UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The syntax and semantics of clause-typing in Plains Cree Cook, Clare Elizabeth 2008

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

Item Metadata

Download

Media
[if-you-see-this-DO-NOT-CLICK]
[if-you-see-this-DO-NOT-CLICK]
ubc_2008_fall_cook_clare.pdf [ 4.84MB ]
[if-you-see-this-DO-NOT-CLICK]
Metadata
JSON: 1.0066456.json
JSON-LD: 1.0066456+ld.json
RDF/XML (Pretty): 1.0066456.xml
RDF/JSON: 1.0066456+rdf.json
Turtle: 1.0066456+rdf-turtle.txt
N-Triples: 1.0066456+rdf-ntriples.txt
Original Record: 1.0066456 +original-record.json
Full Text
1.0066456.txt
Citation
1.0066456.ris

Full Text

THE SYNTAX AND SEMANTICS OF CLAUSE–TYPING IN PLAINS CREE by CLARE ELIZABETH COOK 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 © Clare Elizabeth Cook, 2008  Abstract This thesis proposes that there are two kinds of clauses: indexical clauses, which are evaluated with respect to the speech situation; and anaphoric clauses, which are evaluated with respect to a contextually-given (anaphoric) situation. Empirical motivation for this claim comes from the clause-typing system of Plains Cree, an Algonquian language spoken on the Canadian plains, which morpho-syntactically distinguishes between two types of clauses traditionally called INDEPENDENT  and  CONJUNCT  orders. In the current analysis, the INDEPENDENT order instantiates  indexical clauses, and the CONJUNCT order instantiates anaphoric clauses. After laying out the proposal (chapter 1) and establishing the morphosyntax of Plains Cree CPs (chapter 2), the remaining chapters discuss the proposal in detail. Chapter 3 focusses on the syntax and semantics of indexical clauses (Plains Cree’s INDEPENDENT  order). Syntactically, I show that there is an anti-c-command and an anti-  precedence condition on indexical clauses. Semantically, I show that indexical clauses are always and only evaluated with respect to the speech situation, including the speech time (temporal anchoring), speech place (spatial anchoring), and speaker (referential anchoring). Chapter 4 focusses on the syntax and semantics of anaphoric clauses (Plains Cree’s CONJUNCT  order). Syntactically, I show that anaphoric clauses must always be either preceded or  dominated by some other antecedent clause.  Semantically, I show that the value of  temporal/spatial/referential dependent elements within an anaphoric clause is determined by an antecedent. Chapter 5 turns to the syntactic sub-classification of Plains Cree’s  CONJUNCT  (i.e.,  anaphoric) clauses. I propose that there are three classes: chained clauses, adjunct clauses, and mediated argument clauses. I provide two kinds of diagnostics that distinguish these classes, and explore the consequences of this classification for argument clauses and complementation. Finally, Chapter 6 proposes a semantic sub-classification of Plains Cree’s CONJUNCT (i.e., anaphoric) clauses. I propose that there is a direct mapping between the morphology and the semantics: one complementizer encodes presupposition of the proposition, the lack of a complementizer encodes a-veridicality of the proposition, and one complementizer is semantically unspecified (the elsewhere case). This means that Plains Cree’s clause-typing is fundamentally concerned with how the truth of the proposition is represented.  ii  Table of contents Abstract...................................................................................................................................... ii Table of contents....................................................................................................................... iii List of tables ..............................................................................................................................xi List of symbols and abbreviations ............................................................................................ xii Acknowledgements ...................................................................................................................xv Citation ...................................................................................................................................xvii 1  Indexical versus anaphoric CPs ....................................................................................1 1.1 Proposal: Indexical vs. anaphoric CPs........................................................................1 1.2 Relation to previous work ..........................................................................................3 1.2.1 Connection to the matrix/embedded distinction ...........................................3 1.2.2 Connection to illocutionary force.................................................................3 1.2.3 Parallel between indexical CPs and indexical expressions............................4 1.2.4 Parallel between anaphoric CPs and anaphoric expressions .........................5 1.2.5 Connection to the INDEPENDENT/CONJUNCT contrast ....................................6 1.3 Methodology: Data collection and presentation..........................................................8 1.4 Plains Cree terminology...........................................................................................10 1.5 Layout of the thesis..................................................................................................12  