habitual readings. Taking multiple events to be part of the meaning of the habitual, this may be an avenue of 1 argue that this is an achievement based on the behaviour of this predicate on other diagnostics - e.g., out of the blue sentences and its behaviour with the imperfective. 18  -89-  analysis for what is going on here. That is, the translation of the sentence in (47) suggests another event occurred; this may be due to the use of the imperfective as an indicator of multiple events. A remaining question is whether accomplishments have 'again' readings with -xw. This was unfortunately not tested, though the 'again' translation was never volunteered for accomplishments. I suggest that the contrast between achievements and accomplishments in this matter is most likely due to their contrast in final points in their representations; wa induces both in-progress and habitual readings with accomplishments, but either infelicitous judgements or habitual readings with achievements. Given that habitual readings with accomplishments seem to require context (that is, they are not volunteered readings), the in-progress reading seems to be the more salient of the two. I suggest that while the 'again' reading might be possible with accomplishments, it would require context to force it. With achievements, on the other hand, the habitual reading is the only reading that seems to be available for the imperfective marker if the sentence is judged felicitous. Thus, assuming there is a connection between the habitual reading and the 'again' reading above, the 'again' reading might be expected to be more easily available with achievements than with accomplishments. A s I have not yet elicited these data with accomplishments, I leave this issue for further research. Data with questions provide further evidence of the presence of final points in the denotation of achievement predicates. The examples below show that a dialogue of a sentence containing a perfective achievement followed by a question that asks whether the event has culminated results in an unusual exchange. The entailment of culmination for perfective achievements, and the ungrammaticality of the sequence of sentence and questioning the culmination, is further emphasized by the speaker's comments (# here is meant to indicate that the sequence is infelicitous, not the sentence itself):  (53)  1 9  a.  na k w ' u y kwa RL die DET 'John died'  19  John John  In these questions, I have used different predicates that the ones used in the sentences in the (a) examples; that  is, rather than-asking 'Is John still dying' or 'Is John still finding money'; although I argue that these sentences still test whether or not the final point of achievements are entailed, it could be the case that these sentences do not test event continuation but culmination cancellation.  -90-  b.  #na7-xw  u  wa  7es-kw'uy  kwa  RL-still  Q  IMPERF  STAT-die  DET  John John  'Is John still s i c k ? ' Speaker's comments: Not OK as a question about the previous statement; "Why would you ask a question like that if he is already dead? " 20  (54)  a.  na mekw'-em ti RL find-lNTR DET 'John found some money.'  b.  #na7-xw  u  wa  RL-still  Q  IMPERF  tala kwa money D E T  John John  yelx-t-as ti tala kwa John look.for-TR-3 D E T money D E T John 'Is John still looking for the money?' Speaker's comments: Not OK as a question after the previous statement; "Why is he still looking if he found it? "  The fact that these sequences of statements and questions are unusual indicates that perfective achievement predicates entail culmination in the actual world; moreover, entailment of culmination rather than an implicature of culmination is taken as evidence that a predicate, in this case, achievement predicates, have final points. Some predicates are not clearly members of a particular aspectual class. For example, the predicate leave in English seems to behave as an achievement; data from Skwxwu7mesh, however, suggests that the predicate that corresponds to 'leave' in Skwxwu7mesh (huydT), is not an achievement. The question diagnostic shows that it behaves more like an activity or accomplishment, since the sequence of huyd7 in the perfective, followed by a question that suggests the event might not yet have culminated, is completely grammatical:  (55)  a.  na huya7 lha RL leave D E T ' M a r y left.'  Mary Mary  b.  na7-xw  u  wa  RL-still  Q  IMPERF  i here  'Is Mary still here?'  -91-  lha DET  Mary Mary  Furthermore, the event culmination can be cancelled without inducing any contradiction, as the data below illustrates:  (56)  na  huya7 kw John iw'ayti leave D E T John maybe 'John went away, he might be still here.'  wa7-xw RL-still  RL  wa  i here  IMPERF  21  This contrasts with the behaviour of achievement predicates, which suggests that these tests not only help to determine the representation of predicate classes, but they also act as diagnostics for membership in aspectual classes. The consequence of such a proposal is that the English translation of a particular predicate is not necessarily a clear indication of the aspectual class to which it belongs. The types of tests discussed here are helpful in determining these facts.  22  3.3.3. S k w x w u 7 m e s h achievements have intrinsic final points A summary of the generalizations for Skwxwti7mesh achievements shown in this section is given the table below:  (57)  Achievements: Culmination cancellation and Event  Continuation  TESTI  TEST 2  CULMINATION CANCELLATION  Conjunctions Achievements  EVENT  CONTINUATION  Questions  Conjunctions  Questions  -  X  -  Although the data with questions are not available here, I w i l l assume that the data with conjunctions for both diagnostics is sufficient to generalize. Given that the culmination of achievement predicates is not cancelable, I conclude that the culmination of a perfective achievement is an entailment. The results of the culmination implication test provide evidence that achievement predicates do have inherent final points.  2 0  Kuipers (1967) lists  7es-kw'uy  as meaning either 'sick' or 'dead'. In my elicitation, speakers have used both  these translations. 2 1  It is not clear whether the speaker interprets this as an in-progress event or one that never took place.  2 2  H . Davis (p.c.) notes that directed motion verbs are known to be flexible with respect to their aspectual  behaviour.  -92-  Above I argued that neither activities nor accomplishments have final points in their denotations. T o account for this, I proposed that neither predicate has a final B E C O M E event. This remains uncontroversial for activity predicates (when compared to English), but based on what is reported in the literature, could be considered unusual for accomplishment predicates. Skwxwu7mesh achievement predicates contrast with both accomplishments and 23  activities in the language in that they necessarily culminate, and thus have a culmination entailment. This entailment must be reflected in their representation; thus, I suggest that the representation of Skwxwu7mesh achievements is the same as English achievements. Rothstein suggests the following template for English achievements, where the event consists only of a B E C O M E operator:  (58)  Achievements Xe.(BECOME(P))(e)  This template can account for the Skwxwu7mesh achievement facts since B E C O M E is an event of change; it involves the change from one state, -•()), to another state, §. Thus, since achievement predicates themselves are B E C O M E events, I expect that achievements should entail completion - the event is only understood to have occurred once the change has happened. Thus I argue that Skwxwu7mesh achievements have the same structure as in English. Combining the representation proposed for achievements with Kratzer's (1998) semantics for the perfective, I account for the culmination entailment since the event time of the B E C O M E event must occur within the reference time, as shown below:  (59)  3e[(BECOME(P))(e) A x(e) C t]  The following table summarizes the representation of Skwxwu7mesh achievements, the fact they have a final point, and the result of the final point tests that show they have an entailment of culmination:  Note that we do not actually know how (un)common this is cross-linguistically.  (60)  Achievements: Final points  Representation Achievement  Xe.(BECOME(P))(e)  Final Point  Culmination  •/  Entailment  3.4. Skwxwu7mesh inchoative states The final predicate class to examine is inchoative states. This predicate class most closely resembles Skwxwu7mesh activities in that both predicate types consist of two sub-events, the first of which is an initial B E C O M E sub-event. The predicates differ in their second subevent: activities have D O sub-events and inchoative states have simple P sub-events. Given what I have already said for Skwxwu7mesh activities, Skwxwu7mesh inchoative states are expected to be the same: no final points. I show in this section that inchoative states have no culmination entailment and w i l l thus argue that inchoative states have no final points in their representation. I have not elicited data for the culmination cancellation diagnostic, and thus I focus on event continuation alone.  3.4.1. Skwxwu7mesh inchoative states can be continued A perfective inchoative state can be conjoined with an imperfective clause, without inducing any contradiction; this is illustrated in the example below:  (61)  chen ls.SG  t'ayak' angry  ti natlh D E T morning  i  na7-xw  chen  wa  t'a-t'ayak'  CONJ  RL-still  IB.SG  IMPERF  REDUP-angry  T got mad this morning and I ' m still mad.' •S Context: I was mad the entire time Context: I was mad about the same thing Furthermore, a sentence containing a perfective state can be followed by a question asking whether the state still holds, as illustrated by the sentences below:  -94-  (62)  a.  na t'ayak' RL angry ' M a r y got mad.'  lha  na7-xw u RL-still Q 'Is she still mad?'  wa  Mary Mary'  DET  t'a-t'ayak' REDUP-angry  IMPERF  lha DET  Mary Mary  Given that Skwxwu7mesh inchoative events can be continued without inducing infelicity, I argue that inchoative states have no culmination and thus no final points.  3.4.2. Inchoative states have no final points A summary of the generalizations for Skwxwu7mesh inchoative states shown in this section are given the table below:  (63)  Inchoative States: Culmination cancellation and Event  Continuation TEST 2  TEST 1 CULMINATION CANCELLATION  Inchoative States  Questions ?  Conjunctions ?  EVENT  CONTINUATION  Conjunctions  Questions  For Rothstein, states are the most basic event type in English. Their representations consist of P alone, with no operators:  (64)  English States Ae.P(e)  A t this stage, this representation alone can capture the fact that Skwxwu7mesh inchoative states lack final points. However, although I have not yet motivated why I include a class of "inchoative states" in my classification of Skwxwu7mesh predicates, I w i l l do so in Chapter Three. The representation of inchoative states is as follows:  (65)  Skwxwii7mesh Inchoative States Xe.lt^c^Uej  A ( B E C O M E ( P ) ) ( ) A P(e )] ei  -95-  2  Thus, although the initial B E C O M E sub-event is not part of this discussion, what is relevant here is that the representation above for Skwxwu7mesh inchoative states can also account for the fact that they do not entail culmination and thus have no final points. The following table summarizes the representation of Skwxwu7mesh inchoative events, the fact they lack a final point, and the result of the final point tests that show they have no entailment of culmination:  (66)  Inchoative States: Final points  Representation Inchoative State  Final Point  Culmination  X  N o entailment  A£.3e 3e2[e= (eiUe ) A s  1  2  (BECOME(P))(e,) A P(e )] 2  3.5. Summary and concluding remarks The chart below is a summary of all the data regarding culmination cancellation and event continuation for each predicate class that is discussed in this section:  (67)  Summary: Culmination Cancellation  and Event  TEST 1  TEST 2  CULMINATION CANCELLATION  E V E N T CONTINUATION  Questions  Conjunctions Activities Accomplishments Achievements Inchoative States  Continuation  X  -  -  -  Conjunctions  Questions  •/  •/  X  -  (*=infelicitous; S=fe icitous; - =data not yet tested)  These results have led us to the claim that only achievement predicates have final points in their representations as they are the only predicate class that induce infelicitous judgements in these contexts. A summary of the results of the first two diagnostics for final points and an illustration of how these results inform my representations of Skwxwu7mesh predicates are given in the chart below:  -96-  (68)  Summary: Final points  Representation  Final Point  Culmination  Xe.3e 3e [e= (e ue2) A (BECOME(P))( ) A (DO(P))(e )]  X  N o entailment  Accomplishment  Xe.[DO(P))(e) A [VW' [W' is an inertia world w.r.t. w at the beginning of e -» [3e' [culminates (e') in w' A e causes e'inw']]]]  X  Implicature  Achievement  Xe.(BECOME(P))(e)  Activity  s  1  2  1  ei  2  Entailment  Inchoative State  X  No entailment  ( B E C O M E ^ X ^ ) A P(e )] 2  3.5.1. Testing perfectivity vs. testing telicity These tests have been used by other researchers to argue for a particular meaning of the perfective, not for the presence of final points. For example, Smith (1997) uses culmination cancellation and event continuation as diagnostics for the meaning of the perfective. She argues, for example, that in English, perfective activities are incompatible with assertions of simple continuation and perfective accomplishments are incompatible with assertions of event termination rather than completion; Smith takes these data as evidence that in English, "the perfective imposes an implicit bound", that "the Activity sentence conveys termination" and that the "perfective viewpoint Accomplishment sentences in English semantically convey completion" (p. 68). Adopting Smith's diagnostic, Wilhelm (2003) argues that in Dene Suline, completion is entailed for both perfective activities and accomplishments, leading her to propose that the Dene Suline perfective focuses on a posttime in addition to the entire situation. Koenig and Muansuwan (2000) argue that in Thai, perfective accomplishments are compatible with assertions of culmination cancellation, but they are incompatible with assertions of event continuation. This leads them to call the Thai perfective a "semi-perfective". What all these analyses have in common is the fact that they are arguing that these two tests are a diagnostic for the meaning of the perfective. I argue that these tests are a diagnostic for the presence of final points (see Chapter Four for detailed discussion on the meaning of the perfective in Skwxwu7mesh and cross-linguistically).  -97-  3.5.2. Cross-Salish c o m p a r i s o n The acceptability of canceling the culmination of a control transitive marked predicate was first noted by J. Davis (1978) for Comox (another Coast Salish language), and discussed in greater detail in Watanabe (2003) for the same language. The following data illustrate that canceling the culmination of a control-marked transitive predicate is judged grammatical, but the cancellation of a non-control transitive marked predicate is rejected:  (69)  Comox ?iy and  a.  k'ap-t-ul con cut-CTr-Past lsg.Indc T cut it, but it is not c u t '  b.  * k'ap-9X -an ?iy x a? cut-NTr-lsg.Erg and Neg (T cut it, but it is not cut.') w  w  x a? Neg  k'sp-as cut-3Cnj  w  k'sp-as cut-3Cnj (Watanabe 2003: 205, ex. (18-41a-b)  (70)  a.  x a? Neg  Xspx -as Break-3Cnj  ?iy x a? * Xapx -a-t-as-ul and Neg break-NTr-3Erg ('He broke it, but it did not break.')  Xspx -as Break-3Cnj  Xapx -a-t-as-ul ?iy break-lv-CTr-3Erg-Past and ' H e broke it but it did not break.' w  w  w  w  w  w  (Watanabe 2003: 205, ex. (18-42a-b)  Although Watanabe does not provide an analysis of these data, he does suggest that "encoding completion" may be a more important function of the non-control transitivizer than control. Davis and Matthewson (2001) were the first to show that in St'at'imcets, the culmination of accomplishment predicates can be cancelled without inducing a contradiction, but the culmination of achievements predicates would induce a contradiction:  -98-  St'at'imcets a.  accomplishments  mays-en-lhkanti fix-TR-lSG.su D E T t'u7 but  q'laxan-a, fence-DET  cw7ay t'u7 NEG just  kw-s DET-NOM  tsukw-s-an finish-CAU-lSG.ERG  'I fixed a fence, but I didn't finish.' b.  ts'aqw-an'-lhkan eat-TR-lSG.su t'u7 but  ti DET  n-kiks-a lhkiinsa ku sq'it, lSG.POSS-cake-DET now DETday  qelh-cal-lhkan ku k'wik'wena7 t'u save-ACT-lSG.su DET few until  natcw tomorrow  'I ate my cake today, but I saved a little for tomorrow.'  c.  k'ul'-un'-lhkan  ti  ts'la7-a,  make-TR-1SG.SU  DET  basket-DET  t'u7 just  kw DET  t'u7 but  aoy NEG  tsukw-s finish-3POSS  'I made the basket, but it didn't get finished.' St'at'imcets achievements a.  *mays get.fi xed  ti DET  t'u7 aoy but . N E G  q'laxan-a, fence-DET t'u7 just  kw-s  ka-mays-ts-a  DET-NOM  OOC-fix-3POSS-OOC  'The fence got fixed, but it couldn't get fixed.' Speaker's comments: "Contradiction." b.  *mets ta get.written t'u7 but  pukw-a, D E T book-DET aoy NEG  t'u7 just  kw-s  tsukw-s  DET-NOM  finish-3poss  'The book got written, but it isn't finished.' Speaker's comments: "Contradiction."  -99-  Finally, K i y o t a (2004) observes that in SsncaOsn, accomplishments marked with control transitivizers can be continued without inducing contradiction: (73)  SsncdOan control-marked  accomplishments  l.sg  k a D  le?t fix-CTR  ?i? but  ?awa NEG  san l.sg  la?9  san  AUX  24  w  ta D s  latem table  ssqnax* complete-NCTR  T fixed the table, but I didn't finish i t ' canst bury-CTR ?i? but  san l.sg  ta D  sqexs? dog  ?awa NEG  san l.sg  haynax* finish-NCTR  s  T buried the dog but I didn't finish it.'  (Kiyota 2004)  In the following section, I turn to the next two diagnostics that test for final points: (i) the scope of kilh 'almost' and (ii) the scope of negation.  4. Diagnosing final points: event cancellation vs. event non-completion In this section, I explore how the predicate classes under examination behave with respect to the second two tests: (i) the scope of kilh 'almost' and (ii) the scope of negation. Based on the data in the remaining sections, I show the following results of these two diagnostics for final points. In particular, I illustrate which readings are induced by these two scope tests. It is these results (along with the results of the two diagnostics discussed in §3) that motivate the following claims with respect to presence of final points in the representation of Skwxwu7mesh predicates:  2 4  K i y o t a suggests that predicates marked with the SsncdBgn non-control transitivizer have culmination  entailments as their culmination is not cancelable, though, he admits, his data set is limited.  -700-  (74)  Skwxwu7mesh Final points: Event cancellation vs. Event  non-continuation  Readings induced by scope tests  Final Point  Activity  Event cancellation  X  Accomplishment  Event cancellation  X  Achievement  Event cancellation  </  Inchoative state  Event cancellation  X  The chart above illustrates that the two diagnostics induce event cancellation readings only. Although achievements, in contrast with the other classes, are proposed to have final points (and thus the results for this class might be expected to be different), I show in this section that their representation allows only the one reading.  4.1. Test 3: The scope of Skwxwu7mesh kilh 'almost' There are at least three lexical items in Skwxwii7mesh that correspond to the same meaning as almost in English: kilh, chiyay', cha7i and cha7lha (the latter three of which might be morphologically related (P. Jacobs, p.c.)). A s kilh is the lexical item used most often in to 25  translate English almost (the other words are glossed in different ways), I use it throughout these data. M o r c z y c k i (2002) suggests that English modifiers similar to almost, such as 26  virtually, nearly, damn near, pretty much, not quite, just about, "occur in the same positions, give rise to the same range of readings and impose similar restrictions on the expressions they modify" (p. 55). Thus, while kilh may not exactly correspond to English almost, it is obviously a member of this class of predicates and is expected to show the same effects. With respect to Skwxwu7mesh kilh 'almost', I assume the following:  (75)  2 5  Kuipers glosses the lexical item  chiyay' 2 6  Skwxwu7mesh kilh can modify P, or an operator that takes scope over P; thus the number of readings that kilh w i l l give rise to is dependent on the structure/complexity of the predicate.  kilh  as 'be almost' or 'come close to'. H e glosses  chiyalh  as 'soon' and  as 'a little' or 'to a small degree'.  Preliminary research on Saanich (Masaru K i y o t a p.c. to L i s a Matthewson) suggests that different words for  almost  in that language have different interpretations with respect to these types of data. Future research on  Skwxwu7mesh may reveal some interesting differences between the lexical items, perhaps along the lines of the differences between the English lexical items listed by M o r c z y c k i (2002).  -101-  I will be assuming that each and every predicate (regardless of its structure) must minimally contain P, which represents the idiosyncratic content of a particular lexical item. The ability of kilh 'almost' to take scope over this part of a predicate will not differ among predicates; in other words, each and every predicate will minimally have this reading. The intuition is that the reading that results from this scope is the reading where you do something and it is not quite the perfect version of that event. This is the reading where an event takes place, but it is not interpreted by the speaker as having taken place in an idealized way. Examples for each of the predicate classes are given below (other contexts are possible here):  (76)  a.  John almost sang. Context: John did sing, but (it was so bad) I would not describe what John did to be singing 27  b.  John almost built a house. Context: John did build a house, but the structure of it is so poor that I would never live in it  c.  John almost reached the summit. Context: John did reach a summit, but I know there is another higher peak that he never reached  d.  John almost got scared. Context: John did get scared, but all he gave was a little peep, so he probably wasn't really all that scared - when I get scared I scream out loud  While it may be difficult to get these types of readings for some predicates, the null hypothesis is that they are available for each class of predicate since each predicate minimally has P as part of its denotation. 1 will set aside this reading for each of the 28  Hotze Rullman (p.c.) suggests that these examples may be of the type that X V-ed in one sense by not in another, which might be a metalinguistic use of almost (e.g., He did something that I would almost call 'singing'). As I have not pursued this reading in elicitation, and am setting the reading in (a) aside for the Skwxwu7mesh data, I leave this issue for further research. 2 7  Sevi (1998) proposes a denotation for almost like the one given by Rapp and von Stechow but in which almost can operate on not just worlds but also intervals or standards of precision, depending on context. I think the notion of "standards of precision" relates to the scope over P readings and while I consider this an important part of the meaning of almost, it will not change the analysis here since it will not distinguish between the predicates. 2 8  -102-  predicate classes and focus on the readings that arise from modification of the other structure of the predicates.  29  In this section, I w i l l argue that the number of readings that are induced by kilh in Skwxwu7mesh provide further evidence for the representations I propose:  (77)  Claims: Skwxwu7mesh kilh 'almost'  Predicate Representation  Readings induced by kilh  Activity  A e . B e . B e ^ e ^ U e J A (BECOME(P))(e,) A (DO(P))(e )]  Event Cancellation  Accomplishment  X.e. [DO(P))(e) A [VW' [w' is an inertia world w.r.t. w at the beginning of e -> [3e' [culminates (e') in w' A e causes e' in w'J]]]  Event Cancellation  Achievement  Ae.(BECOME(P))(e)  Event Cancellation  Inchoative States  X.e.3e,3e [e= (eiUe ) A (BECOME(P))(e,) A P(e )]  2  s  2  2  2  Event Cancellation  The readings should be predicted by the structure of the predicates. I suggest that all predicates minimally have an event cancellation reading where almost takes scope over the entire predicate since there is no sensitivity to a particular structure or operator. In these cases, kilh simply cancels the event. However, given that this diagnostic relates to scope, the complexity of the predicate representations (e.g., how many operators or sub-events there are) should dictate how many readings are available. I will show that the other readings that are expected are not distinct readings, but collapse with the event cancellation reading or the scope over P reading. This is particularly relevant with activities and inchoative states, as I show below. In the following sections I examine the results of this diagnostic for each of the predicate classes.  2 9  If it were the case that this reading was not available, I would not attribute this to a problem with the template  since by definition, each predicate must have the idiosyncratic meaning of the lexical items as part of its denotation. T h u s , I would attribute this problem instead to some other pragmatic reason, which I do not want to pursue here.  -103-  4.1.1. Skwxwu7mesh activities: kilh induces event cancellation only When modified by kilh, activity predicates in Skwxwu7mesh have an event cancellation reading only. This is illustrated by the data below where the context that elicits the event non-completion reading is not appropriate:  (78)  a.  kilh na lulum lha Linda almost R L sing D E T Linda ' L i n d a almost/nearly sang.' •SContext: Linda was going to sing but didn't because she was too embarrassed xContext: Linda started singing but lost her voice part way and couldn't finish  b.  kilh na xik'-int-as ta hem'ten ta almost R L scratch-TR-3ERG D E T blanket DET 'He almost scratched the blanket.' •SContext: cat changes his mind and decides not to scratch it  push cat  kilh chen imesh almost l s . S G walk 'I nearly walked.' SContext:  I planned to walk but changed my m i n d  30  This is summarized by the chart below: (79)  Activities: Readings induced by kilh 'almost' TEST 3 kilh ' A L M O S T '  T H E SCOPE OF  Event Cancellation Activities  Event Non-completion X  Skwxwu7mesh activities behave like English activities in these types of clauses - there is no culmination point, thus, there is no possible modification of that culmination point. This suggests that the sentence in (78a) above could be represented as (80b) below where kilh takes scope over the entire event:  3 0  T h e speaker volunteers another sentence as a preferred for this context:  the auxiliary nam'  'go'.  -104-  kilh chen nam' 7imesh, which  adds  (80)  a.  kilh na lulum lha Linda  (=78a)  b.  almost 3 e , 3 e [ e = ( e U e ) ( B E C O M E ( S I N G ) ) ( e ) A (DO(SING))(e )J s  1  2  2  1  2  This representation suggests that there was almost an event of singing, and thus correctly predicts that the Skwxwu7mesh sentence in (a) is acceptable in the event cancellation context (context 1), but not in the event non-culmination context (context 2). There is another scope possibility in this representation, namely the one where almost takes medial scope over D O , as given below:  (81)  3e 3e [e= (e Ue ) ( B E C O M E ( S I N G ) ) ( e ) A almost ( D O C S I N G ) ) ^ ] . s  1  2  1  2  1  This would suggest either a reading where (i) there is an event in which Linda began to almost sing, or (ii) where L i n d a began to sing badly. The former does not seem to be distinguishable from the outer reading since in both cases, nothing actually takes place. The latter is indistinguishable from the scope over P reading, which I have argued is available for every predicate. Crucially, what I do not find is a reading where the event begins but does not finish (the event continuation reading), only the event cancellation reading. The chart below summarizes the available reading for activities modified by kilh, which has contributed to the motivation for the representation of Skwxwu7mesh activity predicates:  (82)  Activities: final points  Predicate Representation Activity  Xe.3e 3e [e=(e \Je2) A . (BECOME(P)Xe,) A (DO(P))(e )] l  2  1  2  Final Point  Readings  X  Event Cancellation  4.1.2. Skwxwu7mesh accomplishments: kilh induces event cancellation only Like activities, Skwxwu7mesh accomplishments modified by kilh yield the event cancellation reading only. This is illustrated by the data below. Speakers often translate the sentence with "started", suggesting that the event cancellation reading is the O N L Y reading available for these predicates:  -105-  (83)  a.  kilh na xel'-t-as almost RL write-TR-3ERG ' M a r y almost wrote a story.'  