Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

kâ-yôskâtahk ôma nêhiyawêwin : the representation of intentionality in Plains Cree Mühlbauer, Jeffrey Thomas 2008

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2008_fall_muehlbauer_jeffrey.pdf [ 5.49MB ]
Metadata
JSON: 24-1.0066442.json
JSON-LD: 24-1.0066442-ld.json
RDF/XML (Pretty): 24-1.0066442-rdf.xml
RDF/JSON: 24-1.0066442-rdf.json
Turtle: 24-1.0066442-turtle.txt
N-Triples: 24-1.0066442-rdf-ntriples.txt
Original Record: 24-1.0066442-source.json
Full Text
24-1.0066442-fulltext.txt
Citation
24-1.0066442.ris

Full Text

kâ-yôskâtahk ôma nêhiyawêwin: THE REPRESENTATION OF INTENTIONALITY IN PLAINS CREE by JEFFREY THOMAS MÜHLBAUER B.A., The University of Wisconsin. 2001  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Linguistics)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  June 2008 © Jeffrey Thomas Mühlbauer, 2008  Abstract This thesis considers the reference system of Plains Cree, an Algonquian language spoken in Canada. I argue that the referential system of this language can be understood as coding distinctions in extentionality; it distinguishes between referents that possess perspectives (‘intentional’) and referents that do not (‘extentional’). With respect to perspectival possession, Plains Cree distinguishes four referential classes: (i) inherently extentional “Inanimate” referents, (ii) contextually extentional “Obviative” referents, (iii) contextually intentional “Proximate” referents, and (iv) unspecified “Animate” referents. I then show that the referential class “Obviative” is decompositional; it is constructed out of components that code referential dependency, which is the confluence of structural ordering and perspectival embedding. Finally, I consider the methodological issues raised by the study of referential types, showing how different data-collection methods interact with the semantics of perspectival possession.  ii  Table of Contents Abstract ...........................................................................................................................................ii Table of Contents ...........................................................................................................................iii List of Tables ..................................................................................................................................x List of Symbols and Abbreviations ..............................................................................................xii Acknowledgements .....................................................................................................................xiii Citation ..........................................................................................................................................xv 1.  Introduction: Making truth happen ...............................................................................1 1.1. Introduction: kâ-yôskâtahk ôma nêhiyawêwin..............................................................1 1.2. How does “truth happen” in Plains Cree? ....................................................................2 1.2.1 The “Calvin and Hobbes” problem in Plains Cree ........................................3 1.2.2 The “6-Degrees of Separation” problem in Plains Cree ............................…5 1.3. Proposal: Individuals related to perspectives ......................................................…....6 1.3.1. Perspective possession .............................................................................….6 1.3.2. Discourse Representation Theory with perspectives ................................…8 1.3.3. Proposal: Four kinds of referents in Plains Cree ....................................... 11 1.3.4. Proposal: Constructing obviation via referential dependency ................…16 1.4. Relation to previous work .....................................................................................….17 1.4.1. Models of subjective meaning ....................................................................17 1.4.2. Perspectives vs. possible worlds .................................................................19 1.5. The Plains Cree language ...........................................................................................21  1.5.1. Sources ..................................................................................................…21 1.5.2. Orthography ................................................................................................23 1.6. Outline of the thesis ...................................................................................................24 2.  The Form, Content, and Context of Animacy...............................................................26 2.1. Proposal …................................................................................................................26 2.2. Previous accounts .....................................................................................................30 2.2.1. Syntactic accounts of Plains Cree ...............................................................30 2.2.2. “Inanimate” as semantically specified ........................................................33 2.2.3. “Inanimate” as extentional ..........................................................................34 iii  2.2.4. Animacy is syntactically determined ..........................................................37 2.2.5. The data set........................................................................................…......38 2.3 The form of animacy: Distributed throughout the syntax .........................................38 2.3.1. Animacy is not specified in the nominal......................................................42 2.3.2. Animacy contrasts are not suppletive..........................................................44 2.3.3. The necessity of homophony ......................................................................46 2.3.4. Animacy coding does not determine the distribution of bare nouns...........49 2.3.5. Animacy coding is not asymmetric..............................................................52 2.3.51 Animacy without nominals.............................................................53 2.3.52 Animacy without verbs...................................................................56 2.3.6. Conclusion: Animacy is distributed.............................................................58 2.4. The content of animacy: “Inanimates” are inherently extentional .............................59 2.4.1. Asymmetries in the interpretation of the two classes .................................62 2.4.2. Abstract nominals are always “Inanimate”..................................................64 2.4.3. Extentionality in the verb system: “Inanimates” can’t think, speak, or feel......................66 2.4.31. Expletive arguments must be “inanimate” ...................................67 2.4.32. Psych verbs prohibit “inanimate” arguments................................69 2.4.33. Reflexive verbs prohibit “inanimate” arguments .........................72 2.4.4. Animates are unspecified ..........................................................................73 2.4.5. Only the “inanimate” form has specification ............................................75 2.4.6. “Inanimate” nominals are inherently extentional .....................................79 2.5. The context of animacy: Relativization to individual perspectives............................81 2.5.1. “Inanimates” as preconditions on embeddings ...........................................82 2.5.2. Relativizing to the Speaker: Changes in belief ...........................................84 2.5.3. Relativization to the Hearer: Effects of common ground ...........................86 2.5.4. Relativizing to third persons: Perspective shifts .........................................90 2.6. Conclusion .................................................................................................................97 3.  The Form, Content, and Context of Obviation.............................................................98 3.1. Proposal: Obviation as contextual extentionality ......................................................98 3.2. Previous accounts .....................................................................................................102 3.2.1. Obviation as discourse, not argument structure.........................................102 iv  3.2.2. Obviation as a construct, not a primitive...................................................105 3.2.3. “Obviative” is specified, “proximate” is contextually determined............105 3.2.4. The meaning of obviation.........................................................................107 3.2.5. The data set ...............................................................................................108 3.3. The form of obviation: no dedicated “obviative” marking .....................................108 3.3.1. The nominal suffix –a................................................................................110 3.3.11 The traditional view: –a marks “obviative”.................................111 3.3.12 The current analysis: –a marks extentional referents...................111 3.3.13 Modelling the “obviative” effects of –a.......................................114 3.3.2. The demonstrative suffix –ihi....................................................................115 3.3.21 The traditional view: –ihi marks “obviative”...............................117 3.3.22 The current analysis: –ihi marks extentional referents.................118 3.3.23 Modelling the “obviative” effects of –ihi.....................................120 3.3.3. The theme sign –ê–....................................................................................120 3.3.31 The traditional view: –ê– marks “third person interactions”........121 3.3.32 The current analysis: –ê– marks extentional objects....................122 3.3.33 Modelling the “obviative” effects of –ê–......................................126 3.3.4. The predicate suffix –im–..........................................................................127 3.3.41 The traditional view: –im– marks “obviative” objects.................128 3.3.42 The current analysis: –im– marks argument disjunction..............128 3.3.43 Modelling the “obviative” effects of –im–....................................133 3.3.5. The predicate suffix –yi–...........................................................................137 3.3.51. The traditional view: –yi– marks “obviative” arguments ...........139 3.3.52. The current analysis: –yi– marks disjoint subjects .....................139 3.3.53. Modelling the “obviative” effects of –yi–....................................144 3.3.6. Summary: The derivative nature of obviation.........................................148 3.4. The content of obviation: contextual extentionality ................................................148 3.4.1. Intentional verbs restricted “obviative” reference ....................................154 3.4.11. “Obviative” referents are restricted with psych verbs ................155 3.4.12. “Obviative” referents are restricted with speaking verbs ...........158 3.4.2. “Obviative” referents lack awareness .......................................................161 3.4.21. If contextual absence, referent is “obviative” .............................162 3.4.22. An “obviative” referent that is present is unaware .....................166 v  3.4.3. The emergence of “proximate” via contrast with “obviative” ..................168 3.4.31. “Obviative” as context for “proximate”.......................................169 3.4.32. Paradigmatic contrast of “animate” with “obviative” .................172 3.4.4. Summary: “Obviative” as contextual extentionality ................................173 3.5. The context of obviation: Speaker knowledge of intentions ...................................174 3.5.1. “Obviative” perspectives force Speaker dissociation ...............................178 3.5.2. “Obviative” perspectives force invented “proximate” ..............................179 3.5.3. Obviation patterns with indirect evidentiality ...........................................182 3.5.31. “Obviative” psych arguments and indirect knowledge ..............186 3.5.32. “Obviative” speakers and indirect knowledge ............................188 3.5.33. “Obviative” referents and the indirect evidential êsa .................196 3.6. Conclusion ...............................................................................................................198 4.  Referential dependency in Plains Cree........................................................................200 4.1. A structural and a semantic condition on referential dependency ...........................200 4.2. Previous work ..........................................................................................................204 4.2.1. Structural conditions on anaphora.............................................................204 4.2.2. Clausal dependency and referential dependency.......................................206 4.2.3. Possession and grammatical obviation .....................................................210 4.2.4. The data set................................................................................................211 4.3. Cross-predicate dependencies and obviation ...........................................................212 4.3.1. Nominal ordering and referential dependency ..........................................213 4.3.11. Nominal orderings in texts ..........................................................215 4.3.12. Elicitation and nominal-nominal ordering ..................................217 4.3.2. The suffix -yi- constructs referential dependency .....................................223 4.3.21. Structural conditions on -yi- ........................................................225 4.3.211. –yi– is c-commanded by its antecedent ........................226 4.3.212. –yi– is linearly preceded by its antecedent....................227 4.3.22. Semantic conditions on –yi– .......................................................231 4.3.221. The antecedent of –yi– must be in previous discourse..232 4.3.222. The antecedent of –yi– must be animate .......................233 4.3.223. The antecedent of –yi– must have a perspective ..........233 4.3.224. Propositional attitude verbs force –yi–..........................236 vi  4.3.23. Conclusion: -yi- constructs referential dependency ....................238 4.4. Predicate-internal dependencies and obviation ........................................................239 4.4.1 Possession as referential dependency ........................................................241 4.4.11. Structural subordination of possessums to possessors ................242 4.4.12. Perspectival embedding with possession ....................................245 4.4.121. “Inanimate” referents cannot be possessors ..................246 4.4.122. Possessors are always existential ..................................247 4.4.123. Possession blocks attitudes towards possessum .......... 