2  Mapping indexical and anaphoric CPs onto Plains Cree’s morpho-syntax ..............15 2.1 Proposal: A one-to-one mapping in Plains Cree .......................................................15 2.2 Diagnostics for CPs in Plains Cree...........................................................................19 2.2.1 Ordering properties....................................................................................20 2.2.2 Peripheral agreement diagnoses CPs..........................................................23 2.2.3 Clause-typing diagnoses CPs .....................................................................31 2.2.4 Pronominal proclitics are complementary with clause-typing.....................34 2.2.5 Interim summary: Verbal complexes are CPs.............................................35 2.3 Diagnosing C vs. spec, CP .......................................................................................36 2.3.1 Selection of complement ...........................................................................37  iii  2.3.2 Substitution (does not) determine distribution............................................40 2.3.3 The significance of non-overtness..............................................................42 2.3.4 Interim summary .......................................................................................47 2.4 The indexical/anaphoric distinction ≠ matrix/embedded distinction..........................48 2.4.1 Negation distinguishes matrix and embedded clauses ................................48 2.4.2 Interrogative cî distinguishes matrix and embedded clauses.......................51 2.5 Summary .................................................................................................................52 3  Indexical clauses: Plains Cree’s INDEPENDENT order .................................................54 3.1 Proposal: The syntax and semantics of indexical clauses..........................................54 3.2 The structural context of indexical clauses ...............................................................56 3.2.1 Indexical clauses must be matrix clauses ...................................................57 3.2.1.1 Embedding predicates do not introduce indexical clauses ............58 3.2.1.2 Subordinating particles do not introduce indexical clauses ..........60 3.2.1.3 Embedded negation does not modify indexical clauses................62 3.2.1.4 Summary: indexical clauses cannot be subordinated....................63 3.2.2 Indexical clauses exclude cross-clausal dependencies ................................63 3.2.2.1 Relative roots..............................................................................63 3.2.2.1.1 Relative roots with predicate modifier antecedents........66 3.2.2.1.2 Relative roots with CP modifier antecedents.................69 3.2.2.1.3 Relative roots with cross-clausal antecedents ................70 3.2.2.2 Spatio-temporal variables must be bound in indexical clauses .....74 3.2.3 Pronominal proclitics are indexical............................................................76 3.2.3.1 Indexical proclitics cannot be bound ...........................................77 3.2.3.2 The absence of third-person proclitics .........................................79 3.2.3.3 Referents are deictically anchored in indexical clauses ................81 3.2.3.3.1 Distribution of overt nominals in a discourse ................82 3.2.3.3.2 Restrictions on switch reference in INDEPENDENT clauses...............................................87 3.3 The semantics of indexical clauses: Indexicality ......................................................90 3.3.1 Temporal deixis: Relating reference time to speech time ...........................90 iv  3.3.1.1 Contrasting temporal interpretations of indexical and non-indexical clauses .......................................94 3.3.1.2 Indexical clauses present statives that hold at T0..........................98 3.3.1.3 Indexical clauses present activities that coincide with T0 ...........102 3.3.1.4 Indexical clauses present telic predicates whose result state coincides with T0 .........................................105 3.3.1.5 Interim summary.......................................................................110 3.3.2 Referential deixis: The role of the speaker in indexical clauses ................110 3.3.2.1 Person-effects: Preference for indexical clauses when talking about self ...........................................................112 3.3.2.2 Indexical clauses are infelicitous in contexts of unconsciousness ................................................114 3.3.2.3 Events in indexical clauses must be directly perceived by the speaker .............................................117 3.3.2.4 êsa has mirative force in indexical clauses ................................122 3.3.2.4.1 The interaction of clause-typing and êsa .....................123 3.3.2.4.2 Mirativity as incongruent experience ..........................125 3.3.2.5 Speaker commitment to the proposition.....................................129 3.3.2.6 Subjective predicates convey speaker’s attitude.........................131 3.3.2.7 Interim summary.......................................................................132 3.4 Summary: Structural and semantic conditions on indexical clauses ........................133 