ta DET  sxuxuyam story  lha DET  Mary Mary  Speaker's comments: she didn't even start b.  kilh chen mikw'-en ta almost l s . S G clean-TR DET 'I nearly started to wash the car.'  tetxwem car  c.  kilh chen p'ats'-an ta almost ls.SG sew-TR DET 'I nearly sewed the blanket.'  hem'ten blanket  kilh na yetl'k'-ant-as ta almost RL paint-TR-3ERG DET 'John almost painted the house.' .  lam' ta house DET  ^Context A :  ^Context B :  John John  xQontext C :  Speaker's comments (re: B): "isn't he already painting it?" I take the speaker's comments for context A above to suggest that for the Skwxwii7mesh sentence to be true of a certain context, no painting may have occurred. Thus, the fact that she describes the picture as though John is  waiting to paint the house, what is crucial is that  he hasn't done any painting. The data in (a-d) above suggests that Skwxwu7mesh accomplishments do not have final points as part of their basic meaning; i f they did, I would expect  kilh to be able to modify them. Further evidence that the event non-culmination reading is N O T available for  Skwxwu7mesh accomplishments is illustrated in the set of sentences given below. The sentence in (a) parallels the sentences above where the event cancellation reading is the available reading. The sentence in (b), on the other hand forces the  context of an event non-  culmination reading using the overt achievement huy-nexw 'finish' as the predicate over  -106-  which kilh takes scope. The only available reading for (b) is that the agent started but did not finish painting the house. When asked i f the two sentences are appropriate in the same context, the speaker clearly identifies that the sentence in (a), with the accomplishment, only has the event cancellation reading:  (84)  a.  kilh chen yetl'k'-an ta almost l s . S G paint-TR DET T almost painted the house.' Context: didn't start painting  b.  kilh chen almost l s . S G  lam' house  huy-nexw finish-TR(LC)  kwi-n-s DET-IPOSS-NOM  yetl'k'-an paint-TR  ta DET  lam' house  T almost finished painting the house.' Speaker's comments: [in response to whether these two sentences mean the same thing?] "no, they are different... [in the first oneJ you hadn't started painting...you said 'almost'" The results of this diagnostics for the class of Skwxwu7mesh accomplishments is summarized below:  (85)  Accomplishments:  Readings induced by kilh 'almost' TEST 3 T H E SCOPE OF kilh ' A L M O S T '  Event Cancellation  Event Non-completion  Accomplishments  X  I propose that, contrary to English, there is no final B E C O M E operator in the representation of Skwxwu7mesh accomplishments. Instead, it is proposed that Skwxwu7mesh accomplishments have culmination implicatures:  (86)  Skwxwu7mesh  Accomplishments  Xe.[DO(P))(e) A [ V W ' [W' is an inertia world w.r.t. w at the beginning of e - » [Be' [culminates (e') in w ' A e causes e' in w']]]]]  -107-  The representation for the sentence in (78d) above, is shown below:  (87)  a.  kilh na yetl'k'-ant-as ta lam' ta John  (=78d)  b.  almost 3e.[DO(PAINT))(e) A [ V W ' [W' is an inertia world w.r.t. w at the beginning of e -*• [Be' [the house gets painted (e') in w ' A e causes e' in w']]]]j  Although speaker's judgements differ for the sentence below, it should be pointed out that sentences containing plural objects might pose a problem for this analysis. For example, the data below show that a sentence containing a plural object might have event noncompletion readings with kilh:  (88)  a.  kilh na pen-t-as ta chanat shaw' ta skwemay' almost R L bury-TR-3ERG DET three bone D E T dog 'The dog almost planted three bones.' 31  S Context A :  S Context B :  ^ Context C:  * Context D :  However, in another elicitation session, the same speaker does not accept the sentence in context C. This seems to be a counter-example, which I cannot account for. I leave this issue for further research.  32  pent is used for both 'bury' and 'plant'.  3 1  T h e words  3 2  T h e data below might also suggest that in fact, the event non-completion reading is unavailable for  accomplishments. T h e speaker comments that the word  i7xw 'all' is not necessary in the sentence as it is in  some sense redundant. I expect it to be redundant here if the sentence without 'all' indicated that none of the bones were buried, which is precisely the event cancellation reading that I expect with accomplishments. If the event non-completion reading were available also, I would not expect the speaker's comments below: (i)  kilh  na  pen-t-as *  i7xw  ta  shew'-shaw'  ta  skwemay'  almost  RL  bury-TR-3ERG  all  DET  REDUP-bone  DET  dog  'The d o g almost planted all those bones'  Speaker's comments: "you don't have to use i7xw" ..."didn't even start [burying]" W h i l e the fact that the speaker comments that i7xw is not necessary may be a comment on the use of any D P here, the fact that the speaker states that no burying has taken place can be taken as evidence enough that this is only an event cancellation reading.  -705-  The chart below summarizes the available reading for accomplishments modified by kilh, which has contributed to the motivation for the representation of Skwxwu7mesh accomplishment predicates:  (89)  Accomplishments:  final points  Accomplishment  Predicate Representation  Final Point  Readings  te.[DO(P))(e) A [Vw' [w' is an inertia world w.r.t. w at the beginning of e ^ [3e' [culminates (e') in w' A e causes e' in w']]]]  X  Event Cancellation  4.1.3. Skwxwu7mesh achievements: kilh induces event cancellation only A s I have shown for Skwxwu7mesh activities and accomplishments, achievements r by kilh yield an event cancellation reading. This is illustrated by the data below:  (90)  a.  kilh  chen  almost l s . S G  33  wi7xw-em fall-iNTR  T nearly fell.' b.  kilh  chen  almost l s . S G  huya7 leave  T nearly went away.' •SContext: I changed my mind c.  kilh  na  almost R L  xelk'-em  ta  skakl  na7  ta  yay'wes  fall-iNTR  DET  baby  L O C D E T bed  'The baby almost fell off the bed.'  3 3  H . Davis (p.c.) suggests that this argument is misleading since these are all what he labels non-instantaneous  achievements. However, some o f these predicates are felicitous in the progressive and thus can be considered non-instantaneous: (i)  na  xe-xelk'-em  ta  skakl  na7  ta  RL  REDUP-fall-INTR  DET  baby  LOC  DET  'He's falling off the bed'  Speaker's comments: "he could be rolling off the bed"  -109-  yay'wes bed  d.  kilh  chen  huy-nexw  almost l s . S G  finish-TR(LC)  kwi-n-s  yetl'k'-an  DET-IPOSS-NOM  paint-TR  ta  lam' DET house  T almost finished painting the house.' Unfortunately, it is not easy to test whether the event non-completion reading is available with achievements as a "partway through" context is quite difficult to set up. For example, in the following sequence where the first sentence suggests that the event was not completed and then the second half of the sentence asserts that the event was completed, it is not possible to ascertain whether the reading in the first clause is the event non-completion reading as it does not seem to be distinct from the event cancellation reading:  (91)  kilh na almost R L ...s-es  kw'ach-nexw-as look.at-TR(LC)-3ERG men  N O M - 3 P O S S just  ta DET  mixalh black.bear  kw'ach-nexw-as  ta  look.at-TR(LC)-3ERG D E T  k w a John... D E T John mixalh black.bear  'He almost saw the bear, then he saw the bear.' The following data was another attempt at setting up the event non-completion context. The second context was meant to suggest that a partial seeing event took place, but it is not clear whether this reading is available, as the speaker volunteers a different sentence:  (92)  a.  kilh chen kw'ach-nexw ta mixalh almost I S . S G s e e - T R ( L C ) D E T bear T nearly saw a bear.' yContext: I heard it but I didn't see it ? Context: I saw the bear's backside only  b.  men chiyay' kwi-n-s just little.bit DET-IPOSS-NOM T saw just a bit of the bear.'  kw'ach-nexw k w i see-TR(LC)  DET  mixalh bear  I w i l l assume, based on these data and my understanding of the near-instantaneous meaning of achievements, that the event cancellation reading is the only reading available for Skwxwu7mesh achievements. A s with English achievements, the event cannot be said to  -110-  have taken place until the entire event takes place as the event is both an initial point and an final point. Although I am using this test as motivation for the representation of predicates, the facts above are consistent with those observed in the previous section (on culmination). There are no other predicted readings as there is only one operator in the representation of an achievement. The results are summarized below:  (93)  Achievements: Readings induced by kilh 'almost' TEST 3 T H E SCOPE OF  Event Cancellation  kilh  'ALMOST'  Event Non-completion  Achievements  X  In § 3 , 1 proposed that the representation of Skwxwu7mesh achievements were just like English achievements - they consist of a B E C O M E event only:  (94)  Skwxwii7mesh  achievements  Ae.(BECOME(P))(e)  The data in this section has reinforced this proposal in that is has shown that the event noncompletion reading is unavailable for achievements. The representation below illustrates the scope possibility for kilh with achievements where almost takes scope over the entire event:  (95)  a.  kilh chen wi7xwem  b.  almost 3e[ B E C O M E ( P ) ( e ) |  (=85a)  Thus, while the proposal is that achievements have final points, these final points do not constitute separate sub-events than any other part of the representation. That is, there is only one relevant reading induced by kilh for achievements since its representation consists of only one B E C O M E event. The chart below summarizes the available reading for achievements modified by kilh, which has contributed to the motivation for the representation of Skwxwu7mesh accomplishment predicates:  -111-  (96)  Achievements; final points  Predicate Representation Achievement  Final Point  Readings  </  Event Cancellation  A.e.(BECOME(P))(e)  4.1.4. Skwxwu7mesh inchoative states: kilh induces event cancellation only The final predicate class under consideration is inchoative states. A s with achievements, trying to elicit a reading other than the event cancellation reading is difficult and as such, the data I have here is limited. However, what I think is clear is that the event cancellation reading is available. The following data illustrates this reading:  (97)  kilh chen t'ayak' almost l s . S G angry 'I almost got angry.' SContext: John did something annoying and then he immediately apologized, so I didn't get angry  In the following example, I attempted to elicit a reading other than the event cancellation reading. Although the speaker says that the sentence is O K (a), she does not repeat it and instead volunteers another sentence that states "a little bit" overtly (b):  (98)  a.  kilh chen lhchiws almost l s . S G tired ?Context: I was almost tired, but I had a coffee and it woke me up  b.  men chiyay' just a.little.bit ' I ' m just a bit tired.'  kwi-n-s DET-1POSS-NOM  lhchiws tired  The set of examples below illustrate the same thing. However, here the elicited sentence in (a) is judged neither felicitous nor infelicitous. Instead, the speaker volunteers the sentence in (b) that includes the word chiyay' (I am not sure as to why the speaker translates the sentence in (b) with "slowly" as it seems somewhat different that the volunteered comments; furthermore, here the predicate contains the inchoative marker):  -112-  (99)  a.  b.  ?kilh na almost R L  ch'i7xw-i7 dry-iNCH  chiyay' kwi a.little.bit DET ' S l o w l y my dress got Speaker's comments:  ta-n  yekway' dress  DET-IPOSS  s-es NOM-3POSS  ch'i7xw-i7 dry-iNCH  ta-n DET-IPOSS  yekway' dress  dry.' "chiyay' means a little dry"  When asked a follow-up question as to whether kilh (as opposed to chiyay') means that the dress was not dry at all, the speaker responds affirmatively. Assuming that these readings are general across other inchoative states, the generalization with respect to inchoative states and the readings induced by kilh is given in the table below:  (100)  Inchoative states: Readings induced by k i l h 'almost' TEST 3 T H E SCOPE OF  Event Cancellation Inchoative States  kilh ' A L M O S T '  Event Non-completion  V  X  I have proposed that the representation for inchoatives is parallel to the structure for activities in that they contain two sub-events, the first being an initial B E C O M E event. However, inchoative states differ from activities in that the second sub-event of an activity is a D O event, while the second sub-event of an inchoative state is a plain state. A s shown above, the reading induced by kilh for inchoative states is an event cancellation reading where kilh takes scope over the entire event '(b):.  (101)  a.  X e . B e ^ e ^ T A U e j ) A ( B E C O M E ( P ) ) ( e ) A P(e )]  b.  almost 3e 3e [e= (e Ue ) A ( B E C O M E ( A N G R Y ) ) ( e I ) A A N G R Y ( e ) ]  1  2  s  1  2  1  2  2  This representation suggests that there was almost an event of being angry, and thus correctly predicts that the Skwxwu7mesh sentence in (97) above is acceptable in the event cancellation context, but the sentence in (98a) is not acceptable in a different context. A s with activities,  -113-  there is another scope possibility in this representation, namely the one where almost takes scope over the second sub-event, as shown below:  (102)  He^te^fouej) ( B E C O M E ( A N G R Y ) ) ( e ) A almost ( A N G R Y ) ( e ) ] . 