249 4.4.124. Possession and speech act participants .........................251 4.4.125. Changing possession changes the Speaker....................254 4.4.13. Conclusion: Possession is referential dependence ......................258 4.4.2. Theme signs and referential dependency...................................................259 4.4.21. The direct theme sign –ê– constructs referential dependency.....262 4.4.211. Structural conditions on –ê– .........................................263 4.4.212. Semantic conditions on –ê– ..........................................265 4.4.2121. –ê– occurs with perspectivally-embedded objects .....................................................266 4.4.2122. –ê– cannot occur when the subject is not a perspective holder ................................272 4.4.213. Conclusion: –ê– constructs referential dependency .....276 4.4.22. The inverse theme –ikw constructs referential dependency ........277 4.4.221. Structural conditions on –ikw .......................................279 4.4.2211. Evidence for IP position of theme sign –ikw...279 4.4.2212. Evidence for structural relations of –ikw ........282 4.4.222. Semantic conditions on –ikw ........................................291 4.4.2221. Semantic restrictions on topic op ....................292 4.4.2222. Perspective-less properties of –ikw .................296 4.4.223. Conclusion: –ikw as referential dependency .................301 4.5. Conclusion ...............................................................................................................301 5.  Eliciting obviation and animacy .................................................................................303 5.1. The place of methodology in linguistic analysis.......................................................303 5.2. Fieldwork in linguistics.............................................................................................303 vii  5.3. Obviation and animacy data fluctuates in elicitation ...............................................304 5.4. A taxonomy of elicitation tasks and their effects on obviation and animacy ..........307 5.4.1. Translation tasks .......................................................................................307 5.4.11. Confounds in translation: Non-correlation and priming .............308 5.4.12. Obviation in translation tasks .....................................................309 5.4.13. Animacy in translation tasks .......................................................310 5.4.2. Judgment tasks ..........................................................................................312 5.4.21. Confounds in judgment tasks: Metalinguistics ...........................314 5.4.22. Judgment tasks and obviation .....................................................315 5.4.23. Animacy in judgment tasks .........................................................316 5.4.3. Supporting judgment tasks: Utterance-in-context tasks ...........................319 5.4.31. Obviation in utterance-in-context tasks ......................................321 5.4.32. Animacy in utterance-in-context tasks ........................................323 5.4.4. Correction tasks .......................................................................................324 5.4.41. Confounds in correction tasks: Un-noticed corrections ..............325 5.4.42. Obviation in correction tasks ......................................................328 5.4.43. Animacy in correction tasks ........................................................329 5.4.5. Analysis tasks ...........................................................................................331 5.4.51. Confounds: Decontextualizing forms, pseudo-linguistics ..........331 5.4.52. Obviation in analysis tasks .........................................................332 5.4.53. Animacy in analysis tasks ...........................................................333 5.5. Elicitation and impoverished context ......................................................................334 5.5.1. The proliferation of simple animates in elicitation ...................................335 5.5.2. Obtaining “obviative” forms in elicitation.................................................338 5.5.3. Contextless obviation = “Inanimate” ........................................................339 5.6. The consultant as linguist .........................................................................................341 5.6.1. The “experimentalist” position: Consultant as data bank .........................341 5.6.2. The “cooperative” position ......................................................................342 5.6.3. The two types of fieldwork and obviation data ........................................343 5.7. Conclusion: Elicitation as performance, not competence ........................................346 6.  Conclusion and Further Research................................................................................347 6.1. Conclusion: Constructing intentionality in Plains Cree ...........................................347 viii  6.2. Approaches to –a via the verbal suffix –i : Plurality and obviation ........................348 6.2.1. The “plural” reading .................................................................................350 6.2.2. The habitual reading .................................................................................350 6.2.3. The irrealis reading ...................................................................................351 6.2.4. Conclusion ................................................................................................352 6.3. –yi– across Cree........................................................................................................353 6.3.1. All Cree languages have a reflex of –yi–...................................................354 6.3.2. –yi– occurs with “obviative” possessors....................................................354 6.3.3. –yi– occurs with intransitive subjects........................................................355 6.3.4. –yi– occurs with transitive subjects...........................................................355 6.3.5. –yi– is insensitive to animacy....................................................................357 6.4. A comparison of the Plains Cree system to Athabaskan..........................................359 6.4.1. Athabaskan yi-/bi- are pronominals, Plains Cree’s are not........................362 6.4.2. Plains Cree –yi– and Athabaskan yi- code disjoint reference……............364 6.4.3. Plains Cree –yi– and Athabaskan yi- differ in transitivity.........................366 6.4.4. A and A’ dependencies. ............................................................................367 Glossary ....................................................................................................................................370 References .................................................................................................................................377  ix  List of Tables Table 1.1. Two conditions on referential dependency ..................................................................17 Table 1.2. The forms of obviation and their dependencies ...........................................................17 Table 2.1. Summary of animacy coding for verb, demonstrative, and nominal .......................... 39 Table 2.2. Summary of animacy shifting for the Severed Head story ......................................... 43 Table 2.3. Forms commonly found in both animacy classes ....................................................... 46 Table 2.4. Summary of predictions for three hypotheses ............................................................ 59 Table 2.5. Semantic gaps among verbs ........................................................................................ 70 Table 2.6. A summary of “inanimate”/“animate” shifting for the severed head story ................ 91 Table 3.1. No dedicated obviation forms ..................................................................................... 99 Table 3.2. “Inanimate” plural and “obviative” across Algonquian ........................................... 113 Table 3.3. Canonical demonstrative organization in Plains Cree ............................................. 115 Table 3.4. Demonstrative organization of S2 ............................................................................ 116 Table 3.5. “Inanimate” plural and “obviative” demonstratives across Algonquian .................. 119 Table 3.6. Independent order distribution of -ê- ........................................................................ 125 Table 3.7. No -ê- in conjunct order ............................................................................................ 125 Table 3.8. Distribution of -im- with “obviative” arguments ...................................................... 127 Table 3.9. Distribution of –yi– with “obviative” arguments ...................................................... 139 Table 3.10. Distribution of –yi– with “inanimate” arguments ................................................... 140 Table 3.11. “Obviative” forms and their patterning .................................................................. 148 Table 3.12. Summary for psych verbs built with –êyim/–êyiht/–êyimo ..................................... 157 Table 3.13. Summary of other psych verb forms ........................................................................157 Table 3.14. Summary of preverb forms ......................................................................................157 Table 3.15. A Summary of psych verbs in 3 texts ......................................................................158 Table 3.16. Speaking verbs and obviation ..................................................................................159 Table 3.17. Intransitive itwê- and obviation .............................................................................. 160 Table 3.18. “Obviative” referents and speaking in a text ...........................................................160 Table 3.19. “Obviative” reference shifts when topic event shifts ...............................................163 Table 3.20. Obviated animal in a story .......................................................................................170 Table 3.21. Chart of obviation for modified re-telling .............................................................. 171 Table 3.22. Categorization of “obviative” with psych verbs ......................................................186 Table 3.23. Transitive it- and obviation ......................................................................................190 x  Table 3.24. Transitive wîhtamaw- and obviation ........................................................................191 Table 3.25. Transitive speaking and obviation in a text ............................................................ 192 Table 3.26. “Obviative” referents speaking in a text ..................................................................194 Table 3.27. Obviatiation in an êsa evidential span .....................................................................197 Table 4.1. “Obviative” constructions meet the conditions of referential dependency ................204 Table 4.2. Nominal-nominal ordering meets the conditions of referential dependency .............214 Table 4.3. “Proximate” referring nouns precede “obviatives” in texts .......................................216 Table 4.4. Nominal-nominal ordering patterns in elicitation ......................................................218 Table 4.5. The suffix –yi– meets the conditions of referential dependency ...............................225 Table 4.6. Possession meets the conditions of referential dependency ......................................242 Table 4.7. The inverse meets the conditions of referential dependency .....................................278 Table 4.8. Restrictions on the identity of topic op in Plains Cree ..............................................292 Table 4.9. Restrictions on the identity of the argument introduced by –ikw ..............................296 Table 4.10. Topic operator and Inverse pro are opposites in their properties ............................301 Table 6.1. A summary of the occurrences of –i ..........................................................................349 Table 6.2. A summary of contexts for plurality ..........................................................................353 Table 6.3. Navajo, Dogrib and Dëne Sųłiné pronominal affixes ................................................359 Table 6.4. Comparing Plains Cree –ikw and –yi– to Athabascan yi- ..........................................361 Table G1. Summary of verbal morphology ................................................................................374  xi  List of Symbols and Abbreviations xy x↔y x∧y ≈ ≠ ∀ ∃ λ ψ  = = = = = = = = =  if x, then y iff (if and only if) x, then y x and y approximately does not equal universal quantifier existential quantifier lambda operator perspective  0 1 1>3  = = =  2 3 3>1  = = =  inanimate agreement first person first person acting on third person second person third person third person acting on first person animate intransitive animate intransitive-transitive actor animate applicative changed conjunct 1 changed conjunct 2 clause phrase demonstrative phrase diminutive direct determiner phrase discourse representation structure discourse representation theory disjoint subject disjoint marker distal epenthetic exclamation extentional extentional phrase  AI AIT ACT AN APPLIC C1 C2 CP DEMP DIM DIR DP DRS  = = = = = = = = = = = = =  DRT  =  DS DSJ DST EP EXCL EXT EXTP  = = = = = = =  FUT GEN HABIT HES II IN INCH INTERJ INT INV IP LOC LP NEG NOM NP  nP NUMP OBV PERSP PL PREV PROX PRX R RECIP RED RESUM RFLX RR SG STAT TA TI UP VP vP  x XP XT  y  = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =  future generic habitual hesitation inanimate intransitive inanimate inchoative interjection intentional inverse inflectional phrase locative local person negation nominalizer noun phrase little noun phrase number phrase obviative person phrase plural previous proximate proximal relation reciprocal reduplication resumptive reflexive relative root singular stative transitive animate transitive inanimate unknown possessor verb phrase little verb phrase variable unidentified phrase extentional variable  xii  Acknowledgements First and foremost, I thank the Plains Cree and Métis people who have contributed their linguistic knowledge to this thesis. This includes Wally Awâsis, who showed up to our first session in an Cleveland Indians jacket (“I’m an Indian – they made the jacket just for me”), Rita Daniels, whose kindness won’t be forgotten, Joseph Deschamps, who has repeatedly provided crucial insights with his knowledge of Cree culture and language, Josephine Small, whose linguistic flexibility and patience has provided many crucial data points, and Dorothy Thunder, who gave me many good