4  Anaphoric clauses: Plains Cree’s CONJUNCT order ..................................................134 4.1 Proposal: Anaphoric clauses ..................................................................................134 4.2 From pronominal to clausal anaphora.....................................................................135 4.2.1 The forms of anaphoric elements.............................................................136 4.2.2 The relation between anaphor and antecedent ..........................................137 4.2.3 Conditions on antecedent-licensing: C-command and precedence............139 4.2.4 Summary: The properties of anaphoric clauses ........................................143 4.3 The distribution of anaphoric clauses .....................................................................144 4.3.1 Anaphoric clauses are subject to precedence and/or c-command..............144 v  4.3.1.1 Anaphoric clauses that must be c-commanded are not subject to precedence ...................................................144 4.3.1.2 Anaphoric clauses with subordinating particles are not subject to precedence ..................................................148 4.3.1.3 Anaphoric clauses that are subject to precedence: Unembedded ê- clauses ...........................................................151 4.3.2 Long-distance precedence of antecedent for anaphoric clauses ................154 4.3.2.1 Out-of-the-blue contexts vs. established contexts ......................156 4.3.2.2 Contrastive focus ......................................................................158 4.3.2.3 The distribution of clause-typing in elicitation: A discourse effect ...................................................................160 4.3.3 Interim summary .....................................................................................164 4.4 Anaphoric dependencies in anaphoric clauses ........................................................165 4.4.1 kî- is subject to c-command and precedence in CONJUNCT clauses............167 4.4.1.1 Precedence without c-command: kî- is anaphoric ......................169 4.4.1.2 Precedence and c-command: kî- is anaphoric.............................172 4.4.1.3 C-command without precedence: kî- is anaphoric......................176 4.4.1.4 No precedence, no c-command: kî- is not anaphoric..................178 4.4.2 -yi is subject to c-command and precedence in CONJUNCT clauses............181 4.4.2.1 C-command and precedence: -yi is licensed ..............................183 4.4.2.2 Precedence without c-command: -yi is licensed.........................184 4.4.2.3 C-command without precedence:-yi is licensed .........................186 4.4.2.4 No c-command, no precedence: -yi is not licensed.....................188 4.5 The cross-linguistic typology of anaphoric clauses.................................................191 4.5.1 Chained clauses are anaphoric clauses .....................................................192 4.5.1.1 The significance of asymmetric marking...................................193 4.5.1.2 Fixed relative order of the anaphoric clause and antecedent clause ..............................................................198 4.5.2 English modally subordinated clauses are anaphoric clauses....................200 4.6 Summary ...............................................................................................................205  vi  5  The syntax of anaphoric clauses................................................................................206 5.1 Proposal: Chained, adjoined, and mediated argument clauses.................................206 5.2 The diagnostics ......................................................................................................208 5.2.1 Exclusion tests.........................................................................................208 5.2.2 Island tests...............................................................................................208 5.3 Applying the exclusion tests...................................................................................209 5.3.1 Linear precedence....................................................................................209 5.3.1.1 Chained clauses must follow their antecedent............................210 5.3.1.2 Adjoined clauses can follow their antecedent ............................211 5.3.1.3 Mediated argument clauses usually (but not always) follow their antecedent ............................................................211 5.3.2 Ability to be a matrix clause based on morpho-syntax .............................213 5.3.2.1 Chained clauses are always capable of being matrix clauses......214 5.3.2.2 Adjoined clauses are not always capable of being matrix clauses ............................................................215 5.3.2.3 Mediated argument clauses are not always capable of being matrix clauses ...............................................215 5.3.3 Prosodification ........................................................................................215 5.3.3.1 Chained clauses require an intonational break ...........................216 5.3.3.2 Adjoined clauses do not require an intonational break ...............216 5.3.3.3 Mediated argument clauses do not require an intonational break ...............................................................217 5.3.4 Interim summary .....................................................................................217 5.4 Applying the island tests ........................................................................................218 5.4.1 Long distance wh-construal must be across mediated arguments .............219 5.4.1.1 Wh-words can be long distance with mediated argument clauses ......................................................221 5.4.1.2 Wh-words cannot be long distance with adjoined clauses ..........222 5.4.1.3 Wh-words cannot be long distance with chained clauses ...........223 5.4.2 Long distance quantifier fronting must be across mediated arguments .....224  vii  5.4.2.1 Mediated argument clauses permit long distance quantifier-fronting..............................................225 5.4.2.2 Adjoined clauses do not permit long distance quantifier fronting ..............................................226 5.4.2.3 Chained clauses do not permit long distance quantifier-fronting..............................................227 5.4.3 Long distance argument expression-fronting............................................227 5.4.3.1 Mediated argument clauses permit long distance argument-fronting ..............................................228 5.4.3.2 Adjoined clauses do not permit long distance argument-fronting ..............................................230 5.4.3.3 Chained clauses do not permit long distance argument fronting...............................................231 5.4.4 (Non-)obligatory switch reference picks out object-oriented clauses ........231 5.5 Consequences ........................................................................................................236 5.5.1 The non-existence of argument clauses....................................................237 5.5.1.1 ‘Subject’ clauses require overt nominal antecedent....................241 5.5.1.2 Clauses have different ordering properties than arguments ........244 5.5.1.3 No predicates subcategorize for clauses.....................................247 5.5.1.4 Predicates that subcategorize for nominals: AIt verbs................250 5.5.1.5 Incorporation is nominal ...........................................................252 5.5.1.6 Summary: Clauses do not sit in argument positions...................254 5.5.2 VP-complementation involves restructuring ............................................255 5.5.2.1 Restructuring involves a single set of agreement .......................256 5.5.2.2 Independent-order agreement is possible...................................257 5.5.2.3 Restructuring allows only one set of temporal marking .............258 5.5.2.4 Restructured clauses introduce a single set of arguments...........259 5.5.2.5 Restructured clauses only permit a single subject ......................261 5.5.2.6 Restructuring preverbs are category-insensitive.........................262 5.5.3 Copy-to-object constructions must be local agreement.............................263  viii  5.6 Summary ...............................................................................................................268 6  The semantics of anaphoric clauses ..........................................................................270 6.1 Proposal: Presuppositional, a-veridical, and unspecified clauses.............................270 6.2 ê- as the unspecified complementizer .....................................................................271 6.2.1 Distributional evidence for ê- being unspecified ......................................272 6.2.2 Interpretational evidence that ê- is unspecified.........................................274 6.3 kâ- as a presuppositional complementizer ..............................................................277 6.3.1 Relative clauses: kâ- and ê.......................................................................278 6.3.2 Wh-questions ..........................................................................................283 6.3.2.1 kâ- wh-questions as presuppositional ........................................285 6.3.3 Temporal modification ............................................................................290 6.3.4 Concessive clauses ..................................................................................292 6.3.5 Correlatives .............................................................................................295 6.4  as a-veridical.....................................................................................................296 6.4.1 Mediated argument clauses split along a-veridicality ...............................299 6.4.1.1 Sensitivity to weak intensional predicates..................................299 6.4.1.2 Sensitivity to weak intensional meanings ..................................301 6.4.1.3 Sensitivity to negation...............................................................303 6.4.1.4 Insensitivity to factive-emotive predicates.................................304 6.4.1.5 Insensitivity to lack-of-speaker-knowledge ...............................305 6.4.2 Adverbial clauses split along a-veridicality..............................................307 6.4.2.1 Irrealis temporal modification ...................................................307 6.4.2.2 Unrealized alternatives..............................................................308 