1  2  This reading is the scope over P reading, which I have argued is available for every predicate (since every predicate minimally contains P, its idiosyncratic lexical content). Crucially, what does not surface is a reading where the event begins but does not finish (the event nonculmination reading), only the event cancellation reading. A summary for inchoative states is given in the chart below. Inchoative states have only event cancellation readings when modified by kilh. This is further evidence for the predicate representation proposed in §3 and further evidence that the representation for inchoative states does not include a final point.  (103)  Inchoative states: final points Final Point Inchoative States  X  Predicate Representation Ae.3e 3e2[e= (e Ue ) A (BECOME(P))(e,) A P(e )] 1  1  Readings Event Cancellation  s  2  2  4.1.5. Summary of test 3 A final summary of the results test 3 which showed the readings induced by kilh with each of the predicates classes is given below:  (104)  Summary Final Point Activity  X  Predicate Representation Ae^e^ezte^^uez) A  (BECOME(P)Xe,) A (DO(P))(e )] 2  Accomplishment  X  Achievement Inchoative States  X  Readings Event Cancellation  Ae.[DO(P))(e) A [Vw' [w' is an inertia world w.r.t. w at the beginning of e - » [Be' [culminates (e') in w' A e causes e' in w']]]]  Event Cancellation  Ae.(BECOME(P))(e)  Event Cancellation  Xe3e 3e [e=\e \Je ) A (BECOME(P)Xe,) A P(e )]  Event Cancellation  l  2  1  2  2  -114-  A potential alternative explanation must be addressed. Rapp and von Stechow (1999) argue that in German, there is only one inner reading for fast 'almost'; they argue that the inner reading for fast is always scalar and the resultative position is never occupied by fast. This contrasts with what has been observed for wieder 'again' in German, where the resultative position can be occupied. It is worth noting that there is variation among speakers/dialects of German and that proposal put forth in Rapp and von Stechow relies on the judgements of the authors only. Given this claim, I need to say something about Skwxwu7mesh, namely, why I don't argue here that kilh is like German fast in that it cannot modify the resultant state, instead of what I do propose for Skwxwii7mesh accomplishment predicates, namely, that there is no final B E C O M E event to modify. I argue that this analysis would require us to stipulate something for Skwxwu7mesh that there is no other evidence for. Furthermore, if I were to 34  assume that the structure of Skwxwu7mesh accomplishments parallels the structure of English-type accomplishments where there is both a D O sub-event and a B E C O M E subevent, but that kilh cannot take scope over the B E C O M E sub-event, I am still left to explain the facts observed in the previous section, namely, the fact that the culmination of an accomplishment is only an implicature since it can be cancelled. While this may still be possible, I then lose a very interesting (and important) relation between the two sets of facts (that is, kilh and the culmination implicature).  4.2. Test 4: the scope of Skwxwu7mesh negation Gillon (2002) shows that there are two patterns of negation in Skwxwu7mesh; she notes that "pattern 1 negation reverses the truth value of the clause introduced by k and can occur with any predicate type" (p. 8). I argue here that negation, like the almost modifier, can modify different sub-events of a given predicate in addition to the entire predicate itself. With this test, I propose that it is possible to diagnose whether or not a predicate has a final point. A s with almost, if negation takes scope over the entire predicate, we get an event cancellation  34  1 have not yet tested the facts with  kiy'dt  'again' in Skwxwu7rnesh, but this might help us make a decision  here.  -115-  reading. If negation takes scope over a final point, an event non-culmination reading is expected. In this section, I will argue that the number of readings that are induced by negation in Skwxwu7mesh provide further evidence for the following representations I propose for predicate classes. In particular, I argue, once again, that only achievements have final points in Skwxwu7mesh:  (105)  Claims: Skwxwu7mesh  negation  Predicate Representation Activity  X e . B e ^ e ^ e ^ e ^ e j ) A (BECOME(P))(e,) A (DO(P))(e )] 2  Accomplishment  Ae.[DO(P))(e) A [ V w ' [w' is an inertia world w.r.t. w at the beginning of e - » [3e' [culminates (e') in w ' A e causes e' in w']]]]  Achievement  Xe.(BECOME(P))(e)  Inchoative States  Xe.3e,3e [e= (e Ue ) A ( B E C O M E ^ ) ) ^ ) A  Readings induced by negation Event cancellation Event cancellation Event cancellation Event  s  2  1  2  P(e )]  cancellation  2  I initially used this diagnostic to test for further evidence as to whether or not Skwxwu7mesh accomplishments have final points. A s such, I have a limited set of data for the other predicate classes. However, I w i l l use the translations given for the other predicate types as a key to the readings available with negation, leaving further exploration to future research.  4.2.1. Skwxwti7mesh activites: negation induces event cancellation only I take the following data from G i l l o n (2002) to illustrate that Skwxwu7mesh activities have event cancellation readings under negation:  35  G i l l o n uses SBJ for subjunctive marking; for consistency, i have glossed these suffixes CNJ for conjunctive marking.  3 5  -116-  (106)  a.  chen  sp'utl'em  ls.SG  smoke  T smoked.' haw  k-an  NEG  IRR-lCNJ  sp'utl'em smoke  T didn't smoke.' (Gillon 2002: 8, ex. 20) (107)  a.  chen  ts'its'ap'  ls.SG  work  T worked.' haw  k-an  NEG  IRR-lCNJ  ts'its'ap' work  T didn't work.' (Gillon 2002: 1, ex. 2) 4.2.2. Skwxwu7mesh accomplishments: negation induces event cancellation only Skwxwu7mesh sentences containing accomplishment predicates yield event cancellation readings with negation, but not event non-culmination. This is shown by the data below where the Skwxwu7mesh sentences are judged felicitous in event cancellation contexts but not in event rion-culmination contexts:  (108)  a.  haw  k-an  xel'-t  ta  NEG  IRR-lCNJ  write-TR  DET  sxwexwiy'am' story  ' I ' m not writing a story.' * Context: you started it but you are not doing it now Speaker's comments: "I'm not going to write a story" haw  k-as  1  mikw'-ent-as  NEG  IRR-3CNJ  PART  wash-TR-3ERG  ta DET  lhxenptn floor  kwa DET  John John  'John never washed the floor.' •SContext: he didn't even start haw  k-an  i  mikw'-en  ta  NEG  IRR-lCNJ  PART  wash-TR  DET  T didn't wash the floor.' •SContext: I decided to go to the beach instead  -117-  lhxenptn floor  d.  haw  k-an  p'ats'-an  ta  NEG  IRR-lCNJ  sew-TR  DET  hem'ten blanket  ' I ' m not sewing the blanket.' •S Context: I planned to but I decided not to...I didn't even start it e.  haw  k-as  i  p'ayak-ant-as  NEG  IRR-3CNJ  PART  fix-TR-3ERG  ta DET  tetxwem car  k w a John D E T John  'John never fixed the car.' SContext: he didn't even start Speaker's comments: "He didn't do it" * Context: he fixed it partway but never finished f.  haw  k-an  yetl'k'-an  ta  NEG  IRR-ICNJ  paint-TR  DET  lam' house  ' I ' m not going to paint this house.' * Context: I started but I didn't finish T o get the part-way (event non-culmination) reading where the event has begun but is not yet complete, speakers volunteer a different sentence where the predicate "finish" is overtly expressed. This is shown by the volunteered sentences below:  (109)  a.  haw  k-as  i  huy-nexw-as  NEG  SBJ-3CNJ  PART  finish-TR(LC)-3ERG  k w i s-es  wa  p'ayak-ant-as ta  DET NOM-3POSS  IMPERF  fix-TR-3ERG  T didn't finish fixing the car.' b.  DET  tetxwem car  36  haw  ek'  k-an  huy-nexw  ta  NEG  FUT  SBJ-lCNJ  finish-TR(LC)  DET  lam' house  T won't finish it.' SContext: I started to paint the house but I quit half-way through These data illustrate that the event non-completion reading is unavailable for Skwxwu7mesh accomplishments, providing further evidence that accomplishment predicates do not have final B E C O M E events over which negation can take scope.  3 6  N o translation was offered by the speaker; I assume this to be the appropriate gloss.  -118-  4.2.3. Skwxwu7mesh achievements: negation induces event cancellation only A s with activities, I take the data below to show that negation induces the event cancellation reading with achievements:  (110)  a.  haw  k-as  i  NEG  IRR-3ERG  PART  tl'ik arrive  'S/he hasn't arrived.' b.  haw  k-an  NEG  IRR-ICNJ  tl'xwenk win  'I did not w i n . ' c.  haw  k-an  i  NEG  IRR-ICNJ  PART  mekw'-em find-iNTR  ti DET  tala money  'I didn't find the money.' d.  haw  k-an  i  NEG  IRR-ICNJ  PART  kw'ach-nexw see-TR(LC)  'I didn't see h i m . ' e.  haw  k-as  i  huy-nexw-as  NEG  IRR-3CNJ  PART  finish-TR(LC)-3ERG  ' H e didn't finish his project.'  The crucial data that are missing here are examples where the event non-culmination context is tested. In other words, I still need to test whether or not these sentences are felicitous in contexts where the events begin but never (or not yet) finish. In (a), the relevant context would be she began to arrive, but did not yet get there. In (b), this would be / started to win, but ended up not winning. I showed in.§3 that Skwxwu7mesh achievements cannot be continued or incomplete in any way; thus, although I have not yet elicited the relevant negative data for negation, I expect that it is not possible to be able to utter a sentence under negation in Skwxwu7mesh where the achievement had begun but not finished. I argue that negation can provide further evidence as to its representation. Since it seems that negation induces only event cancellation readings with achievements, rather than an even non-  -119-  completion reading, I propose that there is only one possible position for negation in the representation for achievements.  4.2.4. Skwxwu7mesh inchoative states: negation induces event cancellation only Finally, inchoative states also occur under negation and have the same results as all three previous predicate classes, namely, event cancellation readings:  a.  (111)  haw NEG  k-an  i  IRR-lCNJ  PART  t'ayak' angry  T didn't get mad.' b.  haw  k-an  i  NEG  IRR-lCNJ  PART  lhchiws tired  ti  stsi7s  DET  today  kw'ay' hungry  ta  mex-mixalh  DET  RED-bear  I didn't get tired today.' c.  haw  k-as  na  NEG  IRR-3CNJ  RL  na7  ta  LOC  DET  tem-t'ikw time-cold  'Bears don't get hungry in the wintertime.' In the data above, the predicates are translated as inchoatives, and not statives. For example, in (a), the speaker translates the sentence as T didn't get mad' and not T am not mad'. I take these data as a clue that these readings involve negation taking scope over the entire event and not just the second sub-event, which is the plain state itself. However, as with achievements above and activities before that,I am missing crucial data here, namely contexts where non-culmination readings are tested.  4.2.5. Summary of test 4 A summary of the negation facts for each of the predicate classes are given in the chart below : 37  Davis (2005) argues that negation is a propositional operator in Salish, and thus it w i l l always have widest scope over the predicate in an embedded clause. A s a result, he states that only one reading for negation is expected. T h i s suggests that the results from the negation diagnostic are only accidentally parallel to the almost data.  -120-  (112)  Summary: the scope of negation TEST 4 T H E SCOPE OF NEGATION  Event Cancellation Activities Accomplishments Achievements Inchoative States  Event Non-completion X X X  •/  X  These generalizations provide evidence for the following predicate representations and claims about the presence of final points in Skwxwu7mesh predicates.  (113)  Claims: Skwxwu7mesh  negation Predicate Representation  Activity  Readings induced by negation  Ae.3e 3e [e= (e ue ) A ( B E C O M E ( P ) ) ( e ) A (DO(P))(e )] s  1  2  1  2  1  2  Accomplishment  Xe.[DO(P))(e) A [VW' [W' is an inertia world w.r.t. w at the beginning of e —» | 3 e ' [culminates (e') in w ' A e causes e' in w']]]]  Achievement  Xe.(BECOME(P))(e)  Inchoative States  Xe.3e 3e [e= (e ue ) A ( B E C O M E ( P ) ) ( e ) A  Event Cancellation Event Cancellation Event Cancellation  s  1  2  1  2  1  P(e )]  Event Cancellation  2  4.3. Summary: event cancellation vs. event non-completion In this section, I have shown the results of two diagnostics that test the presence of final points in the representations of Skwxwu7mesh predicates: the scope of kilh 'almost' and the scope of negation. Although there are still crucial data missing, I summarize the data presented in this section as follows:  -727-  (114)  Skwxwu7mesh  Final points: Event cancellation vs. Event  non-continuation  READINGS INDUCED B Y SCOPE TESTS  Test 3: The scope of kilh  Test 4: The scope of negation  Event  Event non-  Event  Event non-  Cancellation  completion  Cancellation  completion  Final Point  Activity  X  X  X  Accomplishment  X  X  X  Achievement  X  X  S  I n c h o a t i v e state  X  X  X  y  I have argued that the readings (or lack thereof) induced by these diagnostics provide evidence for claim that achievements are the only predicate class in Skwxwu7mesh that have final points. This claim and the proposed predicate representations are summarized again below: (115)  Final Points Predicate Representation  Final Point  Accomplishment  hz3z3e [t=Xt ut ) A (BECOME(P))(e ) A (DO(P))(e )] Xe.[DO(P))(e) A [VW' [W' is an inertia world w.r.t. w at the beginning of e -» [3e' [culminates (e') in w' A e causes e' in w']]]]  Achievement  Ae.(BECOME(P))(e)  Inchoative States  Xe.3e 3e [e= (e ue ) A (BECOME(P))(e,) A P(e )]  Activity  2  x  2  1  X  2  s  1  2  1  2  2  X  X  5. I n t r i n s i c f i n a l p o i n t s v s . p r a g m a t i c a l l y - c o n d i t i o n e d f i n a l p o i n t s  In this chapter, I have argued that on the basis of the results of four tests, achievements are the only predicate class in Skwxwu7mesh that have final points as part of their representation. These final points, I argue, are represented as final BECOME events. Final points, however, can be pragmatically conditioned rather than intrinsic to the predicate. In this section, I will show that contexts where there is no overt culmination cancellation (such as, but not limited to, out of the blue contexts) trigger culmination in Skwxwu7mesh accomplishment predicates.  -122-  5.1. Pragmatically-conditioned final points in out-of-the-blue contexts 5.1.1. Skwxwu7mesh activities: past or present In out-of-the-blue sentences, Skwxwu7mesh perfective activities are translated in the past perfective or present progressive:  (116)  a.  chen  xay-m  ls.SG  laugh-INTR  38  (i) T laughed.' (ii) T am laughing.' b.  chen itut 1 S . S G sleep (i) 'I slept' (ii) T am sleeping.'  c.  na lulum lha slhanay' RL sing DET woman (i) 'The woman sang.' (ii) 'The woman is singing.'  d.  na imesh ta John RL walk D E T John (i) 'John walked.' (ii) 'John is walking.'  In Chapter Four, I show that the perfective is not overtly marked in Skwxwu7mesh, and argue that there is no dedicated perfective morpheme. A progressive or imperfective activity (marked by a progressive or imperfective morpheme) sentence is always translated with the English progressive. When a speaker is given both a perfective sentence and a sentence marked by an imperfective morpheme, speakers offer a past perfective English translation for the perfective sentence. This is illustrated in the data below where the perfective sentence in  I assume that the present readings available are perfective. I suggest that they are translated as English present progressives, because there is no simple present in English (see Bar-el 1998 for discussion on Skwxwu7mesh and Brinton 1988 for discussion on the notion of present perfective). The addition of an adverbial referring to a past time (e.g., kwi chel'aklh 'yesterday') yields a simple past translation.  -123-  (a) yields a past non-progressive reading while the imperfective sentence in (b), marked by •  39  the addition of wa and C V (bolded), yields an in-progress (imperfective) reading:  (117)  a.  na RL  shupn whistle  lha DET  Carrie Carrie  kwi  s-es  tin-tin  ta  DET  NOM-3POSS  REDUP-ring  DET  new'tstn phone  'Carrie whistled when the phone rang.' xContext: Carrie was whistling Speaker's comments: "you said whistled .. .shushpn [reduplicated form] is for whistling" b.  na  wa  shu-shpn  lha  RL  IMPERF  REDUP-whistle  DET  Carrie Carrie  'Carrie's whistling'  5.1.2. Skwxwu7mesh accomplishments: past culminated Accomplishments yield past culminated events in out of the blue contexts. Speaker's comments such as those given in (21a, b, h) emphasize that the event is interpreted as having culminated:  (118)  a.  na xel'-t-as ta sxwexwiy'am' RL write-TR-3ERG DET story ' M a r y wrote a story.' Speaker's comments: "she wrote it...shefinished"  b.  chen  p'ats'-an  ta  hem'ten  ls.SG  sew-TR  DET  blanket  tetxwem  lha DET  Mary Mary  T sewed the blanket.' c.  chen  p'ayak-an  ta  ls.SG  fix-TR  DET  car  T fixed the car.' Speaker's comments: "You already fixed it"  3 9  H . Davis (p.c.) suggests that not all Salish activities are alike; for example, the word in St'at'imcets for  'work' is consistently translated in the past in out of the blue contexts, unlike many other activity predicates. W h i l e I agree that this might be the case in Skwxwu7mesh as well, my fieldwork has not shown this. T h i s would require further exploration.  -124-  d.  na kw'el-ent-as ta RL cook-TR-3ERG D E T 'John cooked the meat.'  smeyts meat  e.  na  pen-t-as  shaw'  ta  RL  bury-TR-3ERG D E T  bone  DET  ta  kwa DET  John John  skwemay' dog  'The dog buried the bones.' f.  chen xewtl'-an ls.SG break-TR T broke the pencil.'  ta DET  xel'ten pencil  g.  chen lhen'-t 1 s.SG weave-TR T wove a basket.'  ta DET  basket basket  h.  chen  ilhen  kwi  shawik  ls.SG  eat  DET  carrot  'I ate a carrot.' Speaker's comments: does it mean that the carrot is finished?  "yes"  5.1.3. S k w x w u 7 m e s h achievements: past culminated The following data show that perfective achievements yield past, culminated readings only. The speaker's comments in (a) reinforce the fact that the speaker perceives the event as completed: (119)  a.  b.  t l ' i k ta John arrive D E T John 'John got here.7'John arrived.' Speaker's comments: "can't mean he hasn't arrived chen  tl'exwenk  ls.SG  win  'I won.' chen  wi7xw-em  ls.SG  fall-INTR  T fell [from above].' d.  na  kw'uy die 'The man died.'  RL  kwetsi swi7ka DEM man  -125-  yet"  chen  mekw'm  ls.SG  find  kwetsi smant DEM rock  'I found a rock.' chen tsixw ta ayalhkw l s . S G reach D E T beach 'I reached the beach.' na xwey ta swi7ka RL appear D E T man 'The man appeared.' na  xelk'-em  ta  skakl  na7/tina7  ta  RL  fall-iNTR  DET  baby  LOC/FROM  DET  yay wes bed  'The baby fell off the bed.'  5.1.4. Skwxwu7mesh inchoative states: past inchoative or present stative Finally, in out of the blue contexts, inchoative states yield past inchoative translations as well as present stative translations:  (120)  a.  chen  t'ayak'  ls.SG  angry  (i) 'I got angry/upset.' (ii) 'I am angry.' b.  chen  kw'ay'  ls.SG  hungry  (i) 'I got hungry.' (ii) 'I am hungry.' c.  chen  lhchiws  ls.SG  tired  (i) 'I got tired.' (ii) ' l a m tired.' d.  na katl' RL cloudy 'it got cloudy.'  -126-  e.  chen  ts'ulh  ls.SG  cold  (i) ' I ' m cold.'  (ii) 'I got cold.' 5.1.5. Summary These results are summarized in the table below:  (121)  Out of the Blue Contexts  Predicate Class Activities Accomplishments Achievements Inchoative States  in Skwxwu7mesh  Out of the Blue Reading  Past/Present Past Culminated Past Culminated Past(inchoative)/Present stative  In this chapter I have proposed that only achievements have intrinsic final points in their representations. The data above show that accomplishments have pragmatically conditioned final points. That is, without any evidence to the contrary, a perfective accomplishment is interpreted as a culminated event. Although Skwxwu7mesh activities and inchoative states have past readings, neither of them represent culminations of the entire event.  5.2. Deriving the Skwxwu7mesh accomplishment culmination implication  40  In this chapter I have argued that Skwxwu7mesh accomplishments do not have intrinsic final points. However, as I show in the previous section, Skwxwu7mesh accomplishments are interpreted as culminated events in out of the blue contexts. In this section, I argue that, following Matthewson (2004) who claims that St'at'imcets accomplishments have culmination implicatures, not entailments, Skwxwu7mesh accomplishments have culmination implicatures. In out of the blue contexts, accomplishments are translated as past culminated events, as shown again below:  T h e analysis presented in this section is due to Matthewson (2004), based on St'at'imcets data, and was then extended to S k w x w u 7 m e s h in Bar-el, Davis and Matthewson (2004).  -127-  (122)  a.  na xel'-t-as ta sxwexwiy'am' RL write-TR-3ERG D E T story ' M a r y wrote a story.' Speaker's comments: "She wrote it.. .she finished."  b.  chen p'ayak-an l s . S G fix-TR 'I fixed the car.'  ta DET  lha DET  Mary Mary  tetxwem car  Speaker's comments: [in response to whether it can mean 'started to fix]"You already fixed it." c.  na kw'el-ent-as RL cook-TR-3ERG 'John cooked the meat.'  ta DET  smeyts meat  kwa DET  John John  However, these culminations can be denied without inducing any contradictions or infelicity. This is shown once again below where the culmination can be cancelled (a), the event can be continued (b) and the culmination can be questioned (c): (123)  a.  na RL  p'ayak-ant-as heal-TR-3ERG  ta DET  John John  ta DET  huy-nexw-as  welh  haw  k-as  i  CONJ  NEG  IRR-3CNJ  PART  finish-TR-3ERG  'He fixed his canoe but he didn't finish (fixing) i t ' b.  snexwilh-s canoe-3POSS  (volunteered gloss)  chen  yetl'k'-an ta lam' i na7-xw chen wa yetl'k'-an paint-TR DET house P A R T RL-still l s . S G I M P E R F paint-TR 'I painted the house and I ' m still painting it.' (volunteered gloss) ls.SG  c.  A:  na cha7-st-as RL make-CAUS-3ERG ' M a r y made a box.'  B:  na  u  huy-nexw-as  RL  Q  finish-TR-3ERG  kwi DET  kw'axwa7 box  lha DET  Mary Mary  ' D i d she finish it?' Crucially, in each of the above examples, an accomplishment is formed by the suffixation of a transitiver, which in each of these cases, is a control transitivizer, which indicates that the agent of the event has complete control over the event denoted by the verb (Thompson 1979,  -128-  Davis and Demirdache 2000, among others).  41  T o account for the difference between English and Skwxwu7mesh where accomplishments in the former language entail culmination but i n the latter only implicate culmination, a number of assumptions must be made. First, following Davis (1997) and Davis and Demirdache (2000), who argue that all roots in Salish are unaccusatives based on the fact that they have single internal arguments, I assume that Skwxwu7mesh accomplishments are derived from unaccusatives (I am refraining from assuming that A L L roots are unaccusatives in Skwxwu7mesh, but restricting myself to accomplishments, that is, transitive marked VPs). The status of Skwxwii7mesh roots as unaccusatives can be seen in examples like the following where the sentences are interpreted with only one internal argument: (124)  42  a.  na kw'el ta smeyts RL getcooked D E T meat 'The meat is cooked / got cooked.'  b.  na yetl'k' ta lam' RL get.painted D E T house 'The house got painted.'  c.  na p'ayak ta snexwilh RL getfixed D E T canoe 'The canoe got fixed.'  d.  na  xwekw RL use 'That was used.'  tay' DEM  In general these sentences are interpreted in the past; however, as can be seen in the  4 1  T h i s discussion is confined to the prototypical control transitivizers  -Vnl-Vt  in Skwxwu7mesh, glossed T R f o r  'transitive'. There are other transitivizers which could be classified as 'control' (see Jacobs 1999). A s for the contrast between - Vn and —Vt in Skwxwu7mesh, Kuipers (1967:69) suggests that there may at some point have been a distinction, but it is not productive and that they "may be regarded as non-automatic allomorphs of one transitivizer". 4 2  Bare roots are often difficult to elicit and obtain judgements on in S k w x w u 7 m e s h ( H . Davis (p.c.) states that  the same is true for St'at'imcets, but that speakers have a lot of practice with them). Speakers prefer transitive versions of the predicates and to refer to an agent, even if the agent is not known. A n additional piece of evidence is missing here, that is, an example showing that an oblique external argument is impossible. I will have to leave this for future research.  -129-  translation for (a) above, a resultant state reading is also available for (at least some of) thes predicates.  43  These unaccusatives are achievements, and like Skwxwu7mesh achievements, they have culmination entailments. Thus, canceling the culmination of a bare root results in a contradiction. This is seen in the examples below:  (125) a. *  na RL  kw 'el get.cooked i CONJ  ta DET  na7-xw RL-Still  smeyts meat wa  kw'el-t-as  IMPERF  COOk-TR-3ERG  Speaker's comments: "You're saying it's cooked but they're still cooking it!.. .Why would you keep on cooking it? It's cooked!. ..Unless she thought it wasn't quite cooked, then she might put it back in the oven."  b. *  na  yetl'k' ta lam' i na7-xw wa yetl'k'-ant-as get.painted DET house C O N J RL-still I M P E R F paint-TR-3ERG 'The house got painted and it's still being painted.' Speaker's comments: [laughs] "no good"  RL  The denotation of a bare root is as follows: (126) a.  b.  na yetl'k' ta RL get.painted DET 'The house got painted."'  lam' house (volunteered gloss)  [[na yetl'k' ta lam' ] ] = Xe [the house gets painted in w (e)] w  T o summarize, the following are the facts to be accounted for: (i)  control-marked transitive V P s are derived from unaccusative roots by the suffixation of a control transitive (-Vt or-Vn)  (ii)  unaccusative roots have culmination entailments  (iii)  control-marked transitive V P s have culmination implicatures  It may not be the case that this is true for all predicates of this type.  -130-  The control transitivizer not only introduces the agent's control over the event (and possibly also the agent itself, but I w i l l set this aside for now), but must also be responsible for removing the culmination entailment, the requirement that the event culminate in the actual world, which is part of the meaning of the achievement from which the accomplishment is derived. T o account for this, I draw on the modalized approach often used for the progressive as in Dowty 1977. Take the following example of the English progressive in (a) and its truth conditions in (b): (127)  a.  Mary is building a house.  b.  Mary is building a house is true in a world w at an interval / iff in every inertia world w' for w at / this interval / is a subinterval of a larger interval where Mary build a house is true.  Drawing on L e w i s ' s notion of the "natural course of events", Dowty defines inertia worlds as follows:  (128)  Inertia Worlds (Dowty  1979:148)  Worlds which are exactly like the given world up to the time in question and in which the future course of events after this time develops in ways most compatible with the past course of events. A schematic illustration is given below:  (129)  Event Time  <—  ^  Reference  Timp  ^  Extending the notion of inertia worlds to the control transitive in Skwxwu7mesh (and Salish in general), I propose the following denotation for the control transitivizer:  -131-  (130)  [[ C O N T R O L . T R A N S  ]]  w  = Xf G D<]  s  t  >  [Xe [e is controlled by its agent in w  A VW'  [w' is an inertia world w.r.t. w at the beginning of e -» [3e' [f (e')(w') A e causes e' inw'llll This suggests that the control transitivizer takes a culminating verbal root (that is, an achievement with a requirement that it culminate in the actual world), and says that (i) the event is controlled by its agent and (ii) in all inertia worlds, the event leads to the culmination expressed by the root. A p p l i e d to a Skwxwu7mesh example, this is seen below: (131)  a.  chen yetl'k'-an ta l s . S G paint-TR DET 'I painted the house.'  lam' house  b.  [[chen yetl'k'-an ta lam' ] J = Xx Xe [x is the agent of e A e is controlled by x in w A [ V w ' [w' is an inertia world w.r.t. w at the beginning of e -*• [3e' [the house gets painted in w ' (e') A e causes e' in w']]]]] W  What (131) mean is that 'I was the agent that was in control of an event which in all inertia worlds causes the house to get painted.' The analysis for perfective Salish accomplishments differs from Dowty's analysis of the English progressive because this analysis allows the inertia worlds to branch off at the beginning of the event, rather than at the end of the reference time. This is schematically illustrated below:  (132)  Reference Time  <  F.vcnt Tinjo  ^  p.  Thus, Skwxwu7mesh perfective accomplishments have different truth conditions from English progressives. In the English sentence / was fixing a fence, the reference time is contained within the event time, and the event cannot culminate i n the actual world within the reference time. In the case of the Skwxwu7mesh perfective accomplishment, the event time is inside the reference time (as with any perfective), thus the event may or may not  -132-  culminate in the actual world within the reference time. This accounts for the fact that Skwxwu7mesh accomplishments do not entail culmination; however, recall that Skwxwu7mesh accomplishments do culminate in out of the blue contexts and thus have been argued to have a culmination implicature which still needs to be accounted for. I argue here that this implicature of culmination arises because in all inertia worlds, the event culminates; in the absence of other information, the hearer assumes that the 'normal' course of events (which is culmination in the case of accomplishments) takes place. Given that I draw on the modalized analysis of the English progressive in order to account for the Skwxwu7mesh accomplishment facts, a question that arises is why the English progressive (which also involves inertia worlds) does N O T implicate culmination. That is, the sentence in (a) below does not implicate (b):  (133)  a.  John was building a house,  b.  John built a house.  Perfective/imperfective Entailment of Culmination  John was building a house  John built a house  Imperfective  Perfective  X  Implication of Culmination  N/A  In English there is a contrasting form (the perfective) which entails culmination, namely (b) above; thus, to express culmination, (b) will always be a better way of doing so. A s a result, the progressive i n (a) lacks a culmination implicature. Preliminary data show that there are perfective transitive accomplishments which entail culmination in Skwxwu7mesh; these take a non-control/limited-control transitivizer. A t this point it is not clear to me what the status of these predicates is with respect to aspectual distinctions. What I will commit myself to now is that it holds in Skwxwu7mesh that for any agent in controfof an event, there is no contrasting perfective control marked form that entails culmination. Turning now to the tense effect that is observed with Skwxwu7mesh  -133-  accomplishments, recall that perfective accomplishments are usually translated in the past:  (134)  a.  na xel'-t-as ta sxwexwiy'am' RL write-TR-3ERG D E T story ' M a r y wrote a story.' Speaker's comments: "She wrote it...she finished."  b.  chen  p'ayak-an  ta  tetxwem  ls.SG  fix-TR  DET  car  lha DET  Mary Mary  'I fixed the car.' Speaker's comments: [in response to whether it can mean 'started tofix]"You already fixed it."  c.  na kw'el-ent-as RL cook-TR-3ERG 'John cooked the meat.'  ta DET  smeyts meat  kwa DET  John John  The question that arises is why is this the case given that the truth conditions that have been assigned to the control transitive do not force a past event time. This issue can be dealt with in the same way as the English progressive. In Skwxwu7mesh, there is a contrasting form (containing the imperfective morpheme) which is a better way to express a present progressive interpretation (since it explicitly makes sure the event is not completed within the reference time). A minimal pair is given below (imperfective sentences are discussed in further detail in Chapter Five):  (135)  a.  chen lhen'-t l s . S G weave-TR 'I made a blanket.'  ta DET  hem'ten blanket  b.  chen  wa  ta  ls.SG  IMPERF  DET  lhen'-t weave-TR ' I ' m making the blanket.'  hem'ten blanket  Therefore, the preferred interpretation of a perfective accomplishment sentence will be in the past. Note, however, that, as predicted, present progressive translations of perfective accomplishments are sometimes available. This can be seen in the following dialogue where the perfective form of the accomplishment predicate is used, but the context has forced an imperfective interpretation of the sentence. While in some cases speakers will change the  -134-  sentence to include the imperfective morpheme when repeated, speakers w i l l translate the bolded predicate as imperfective: (136)  A:  na encha lha RL where D E T 'Where is Carrie?'  B:  na7 ta lam'-s LOC DET house-3poss ' A t her house.'  A:  na  wa  RL  IMPERF  Carrie Carrie  chanem do.what  lha  Carrie Carrie  DET  'What is she doing?' na  p'ats'-ant-as ta  yekway'-s  RL  sew-TR-3ERG D E T  dress-3POSS  'She's sewing her own dress.'  (volunteered gloss)  6. Conclusion In this chapter I have motivated the absence/presence of intrinsic final points in the representations of Skwxwu7mesh predicates on the basis of the results of four diagnostics: (i) culmination cancellation, (ii) event continuation (iii) the scope of kilh 'almost' and (iv) the scope of negation. I argue that these tests show that activities, accomplishments and inchoative states do not have intrinsic final points, while achievements have intrinsic final points. These results and claims are summarized in the charts below:  (137)  Culmination  Cancellation  and Event  Continuation  TEST 1  TEST 2  CULMINATION CANCELLATION  Conjunctions Activity Accomplishment Achievement Inchoative State  Questions  Questions  Final Point  >/  X  E V E N T CONTINUATION  Conjunctions  X X  -  -  X  (*=infelicitous; S=felicitous;  -135-  -  S X  - =data not yet tested)  (138)  Event cancellation  vs. Event  non-continuation  READINGS INDUCED B Y SCOPE TESTS  Test 3: The scope of kilh Event Cancellation  Event noncompletion  Activity  Test 4: The scope of negation Final Event Cancellation  Event noncompletion  Point  X  X X  X  Accomplishment  •/  X  </  X  Achievement  X  •/  X  Inchoative state  X  (139)  Summary: Initial  X  Points Predicate Representation  Activity  X  te.ae^eate^fouea) (DO(P))(e )]  Final Point  A (BECOME(P))(e ) A  X  1  2  Accomplishment  Xe.[DO(P))(e) A [VW' [W' is an inertia world w.r.t. w at the beginning of e -> [Be' [culminates (e') in w ' A e causes e ' i n w']]]]  X  Achievement  Ae.(BECOME(P))(e)  •  Inchoative States  XcBe^ej^CejUej) A ( B E C O M E ( P ) ) ( e , ) A P(e )] 2  X  I also have shown that judgements of sentences that lack any overt cancellation indicate that accomplishments in Skwxwu7mesh, while they lack intrinsic final points, have culmination implicatures. This implicature arises with control marked transitivizers and is derived by the addition of modality in the representation of Skwxwu7mesh accomplishments.  -136-  Chapter 3: Intrinsic Initial Points  Events take place in a temporally ordered fashion and can, in many cases, be segmented into a series of smaller temporal periods.. .an ONSET, a NUCLEUS,  and a CODA. Each of these can be individually described, and  each can be viewed in relation to some other segment; it is sometimes difficult to draw a strict line of demarcation between them, hence the slippery nature of the verbs that characterize these different time periods. (Freed 1979: 30)  1. On the relevance of initial points Initial points are rarely, if ever, addressed in discussions of aspectual predicate classes. Although initial points may be implicitly assumed, the focus in the literature on the classification of predicates according to their aspectual status has been on their final points. A s noted in Chapter One, Smith (1997) suggests that "the initial endpoints of events are natural, since they represent a change from a state of rest" (p. 22). She proposes that activities and accomplishments have initial points, contrasting with states that do not; the temporal schemata informally representing initial points is illustrated below (initial points 1  are in boldface; see Chapter One for further discussion):  (1)  a.  Activities:  I  b.  Accomplishments  I  c.  Semelfactives  E  d.  Achievements  e.  States  ^Arb  •E  R  (F)  (I)  (p. 23-32) A s in Chapter T w o where I use the term "final points" to refer to final B E C O M E events in the representation of a predicate, in this chapter, I use the term "initial points" to  1  It is not clear whether Smith assumes semelfactives and achievements (both events) to have initial points; this  would be dependent on her definition of initial points.  -137-  refer to initial B E C O M E events. Once again, the term "intrinsic" here is used to refer to the 2  claim that final points are represented in the templates of aspectual classes, not that final points are necessarily primitive.  1.1. The presence of intrinsic initial points in Skwxwu7mesh predicates In the same way that a final point is represented as a final B E C O M E event in the representation of a predicate, I suggest that an initial point corresponds to an initial change of state, represented by an initial B E C O M E event in the denotation of a given predicate. On the basis of two diagnostics: (i) the readings induced by punctual clauses/adverbials and (ii) the readings induced by the auxiliary mi, I propose the following classification of final points in the meaning of Skwxwu7mesh predicates (initial points are in boldface):  (2)  Skwxwii7mesh predicates: Initial points  Initial Point  Representation Xe.3e 3e [e= (e ue ) A (BECOME(P))(e,) A  Activity  s  1  2  1  2  (DO(P))(e )j  swim, rest, laugh  2  write a book, fix the car  A,e.[DO(P))(e) A [ V W ' [W' is an inertia world w.r.t. w at the beginning of e -> [3e' [culminates (e') in w ' A e causes e' in w']]]]  Achievement  Xe.(BECOME(P))(e)  Accomplishment  win, arrive, find a rock  Inchoative  Xe.3e 3e [e= (e ue ) A (BECOME(P))(e ) A  (get) angry, (get) cloudy  P(e )J  s  1  2  1  2  1  2  Rothstein does not address initial points in her discussion of the representation of English predicates. Comparing predicates in the two languages, it is activities and inchoative states that are the most contrastive with respect to initial points. That is, initial points are not represented in either class in English, but are proposed to be in the representation of the parallel Skwxwu7mesh predicates.  2  K i y o t a (2005) proposes a slightly different approach to analyzing initial points by using a B E G I N operator.  His template for SancaSan activities is as follows: (i)  A c B e ^ l e ^ U e . )  A ( B E G I N ( P ) ) ( e , ) A (DO(P))(e)  -138-  =e ] 2  (3)  Activities: English vs. Skwxwu7mesh a.  English Xe.(DO(P))(e)  b  Skwxwu7mesh te3e 3e [e= (e. ut ) ,  l  (4)  2  1  2  A . ( B E C O M E ( P ) ) ( e ) A (DO(P))(e )] 1  2  (Inchoative) States: English vs. Skwxwu7mesh a.  English Xe.P(e)  b.  Skwxwu7mesh ' f X e . B e ^ e . t e ^ C e ^ e ^ A ( B E C O M E ( P ) ) ( e ) A P(e )] 1  2  A s the representations above show, English activities and states have a simple structure, while the parallel classes in Skwxwu7mesh are more complex. Although the second subevent of each of the Skwxwu7mesh predicates are the same as their simpler English counterparts, the crucial difference is the presence of an initial point, represented as an initial B E C O M E sub-event.  1.2. Outline of this chapter This chapter is organized as follows. I begin with an overview of and motivation for the two diagnostics used in this chapter to test for the presence of intrinsic initial points in Skwxwu7mesh (§2). I then show the results of the first diagnostic (the effect of punctual clauses and adverbials) by providing the relevant data for each predicate class and argue that this test provides evidence that while Skwxwu7mesh activities, achievements and inchoative states have intrinsic initial points, accomplishments in the language do not (§3). Turning to the second diagnostic (the effect of the auxiliary mi 'come'), I again present the relevant data for each predicate class and show the readings induced by the combination of these predicates with mi and argue that there is further evidence for the presence of intrinsic initial points in each of the predicate classes except for Skwxwu7mesh accomplishments (§4). I end this chapter with a discussion of some alternatives to the proposal and show what would have  H e states that the first sub-event of activities is "not a change of state but it is rather a beginning or initial point of an event which is indicated with a proposed operator B E G I N " (p. 23).  -139-  to be proposed in order to account for the Skwxwu7mesh facts and that, ultimately, these solutions are no better than the one proposed (§5).  2. Diagnosing initial points I use the following diagnostics to argue for the above claims regarding the presence of initial points in the representations of Skwxwu7mesh predicates:  (5)  Diagnostics for initial points a.  Punctual clauses and punctual adverbials  b.  The auxiliary m i ' c o m e '  The first diagnostic (a) tests for inceptive readings with punctual clauses and adverbials, while the second diagnostic (b) tests for an initial change of state. In this section I examine these diagnostics in detail, motivating their use as diagnostics for initial points, and showing how they are replicated in Skwxwu7mesh.  2.1. Test 1: readings induced by punctual clauses/adverbials The effect of punctual clauses and adverbials has been discussed in the literature; however, the documented readings differ from author to author, specifically with respect to accomplishments (see Smith 1997, Bertinetto 2001, and Terry 2004, for some examples), and there does not seem any strong argument as to why the readings observed arise. In this section, I sketch out some of the readings that are induced by English punctual clauses/adverbials that have been reported in the literature, and outline the proposals that have been put forth to account for them. I w i l l show that those proposals that attempt to explain how the readings are induced have difficulty accounting for the contrast between activities and accomplishments in English. 