questions to ask about the language. Toni Cardinal’s exceptional linguistic abilities were crucial to the construction of the generalizations that are the backbone of this work. Repeatedly, and over many years of work, she turned her intuitive understanding of the Plains Cree language into clear, concise descriptions that far exceeded any analytic work done on these aspects of her language. In addition, I would also like to thank speakers and community members, including Beverly Crier, Judy Louis, Roy Louis, Barbara MacLeod, Tom McCallum, Jean Okimâsis, Maryann Palmer, Solomon Ratt, Henry Rain, Wayne Roan, Blanche Steinhauer, and Simon Threefingers. I know it is hard to listen to ignorance speaking with unearned authority, and I appreciate your forbearance and compassion. kinânaskomitinâwâw. My supervisor, Rose-Marie Déchaine, is the reason I survived UBC academics, the expert hand behind all my successful grants and any papers you liked, and the main reason my thesis is at all coherently written (You readers think this is bad – you should have seen Draft #2!). From the admissions process until I filed this thesis, I cannot think of any stage that she was not involved in. I am greatly indebted to her for her time, her energy, and her ideas. The other two members of my committee, Guy Carden and Hotze Rullmann have also been of central importance to this work. Guy is one of the great linguists, and his sharp questions and encouragement have been much enjoyed. I am grateful that he took time for me. Hotze has suffered my semantic excesses (and their ensuing chaos) more cruelly than anyone, and any clarity in the final semantic discussion is to his credit. He has given me significant amounts of his time, carefully reading many drafts and patiently trying to pull me from the mud. I would like to thank the four additional members of my examining committee: Andrew Irvine, Katarzyna Jaszczolt, Leslie Saxon, and Geoff Williams for their voluntary and insightful contributions to this work. I would especially like to thank Kasia and Leslie for going through my thesis in such fine detail, and the final outcome has been much improved by their ideas. In addition, Tamina Stephenson was a silent member of my examining committee, and her open mind and flexibility provided me with much help when it was most needed. H.C. Wolfart has suffered through both my theoretical and editorial contortions with much patience (Chris: Please don’t look at the references to this thesis). His scholarship, fueled by a sincere interest in the language, has provided an excellent role-model for what linguistic work on a language ought to be and how it ought to proceed, and his academic manner represents the best standard of what I had hoped Academia would be like. I owe Monica Macaulay a great thank you for introducing me to Algonquian languages. Out of her own good will and entirely on her own time, she spent many hours working through texts, discussing Bloomfield’s grammar, and going through teaching materials, as well as literally driving me up to the reservation. She also gave me much advice about graduate school, and introduced me to fieldwork methodology via Menominee and isiZulu. Thanks to those who shared their language knowledge with me, including Oládiípò Ajíbóyè (Yorùbá), Beatrice and Melinda Bullshields (Blackfoot), Adriano Barbosa (Brazilian Portuguese), Seok Koon Chin (Mandarin, Cantonese, Hokkien), Kerim Demirci (Turkish), Peter Jacobs (Squamish), Bekisizwe Ndimande (isiZulu), and Lena Russell (Blackfoot). xiii  This work has been supported by the Jacobs Fund, Phillips Fund, UBC, and SSHRC. Without Edna Dharmaratne, nothing would have gotten done, I would have never gotten any money, and I would have starved in the rain. All that, and she’s a nice person. Thanks to my teachers at UBC, including Gunnar Hansson, Lisa Matthewson, Doug Pulleyblank, and Martina Wiltschko. Martina’s teaching helped clarify a great deal of syntactic theory to me, and Lisa’s semantic instruction and advising has been of great assistance. Thanks also to the Algonquian scholarly community, including Eleanor Blain, Phil Branigan, Amy Dahlstrom, Inge Genee, Ives Goddard, Marguerite MacKenzie, Jean Okimâsis, Arden Ogg, David Pentland, Marie Odile-Junker, Nicole Rosen, and Arok Wolvengrey. They have sat through my talks, shared their work, offered me help, and have still somehow not given up hope. I am fortunate to have worked with so many talented people. I would like to thank the scholars of ancient languages at the University of Wisconsin, Jack McKeown, Victoria Pagan, Barry Powell, Richard Ringler, and Andrew Wolpert, who gave me a clear picture of the first and best model of linguistic scholarship, and much laughter. Thanks also to my linguistics teachers at the University of Wisconsin, including MarieHelene Côte, Mürvet Enc, Yafei Li, Matt Pearson, and Manindra Verma. In particular, I have vivid memories of Matt’s brutal tests and Mürvet’s fantastic rants (and fashion sense!). Thanks to my teachers at the University of Wisconsin extension in Marathon County, Harlan Grinde, Theo Koupelis, Keith Montgomery, and Peter Okray, who spent a great deal of time and energy teaching me, and introduced me to ideas that still inform my perspective. Thanks to the graduate student group for putting up with me: Oládiípò Ajíbóyè, Solveiga Armoskaite, Leszek Barczak, Fiona Campbell, Mario Chavez-Peon, Yunhee Chung, Ramona McDowell, Atsushi Fujimori, Diana Gibraiel, Carrie Gillon, Fusheini Hudu, Yoko Ikegami, Peter Jacobs, Masaru Kiyota, Karsten Koch, John Lyon, Calisto Mudzingwa, Nahal Namdaran, Jeremy Perkins, Sugunya Ruangjaroon, Scott Shank, Tanya Slavin, Dominique Quis, Ian Wilson, Rachel Wojdak, and Florence Woo. I share your pain; I’m either glad you’re done or hope it will be over soon. Ryan Waldie told me about fairies in linguistics and botanical varieties in Vancouver, while also fixing my busted semantics. Better in a gutter than on a pedestal. Chin Seok Koon has helped the teacher of thieves ( ) out of many of his worst scrapes, and has gracefully endured swashbuckling and barbarity from the Western Pirates. Many people showed me hospitality and kindness through these years, including Solveiga Armoskaite and family, Joe and Nella Cook, Rita Daniels, Denis Déchaine and family, Lucienne Déchaine, Rose-Marie Déchaine, Joseph Deschamps, the Demirci family, Tom Nook, Poncho, Paola Quintanar, Totakeke, and Juliana Wolfart. For still being willing to admit my acquaintance after 6 years of insanity, I would like to thank Kerim Demirci and David Tang. Kerim provided me with the only real break I had during this work, and David has somehow managed to eat most of my ice cream. I owe my interests and abilities to my parents, Gerald and Suzanne Muehlbauer, who have steadily tried to support my intellectual interests since I was in three-cornered pants. Topics have shifted from dinosaurs to languages, but I’m not sure much else has changed. Thanks also to the rest of my family, including Alma Johnson, my grandparents Robert and Elizabeth Muehlbauer, Kathy Ritenour, who has lavished attention on me all my life, Jane “Pufferfish” Ritenour, Ned and Debbie Wicker (and feline and canine housemates), who provided much encouragement over many years, and the Muehlbauer and Klann clans. Thanks for not being academics, and for trying to understand why on earth someone would do such a thing. (Answer: I still don’t know!) Finally, there’s Clare. You never can tell with porcupines, can you? xiv  Citation ‘Most men will not swim before they are able to.’ Is not that witty? Naturally, they won’t swim! They are born for the solid earth, not for the water. And naturally they won’t think. They are made for life, not for thought. Yes, and he who thinks, what’s more, he who makes thought his business, he may go far in it, but he has bartered the solid earth for the water all the same, and one day he will drown.” – Hermann Hesse Steppenwolf  xv  Chapter 1 Introduction: Making truth happen 1.1. Introduction: kâ-yôskâtahk ôma nêhiyawêwin When I asked a Plains Cree (Algonquian, Northern Plains) speaker about the linguistic structures considered in this thesis, I was told ‘kâ-yôskâtahk ôma nêhiyawêwin,’ literally ‘The Plains Cree language is soft.’1 (1)  ᑳ ᔫᐢᑳᑖᕁ ᐆᒪ ᓀᐦᐃᔭᐍᐏᐣ. kâ-yôskâtahk ôma nêhiyawêwin. kâ-yôsk=ât =an-k aw =ima nêhiyaw=ê =win C2-soft =by.air=II -0 PRX=IN.SG cree =AI=NOM ‘The Plains Cree language is soft.’2  Here, the use of the concept of ‘soft’ (yôsk-) is meant to convey a particular philosophy about the purpose of communication and the way it ought to proceed. As Lightning (1996:62) explains it, the Speaker makes themselves “vulnerable and open” to their audience, not trying to impress their assertions on the audience by force of personality or argument. Rather than speaking to convince (cf. Aristotle’s ‘Rhetoric’), a Plains Cree speaker’s communicative intent is understood as coming from a need to express what they believe is the truth, individualized to the particular audience they are speaking to (Lightning 1996:63). The mechanisms of discourse, then, are organized to set up the proper events for “truth to happen” (Lightning 1996:63). This means that accuracy in representation, both of the Speaker’s beliefs and the beliefs they convey from others, is absolutely crucial to the Speaker’s goals for the discourse; the Hearer(s) must trust the Speaker, and the Speaker must be trustworthy. The Plains Cree language is ‘soft,’ then, because its users prize the “mutual-thinking” (Lightning 1996) that develops between Speaker and Hearer through careful representation, and they value this over the force of logical or charismatic persuasion. This means that, to a speaker of Plains Cree, the grammatical material covered in this thesis relates to how the Speaker “makes truth happen.”  1 2  Thanks to Joseph Deschamps, ᐅᑭᒫᐤ ᑳᐹᐏᐤ, for this form and discussion of its meaning. ᑭᓇᓈᐢᑯᒥᑎᐣ. Thanks to H.C. Wolfart (p.c.) for help with the composition of this form.  1  1.2. How does “truth happen” in Plains Cree? This thesis considers the construction and maintenance of “truth happenings” in Plain Cree. By “truth happenings,” I mean the way that a speaker of Plains Cree ties together some proposition (e.g. a sentence) with a means to evaluate it. As is demonstrated in this thesis, Plains Cree possesses a rich set of grammatical mechanisms that the Speaker can manipulate to construct these representations of truth. I claim that there are two basic ingredients to the “truth happening” process in Plains Cree: (i)  A proposition  (ii)  A thinker.  That is, truth “happens” when a proposition is held by someone; without this “someone” there can be no truth. Making truth happen amounts to connecting propositions to thinkers. Truth, then, is the result of relativizing a proposition to a person; it is inherently relational or individualized in the sense of Lightning (1996). If “truth happening” is the process of connecting a proposition to a thinker, we expect that Plains Cree grammar will be concerned with the identification of thinkers, and the linking of propositions to these thinkers. In particular, we expect two kinds of operations to be at play: (i)  Identifying thinkers  (ii)  Coding the relation of these thinkers to propositions.  The first process (identifying thinkers) is the primary of the two; it does not matter what the relation of a thinker is to a proposition if there is not yet a thinker. In studying the process of “truth happening” in Plains Cree, then, we should begin by looking at mechanisms that are used to pick out thinkers. This is the core topic of this thesis. To illustrate these two operations in Plains Cree, let us consider two limited cases of “truth happening.” In the first case, the “Calvin and Hobbes”3 problem (§1.2.1), we see two things: (i)  There are grammatical forms that distinguish referents that can “think” from referents that can’t.  (ii)  These grammatical forms can only be interpreted once a decision has been made about who thinks that the referent can think.  3  The characters and comic strip “Calvin and Hobbes” are copyright Bill Watterson. Its use here is for scholarly and illustrative purposes only, and is thus allowed under “Fair Use” copyright laws.  2  Without this relativization, the grammatical terms still have only the status of propositional elements. In the second case, the “6 degrees of separation” problem (§1.2.2), we see two things: (i)  There are grammatical forms that distinguish which referents are thinking in the given context.  (ii)  These grammatical forms are always relativized to what the Speaker knows.  Plains Cree grammar, then, possesses a rich set of devices for discriminating those who can make “truth happen” from those who can’t. The rest of the thesis is concerned with the proper description and modeling of these phenomena.  1.2.1 The “Calvin and Hobbes” problem in Plains Cree Anyone that has been to a book store or read a newspaper in the last 20 years will be familiar with the comic strip “Calvin and Hobbes” by Bill Watterson. Calvin is a young boy who has a best friend that is a tiger named Hobbes. Being anthropomorphic, Hobbes can walk and talk just like a human, and he goes on many adventures with Calvin. However, at other times in the comic strip, Hobbes is simply an inert stuffed animal. This means that there are two versions of Hobbes presented in the comic strip: (i)  Hobbes is an anthropomorphic tiger.  (ii)  Hobbes is a stuffed toy.  This I term the “Calvin and Hobbes” problem – how are we to understand the changing status of Hobbes? In the Plains Cree language, there are two grammatical classes for nominals, let us call them “A” and “B” for now, and these two classes map directly onto the Calvin and Hobbes problem. Consider a context in which Calvin has lost Hobbes out in the woods, and his father went out to look for him. After finding Hobbes, the stuffed animal, out behind some bushes, the father puts it in bed with Calvin. Upon waking, Calvin runs to inform his mother that Hobbes came back from his evening wanderings. In responding, the mother must either mark Hobbes with grammatical category “A” (2a) or grammatical category “B” (2b).4  4  This context is an actual comic strip by Bill Watterson, and was used to elicit the Plains Cree forms.  3  (2)  a. CATEGORY A ᑯᐦᑖᐏᐩ ᑮ ᒥᐢᑲᑦ Hobbes. kohtâwiy kî-miskam Hobbes. k-ohtâwiy kî- m =isk -am Hobbes 2-father PREV-find=by.body.TI-TI Hobbes ‘Your fatherAN found HobbesIN.’  (Presented S2)  b. CATEGORY B ᑯᐦᑖᐏᐩ ᑮ ᒥᐢᑲᐍᐤ Hobbesᐊ. kôhtâwiy miskawêw Hobbesa. k-ohtâwiy m =iskaw -ê -w Hobbes-a 2-father find=by.body.TA-DIR-3 Hobbes-XT5 ‘Your fatherPROX found HobbesOBV.’  (Presented S2)  If the mother marks Hobbes with grammatical category “A,” which is shown by the –isk–am set of suffixes on the verb in (2a), she is saying that Hobbes is a stuffed animal. If she instead chooses grammatical category “B,” which is shown by the combination of the –iskaw–â–w suffixes on the verb and the –a on the noun (2b), she is saying that Hobbes is a walking, talking tiger. For example, if the “A” form is used, subsequent discourse cannot allow Hobbes to speak, walk, or believe anything (e.g. being happy about being found) (3a), while the “B” form can allow Hobbes to talk, walk, and believe things (e.g. being happy about being found) (3b). (3)  a. CATEGORY A # ᐁᐦᐋ, ᑯᐦᑖᐏᐩ ᑮ ᒥᐢᑲᑦ Hobbes; ᐁ ᒥᐩᐍᔨᐦᑕᒥᕁ. # êhâ, kôhtâwiy kî-miskam Hobbes; ê-miywêyihtahk. êhâ k-ôhtâwiy kîmi =sk -am Hobbes ê- miyw=êyiht -am-k yes 2-father PREV-find=by.body.TI-TI Hobbes C1-good=by.mind.TI-TI -3 Intended: ‘Yes, your fatherAN found HobbesIN; [Hobbes] was happy.’ (Presented S2) b. CATEGORY B ᐁᐦᐋ, ᑯᐦᑖᐏᐩ ᑮ ᒥᐢᑲᐍᐤ Hobbesᐊ; ᐁ ᒥᐩᐍᔨᐦᑕᒥᔨᐟ. êhâ, kôhtâwiy kî-miskawêw Hobbesa; ê-miywêyihtamiyit. êhâ k-ôhtâwiy kîmi =skaw -ê -w Hobbes-a ê- miyw=êyiht -am-yi-t yes 2-father PREV-find=by.body.TA-DIR-3 Hobbes-XT c1-good=by.mind.TI-TI-DS-3 ‘Yes, your fatherPROX found HobbesOBV; heOBV was happy.’ (Presented S2)  In this situation, then, the mother must choose between representing her own belief about Hobbes and that of her child; does she use the “A” form, which corresponds to what she herself sees, or does she use the “B” form, which corresponds to what her son thinks of his tiger?  5  For an explanation of ‘extentional’ (XT), see §1.3.1. below.  4  The Plains Cree grammatical categories “A” and “B,” then, do two important things relative to the current discussion: (i)  They discriminate between non-thinkers (“A”) and potential thinkers (“B”).  (ii)  They require that someone think they are non-thinkers or thinkers.  As such, these two grammatical categories are intimately involved in making “truth happen” in Plains Cree. These two categories are the focus of Chapter 2.  1.2.2 The “6-Degrees of Separation” problem in Plains Cree In considering our relation to famous or important people, we sometimes count in terms of our relation to people who have been related to famous people. For example, my uncle once wrestled Hulk Hogan. Counting from The Hulk to me, I can say that I am separated from him by one degree of separation (i.e. my intermediate Uncle Ned). Of course, I can’t say that I “know” The Hulk – only that I know someone who knew him. This game of counting intermediate people has sometimes been done as a logic puzzle – as in the hobby of connecting actors to other actors until one of them has starred in a movie with Kevin Bacon (i.e. the “Six Degrees of Separation from Kevin Bacon” game). Suppose that I am speaking Plains Cree, and I want to tell you something that Hulk Hogan said to my Uncle. In this language, I have a choice of three ways to present this (4). (4) a. ᐁᑯᓯ ᐃᑘᐤ ᓈᐯᐤ. êkosi itwêw nâpêw êkosi it =wê-w nâpêw so thus=AI -3 man ‘That’s what the manAN said’  (Volunteered S4, Presented S2,S3)  b. ᐁᑯᓯ ᐁ ᐃᑘᐟ ᓈᐯᐤ. êkosi ê-itwêt nâpêw êkosi ê- it =wê-t nâpêw so C1-thus=AI -3 man ‘That’s what the manAN said’  (Volunteered S4, Presented S2,S3)  c. ᐁᑯᓯ ᐁ ᐃᑘᔨᐟ (ᐁᓴ) ᓈᐯᐘ. êkosi ê-itwêyit (êsa) nâpêwa êkosi ê- it =wê-yi -t êsa nâpêw-a so C1-thus=AI -DS-3 EVID man -XT ‘That’s what the manOBV said’  (Volunteered S4, Presented S2,S3)  5  While the form in (4a) indicates that I heard the Hulk myself, and the form in (4b) is noncommittal on how I heard about this, the final form, in (4c), is what I would say if I wanted to make it clear that I wasn’t there to hear Hulk Hogan tell this – it was something he said to someone else, that I’m passing on. By marking the noun nâpêw ‘man’ with the suffix –a, and putting the suffix –yi– in the verb, I am telling you that I don’t know this Hulk Hogan fellow personally, and thus I can’t vouch for this directly. You’d have to go check with my Uncle to make sure that this was really what The Hulk said. Thus, Plains Cree has a way to do this “degree-counting” right in its grammar, by specially-marking the noun and verbs that relate to it.  1.3. Proposal: Individuals related to perspectives In this thesis, I attempt to model “truth happening” in Plains Cree. In doing this, it is important to remember the distinction between a phenomenon (Greek: φαινóµενον ‘that which appears’) and the model of the phenomenon. I am not claiming that the model I am constructing is, in any sense, ‘real.’ Rather, it is a constructed representation that covers the observable facts and gives us a systematic way to look for new facts to model. In other words, the model is a map, not a territory (cf. Korzybski 1958, Bateson 1971).  1.3.1. Perspective possession In this thesis, I take “truth happening” to be the relation of individuals to propositions. This is an inherently relativized process; the truth of a proposition is relative to some individual. In particular, the truth of a proposition is relative to the individual’s perspective (in the sense of Kölbel 2002) – the way that individual sees the world. The process of “truth-happening,” then, is the process of mapping a proposition into some individual’s perspective. Perspectives can be thought of as analogous to other possessed entities (Kölbel 2002); each individual owns one, and uses it to evaluate the truth of propositions. In a sense, it is like a body part that the individual takes with them wherever they go. Parallel to possession, then, we can define a perspective and an individual as being related by possessor relation ‘R’ (Higginbotham 1983). (5)  a. [[John’s dog]] = R(John,dog) = There is a relation between John and the dog. b. R(x,ψ) = There is a relation between an individual x and a perspective ψ.  6  In most cases, the relation between the possessor and the perspective is unspecified, analogous to normal possession. However, there are ways to further restrict this relation, by introducing predicates that explicitly restrict this unspecified ‘R’ relation, just as with kinship terms (Burton 1995). With perspectives, these predicates are typically called “propositional attitude” predicates (Russell 1918), and comprise predicates like “think,” “feel,” or “say.” (6) a. [[John’s mother]] = mother-of(x,John) = There is a mother relation between John and x.  (Burton 1995)  b. [[John thinks]] = think(John, ψ) = There is a think relation between John and ψ. A perspective, then, is a domain that an individual possesses with which to evaluate the truth of a proposition. If the ‘R’ relation of perspective possession is further specified by the addition of predicates, we could expect a language to grammatically separate perspective possession from the relation to the perspective. In fact, in Plains Cree, many of these propositional attitude predicates are built off of one element – the suffix –êyiht ‘by mind.’ Thus, to ‘believe’ is tâpwêwakêyihtam ‘to hold it true in the mind,’ while to ‘think’ is itêyihtam ‘to do thus with it in the mind.’ (7)  a. ᑖᐻᐘᑫᔨᐦᑕᑦ. tâpwêwakêyihtam. tâpwêwak=êyiht -am true =by.mind.TI-TI ‘S/heAN believes in itIN, holds it to be true.’ b. ᐃᑌᔨᐦᑕᑦ. itêyihtam. it =êyiht -am thus=by.mind.TI-TI ‘S/heAN thinks thus of itIN.’  Plains Cree, then, could be thought of as coding the having of a perspective (-êyiht) and the relation to the perspective (e.g. tâpwê- ‘true’) as separate linguistic elements, which crucially exist in an structurally-conditioned, asymmetric relation to the holder of perspective.6 The concept of perspective possession provides a way to model this pattern. 6  This asymmetry presents challenges for accounts that treat the perspective holder, the perspective, and the relation to the perspective as an undifferentiated triple (e.g. Kamp 1990). While the current account does not provide a full  7  The definition of perspective possession adopted here gives rise to two kinds of individuals: (i)  Those that hold a perspective,  (ii)  Those that don’t.  For perspective-holders, I use the term “intentional,” because they have “intentions,” which means that they have mental processes directed at something (Brentano 1874). For individuals that do not possess a perspective, I adopt the term “extentional,” since they do not have mental processes directed at anything.7 “Intentionality,” then, is the property of possessing a perspective.  1.3.2. Discourse Representation Theory with perspectives To model the contexts for perspective possession in Plains Cree, I employ a modified form of Discourse Representation Theory (DRT: Kamp 1981) that is built off of the work of Farkas (1992) and Smith (2004). In this model, each Discourse Representation Structure (DRS: symbolized by a ‘box’ in the representation) represents a perspective (the domain in which the truth of a proposition is evaluated). This perspective is necessarily ‘anchored’ to an individual (Fillmore 1971, Ruwet 1982, Kölbel 2002, Lasersohn 2005, Stephenson 2007), creating a perspective-individual pair (e.g. R(x,ψ)). This perspective-individual pair is represented in the DRS through the following conventions: (i)  The perspective is represented by the ‘box’ of the DRS itself.  (ii)  The individual that possesses this perspective is marked above the upper-left corner of the box.  (iii)  The relation between the individual and the perspective is represented by the predicate in angle brackets to the right of the individual. x <R>  (8)  = R(x,ψ)  model of this asymmetry, it is amenable to future work that considers the compositional nature of propositional attitudes. 7 Thanks to Hotze Rullmann (p.c.) for suggesting this term.  8  Elements embedded within a perspective are thus placed inside the corresponding ‘box’ of the DRS, and variables related to this perspective are introduced in the frame at the top of each ‘box.’ (9)  Jeff: The sun feels good.8 Jeff <say> x  = say(Jeff,ψ)  sun(x) feel.good(x)  The relation between the individual and proposition is neutrally unspecified If the Speaker wishes to represent the perspective of some other individual, a second layer of embedding can be added.9 (10)  Clare: Jeff thinks the sun feels good. Clare <say> x Jeff  = say(Clare,ψ)  sun(x) think(Jeff) Jeff <think> x  = think(Jeff,ψ')  sun(x) feel.good(x)  Of course, the DRS in (10) is not a complete representation of the utterance in (10); I have abstracted away from places and times. For the purposes of understanding the reference-typing 8  Regarding the argument structure of propositional attitude verbs, Cook (2008) provides evidence that propositions (syntactically CPs) are never the objects of verbs. This is the reason for the treatment of predicates like say in the formalism. 9 Although the implementation here is similar in method to that considered by Kamp (1990), which would model “intentional” as an “external anchor” and “extentional” as an “internal anchor,” the two accounts differ with respect to their ability to model multiple layers (>2) of embedded perspectives, which is crucial for the Plains Cree data considered in Chapter 3.  9  phenomena in Plains Cree, there does not appear to be a need to distinguish between times and places, although a full model of Plains Cree perspectival meanings will necessarily require their inclusion. In the study of obviation phenomena (Chapter 3), I employ the context variable C of Partee (1989), which can be conceived of as a collapsing of place and time together. This is outlined in section 1.3.3 below. The use of a DRS to represent a perspective implicates several alterations to the standard DRT developed by Kamp (1981). In particular, the typical DRT convention of carrying referents into embedded boxes cannot here be maintained; to do so would indicate that any referent known to the Speaker would also be known to the possessor of embedded perspectives. Altering this convention allows for the modeling of differences in perspectives regarding referents. For example, the “sun” in the perspective of Clare in (10) is repeated in the embedded perspective of Jeff. This represents that both Clare and Jeff have the same referent in mind. If the variable of the embedded perspective were switched (e.g. to y), then Clare and Jeff would have different versions of the “sun” in mind, as schematized in (11). (11)  Clare: Jeff thinks the sun feels good. Clare <say> x Jeff  = say(Clare,ψ)  sun(x) think(Jeff) Jeff <think> y  = think(Jeff,ψ')  sun(y) feel.good(x)  The ban on referents carrying into embedded DRS structures also has implications for existential quantification. In typical DRT frameworks, the referent is taken to exist within the DRS it is located in. In the current framework, this existence only holds of a perspective. For example, consider a context in which Clare and Jeff were walking in the woods together and they saw a  10  large creature run by. Jeff claims this creature was a hodag,10 while Clare thinks it was merely an exceptionally large dog. We can model this as in (12), where the predicate ‘large’ is located in both perspectives, while the predicate ‘hodag’ is located only in the perspective possessed by Jeff. (12)  Clare: Jeff thinks he saw a large hodag. Clare <say> x Jeff  = say(Clare,ψ)  think(Jeff) large(x) Jeff <think> x  = think(Jeff,ψ')  see(Jeff,x) hodag(x) large(x)  This represents that Clare is only committing to the referent’s bigness, while the idea that it was a hodag is entirely Jeff’s. Within this model, “Truth-happening” is understood as the process of connecting a proposition to a perspective-individual pairing. This connection provides the means for an individual to evaluate the proposition.  1.3.3. Proposal: Four kinds of referents in Plains Cree I propose that Plains Cree can be understood to refer to four kinds of individuals, defined in terms of perspectival possession (i.e. intentionality): (i)  Individuals that can never possess a perspective.  (ii)  Individuals that cannot currently possess a perspective.  (iii)  Individuals that currently possess a perspective.  (iv)  Individuals unspecified for perspective possession.  10  A ‘hodag’ is a legendary beast that is thought to live in Northern Wisconsin woodland areas, particularly around Rhinelander, Eagle River, Minocqua, and Hayward.  11  The first kind of individuals, those that can never possess a perspective, correspond to the grammatical category used to mark Hobbes as a stuffed animal in example (2). This category is termed “inanimate” in the literature (cf. Wolfart 1973). (13)  CATEGORY (I) = INANIMATE ᑯᐦᑖᐏᐩ ᑮ ᒥᐢᑲᑦ Hobbes. kohtâwiy kî-miskam Hobbes. k-ohtâwiy kî- m =isk -am Hobbes 2-father PREV-find=by.body.TI-TI Hobbes ‘Your fatherAN found HobbesIN.’  (Presented S2)  Here, the verb is coded with two forms that pick out individuals that never possess perspectives: (i)  The suffix –isk ‘done by the body to an inanimate thing’  (ii)  The suffix –am ‘inanimate object’  I propose that this category marks individuals that are never able to possess a perspective; they are inherently extentional. This is represented in the formalism as in (14). (14)  EXT(x) ↔ ψ y R(y,ψ)  x ≠ y) x is Extentional if and only if for all Perspectives ψ and all individuals y, if there is a relation R between individual y with perspective ψ, then x is not y.  As such, referents marked with this category will be unable to speak, think, or feel, since those are predicates that introduce a relation between a perspective and an individual (§1.3.1). Modeling the difference in who believes the referent is extentional thus reduces to a differential embedding of the extentional-marking predicate. (15)  ᑯᐦᑖᐏᐩ ᑮ ᒥᐢᑲᑦ Hobbes. kohtâwiy kî-miskam Hobbes. k-ohtâwiy kî- m =isk -am Hobbes 2-father PREV-find=by.body.TI-TI Hobbes ‘Your fatherAN found HobbesIN.’  (Presented S2)  12  a. CONTEXT 1: EXTENTIONAL W.R.T. SPEAKER11 Speaker <R> x Hobbes father(x,Hearer) EXT(Hobbes) find(x,Hobbes)  b. CONTEXT 2: EXTENTIONAL W.R.T. HEARER Speaker <R> x Hobbes father(x,Hearer) find(x,Hobbes)  Hearer <R> Hobbes EXT(Hobbes)  Here, the location of the EXT predicate in the perspective of one or the other perspective represents who conceives of Hobbes as extentional. This category is considered in detail in Chapter 2. The second kind of individuals, those that are currently unable to possess a perspective, are signified by the structure used to mark nâpêw ‘the man’ in example (4c), termed “obviative” in the literature (cf. Wolfart 1973). (16)  CATEGORY (II) = OBVIATIVE ᐁᑯᓯ ᐁ ᐃᑘᔨᐟ (ᐁᓴ) ᓈᐯᐘ. êkosi ê-itwêyit (êsa) nâpêwa êkosi ê-itwê -yi -t êsa nâpêw-a so C1-say-DS-3 EVID man -XT ‘That’s what the manOBV said’  (Volunteered S4, Presented S2,S3)  11  I am here treating proper names like other variables. This departs from treatments of proper names as ‘rigid designators,’ in the sense of Kripke (1980), but allows for the modeling of different perspectives on a referent identified by a proper name.  13  Here, the nominal bears a suffix –a, and the verb carries a special suffix –yi–. I propose that this construction codes that the individual is unable to possess a perspective relative to the specified context; the individual is contextually extentional. (17)  EXT(x,C) ↔ ψ y R(y,ψ,C)  x ≠ y) x is extentional at context C if and only if for all perspectives ψ and all individuals y, if there is a relation R of y with ψ at context C, then x is not y.  This contextual extentionality is always relative to the Speaker’s perspective, and means that “obviative” referents are unable to be assigned a perspective by the Speaker. (18)  a. ᐁᑯᓯ ᐁ ᐃᑘᔨᐟ (ᐁᓴ) ᓈᐯᐘ. êkosi ê-itwêyit (êsa) nâpêwa êkosi ê-itwê -yi -t êsa nâpêw-a so C1-say-DS-3 EVID man -XT ‘That’s what the manOBV said’  (Volunteered S4, Presented S2,S3)  b. WELL-FORMED DRS FOR 18A Speaker <R> xC say(x) man(x) EXT(x,C) c. ILL-FORMED DRS FOR 18A Speaker <R> x say(x) man(x) EXT(x,C) x <say> say(x)  14  Here, the well-formed DRS is one in which there is no perspective assigned to the man, who is marked as obviative (18b). Introducing an embedded perspective for this obviative referent is illformed (18c). Modeling “obviation,” then, reduces to modeling a contextual ban on perspective possession. This category is considered in detail in Chapter 3. The third kind of individuals is those that currently possess a perspective, termed “Proximate” in the literature. This kind of referent is created by contrast between the “Obviative” form and a normal nominal, rather than by anything special about the grammatical forms involved. For example, in (19), the perspective of nâpêw ‘man’ is represented, as shown by the explanation offered by the consultant. (19)  CATEGORY (III) = “PROXIMATE” ᐁ ᐚᐸᒫᐟ ᓈᐯᐤ ᐃᐢᑵᐘ ê-wâpamât nâpêw iskwêwa ê-wâp=am -â -t nâpêw iskwêw-a c1-see=by.eye.TA-DIR-3 man woman-XT ‘The manPROX sees the womanOBV.’ COMMENT (S2): “In this example, you’re hearing what the man has to say about it.” Speaker <say> xy man(x) EXT(y)  x <R> y woman(y) see(x,y)  The “Proximate,” then possess a perspective for the evaluation of the truth proposition. Being a function of contrast with the “Obviative,” this category is also considered in Chapter 3. The last kind of individuals is the elsewhere case, termed “Animate” in the literature. This kind of individual has neither intentional or extentional properties, as shown by the pair of examples in (20). 15  (20)  CATEGORY (IV)= “ANIMATE” a. ᓂᐚᐸᒫᐤ ᐊᓇ ᐊᐚᓯᐢ niwâpamâw ana awâsis ni-wâpam -â -w an=a awâsis 1- see=by.eye.TA-DIR-3 DST=AN.SG child ‘I seeAN thatAN child / ribbon.’ Comment (S2): “I don’t know if the child knows about this or not.” b. ᓂᐚᐸᒫᐤ ᐊᓇ ᓭᓇᐹᐣ niwâpamâw ana sênapân ni-wâpam -â -w an=a sênapân 1- see=by.eye.TA-DIR-3 DST=AN.SG ribbon ‘I seeAN thatAN child / ribbon.’ Comment (S2): “The ribbon couldn’t know about this.”  I consider this category in Chapter 2.  1.3.4. Proposal: Constructing obviation via referential dependency Turning from the kinds of individuals coded in Plains Cree to the manner in which these individuals are coded, I propose that the three categories of individuals are not grammatically equal. In particular, I argue that “inanimate” and “animate” are coded by dedicated forms (Chapter 2), while obviation, by contrast is constructed out of other resources available to the grammar (Chapter 3). Animacy, then is a primitive property of Plains Cree grammar, while obviation is not. In constructing obviation, I propose that Plains Cree recruits elements that code referential dependency (i.e. the process of making one referent dependent on another for its interpretation). This referential dependency can be understood as the satisfaction of two conditions: (i) a structural condition (c-command/linear precedence, shown on the left side of the table) and (ii) a semantic condition (perspectival embedding, shown on the right side of the table).  16  STRUCTURAL CONDITION 1 V y V  SEMANTIC CONDITION  V  x  V  2 PRED1  PRED2 5 5 y x Table 1.1.: Two conditions on referential dependency Depending on the configuration that an obviative referent is used in, a different piece of this referential dependency system will be recruited. In Table 1.2, we see that there are five different configurations for obviation, and they correspond with five different dependency types. FORM DEPENDENCY TYPE Noun-Noun Nominal to previous nominal -yiSubject to previous referent Possession Possessum to Possessor -êObject to subject -ikw Subject to Topic Table 1.2.: The forms of obviation and their dependencies Obviation, then, is just the application of referential dependency operations to animate referents. The more general property of Plains Cree is referential dependency. This is explored in detail in Chapter 4.  1.4. Relation to previous work In this section, I consider the ways in which the current model relates to previous work on these topics. In addition, each chapter has its second section devoted to previous work specific to those topics (i.e. §2.2, §3.2, §4.2, §5.2).  1.4.1. Models of subjective meaning Models of meaning in natural language have developed from a philosophical tradition that sought to develop an explicit language for expressing philosophical concepts. Since Aristotle, this philosophical tradition was crucially concerned with constructing arguments that entail 17  contradictions (Hume 1748, Tarski 1944), and numerous proposals were made to pare down linguistic forms to their most abstract, objective forms, in order to reduce ambiguity. Imported into work on non-philosophical language (i.e. natural language), this desire for objective, nonambiguity was maintained (Cresswell 1985:5). This had the result that semantic analyses of natural language tended to focus on contradiction as a means to define the boundaries of meaning in the formal semantic sense (i.e. “It is raining” entails that “it is not raining” cannot also be true). The judge for these contradictions was taken to be someone external to the speechact – the knower of “how it actually is” out there in the world (Tarski 1944). While this convenient way of talking about truth has proven useful in the description of meaning in natural language, it has long been noticed that numerous components of the natural languages considered (e.g. English, French, etc.) do not inherently entail contradictions, even when supplied with a specific time and place (e.g. Austin 1962). This includes, minimally, linguistic forms like questions (“May we come in?”), commands (“Open the door!”), and conditionals (“Forget to close the door and you’ll be sorry!”)(Fillmore 1975, Karttunnen 1977). Further, it was noticed that some of these elements in natural language are dependent on a specific individual involved in the speech act (cf. Searle 1965, Cresswell 1985, etc.). For example, an element like the English word “local” could be true when applied to the University of British Columbia for a speaker in Vancouver, but simultaneously untrue for a speaker in Boston (cf. Fillmore 1975, Mitchell 1987). Thus, the move has been away from the calculation of linguistic meaning solely in terms of contradiction-laden, individual-independent “objective” meaning and towards relativized, individual-dependent “subjective” meaning. One attempt at formalizing these “subjective” contexts was Farkas (1992), which focussed on modeling the distribution and meaning of Romance subjunctives. To model these, Farkas (1992) employed a DRT model in which an individual was paired with a world. More recently, Smith (2003) has employed DRT to model point of view effects in English discourse. The current model takes much of its mechanisms from these works, though exchanges worlds for perspectives. Recent work on the formalization of subjective meaning has focussed on predicates that trigger so-called “faultless disagreement” (Kölbel 2002, Lasersohn 2005, Stephenson 2007). These are contexts in which one person can utter a proposition, and the other person can disagree without saying the first person was wrong.12  12  I here cite the judgements of Stephenson (2007). I myself cannot utter “No, it isn’t!” in this context.  18  (21)  Speaker A: This cake is tasty. Speaker B: No, it isn’t!  (Stephenson 2007)  Lasersohn (2005) and Stephenson (2007) analyze these forms by appealing to the notion of a “judge.” This judge is an individual who judges the proposition to be true relative to a time and world. Being thus relativized, these propositions are inherently “subjective.” By employing a parallel relativization mechanism, the current work can be considered an extension of that done by Lasersohn (2005) and Stephenson (2007), but is much closer in its implementation to that of Farkas (1992). In all three accounts, truth evaluation is relativized to a particular individual, although some of the other particulars differ (e.g. perspectives vs. sets of worlds). Setting aside these more minor notational and philosophical differences, the important difference between the current model and these other works is that, whereas these other accounts relativize only certain predicates with respect to an individual, the current account relativizes all propositions to a perspective-holder (minimally, the Speaker).  1.4.2. Perspectives vs. possible worlds In this thesis, I employ the notion of a “perspective” for the evaluation domain of a proposition’s truth. While this follows Kölbel (2002), it departs significantly from the treatments typically used for the formal semantics of natural language. This means that a comparison of the current theory to other work is in order. It should be noted at the outset that none of the reasons for employing “perspectives” are absolutely vital to the current work, representing instead ontological and philosophical disagreement, and thus a semanticist that uses a possible worlds framework can readily translate the current formalisms into those more comfortable to them (i.e. exchange “perspective of x” for “set of worlds epistemically accessible to x” or “doxastic alternatives of x” and other modifications as necessary). In most theories of meaning, the domain of evaluation for propositions is a “world” (Hintikka 1962, Kripke 1963). What exactly is in a “world” appears to be a matter of some debate, but the canonical view is expressed by Hintikka (1962): a world contains a truth value for every proposition. Sets of worlds, then, are sets of sets of truth values for propositions. Relevant to our current discussion, this kind of system was used by Hintikka (1962) to model belief. For example, were we playing poker, and I wanted to guess the set of cards in your hand, I could write down every possible set of cards you could have, writing down one possible 19  hand on each slip of paper. As the game progresses, I could continually narrow down this set of papers based on new information, until I restrict the set papers to only one or two slips of paper. These would be the “worlds” that still depict possible states of the cards in your hand (i.e. epistemic alternatives; Hintikka 1962). Belief, then, is modeled as the continual narrowing of the set of worlds, based on additional knowledge. To believe something is to know the set of worlds that are accessible given your current belief. This is sometimes called an individual’s “doxastic alternatives.” The core difference between a possible worlds model like this and the perspective model I am here using lies in the notion of exhaustivity; worlds are taken to be exhaustively defined but perspectives are not. That is, a world has a truth value for every proposition in it, whereas a perspective only has a truth value for those propositions that are given to it. This has the result that individuals can only make reference to sets of worlds (since they do not know the truth value of every proposition in existence), but can make reference to a single perspective (since all they have to know is the set of propositions in it). Worlds, then, are always being restricted into a more narrow set (as the beliefs of the referent increases), whereas perspectives are always being specified with more belief content. In this sense, perspectives are something like “partial worlds” or “situations,” depending on the model employed (Barwise 1981). The advantage of the perspective model is this orientation towards specification. Rather than defining the acquisition of new information as the reduction in number of the set of worlds in the individual’s set of doxastic alternatives, we can define “belief” as the addition of some proposition to some individual’s perspective. More precisely, we can say that an individual possesses different perspectives for different kinds of propositional attitudes; there are perspectives filled with the things they “think” versus perspectives filled with the things they “believe” or “doubt.” Thus, there is the perspective, and then there is the relation to this perspective. This allows us to think carefully about the ways that propositional attitudes are constructed in natural languages like Plains Cree, where these predicates often have two elements (e.g. –êyiht ‘by mind’ and tâpwê- ‘true’; §1.3.1).  20  1.5. The Plains Cree language Plains Cree is an Algonquian language originally spoken in the northern Plains of North America by approximately 30,000 speakers.13 Speakers are typically in their 50’s or older, with some reports of children acquiring the language in the less-populated northern areas (Northern Alberta and Northern Saskatchewan). Since the language has failed to transfer to between two and three generations of speakers in most communities, it is safe to conclude that the language is now in a moribund state. Plains Cree is part of a larger language group of Central Algonquian languages sometimes called the “Cree Dialect Continuum,” which stretches from Labrador to Alberta. The precise division between dialects and languages is difficult, but the speakers I work with completely lose intelligibility around James Bay (Moose Cree). Speakers are unable to recognize the more eastern varieties (e.g. Montagnais, Innu-Aimun) as a related language. The languages share basic lexical forms, and much morphology, but the phonology, morphophonology, and syntax are significantly different.  