6.4.2.3 ‘before’ clauses ........................................................................309 6.4.2.4 Antecedents of conditionals ......................................................312 6.4.2.5 Purpose clauses .........................................................................313 6.5 Summary ...............................................................................................................314  7  Conclusions ................................................................................................................317 7.1 The syntax and semantics of clause-typing in Plains Cree ......................................317 ix  7.2 The parallels between CPs and DPs .......................................................................319 7.3 (Im)possible analyses of kî.....................................................................................319 7.3.1 kî- marks disjunction and precedence.......................................................320 7.3.2 kî- is not a deictic past tense ....................................................................322 7.3.3 kî- is not a perfect ....................................................................................324 7.3.4 kî- is not a perfective ...............................................................................328 7.4 Deconstructing modality: Clause-typing, irreality, and kî .......................................330 7.4.1 The role of clause-typing: Circumstantial vs. deontic modality ................331 7.4.2 kî- has existential modal force under negation..........................................334 7.4.3 Negation widens possible interpretations .................................................335 7.4.4 Embedding neutralizes modal distinctions ...............................................336 7.4.5 Summary.................................................................................................339 7.5 Variation in clause-typing across Algonquian ........................................................340 7.5.1 Variation of the pronominal proclitics: Plains Cree vs. Blackfoot ............340 7.5.2 Variation in initial change: Plains Cree vs. Blackfoot vs. Ojibwe.............343 References ..............................................................................................................................345  x  List of tables Table 1.1 Indexicality vs. embedding ..........................................................................................3 Table 1.2 Mapping the indexical/anaphoric contrast onto INDEPENDENT and CONJUNCT ...............7 Table 1.3 The CONJUNCT modes ................................................................................................12 Table 2.1 Left-edge clause-typing proclitic determines distribution of clause ............................34 Table 2.2 Diagnostics for specifiers vs. heads............................................................................36 Table 2.3 Person marking in INDEPENDENT and possession paradigms.......................................37 Table 3.1 Summary of the INDEPENDENT order paradigm in Plains Cree ....................................54 Table 3.2 Distribution of subordinators by clause-type in Plains Cree........................................60 Table 3.3 Predicate modifier antecedents occur in both clause-types .........................................67 Table 3.4 CP-modifier antecedents are possible in both clause-types .........................................69 Table 3.5 No cross-clausal antecedents for indexical INDEPENDENT clauses ...............................71 Table 3.6 Distribution of spatio-temporal proform variables by clause-type...............................74 Table 3.7 Person marking in INDEPENDENT clauses vs. possessed nominals ...............................77 Table 3.8 (Un)attested co-occurrence of -yi in INDEPENDENT order (Wolfart 1973 .....................88 Table 3.9 Interpretation of INDEPENDENT clauses by aspectual class ..........................................94 Table 3.10 Interpretations of êsa .............................................................................................123 Table 4.1 Structures which meet the c-command condition .....................................................139 Table 4.2 Structures which do not meet the c-command condition...........................................140 Table 4.3 Structures which meet the precedence condition ......................................................140 Table 4.4 Structures which do not meet the precedence condition............................................141 Table 4.5 Structures which are undefined for antecedent licensing of anaphoric elements........141 Table 4.6 Clause-typing in discourse-initial position ...............................................................161 Table 4.7 Sequencing of clause-types by discourse-type..........................................................161 Table 4.8 Deictic vs. anaphoric dependents .............................................................................167 Table 4.9 Distribution of -yi by anaphoric configuration..........................................................183 Table 4.10 Anaphoricity vs. embedding ..................................................................................191 Table 5.1 Morpho-syntactic