1 will then motivate my proposal that the effect 3  of punctual clauses/adverbials is a test for the presence of initial points in the meaning of predicates and show how this works for each Skwxwu7mesh predicate class.  See Chapter Six for a detailed discussion of English activities and accomplishments.  -140-  2.1.1. On the reported effect of punctual clauses/adverbials Smith argues that predicates with durative features (for her, accomplishments, activities and states) have an inceptive interpretation with momentary adverbials (what I call punctual adverbs), such as at noon, at 5 o'clock exactly, but that they have "direct" interpretations with instantaneous predicates (those with a [-Durative] feature, such as achievements, as in (d) and semelfactives, as in (e)):  4  (p. 42; ex. 9a)  a.  They ate dinner at noon.  ACCOMPLISHMENT  b.  He pushed the cart at noon.  ACTIVITY  (p. 45; ex. 18e)  c.  The baby was asleep at noon.  STATE  (p. 47; ex. 27a)  d.  The clock struck at noon.  ACHIEVEMENT  e.  She tapped his shoulder at noon.  SEMELFACTIVE  (p. 42; ex. 9b) (p. 46; ex. 23a)  A s Smith suggests, the inception of the eating dinner event is interpreted as having taken place at noon in (a), the inception of the pushing the cart event is interpreted as having taken place at noon in (b) and although she does not state it, she predicts that the inception of the asleep state should be interpreted as having taken place at noon in (c). W h i l e I believe this reading to be possible under a specific context (e.g., I was trying to get the baby to sleep for a long time by reading her books and singing her songs and she was asleep at noon), the preferred reading here is almost certainly the stative reading (the baby fell asleep before noon, and was already asleep at noon). Turning to w/zen-clauses, Smith argues that they impose no particular temporal relation on situations and thus if one of two clauses is a when-c\mse, the events can be simultaneous, overlapping or successive, depending on the viewpoint (perfective or imperfective) of both clauses and the situation type of the predicates i n both clauses. Since I am focusing on punctual events here, let us look at her account of when-cl&uses consisting of instantaneous events. Smith suggests that they can produce either a successive/simultaneous interpretation (a) or an overlapping interpretation (b), which is dependent on the viewpoint and the predicate class. The examples are given below with a discussion and schematic illustration to follow:  -141-  (7)  (8)  a.  Mary swam when the bell rang.  ACTIVITY -  PERFECTIVE  b.  Mary was swimming when the bell rang.  ACTIVITY -  IMPERFECTIVE  a,  B i l l left when the bell rang.  ACHIEVEMENT -  PERFECTIVE  b  B i l l was leaving when the bell rang.  ACHIEVEMENT -  IMPERFECTIVE  (Smith 1997, p. 64, ex. 6)  With a perfective activity predicate, as the one in (7a), it is only the initial point of the activity (the inception of the swimming event) that is either simultaneous with the bell ringing, or successive (the bell rings and then Mary starts to swim) . However, with an 5  imperfective activity, as in (7b), the events are overlapping: Mary was already swimming when the bell rang. These two sentences are illustrated below (the top horizontal line is the time line and the bottom horizontal line is the event time of the matrix clause):  (9)  a.  Mary swam when the bell rang The bell R I N G  4  5  1 assume that by "direct" interpretation, Smith means that two instantaneous events occur simultaneously. Both readings may not be available for all speakers; I have not yet tested thisi I do not believe that this impacts  on the analysis presented here.  -142-  b.  Mary was swimming when the bell rang The bell R I N G  Mary S W I M  With a perfective achievement predicate, like the one i n (8a), the entire two events happen either successively (the bell rings first and then B i l l left just after) or simultaneously (both events occur at the same time). With an imperfective achievement, as in (8b), the two events are overlapping: B i l l was in the middle of leaving when the bell rang. The reading in (a) is illustrated below:  (10)  Bill left when the bell rang The bell R I N G Bill L E A V E  Thus the sentence in (8b) differs from (7b) in that it is not the inception of the matrix event that co-occurs with the bell ringing, but the two events in their entirety occur simultaneously or immediately consecutively. In Smith's words, "The perfective is taken as an inceptive in [(7a)] because swimming is a durative event. It is not plausible that a swimming event occur in its entirety at the same time as a bell ringing, but entirely plausible that it begin at that time" (p. 65). This does not explain why accomplishment predicates do not necessarily receive an inceptive reading as it seems to be the case that it is more plausible that they would begin at a certain time rather than the entire event occurring at a particular time; however, as is shown by the  -143-  English study in Chapter Six, the latter is the typical reading of accomplishment predicates (the "all at once" or "instantaneous" reading).  6  Smith uses these data to argue that w/ien-clauses show that the imperfective viewpoint does not semantically include endpoints; the sequential reading does not arise for sentences with an imperfective viewpoint in the main clause since the endpoints are not visible in this viewpoint. I extend this and argue that it also tests the membership of a predicate in an aspectual class, much like the punctual adverb. Thus, in my terms, the addition of a w/ien-clause to a main clause that contains a perfective predicate tests whether an event has an initial point. Moreover, while I agree that the readings associated with punctual adverbials are related to the predicate class, I argue that it is not the durative or instantaneous feature that forces a particular reading, but the presence of initial points of the events that does so. Notice that many accomplishment predicates do not have an inceptive interpretation with punctual clauses/adverbials. Often these predicates are odd with these adverbials and for many speakers the intended meaning is that the entire event happened at noon:  (11)  a.  #She knit a sweater at noon.  b.  #She wrote the book at noon.  c.  #She built the house at noon.  I suggest that the claim that this is an initial point test will not only account for the facts but will provide an explanation of the facts. Rothstein (2004) suggests that the phrase  at a time  temporally locates an event at a  particular point in time. Examining each aspectual class with a punctually locating expression, she suggests that this type of phrase classifies states and achievements together, presumably because predicates from both classes can occur with such expressions without shifting the meaning i n any way. Rothstein suggests that states can hold at instants because they are "totally homogeneous" (a-b); achievements can be punctually located since they are instantaneous (c). She then suggests that "when an activity occurs with a punctual adverb, the  6  In that chapter I also discuss the inceptive reading that seems to be available for accomplishments in certain  contexts and suggest a possible explanation.  -144-  effect is to assert that the activity began at the temporal point given, presumably since this is the only privileged instantaneous event available" (p. 25). A s for accomplishments (e) Rothstein states that they do not have an inceptive reading, but does not explain why (e) is infelicitous:  (12)  a.  A t that moment, John believed in miracles.  STATE  b.  Mary was happy at midnight.  STATE  c.  The guests arrived at midnight.  ACHIEVEMENT  d.  John ran at 9 p.m.  ACTIVITY  e.  #Mary painted a picture at midnight.  ACCOMPLISHMENT  (Rothstein 2004:25; ex. 41)  Unlike Smith, who argues that semelfactives are a separate predicate class, Rothstein suggests that activities have semelfactive "single event" uses. This is seen clearly with punctual clauses/adverbials:  (13)  a.  John jumped at ten o'clock.  b.  Mary winked at twelve o'clock to remind me to make the phone call. (Rothstein 2004:184; ex. 3a-b)  Rothstein might argue that English accomplishments have two privileged points - an initial and a final, and thus might predict an inceptive or culminating reading. However, she might also say that since English accomplishments have more than one "point", there is no privileged point, in which case that might account for the infelicity. Since Rothstein does not explain this any further, I can only speculate on how she would account for the accomplishments; however, as it stands, it is not clear how the contrast between activities and accomplishments can be accounted for. L i k e Smith, Bertinetto (2001) appeals to the notion of durativity to contrast achievements, states and accomplishments. He argues that while the achievement in (a) suggests that the event occurred precisely at that moment (noon), the state in (b) and the accomplishment i n (c) are unnatural and that they can "at most indicate (depending on the  -145-  situation) the initial or final boundary of the event" (p. 179) (the notions are those used by the author):  (14)  a.  John reached the top of the mountain at noon:  ACHIEVEMENT  b.  ??John liked the music at midnight two days ago.  STATE  c.  ??John wrote his dissertation at 5 o'clock last Tuesday.  ACCOMPLISHMENT  Bertinetto does not refer to activities in the context of punctual adverbials, though they, like states and accomplishments and in contrast to achievements, are assigned a durative feature in his system. A s I noted in regards to Smith's analysis above, the durative feature alone cannot account for the contrast in readings/judgements of predicates from different classes induced by punctual clauses. Finally, Terry (2004) presents the following data that illustrates the contrast between achievements and accomplishments; he shows that in English only events that could "reasonably occur during a very short period of time" are possible answers to the question 'What happened while Esther was entering the room?' (boldface has been added):  (15)  What happened while Esther was entering the room?  (16)  a.  Eugene dropped cake.  b.  Eugene started eating the cake.  c.  Eugene finished eating the cake.  d.  ? Eugene ate the cake.  e.  ?? Eugene wrote his dissertation. (Terry 2004: 77; ex. 79)  Terry states that the sentences in (d) and (e) above are both strange as answers to the above question. A s he notes, (d) is "distinctly odd, forcing an interpretation under which Eugene ate an entire cake while Esther was opening he door and walking into the room - surely an exaggeration. This sentence...cannot mean that Eugene started eating, finished eating, or continued eating the cake. The entire cake-eating event must be contained within the topic  -146-  time", (p. 77). The same is true for the sentence in (e) above which, again, according to Terry, "forces the pragmatically odd reading under which Eugene writes an entire dissertation during what any graduate student knows to be an unreasonably short period of time. This sentence cannot mean that Eugene was simply working on his dissertation, writing, for instance, the very first sentence" (p. 77). I will show in Chapter Six that these judgements closely parallel the judgements of my English consultants. L i k e Bertinetto, Terry does not refer to the readings induced by activities in this context. Thus, I can only assume that Terry would say that the sentence Eugene ran would be a reasonable answer to the question in (13) above, and that it would meant that Eugene began running at the interval that Esther entered the room. Moreover, I assume that Terry would argue that this sentence is a reasonable answer to the question since beginning to run is an event that can "reasonably occur during a very short period of time". However, like the other analyses discussed thus far, there does not seem to be any further explanation as to the contrast between activities and accomplishments with respect to the readings induced. A summary of these reported effects is given in the table below:  (17)  The reported effect of punctual clauses/'adverbials: summary  Smith Rothstein  Bertinetto  Activities  Accomplishments  Achievements  States  |+Durati\e| inceptive Privileged instantaneous event inceptive |+l)urati\e|  |+Durati\e| inceptive  [-Durativel direct Instantaneous instantaneous  |+Durative J inceptive/stati ve  |-Durative| instantaneous  |+Durati\c|  #  | * l >III.IIIW|  inceptive, \tafi\c  •.,(inceptive," •" culminating)-  (iiu epme, ,., *• culminating)  Terry  1 lomogeiU'iiN  -  -  #  instantaneous  I suggest that the table above raises three important issues: first, it should be clear that researchers do not agree as to the readings induced by punctual clauses/adverbials. For example, while Smith argues that accomplishments are felicitous in these contexts, and that they induce inceptive readings. Bertinetto, Rothstein and Terry argue that accomplishments  -147-  are infelicitous in these contexts; Bertinetto suggests that this can only have an inceptive or culminating reading, neither of which is possible. A s for states, Smith and Rothstein suggest that they have both inceptive and stative readings with punctual clauses; Bertinetto suggests that states are infelicitous with punctual clauses and at the most can have inceptive or culminating readings. Second, the table above shows that the ways in which researchers account for the readings they propose, and as a result, the ways in which they group predicate classes, differ. For example, Smith and Bertinetto both propose that activities, accomplishments and states have a [+Durative] feature and thus all behave the same in this context, while achievements have a [-Durative] feature and behave differently . However, Rothstein argues that 7  achievements and states pattern together in that they can both "be punctually located"; for her, state are homogenous and hold at instants, which is meant to explain why they can get both inceptive (12a) and stative readings (12b). Achievements for Rothstein are instantaneous changes of states and thus, as she puts it, can be punctually located. Finally, what the above chart and previous discussion shows is that these analyses do not seem to be able to account for the contrast in readings induced by punctual clauses. For example, Smith's durative feature alone does not seem to be able to explain why accomplishments also have instantaneous readings and are often judged infelicitous, nor can the durative feature alone explain why states have both inceptive and stative readings with punctual clauses. Rothstein's privileged instantaneous event which is meant to account for the inceptive reading of activities, does not explain the infelicity of accomplishments. Moreover, since for her, activities are homogenous as well, there is no account here for why states have both inceptive and stative readings, while activities only have inceptive readings. Although Bertinetto and Terry do not discuss activities, it does appear that neither of them will be able to account for the contrast between activities and accomplishments in these contexts. I suggest that the analysis I propose can account for the readings induced by punctual clauses, and the contrast across predicate classes in Skwxwu7mesh, by appealing to the  7  E v e n though Smith and Bertinetto's systems use the same feature here, they actually make different claims  to the readings punctual clauses induce.  -148-  presence of initial points in the predicate representations. In the next section, I show how this test works in Skwxwu7mesh.  2.1.2. Punctual clause/adverbials in Skwxwu7mesh The types of examples I use to examine these effects in Skwxwu7mesh are shown below. Punctual clauses are nominalized clauses that behave like w/ien-clauses i n English (they are translated as such by Skwxwu7mesh speakers). Punctual adverbials are locative phrases that behave like at clauses in English (and are translated as such by speakers). Example sentences are shown below:  (18)  a.  chen . . .  kwi-n-s  kw'ach-nexw kwa  ls.SG ...  DET-IPOSS-NOM  see-TR(LC)  DET  John John  'I X - e d . . . when I saw John.' b.  chen...  na7  ta  an'us-k  lS.SG...  LOC  DET  2-O'clock  T X - e d at 2 o'clock.' I argue that the readings that result from these test sentences provide evidence for the presence/absence of initial points in predicate representation.  2.2. Test 2: readings induced by mi 'come' The Skwxwu7mesh morpheme mi can be used as a main predicate or as an auxiliary. This is observed by both Kuipers (1967) and Currie (1997), and has been observed in recent elicitation as well. Kuipers refers to mi as a directional clitic or a full verb. He states that mi sometimes corresponds to English 'come' in expressions like 'come have a look' or 'come on, do it', and in other cases, it functions as a more general direction-indicator. The examples below illustrate that, as a main predicate, mi is translated as 'come':  (19)  a.  mi  chexw  come  2S.SG  ' C o m e here.'  -149-  b.  chen  lhk'i7s ti natlh kwi-s l s . S G know DET morning DET-NOM T knew this morning that you were coming.'  mi come  When it behaves as an auxiliary, mi combines with the main predicate and can yield a number of different readings. For example, in some cases, mi yields a 'directional' meaning of the predicate. This is shown in the data below where mi is translated as 'come' and add a directional meaning to the predicates uys 'enter' and t'ukw' 'go home':  (20)  a.  b.  mi chexw come 2 s . S G 'Come inside.'  uys enter (Kuipers 1967:161)  mi t'ukw' come go.home 'Come home.'  (Kuipers 1967:257)  In Bar-el (2003b), I sketched out a first pass at an analysis that accounts for the readings associated with mi when it combines with verbs of different classes. I argued that (i) mi foregrounds the initial point of an event, but does not introduce one, and (ii) i f the event has no initial point, mi contributes the meaning 'come' to create a complex predicate. The analysis presented here w i l l attempt to take that proposal further and show how the facts with the auxiliary mi provide evidence for the existence of initial points. Focusing on the auxiliary behaviour of mi, I claim in this chapter that the meaning associated with the complex predicate made up of mi and a verbal predicate is dependent on the aspectual class of that predicate. Thus, these data provides further evidence for the classification of Skwxwu7mesh predicates and the presence of (initial) B E C O M E events in predicate representations. I propose that mi picks out B E C O M E i n the denotation of a predicate to yield an inchoative reading.  -150-  3. Test 1: the readings induced by punctual clauses/adverbials in Skwxwu7mesh In this section, I explore the readings induced by punctual clauses and adverbials in Skwxwu7mesh. Based on the data in the remaining sections, I show the following results of this diagnostic which I suggest provide evidence for the following claims with respect to the presence of initial points i n the representation of Skwxwu7mesh predicates:  (21)  Skwxwu7mesh Initial points: Punctual  clauses/adverbials  Readings with Punctual Clauses/Adverbials Activity  Initial Point  inceptive  Accomplishment Achievement Inchoative state  inceptive, medial, culminating  X  entire event (instantaneous)  •/  inchoative  8  A s shown in Chapter One, I assume the meaning of the perfective is such that the event time is situated inside the reference time (see K l e i n 1994 and Kratzer 1998). The schema is shown again below:  (22)  [REFERENCE T I M E [EVENT TIME] I  I follow Currie (1997) who argues that postverbal adverbs in Skwxwu7mesh unambiguously have topic time readings (reference time, in the terminology employed here). The when 9  clause/adverbial indicates the reference time while the matrix clause indicates the event time; thus, the time at which the event of the matrix clause takes place must be at  10  or inside the  time at which the event of the when clause takes place. If the event denoted by the when  8  Smith (1997) states that  inchoatives present the coming about of a state, while inceptives present the entry into  an event. Kinkade and K i y o t a (1996) note that "[tjhe beginning of change is properly labeled 'inchoative' or 'inceptive'" (p. 236) and K i n k a d e (p.c.) suggests that the two terms refer to the same thing. I use the term inceptive in the predication for predicates with a D O operator and the term inchoative in the prediction for predicates without a D O operator. I think this is a terminological issue rather than theoretical one. T h e point is to note the "initial point" readings, which seem to correspond to either inchoative or inceptive, depending on.the meaning of the predicate. 9  1 0  Currie does not examine punctual adverbs, although I assume her claim to hold for these as well. "at" is K l e i n ' s term and is often appropriate for momentary events.  -151-  clause is punctual, the only way the matrix clause could be at or inside the event of the when clause is i f it the matrix clause was also punctual. This is illustrated below:  (23)  [REFERENCETIME  [EVENTTIME]  I  ]  I  when clause  matrix clause  (punctual)  (punctual)  A remaining question is then how to deal with predicates that are not themselves punctual. In this chapter I motivate the claim that activities, achievements and inchoative states have initial points; however, only achievements can be considered "punctual". I have to then account for how activities and inchoative states (both of which have initial points, but are not punctual) become punctual. I appeal to Rothstein's (2004) shift operation that she claims is triggered by the progressive, and shifts achievements into accomplishments. She states that achievements are near instantaneous events that are over as soon as they have begun; thus they are not expected to occur in the progressive, but sometimes they do. Examples are given again below:  (24)  a.  The old man is dying,  b.  The plane is landing. (Rothstein 2004:36, ex. d-e)  Rothstein proposes that in the progressive, an achievement is shifted into an accomplishment where the culmination of the accomplishment is the original achievement:  (25)  SHIFT(VP  punctual  11  ) : Xe.(BECOME(P))(e) -> '  Xe-ae^eate^fauej) A ( D O t a ) ) ^ ) A ( B E C O M E ( P ) ) ( e ) A Cul(e)=e ] 2  2  " These are simplified representations that leave out the thematic arguments. See Rothstein (2004) for a detailed discussion.  -752-  I propose here that in Skwxwu7mesh, punctual clauses/adverbials can trigger a shift operation where an achievement is derived from an activity or an inchoative state. Once this shift operation takes place, the activity or state, which did not start off as a punctual event, becomes a punctual event. A n illustration of a shifted activity is given below:  (26)  S H I F T (Xe.3e 3e [e= (e ue ) A ( B E C O M E ( P ) ) ( e , ) A (DO(P))(e )]) — s  1  2  1  2  2  = X e . ( B E C O M E (P))(e)  These representations are discussed in detail in the relevant sections below.  3.1. Skwxwu7mesh activities 3.1.1. Punctual clauses induce inceptive readings In Skwxwu7mesh, the addition of a punctual clause to an activity yields an inceptive reading of the main event. This is emphasized by the speaker's comments that the embedded event is in some sense the cause of the main event, as in (26a). Furthermore, the medial reading is unavailable for basic predicates in these structures, as shown explicitly in (26c), illustrating again that the inceptive reading is the only reading available (the predicate is bolded and the "when"-clause is bracketed):  (27)  a.  12  chen  xay-m [kwi-n-s kw'ach-nexw kwa John] laugh-iNTR [DET-IPOSS-NOM see-TR(LC) DET John] T laughed when I saw John.' Speaker's comments: It's because you saw John that you started to laugh ls.SG  b.  c.  na itut [ ta John RL sleep [ D E T John 'John fell asleep at two.'  na7 LOC  chen  lulum [ k w i  s-es  ls.SG  sing  NOM-3POSS  [ DET  ta DET  7an'us-k ] 2-o'clock]  t l ' i k ' ta arrive D E T  John ] John ]  T sang when John got here' * T was singing when John got here.'  1 2  W/zen-clauses in S k w x w u 7 m e s h are expressed as nominalized clauses, introduced by the determiner kwi.  the appendix for further discussion.  -755-  See  d.  chen  l u l u m [ y e w ' a n i w i l h kwi-s-es l s . S G sing [ before DET-NOM-3POSS T started singing before Peter left.'  huya7 ta leave D E T  Peter] Peter ]  The facts are confirmed further when comparing the perfective and imperfective forms of the predicates. I propose that wa is the imperfective marker and that the C V - reduplicant is the progressive maker (see Chapter Five); the data below show that the imperfective or progressive forms yield on-going interpretations with punctual embedded clauses while the perfective forms yield inceptive readings, even when the translation does not explicitly say so. For example, in the (a) example below (the perfective), the event of the speaker crying begins at the moment of (or immediately following) John's arrival; in the (b) example (the imperfective, indicated by the addition of wa), however, as the speaker's comments indicate, the event of the speaker crying had begun prior to John's arrival and is in progress at the moment when John enters:  (28)  a.  chen  xaa-m  kwi  s-es  ls.SG  cry-iNTR  DET  NOM-3POSS  tl'ik kwa arrive D E T  John John  ' I cried when John got here.' b.  chen lS.SG  wa IMPERF  xaa-m cry-INTR  kwi  s-es  DET  NOM-3POSS  tl'ik kwa arrive D E T  John John  T was crying when John got here.' Speaker's comments: "You say this if you had been crying for a while before" (29)  a.  na RL  kway ta hide DET  sta7uxwlh children  kwi  s-es  DET  NOM-3POSS  tl'ik lha arrive D E T  Mary Mary  'The children hid when Mary came.' SContext: Mary arrived, then children went to hide  -154-  b.  na  wa  RL  IMPERF  kway ta sta7uxwlh hide D E T children  kwi  s-es  DET  NOM-3POSS  tl'ik lha arrive D E T  Mary Mary  'The children were hiding when Mary came.' (30)  a.  na RL  shupn whistle  lha DET  Carrie Carrie  kwi  s-es  tin-tin  ta  DET  NOM-3POSS  REDUP-ring  DET  new'tstn phone  'Carrie whistled when the phone rang.' ^Context: Carrie was whistling Speaker's comments: "You said whistled .. .shushpn [reduplicated form] is for whistling"  13  b.  na  wa  RL  IMPERF  shu-shpn REDUP-whistle  lha DET  kwi  s-es  tin-tin  ta  DET  NOM-3POSS  REDUP-ring  DET  Carrie Carrie new'tstn phone  'Carrie was already whistling when the phone rang.' (31)  a.  chen  itut  kwi  s-es  huy-nexw  ta  ls.SG  sleep  DET  NOM-3POSS  finish-TR(LC)  DET  14  sxwexwiy'am story  'I went to sleep when he finished the story.'  b.  chen  wa  lS.SG  IMPERF  i-7tut REDUP-sleep IS  kwi  s-es  huy-nexw  ta  DET  NOM-3POSS  finish-TR(LC)  DET  sxwexwiy'am story  'I was sleeping when he finished the story.' (32)  a.  chen  xwitim  kwi  s-es  tin-tin  ta  ls.SG  jump  DET  NOM-3POSS  REDUP-ring  DET  new'tstn phone  'I jumped when the phone rang.' Speaker's comments: "Maybe it startled you"  1 3  See Chapter F i v e for the c l a i m that the C V - reduplicant is the Skwxwu7mesh progressive marker.  1 4  T h e speaker only offered "already whistling" for a gloss; I take this translation to be an appropriate one.  1 5  In the form for sleep there is an initial glottal stop that is not represented in the orthography.  -155-  b.  chen  wa  xwi-xwitim  lS.SG  IMPERF  REDUP-jump  kwi  s-es  tin-tin  ta  DET  NOM-3POSS  REDUP-ring  DET  new'tstn phone  T was jumping when the phone rang.' SContext: phone did not make me jump but I was jumping maybe for exercise and then the phone rang (33)  a.  chen  tkwaya7n  ta  ls.SG  hear  DET  kwi  s-es  DET  NOM-3POSS  slulum song tl'ik kwa arrive D E T  John John  T heard the song when John got here.' •SContext: John was singing the song (I only heard it once he got here) * Context: I could hear it before John arrived - Speaker's comments: "I guess it's possible, but only if you were a spiritual dancer" b.  chen  wa  tkwaya7n  ta  ls.SG  IMPERF  hear  DET  kwi  s-es  DET  NOM-3POSS  tl'ik kwa arrive D E T  slulum song John John