1.5.1. Sources Plains Cree has a strong descriptive tradition, stretching back to the 1840’s (Howse 1844). I here offer a review of available materials that the reader can refer to for particular forms and discussion. I have also included a glossary section at the end of the thesis, which identifies and defines every form used in this thesis. Using this glossary in concert with other grammatical materials should provide the necessary tools to understand the Plains Cree data presented here. There are three major grammatical descriptions of Plains Cree; Howse (1865), Lacombe (1874), and Wolfart (1973). Of these, Wolfart (1973) is the most accessible, and provides a discussion and analysis of the earlier two works. It represents a clear statement of the principles of grammatical description that Hockett (1966) and Bloomfield (1962) developed, and is thus of interest for theoretical as well as practical reasons. 13  Note that these estimates are quite old (SIL: 1982).  21  Beyond these more formal grammatical descriptions are those aimed at a less linguistically-trained audience, written with the intent of introducing beginners to the language, training teachers, and instructing students of the language. In particular, the grammatical descriptions of Wolfart and Carroll (1981), Ahenakew (1987), and Okimâsis and Ratt (1999) are useful sources. The first book is a succinct introduction to the most notable features of Plains Cree grammar. In addition to presenting user-friendly grammatical discussion, the latter two books offer the insights of native Plains Cree speakers, which are sometimes found nowhere else (for example Ahenakew’s discussion of ‘factive’ ôma). In terms of dictionaries, there are three major ones to choose from: LeClaire and Cardinal (1998), Wolfart and Ahenakew (1998), and (Wolvengrey 2001). In work with modern Plains Cree speakers, the 2-volume dictionary by Wolvengrey (2001) is the most useful. It serves as a fully-functional dictionary, comparable to an modern dictionary for Spanish or German. Wolfart and Ahenakew (1998), which is constructed from corpus work, is an excellent companion to published texts, but can only offer words that have been recorded there (e.g. if the texts have no instance of a word for “high-heeled shoes,” the dictionary will not have an entry for this word.) Plains Cree has one of the best text collections of any language indigenous to North America. Starting with Bloomfield’s work (1930, 1933), high-quality textual work has been consistently produced. Beginning in the late 1980s, the partnership of Ahenakew and Wolfart produced a set of Plains Cree-language books (1993, 1997, 1999, 2000) that provide a significant body of high-quality data, complete with translation, commentary, and glossary. More recently, Wolvengrey (2007) has begun an initiative to publish sets of texts. Formal linguistic work on Plains Cree has focussed largely on the complex morphosyntax of the language. Relevant work that deals with morphosyntax includes Dahlstrom (1986), Dryer (1996), Hirose (2000), and Déchaine (1999, 2003, 2008). The complex particle system of Plains Cree has been described in detail by Ogg (1991). Clausal structure has been considered by Blain (1997), Long (1999), and Cook (2008). Obviation phenomena has been considered by Wolfart (1978), Russell (1991, 1996), and 22  Mühlbauer (herein). Other grammatical features that have been considered are pronouns (Blain 1995), weather verbs (Blain 1987), incorporation (Wolfart 1971), possession (Mühlbauer 2004, 2007), word order (Mühlbauer 2003, Wolvengrey 2007, Déchaine 2007), demonstrative ordering (Wolvengrey 2003), preverbs (Cook 2006, Wolvengrey 2006), relative roots (Cook 2003), evidentials (Blain et al. 2006, Blain & Déchaine 2002, 2007), and reduplication (Ahenakew & Wolfart 1983). Phonetic and phonological discussions include treatments of prominence (Mühlbauer 2006), ablaut (Wolfart 1973), epenthesis and deletion (Wolfart 1973), syllable structure (Cook 2003), diminutive palatalization (Hirose 1999), prosody (Cook 2006), and phrasing (Cook & Mühlbauer 2005).  1.5.2. Orthography There are two orthographies for Plains Cree: (i) the standard roman orthography (SRO), and (ii) the syllabic orthography. I here offer a short guide to understanding these systems, as they relate to the current thesis. The standard roman orthography’s main notable features are the circumflexes over vowels, which represent vowel quality contrasts (e.g. a [ә] vs. â [a], i [ɪ] vs. î [i], o [ʊ] vs. ô [u]), the use of ‘c’ for the palatal [ts], and the use of pre-consonantal ‘h’ to code pre-aspiration. It should be noted that this orthography aims for a highly phonemicized spelling system, and does not closely approximate the surface forms of the language as it is currently spoken. As such, it does not record the wide dialect variation found within Plains Cree. For example, one consultant (S2) turns all [ks] clusters into geminate [ss], often turns CVC sequences into [ʔ]C, and metathesizes all [wa] sequences to [ɔ]. Using the standardized orthography, none of this is written down – just as none of the dialect variation of English is captured in the orthography (e.g. you can’t tell by reading this that I often drift into a heavy Northern Wisconsin accent). The data in this thesis, then, is not suitable for phonological analysis. The syllabic writing system is favored by many speakers of the language. It represents the same basic vowel and consonant contrasts as the standard roman orthography, but does so syllabically. The shape of the symbol codes the consonantal value, while its orientation codes the vowel. Thus, ᐯᐱᐳᐸ all represent different combinations of [p] with a vowel. This orthography has been used in this thesis because of its value for speakers of the language. Detailed explanations of this system can be found in the dictionary of Wolvengrey (2001), and elsewhere. 23  1.6. Outline of the thesis There are five chapters in this thesis, as well as a glossary. They are organized in the following way. Chapter 2 considers the form, content, and context of the two animacy classes in Plains Cree. I argue there that the formal organization of these two classes can be best understood as dedicated coding that is distributed across all syntactic positions that code reference (e.g. nominal and argument structure). I then argue that the inanimate class is inherently extentional in meaning, while the inherently animate class has no inherent content at all. Contextually, these two classes of nominals are manipulated to code different individual’s perspectives about the referent. Chapter 3 considers the form, content, and context of obviation. Rather than being a basic grammatical category of Plains Cree, I argue that obviation is best understood as the result of using several independent kinds of morphosyntax to construct the referential category “obviative.” In terms of its specific referential properties, I analyze the obviative as denoting referents that are contextual extentionality. This referential information is then situated within the Speaker’s perspective as a filter on potential perspective embeddings for this referent. In Chapter 4, I turn to the forms used to code obviation. I propose that Plains Cree constructs obviation out of forms that code the more general property of referential dependency. This referential dependency obeys a structural and a semantic condition that is operative across Plains Cree grammar. Obviation, then, is a by-product of more basic (morpho)syntactic operations of Plains Cree. In Chapter 5, the focus shifts from analysis of the data to how the data was obtained. Here, I consider the ways that different kinds of data collecting methods affect the outcome, and catalogue how each kind of method interacts with obviation and animacy. From this consideration, I argue that all data is good, so long as we think carefully about the context of its collection. In Chapter 6, I conclude the discussion and consider its implications. In particular, I consider approaches to the plural/obviative parallels, the behaviour of the suffix –yi– across the Cree languages, and the similarities and differences of the Plains Cree system described here to that shown in Athabaskan languages. Because Algonquian linguistics has its own terminology, I have included a glossary at the end of the work. This glossary contains every term and every gloss used in this thesis, and what 24  they mean. It is intended to not only make the thesis more accessible to a non-Algonquianist audience, but to also help make the entire field of Algonquian linguistics accessible to nonspecialists.  25  Chapter 2 The Form, Content, and Context of Animacy 2.1. Proposal In this chapter, I consider the form, content, and context of the two basic referential classes of Plains Cree, traditionally termed “Animate” and “Inanimate.”1 (1)  a. “ANIMATE” ᓂᐚᐸᒫᐘᐠ ᐊᓂᑭ ᒪᐢᑭᓯᓇᐠ niwâpamâwak aniki maskisinak ni-wâp=am -â -w-ak an =iki maskisin-ak 1-bright=by.eye.TA-DIR-3-PL DST=AN.PL shoe -PL ‘I seeAN thoseAN shoesAN.’ b. “INANIMATE” ᓂᐚᐸᐦᑌᐣ ᐊᓂᐦᐃ ᒪᐢᑭᓯᓇ niwâpahtên anihi maskisina ni-wâp=aht -ê -n an =ihi maskisin-a 1-bright=by.eye.TI-TI-LP DST=XT shoe -XT ‘I seeIN thoseIN shoesIN.’  (Presented S2)  (Presented S2)  The form in (1a) shows 4 distinct morphemes that code “Animate” referents (–am–, –â–, –ak, and –iki), and these are mirrored by 4 distinct morphemes in (1b) that code “Inanimate” referents (–aht–, –ê–, –a, and –ihi). After introducing the proposal, I review the previous literature on animacy in Plains Cree, paying special attention to how the current account relates to these others (§2.2). I then argue that animacy does not have a dedicated locus in the syntax of Plains Cree (§2.3). Rather, animacy is coded in all places that reference is coded. In the verb system, these locations minimally include verbal argument positions (e.g. –am–, –â–, –ê–). Following the syntactic work of Hirose (2000), Déchaine (2003), and Déchaine & Reinholtz (2008) of the verb system of Plains Cree, I schematize this as in (2).  1  While my definition of these two classes is new, I have here adopted the traditional terms for them. I used scare quotes in this thesis to show that the terms themselves do not carry any ontological significance. I have maintained this terminology because these are the names used throughout the literature; if I were to change terminology, comparing different approaches would be made much more difficult.  26  (2)  niwâpahtên ni-wâp=aht -ê-n 1- see =by.eye.TI-TI-LP ‘I see itIN’ CP 3 C IP 3 I vP 3 wâpvP 3 pro 3 v VP -aht3 [EXT] pro 6 V -ê[EXT]  In the nominal system, referential positions minimally include demonstrative positions (e.g. –iki, –ihi), and determiner-related positions (e.g. –a, –ak).2 (3)  anihi maskisina an=ihi maskisin-a dst=XT shoe -XT ‘Those shoesIN’ DEMP 3 DEM EXTP an3 pro 3 EXT NUMP -ihi 3 [EXT] NUM NP -a 6 [EXT] maskisin  Animacy, then, is distributed across the syntax of Plains Cree.  2  For a consideration of the syntax of demonstratives, please turn to §3.3.2.  27  In section four, I argue that the concept of “animacy” is more precisely defined in terms of extentionality. I show that the two sets of forms (“Inanimate” and “Animate”) map on to the semantic property of extentionality in the following way: (i)  The grammatical class traditionally called “Inanimate” specifies that the referent never possesses a perspective (i.e. are inherently extentional; [EXT]).  (ii)  The grammatical class traditionally called “Animate” is unspecified in its content; it does not specify anything about a referent’s extentionality or intentionality.  Thus, Plains Cree has a privative opposition between a form that is specified for extentional content (the “Inanimate”) and a form that is unspecified (the “Animate”). (4)  a. “INANIMATE” FORM : [EXT] b. “ANIMATE” FORM  : [Ø]  I define an extentional referent as one that cannot be paired with a perspective (cf. §1.3.1). (5)  EXT(x) ↔ ψ y R(y,ψ)  x ≠ y) x is Extentional if and only if for all Perspectives ψ and all individuals y, if there is a relation R between individual y with perspective ψ, then x is not y.  Applying this to the two referential classes of Plains Cree, I claim that the “Inanimate” forms will have the content of (6), while the “Animate” forms will have the content of (7). (6)  “INANIMATE” FORM = λx · [EXT(x) ⋀ PRED'(x)] x, such that x is extentional and x is a member of the set of referents denoted by the predicate.  (7)  “ANIMATE” FORM = λx · [PRED'(x)] x, such that x is a member of the set of referents denoted by the predicate. Finally, in section five I argue that these form-content pairs bear contextual information:  a referent’s extentionality is always evaluated within the perspective of some referent. This may be any of three possible perspective possessors: (i)  The Speaker  (ii)  The Hearer  (iii)  Some prominent third-person (Kuno’s 1972 “Discourse Perspective”)  To adequately model this, I utilize a modified version of the Discourse Representation Theory (DRT: Kamp 1981) proposed by Farkas (1992) and Smith (2003). In this formulation, Discourse 28  Representation Structures (DRS) are taken to represent perspectives (Kölbel 2002). Consider the Calvin & Hobbes example (§1.3.1). (8)  HOBBES IS “INANIMATE” ᑯᐦᑖᐑ ᒥᐢᑲᑦ ᐦᐋᐱᐢ kohtâwiy miskam Hobbes k-ohtâwiy m=isk -am Hobbes 2-father find=by.body.TI-TI Hobbes ‘Your father foundINAN Hobbes.’  (Translated S2)  Using the proposed model, we can represent the different individual’s perspectives in the following way: (i)  When Hobbes is “Inanimate” to Speaker, [EXT(Hobbes)] is embedded in the Speaker’s perspective. Speaker <say> x Hobbes father(x,Hearer) EXT(Hobbes) find(x,Hobbes)  (ii)  When Hobbes is “Inanimate” to the Hearer, [EXT(Hobbes)] is embedded in the Speaker’s representation of the Hearer’s perspective. Speaker <say> x Hobbes father(x,Hearer) find(x,Hobbes)  Hearer <R> Hobbes EXT(Hobbes)  29  (iii)  When Hobbes is “Inanimate” to some third person (y), [EXT(Hobbes)] is embedded in the Speaker’s representation of that referent’s perspective. Speaker <say> x y Hobbes father(x,Hearer) find(x,Hobbes) y <R> Hobbes EXT(Hobbes)  The modeling of the contextual component of animacy reduces to differential embedding within perspectives.  