classification of anaphoric CONJUNCT clauses ..............................206 Table 5.2 Diagnostic 1: Subjection to precedence....................................................................209 Table 5.3 Diagnostic 2: Matrix capability................................................................................213 Table 5.4 Only potential matrix clauses can be chained clauses ...............................................214 xi  Table 5.5 Adjoined clauses need not have potential to be matrix clauses .................................215 Table 5.6 Mediated argument clauses need not have potential to be matrix clauses..................215 Table 5.7 Diagnostic 3: Intonational break ..............................................................................216 Table 5.8 Only mediated argument clauses allow long-distance wh-words ..............................220 Table 5.9 Only mediated argument clauses allow long-distance quantifier-fronting .................224 Table 5.10 Only mediated argument clauses allow long-distance argument expressions ..........227 Table 5.11 Restructuring preverbs in Plains Cree ....................................................................256 Table 6.1 ê-clauses are anaphoric in unembedded contexts......................................................273 Table 6.2 ê-clauses are non-presuppositional in relative-clause contexts..................................273 Table 6.3 ê-clauses are veridical in mediated argument-clause contexts...................................273 Table 6.4 Asymmetries between two wh-clause types .............................................................286 Table 6.5 Clause-typing and kiyâm..........................................................................................293 Table 6.6 Comparison of Plains Cree simple CONJUNCT and Romanian subjunctive ................300 Table 6.7 Classification of predicates by clause-type of subordinate clause .............................300 Table 6.8 Unselective predicates .............................................................................................301 Table 6.9 Factive-emotive predicates in Plains Cree ................................................................305 Table 7.1 Patterning of kî- relative to the English perfect.........................................................325 Table 7.2 Patterning of kî- relative to perfective aspect............................................................328 Table 7.3 Modal interpretations...............................................................................................331 Table 7.4 Mapping of Plains Cree forms to modal interpretations............................................331 Table 7.5 Clause-typing affects modal base.............................................................................332 Table 7.6 kî- codes existential force in modal contexts ............................................................333 Table 7.7 Modal base vs. quantification in Plains Cree ............................................................339 Table 7.8 Modal interpretations in Plains Cree ........................................................................340 Table 7.9 Diagnostics for determining the position of pronominal proclitics............................341 Table 7.10 Properties of initial change in Plains Cree, Blackfoot, and Ojibwe .........................343  xii  List of abbrevations and symbols Abbreviations 0 1 1>2 1>3 2 2>1 2>3 3 3>1 3>2 AN APPL BEN C1 C2 CONJ CONN COME DEIC DEM DIR DJ DS DUB DUR EMPH EPEN EVID FUT GO HAB HES HORT IC IMP INAN INCEP INV IRR PERF  = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =  inanimate (agreement) 1st person 1st person acts on 2nd person 1st person acts on 3rd person 2nd person 2nd person acts on 1st person 2nd person acts on 3rd person 3rd person 3rd person acts on 1st person 3rd person acts on 2nd person animate applicative benefactive changed conjunct 1 changed conjunct 2 conjunctive connective directional towards origo deictic demonstrative direct disjoint argument different subject dubitative durative emphatic epenthetic evidential future direction away from origo habitual hesitation hortative internal change imperfective inanimate inceptive inverse irrealis perfective  PV INC INDIC INT IRR LOC MIDST MOD NEG NOM OBV ORIG PL Q QUANT PL PREV PV RAT RED REFLX REL REMP RR SAP SG SIM SUBJ SUBJ1 SUBJ2 TEMP TOP USC  = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =  VAI VII VTA VTI WH  = = = = =  preverb marker inceptive indicative intend irrealis locative on-going action modifier negation nominalizer obviative origin plural interrogative quantifier plural previous preverb rationale reduplication reflexive relative clause marker remote past relative root speech act participant singular simultaneous subjunctive subjunctive 1 subjunctive 2 temporal topic marker unspecified subject construction verb, animate intransitive verb, inanimate intransitive verb, transitive animate verb, transitive inanimate wh-word  xiii  Symbols  * ! # ∃ ∀ ¬ < [] CP DP F IP Op p S s T t vbl VP  = = = = = = = = = = = = = = = = = = = =  string is ill-formed string is well-formed, but cannot have the intended interpretation string is well-fomed, but infelicitous in the given context existential quantifier universal quantifier negation temporally precedes constituency brackets