2.2. Previous accounts In order to understand the discussion of animacy that follows, it is necessary to consider its relation to previous work. In particular, it is necessary to understand previous work on the syntax of Plains Cree, which I review in section 2.2.1. While the current proposal is built from the significant body of previous work on this topic, it generally departs from previous accounts in four crucial ways: (i)  The treatment of “Inanimate” as the marked member of the opposition (§2.2.2)  (ii)  The equating of the “Inanimate” class with extentionality (§2.2.3)  (iii)  The treatment of animacy as a syntactic element (§2.2.4)  (iv)  The data set considered (§2.2.5)  2.2.1. Syntactic accounts of Plains Cree Algonquian languages have a great deal of morphology, which has attracted much interest from many of the most influential scholars of the last century, including Leonard Bloomfield, Mary Haas, Ken Hale, Charles Hockett, C.C. Uhlenbeck, and Morris Swadesh. This work can be (very 30  roughly) broken into two types: (i) American Structuralist accounts and (ii) Generative Grammar accounts, with work in both frameworks continuing to present. In his account of the grammar of Menominee (Central Algonquian, Wisconsin), Leonard Bloomfield (1962) outlines a positional account of the morphosyntax of the verb and nominal which has been influential for all subsequent work on Algonquian. In this account, the system is conceived of as having slots, for which morphemes were specified. When two morphemes were shown to be in complementary distribution, they were assigned the same slot. Following this method, Bloomfield constructs a description of Menominee that includes 10 suffixal positions and two prefixal positions. Subsequent work in Plains Cree (e.g. Wolfart 1973, Dahlstrom 1986) has posited a similar system. One of the crucial generalizations that has come out of this work is the decompositional nature of the Algonquian stem. In particular, verbs are composed of multiple morphemes, conveniently labeled “Initials,” “Medials,” and “Finals” based on their order of occurrence. A sample for Plains Cree is shown in (9). (9)  ᐑᓴᑭᒋᐦᒌᓀᐤ. wîsakicihcînêw. wîsak -icihciy-in-êpain -hand-by.hand-dirINITIAL MEDIAL FINAL AFFIX ‘s/he hurts his/her hand with his/her hand.’  -w -3 AFFIX (Volunteered S3)  The affix that immediately follows the final in transitive verbs (–ê– above) is labeled a “theme sign.” Building on these generalizations about the decompositionality of verbal predicates, Hirose (2000) observes that verbs in Plains Cree canonically have one morpheme for each of their arguments; intransitive verbs have one affix beyond the root (“Initial”), transitive verbs have two affixes, and di-transitive verbs sometimes have three.3 Modeling this in a generative syntactic framework (Chomsky 1982, etc.), Hirose proposed that the affixes that code transitivity be located in the two heads of a split VP, each of which introduces an argument (taken to be pronominal in nature; pro). The root is introduced in the complement position of the lowest verb phrase. This is all schematized in (10).  3  This is, of course, a gross characterization. There are numerous cases of single-morpheme verbs in Plains Cree – a trend which appears to be generalized in Blackfoot.  31  (10)  niwâpahtên ni-wâp=aht -ê-n 1- see =by.eye.TI-TI-LP ‘I see itIN’ vP 3 pro 3 v VP -ê3 pro 3 V ROOT -ahtwâp-  To account for the linear order of these affixes, Hirose (2000) posited a cyclic raising mechanism; the ROOT raises to the left of the first verbal head V, and then these two raise to the left of the second verbal head v, giving the linear order ROOT > V > v. Considering Algonquian verb syntax in the context of typological work, Déchaine (2003) proposes modifications to Hirose’s (2000) treatment in order to model the relation between Algonquian verbal structure and that of other languages (e.g. English, Semitic, Salish). In particular, she reorders the morphemes and locates the root as an adjunct to the outer verb phrase (vP in transitive structures). (11)  niwâpahtên ni-wâp=aht -ê-n 1- see =by.eye.TI-TI-LP ‘I see itIN’ vP 3 ROOT vP wâp- 3 pro 3 v VP -aht3 pro 6 V -ê-  This structure linearizes via phrasal encliticization; each phrase is spelled out as SPEC > HEAD > COMPLIMENT, yielding a linear order ROOT > v > V for transitive verbs. This is the version of Plains Cree argument structure that is adopted in the current thesis. 32  In contrast to the verbal system, the nominal system of Algonquian has received relatively little attention since the first detailed descriptions of Bloomfield (1962), Hockett (1966), and Wolfart (1973). For Plains Cree, the only generative account of nominal morphosyntax that I am aware of is Déchaine (1999). Working from the generalizations made by Hockett (1966), Déchaine argues that affixation in the nominal domain can be represented as in (12). (12) nimaskisininâna ni-maskisin-nân-a 1-shoe -1pl -XT ‘Our shoes’  (Presented S2)  DP 2 D PERSP ni– 2 2 PERS NumP –nân 2 2 NUM NP –a 2 2 N maskisin Arguing on the basis of selectional restriction, Déchaine locates the suffixes for possessor number (e.g. –nân) in the head of PersP, introduced in the compliment of D, where the possessor prefix is located (e.g. ni–). Elements that code plurality (e.g –a) are introduced in NumP related to the NP itself (e.g. maskisin). To linearize this, Déchaine employs two mechanisms: (i) cyclic movement of the NP to the spec of PersP, and (ii) phrasal encliticization. Together, this yields the linear order D > NP > PERS > NUM. The basics of this model are adopted for this thesis, although some modifications will be necessary, which are outlined as introduced.  2.2.2. “Inanimate” as semantically specified Most analyses of the “Animate/Inanimate” distinction in Algonquian languages take “Animate” to be the specified, contentful member of the contrast. This approach is exemplified by 33  Dahlstrom (1995:64-65), who builds a picture of “Animate” in terms of semantic prototypes. In some form or other, most formal linguists have adopted this view, particularly those linguists who have appealed to a “person hierarchy,” wherein referents are ranked based on their semantic properties, with [±ANIMATE] playing a crucial role (e.g. Blain (1997), Junker (2005), Bliss (2005), Ritter & Wiltschko (2007), etc.). By contrast, the current account treats “Animate” as having no semantic properties, while “Inanimate” is specified. However, Hockett (1966:62) argues that the systematic shifting of “Inanimate” referents to “Animate” (but not the reverse) points to “Inanimate” as the contentful member of the pair, a position that is also echoed by some other linguists (e.g. Wolfart 1973).4 The specified/unspecified contrast, then, is based on a combination of discoursal and grammatical properties. This account is further developed in the current proposal. A third position, proposed most clearly by Goddard (2004) takes the specification relation to be different along different dimensions. On this account, the specification of an animacy form is a function of the context it is used in. For Goddard (2004), “Animate” is unspecified semantically, being “a function of contrast with inanimate,” (Goddard 2004:224), but “Inanimate” is the general member in many morphosyntactic constructions.5 Put in the terms of this thesis, “Inanimate” is unspecified in form, but specified in semantic content. Focusing on semantic specification, the current account is only directly at odds with accounts that adopt “Animate” as the semantically-specified class (e.g. Dahlstrom 1995, Ritter & Rose 2005). By contrast, the current proposal is consistent with Hockett’s (1966) generalizations; given the semantically-specified status of “Inanimate,” it should be easier for nominals to slip into the unspecified class (“Animate”) than the other way around.  2.2.3. “Inanimate” as extentional One can identify three types of semantic proposals that have been put forward to account for animacy patterns in Algonquian. As will be seen, the current proposal, while not completely converging with any of them, assembles the core observations of each line of thought into one unified analysis. One approach to animacy which I call the “concrete” proposal analyzes Algonquian animacy in terms of the semantic properties that are directly accessible to speakers and 4  See Valentine (2000:118-119) for a detailed discussion of animacy shift in Ojibwa (Central Algonquian). Note, however, that most of Goddard’s (2004) morphosyntactic data is from Fox (Central Algonquian). The facts gathered from that study do not all have direct correlates in Plains Cree, although many do. 5  34  observers. Proposals of this kind fall into two camps: (i) animacy codes a “living/non-living” distinction, and (ii) animacy codes a culturally-defined concept of “power.” The first proposal is the earlier of the two, with its first form coming in 1634 by LeJeune (Thwaites 1896-1901:7.2223), and being subsequently adopted by Eliot (1666). For example, in his grammar of Massachusset (Eastern Algonquian), Eliot (1666), states “The Animate form or declension is when the thing signified is a living Creature …The Inanimate form or declension of Nouns, is when the thing signified is not a living Creature” (Eliot 1666:10). Of course, even in these early treatments, linguists were well aware that there were significant groups of nouns that were grammatically animate but could not be said to be alive. For example, articles of clothing like socks (13a) and non-living things like dead snakes (13b) are coded as “Animate” in Plains Cree. (13)  a. ASIKAN ‘SOCK’ = “ANIMATE” ᓂᐚᐸᒫᐤ ᐊᓇ ᐊᓯᑲᐣ. niwâpamâw ana asikan. ni-wap=am -â -w an=a asikan 1- see =by.eye.TA-DIR-3 DST=AN.SG sock ‘I see that sockAN.’  (Presented S2)  b. DEAD ANIMALS = “ANIMATE” ᓂᐚᐸᒫᐤ ᐊᓇ ᑭᓀᐱᑯ ᐁ ᓂᐱᐟ. niwâpamâw ana kinêpik ê-nipit. ni-wap=am -â -w an=a kinêpikw ê- nip=i-t 1- see =by.eye.TA-DIR-3 DST=AN.SG snake C1-die=AI-3 ‘I see that dead snakeAN.’  (Presented S2)  The recognition of this descriptive inadequacy caused a disjunction among theorists, such that one group pursued a more detailed concrete analysis, while others moved away from concrete treatments entirely (see below). Among the remaining concretists, perhaps the most well-known is the work of Darnell and Varnek (1976), which claims that animacy “deals with power to maintain and balance the universe and to interact with persons and other interactive beings.” Because accounts of this kind have invariably settled on animate as the marked member, assigning it semantics of the sort just described, these accounts have been widely criticized for being forced to make numerous ad hoc claims. For example, what is “powerful” to a Cree speaker about socks (14a), but not hats (14b)?  35  (14)  a. ‘SOCK’ = “ANIMATE” ᐆᑭ ᐊᓯᑲᓇᐠ ôki asikanak aw=iki asikan-ak PRX=AN.PL sock -PL ‘these socksAN’  (Presented S2)  b. ‘HAT’ = “INANIMATE” ôhi astotina aw=ihi astotin-a PRX=XT hat -XT ‘these hatsIN’ A second approach, conveniently termed the “abstract” approach, analyzes Algonquian animacy as coding a kind of semantics that is an abstraction from the observable facts. An account of this kind is suggested by early French-speaking linguists (e.g. Nicolas 1672, in Daviault 1994), who divided nominals between a class called “noble” and “ignoble.” Most often, grammarians defined these terms similarly to “animate” and “Inanimate” (i.e. living/non-living), but sometimes more abstract discussion was offered, referring to concepts like “esteem” or “objects of consideration.” Goddard (2004) takes this initial step towards a more abstract analysis further, arguing that a useful way to understand Algonquian animacy is to strip the “noble/ignoble” dichotomy of its notions of “esteem” and “living,” leaving only the more abstract notion of “high” versus “low.” Many of the linguists that ascribe to a “person hierarchy” (e.g. Blain 1997, Déchaine and Reinholtz 1999, Junker 2005) must adopt, in some form, this kind of an “abstract” approach to the semantics of animacy. Where the previous accounts attempted to ascribe the semantic value to the two classes, analyses of the third kind, which I label “grammar” approaches, treat Algonquian animacy as a purely grammatical distinction, with no meaning ascribed to the classes at all. Proponents of this approach point to the apparent inconsistencies in the semantic approach, and draw parallels to the gender systems of Indo-European, which are assumed to be strictly grammatical devices. Frantz (1995), for example, explains that “grammatical gender is a classification of the noun stems themselves, not of the entitites to which they refer” (Frantz 1991:8). Most linguists that work within a formal framework have adopted this kind of approach (e.g. Blain 1997, Hirose 2000, Mühlbauer 2007). The current account agrees with parts of each of these three basic positions, but is not classifiable as any one of these approaches. Like the “concrete” position, I take the animacy 36  contrast to have a clear, defineable set of semantics that bear some resemblence to the description offered by Darnell and Varnek (1976). However, the specific definition I have offered is more abstract; animacy situates references within the discourse constructed by the Speaker, placing restrictions on their ability to host embedded propositions (i.e. “Inanimate” referents are extentional). This means that the “Inanimate” class of referents cannot have a mental state directed at something (i.e. are extentional). This more abstract characterization subsumes previous attempts to characterize “Inanimate” as “lacking power” or being “ignoble.”  2.2.4. Animacy is syntactically determined It is often assumed that the division between “Animate” and “Inanimate” is a lexical division. That is, animacy is an inherent property of stems; stems are either inherently “Animate” or inherently “Inanimate” (cf. Frantz 1995). This can most easily be seen by checking glossing (e.g. Bloomfield 1962, Wolfart 1973, Valentine 2001) and dictionaries (e.g. Wolvengrey 2001), which universally code nominals as either inherently “Animate” or “Inanimate.” This is done so systematically that a nominal form known to regularly appear with both “Animate” and “Inanimate” marking is given two lexical entries in the dictionary. Consider as an example the entry for mistikw ‘tree/stick’ in Wolfart and Ahenakew’s (1998) dictionary of Plains Cree. (15)  a. mistikw- NA tree, post [sic] [sic:NA] b. mistikw- NI stick, pole, post, log, wooden rail [sic:NI] (Wolfart & Ahenakew 1998:55)  Here, the lexical item mistikw has been entered twice, once under “Animate” (NA) and once under “Inanimate” (NI), with notes about speakers shifting from one form to the other (e.g. there is apparently a text example of mistikw occurring in “Animate” contexts but denoting a ‘post’). This set of lexicographic conventions has been standard for several hundred years. When formal treatments are developed that involve animacy (e.g. Hirose 2000, Bruening 2001, Branigan & MacKenzie 2002), it is assumed that the nominal is inherently coded for animacy, and the other morphology agrees with it. Syntactic analyses of this kind include both Minimalist (e.g. Bruening 2001) and Relational Grammar treatments (e.g. Frantz 1991). In his discussion of animacy in Menominee (Central Algonquian, Wisconsin), Bloomfield (1962) takes a different view. Bloomfield describes the nominal maeqtek as “Animate” in its meaning of ‘tree.’ However, he notes that “the same word” can be used in 37  “Inanimate gender” to describe a ‘stick’ (Bloomfield 1962:29). By treating the same nominal as occurring in two different “gender” contexts, Bloomfield is treating the nominal as independent of its animacy specification. In treating animacy as a part of the syntax of Plains Cree rather than the lexicon, then, the current account follows this kind of approach.  2.2.5. The data set The current analysis brings to light six new generalizations about animacy: (i)  The lack of psych-predicates inflected for “Inanimate” referents (§2.4.3.1)  (ii)  The restriction of reflexive-marked verbs to “Animate” referents (§2.4.33)  (iii)  Speaker judgments about the non-specificity of “Animate” coded referents (§2.4.4)  (iv)  The correlation between Speaker belief and animacy coding (§2.4.2)  (v)  The ability of animacy coding to be manipulated for common-ground effects (§2.4.3)  (vi)  The ability of animacy coding to be manipulated for point-of-view effects in narrative contexts (§2.4.4)  Any subsequent account of animacy will have to expand to cover this new data set.  2.3. The form of animacy: Distributed throughout the syntax A consideration of Plains Cree’s grammatical forms shows a basic distinction between “Animate” forms and “Inanimate” forms. The “Animate” structure in (16a) shows “Animate” marking within the verb stem (-am-), in its argument-structure (-â-), and in its pronominal agreement (-w-ak). The demonstrative shows “Animate” marking in its suffixes (-iki). Finally, the nominal shows animate marking in the suffix attached for plurality (-ak). A comparison of these animate forms in (16a) with their “Inanimate” counterparts in (16b) shows the contrast clearly (e.g. -am- vs. -aht- within the verb stem). (16)  a. “ANIMATE” MORPHOLOGY ᓂᐚᐸᒫᐘᐠ ᐊᓂᑭ ᒪᐢᑭᓯᓇᐠ niwâpamâwak aniki maskisinak ni-wâp=am -â -w-ak an =iki maskisin-ak 1-bright=by.eye.TA-DIR-3-PL DST=AN.PL shoe -PL ‘I seeAN thoseAN shoesAN.’  (Presented S2)  38  b. “INANIMATE” MORPHOLOGY ᓂᐚᐸᐦᑌᐣ ᐊᓂᐦᐃ ᒪᐢᑭᓯᓇ niwâpahtên anihi maskisina ni-wâp=aht -ê -n an =ihi maskisin-a 1-bright=by.eye.TI-TI-LP DST=XT shoe -XT ‘I seeIN thoseIN shoesIN.’  (Presented S2)  The four domains of animacy coding are summarized in Table 2.1 below. VERB DEMONSTRATIVE NOMINAL ROOT FINAL THEME ROOT AFFIX ROOT AFFIX ANIMATE wâp am â an iki maskisin ak INANIMATE wâp aht am an ihi maskisin a Table 2.1. Summary of animacy coding for verb, demonstrative, and nominal Within models of syntax built on the generative program of Chomsky (1982), there are at least three possible ways to implement a model of animacy forms in Plains Cree, differing in where they locate animacy in the grammatical structure: (i)  HYPOTHESIS ONE: A “lexical” model. The locus of animacy is the nominal root. The feature [±EXTENTIONAL] is a specified on nominal roots; animacy does not have any structural locus.  (17)  a. nâpêw ‘man’  : [-EXT]  b. maskisin : [+EXT] ‘shoe’ These roots are then merged into the syntax bearing this [±EXTENTIONAL] feature, where their inherent properties trigger agreement with other elements in the syntax (cf. Hirose 2000, Bruening 2001). (18)  a.  DEMP qi DEM N awa nâpêw [-EXT] [-EXT]  39  b.  DEMP qi DEM ôma [+EXT]  (ii)  N maskisin [+EXT]  HYPOTHESIS TWO: A “local” model. Animacy is confined to a single, dedicated head in the syntax. If this head is in the nominal syntax, it is a classifier feature introduced external to the nominal root (18) (Mühlbauer 2007).  (19)  a.  CLASSP qi CLASS Ø [-EXT]  b.  N nâpêw [-EXT]  CLASSP qi CLASS N Ø maskisin [+EXT] [+EXT]  If this head is instead located in the verbal syntax, it restricts the argument structure (e.g. Theta-roles) of the verb. (20)  a.  VP qi ARG  Ø [-EXT] b.  V -am[-EXT]  VP qi ARG  Ø [+EXT]  V -aht[+EXT]  On either view, animacy is reduceable to a choice in the content of a single, dedicated position. (iii)  HYPOTHESIS THREE: A “distributed” model. Animacy is coded in all positions that referential distinctions are coded, with no preference for one position over another.  40  (21)  ᓂᐚᐸᐦᑌᐣ ᐆᐦᐃ ᒪᐢᑭᓯᓇ. niwâpahtên ôhi maskisina. ni-wâp=aht -ê-n aw=ihi maskisin-a 1- see =by.eye.TI-TI-LP PRX=XT shoe -XT ‘I see these shoesIN.’  (Presented S2)  qp CP DEMP 3 3 C IP awEXTP 3 3 I vP pro 3 3 EXT NUMP ROOT vP ihi 3 wâp3 [EXT] NUM NP pro 3 -a 6 v VP [EXT] maskisin -aht3 [EXT] pro 6 V -ê[EXT] Animacy syntax, then, is a distributed set of operations aimed at coding a referential property. In deciding between these different models, it is important to keep in mind that any set of data can be accounted for within any syntactic theory, if we allow enough additional mechanisms. This means that a discrimination between the three possible models involves the reader’s presuppositions about the kinds of structures available to the grammar. A “costly” stipulation to one analysis may be taken for granted in another, meaning that we must have a notion of “cost” before we begin to choose among models. In the following discussion, I consider a stipulation to be more “costly” when it invokes a mechanism that has no surface exponent in the grammar. Thus, the more surface-true a mechanism is, the more desireable it is. With a notion of “cost” now in hand, let us consider what kind of data each of the three analyses covers, and what kind of data it cannot cover. The lexical account requires three things, which are different from the other two accounts: (i)  Animacy cannot be actively manipulated (§2.3.1).  (ii)  Animacy manipulations can only be accomplished by suppletion (§2.3.2).  (iii)  Plains Cree has a high degree of homophony (§2.3.3). 41  As discussed in §2.3.1 through §2.3.3, the lexical model is undesireable on all three counts. Setting this model aside, we can then consider the requirements of the other two kinds of models. These models differ in several respects: (i)  whether or not bare nominals must occur in the scope (i.e. c-commanded by or linear preceded by) of an element that codes animacy (§2.3.4)  (ii)  whether or not verbal coding can function in the absence of nominal coding (§2.3.5)  (iii)  whether or not nominal coding can function in the absence of verbal coding (§2.3.6)  When the different models are compared to the attested language patterns, along with a notion of cost, it is clear that the distributed hypothesis covers the data with the least amount of extra stipulations.  2.3.1. Animacy is not specified in the nominal If nouns are specified for animacy, then a shift in animacy value should not be possible. This is because an element that lacks structural expression cannot be manipulated via structural operations. Put another way, we expect that the animacy value of a noun - [+EXTENTIONAL] or [-EXTENTIONAL] – should be a stable property.6 However, it turns out that animacy values may be contextually manipulated. As we have already seen several times, speakers are capable of manipulating animacy coding. For example, although the typical animacy value for –skât ‘leg’ is “Inanimate,” in contexts where the Speaker views legs as intentional, the nominal can be marked as “Animate.” This is illustrated in (22), with (22a) having the typical “Inanimate” value (niskâta ‘my Inanimate legs’), and (22b) having the animate value (niskâtak ‘my animate legs’). (22) a. “INANIMATE” LEGS ᓂᐢᑳᑕ ᐆᐦᐃ niskâta ôhi ni-skât-a aw =ihi7 1-leg-XT PRX=XT ‘TheseIN are my legsIN’ (Judgment by S4) CONTEXT (S4): Pointing to legs to identify the proper word for them. 6  This assumes that all nominals are lexically specified to the same degree (i.e. all nouns are either specified for animacy, or all nouns are underspecified). Because the evidence shows that any noun in the language can have its animacy value manipulated, there is no evidence for class cleavage among nominals in terms of their animacy. This makes the assumption of uniform specificity warranted. 7 This parsing is based on stem-internal phonology; aw + i  ô. There are places where this alternation is still unstable, and thus provides crucial evidence: âsawinamaw- (Minde 1998) vs. âsônamaw- (canonical).  42  b. “ANIMATE” LEGS ᓂᐢᑳᑕᐠ ᐆᑭ niskâtak ôki ni-skât-ak aw =iki 1- leg -PL PRX=PL ‘TheseAN are my legsAN’ (Judgment by S4) CONTEXT (S4): The speaker’s legs are out of his control, acting on their own. They are kicking him or other people. An example of the manipulation of these animacy contrasts in a discourse is shown by Louis Moosomin’s telling of the ‘Rolling Head’ story (in Bloomfield 1930:§1).8 In the version of this story told by Moosomin, a woman’s head has been cut off by her husband, but continues to be inhabited by the woman’s mind. The head rolls along the ground, pursuing its children, who flee from it (see §2.5.3 for a more detailed discussion). Throughout the story, the animacy value of the head shifts back and forth from “Inanimate” to “Animate.” A summary of the locations of “Inanimate” and “Animate” shifting for the severed head is given in the table below. Note that in normal Plains Cree speech, the nominal -stikwân ‘head’ is usually framed with “Inanimate” morphology (i.e. its ‘cultural default’ is “Inanimate” as in Jaszczolt 2004). Line Event SEVERED HEAD ANIMACY i Severed Head opens eyes INANIMATE ii Severed Head speaks MIX … iv Severed Head asks Utensils ANIMATE … x Boy sees Severed Head talking INANIMATE … xiii Severed Head stopped ANIMATE … xv Only Severed Head going INANIMATE … xvii Severed Head passes fire ANIMATE … xxxii Boy cries that Severed Head kill Boy INANIMATE … xxxiv Severed Head sees Great Serpent ANIMATE Table 2.2. Summary of animacy shifting for the severed head story  8  This is a story common to Algonquian peoples across North America (Cowan 1980).  43  As can be seen, the animacy classification of the severed head changes seven times in the span of approximately 40 clauses. These shifts are considered in detail in section 2.5.3 below, where I argue that they code shifts in the perspective represented in the discourse. From this evidence, we should conclude that there is active manipulation of the “Inanimate/Animate” contrast to code referential distinctions. The fact that animacy values can shift according to the perspective represented indicates that nounrs are not inherently specified for animacy. This runs counter to the expectations of a lexical treatment of animacy, but is expected by both of the other syntactic accounts.  2.3.2. Animacy contrasts are not suppletive If animacy is a property of nominal stems rather than a property of the syntax, as the lexical theory posits, speakers should not be able to access it to alter its content. Not having a position in the grammatical structure of the language, speakers would have to perform substitution operations, changing one nominal form for another, in order to change animacy values. This would follow patterns of suppletion seen elsewhere in the language, for example the stem miy‘give’ encompasses both the root and final of the verbal complex. (23)  a. TRANSITIVE VERB WÂPAM- = 2 MORPHEMES (WÂP- + -AM-) ᐚᐸᒣᐤ. wâpamêw. wâp=am -ê -w see =by.eye.TA-DIR-3 ‘S/hePROX sees him/herOBV.’  (Presented S2)  b. TRANSITIVE VERB MIY- = 1 MORPHEME (MIY-) ᒥᔦᐤ. miyêw. miy -ê -w give.TA-DIR-3 ‘S/hePROX gave it to him/herOBV.’  (Presented S2)  When the transitivity of this verb form is altered, Plains Cree has to employ completely different verb stem, as shown in (24), where the transitive stem is miy- (24a) and the intransitive (AIT) stem is mêki- (24b).  44  (24)  a. TRANSITIVE VERB MIY- ‘GIVE’ ᒥᔦᐤ. miyêw. miy -ê -w give.TA-DIR-3 ‘S/hePROX gave it to him/herOBV.’ b. INTRANSITIVE SUPPLETIVE FORM MÊKI- ‘GIVE’ ᒣᑭᐤ. mêkiw. mêki -w give.AI-3 ‘S/heAN gives things.’  (Presented S2)  (Presented S2)  Thus, whenever transitivity is a property of stems instead of affixes, the stem has to be changed when transitivity changes. Supposing that animacy, like transitivity, were a property of stems, predicts that, whenever there is a shift in animacy, there ought to be suppletion of one stem for another. For example, kinêpikw ‘snake’ (“Animate”) ought to change to (e.g.) sâpâwisk ‘snake’ (“Inanimate”). We can find no evidence for suppletion in Plains Cree. When a speaker alters the animacy of a referent, they employ the same nominal stem but with different morphosyntax. (25)  a. “ANIMATE” ‘SHOE’ = maskisin + [-si-]v + [-a]DEM ᐊᐘ ᒪᐢᑭᓯᓇ ᒦᐦᑯᓲ awa maskisin mîhkosiw aw =a maskisin mîhkw=si-w PRX=IN.SG shoe red =AI-3 ‘ThisAN shoe is redAN.’ CONTEXT: Magic shoe that talks to speaker.  (Presented S2)  b. “INANIMATE” ‘SHOE’ = maskisin + [-â-]v + [-ima]DEM ᐆᒪ ᒪᐢᑭᓯᓇ ᒦᐦᑳᐝ ôma maskisin mîhkwâw aw =ima maskisin mîhkw=â-w PRX=IN.SG shoe red =II-3 ‘ThisAN shoe is redAN.’ CONTEXT: Normal shoe laying on the floor.  (Presented S2)  This runs counter to the lexical analysis, which expects that one form or the other of this pair ought to employ a different nominal stem, in parallel to the suppletion pattern seen in verbs.  45  2.3.3. The necessity of homophony Plains Cree appears to be a system that has numerous nominal forms that are readily able to occur with both “Animate” and “Inanimate” marking. FORM “INANIMATE” “ANIMATE” asiniy bullet rock âtayôhkan sacred story spirit being cikâhkwân lance gambling toy kayâsîyâkan old dish Old Dish (person) kistikân garden seed maskipitôn twisted mouth Twisted Mouth mistikw stick tree mistikwâskisin wooden shoe Dutch person mitâs leggings pants piwâpiskwastotin steel helmet German wâpistikwân white head person with white hair wâpimin white berry white bead wâposwayân rabbit hide The Rabbitskin People Table 2.3. Forms commonly found in both animacy classes 9 In lexicographic work on Plains Cree, these forms are listed in the dictionary twice; once under “Animate,” and once under “Inanimate” (cf. Wolfart & Ahenakew 1998, Wolvengrey 2001, and all other lexical work done on the language). (26)  a. âtayôhkan ᐋᑕᔫᐦᑲᐣ NA spirit being, spirit power, spirit guardian, spirit animal b. âtayôhkan ᐋᑕᔫᐦᑲᐣ NI sacred story; legend [cf. âtayôhkêwin]  (Wolvengrey 2001)  The question that arises is whether this lexicographical homophony should be carried over into the analysis of animacy. A number of nominal forms in Cree are found to alter the kind of referent they denote when framed with each of the two sets of nominal morphosyntax. For example, kistikân is equivalent to English ‘farm’ when the “Inanimate” set of morphology is applied (27a), but ‘grain’ when the “Animate” set is used (27b).  9  Note that the pattern of animate forms correlating with proper names is only accidental; inanimate forms can easily be used as proper names as well. For example, the Plains Cree name for Blaine lake, Saskatchewan is sîhwîhtâkanisâkahikan (lit: salt-lake), a form that is invariably inanimate.  46  (27)  a. KISTIKÂN AS “INANIMATE” = ‘FARM’ ᐁᑮᒫᐦᒥᓵᑭ ᐁᑯᓂ ᑭᐢᑎᑳᓇ ê-kî-mâh-misâki êkoni kistikâna …  ê- kî- mâh-mis=â-k-i êkoni kistikân-a C1-PREV-RED-big =II-0-PL RESM farm -XT ‘thoseIN farms were bigIN…’  (Minde 1997:§72)  b. KISTIKÂN AS “ANIMATE” = ‘GRAIN’ ᓂᑲᓂᑐᐙᑕᐙᑳᐣ ᐊᐘ ᑭᐢᑎᑳᐣ nika-nitaw-âtâwâkân awa kistikân … ni-ka- nitaw-atâwaw=ikê -n aw =a kistikân 1- FUT-go- sell =GEN-LP PRX=AN.SG grain ‘I will go and sell thisAN grain…’  (Ahenakew 2000:§2.3)  A similar phenomenon occurs when a nominal is used as the proper name of a person. Thus kîskihkômân refers to a kind of ‘cut off knife’ when marked “Inanimate” (28a), but refers to a particular person named ‘Cut Knife’ (a Sarcee chief) when marked with the “Animate” set (28b). (28)  a. KÎSKIHKÔMÂN ‘CUT KNIFE’ AS “INANIMATE” ᐊᓂᒪ ᑮᐢᑰᐦᑰᒫᐣ anima kîskihkômân an =ima kîski-mohkômân DST=IN.SG cut -knife ‘thatIN cut knife’  (Wolvengry 2000, S2)  b. KÎSKIHKÔMÂN ‘CUT KNIFE’ AS “A