complementizer phrase determiner phrase function inflectional phrase operator proposition speaker situation variable time trace variable verb phrase  xiv  Acknowledgements Plains Cree is a beautiful language. My heartfelt thanks go first to the people that speak it who have been so patient and generous with me over the past five years as I tried both to learn it and to learn about it. I first heard Plains Cree when I took a conversational class at the University of Alberta; Dorothy Thunder always encouraged my efforts and made the class fun. Wally Awasis not only put up with my really long (and often bizarre!) constructed sentences, but told great jokes and sang beautiful songs. Josephine Small shared her time generously amid a schedule that would already be too much for most people. I appreciated Rita Daniels’ warm hospitality and openness to almost total strangers when we visited her (including the bannock lessons). She told lovely stories and made Plains Cree feel ‘real.’ The early morning conversations and work sessions with Joseph Deschamps in Alberta were the basis for many of the ideas in this thesis, and also made life seem better. Toni Cardinal always seemed to know what question I was trying to ask, and even better, how to answer it. Her clear thinking, patience, and steadfastness brought this work together. I hope some of what they have tried to teach me is captured in these pages. Amidst an amazingly busy schedule, my research supervisor, Rose-Marie Déchaine, spent much time and energy answering my questions, explaining concepts (it always seemed so crystal clear when you said it!), playing devil’s advocate, and helping me work through the problems (obvious and not-so-obvious) of my analysis. She also taught me about a theoretical linguist’s life – guiding me through everything from writing grants, abstracts and papers, to fieldwork, to giving talks, to surviving. Thank you. Thanks also to my other thesis committee members, Lisa Matthewson and Martina Wiltschko. Lisa carefully read every draft I gave her, no matter how sketchy, and gave detailed comments. I also particularly appreciated her enthusiasm for data and trying to understand how the language worked, even when we had no idea what was going on. Martina’s lectures on syntax first made me think I could understand it. Her appreciation of new ideas encouraged me to think about my data in fresh ways, and her insistence on clarity – both in analysis and in writing – definitely made this thesis more readable. My thesis benefited from being read by the three members of the external examination committee. Amy Dahlstrom, Nancy Hedberg, and Jessica de Villiers were generous enough to take the time to read the thesis and asked useful questions about all aspects of the thesis, from theoretical implementation to methodology to implications of the analysis for Plains Cree and other languages. Despite his more unofficial status, H.C. Wolfart played a key role in this thesis and in my development as a linguist. He has always been interested in both the big issues and the details (contentful and editorial), and he has been a sympathetic and encouraging audience even when he vehemently disagreed with the crazy things I was doing. He also set an example for how corpus linguistics can and should be done. I am also grateful to my undergraduate advisor Monica Macaulay. She first introduced me to Algonquian languages, taught my first class on fieldmethods, gave me suggestions about suitable graduate schools to consider and straight-forward practical advice about navigating it once I started, and oversaw my first conference presentation. I met and have benefited greatly from talking to a number of other Algonquian linguists. Thanks to Eleanor Blain, Phil Branigan, Carrie Dyck, Inge Genee, Jean Okimâsis, Charlotte Reinholtz, Nicole Rosen, and Arok Wolvengrey. More generally, thanks to audiences at ACAL,  xv  the Algonquian conference, ASA, CLA, NWLC, SSILA, WSCLA, and UBC for feedback on talks given in the development of this work. My early teachers at UW-Madison – Murvet Enc, Marlys Macken, Jack McKeown, Victoria Pagan, Matt Pearson, Bozena Tieszen, and Andrew Wolpert – gave me a love for language and linguistics. Without the UBC faculty – Guy Carden, Henry Davis, Bryan Gick, Gunnar Hannsson, Doug Pulleyblank, Hotze Rullmann, Pat Shaw, Joe Stemberger, and Eric Vatikiotis-Bateson – the linguistics department would not be nearly what it is. Edna Dharmaratne, the department secretary, helped make sure I got through the many paperwork hoops of a graduate degree, including getting registered and getting paid. I am very grateful for all the times she went the extra mile for me and others; I don’t know what I would have done without her. Fortunately, I didn’t have to go into financial debt to support going to school. This research was supported by a Theodore E. Arnold Fellowship, a University Graduate Fellowship, and a Josephine Berthier Fellowship. Thanks to the taxpayers of Canada. I was part of the largest incoming graduate class at UBC linguistics ever, and we were quite the group. Solveiga Armoskaite made lovely dinners and I enjoyed doing elicitation with her. Fiona Campbell was the first person in the department I met; she introduced me to Benny’s. Ramona MacDowell knew how to make decisions. I will not forget Jason Brown, Peter Jacobs, Jong-won Kim, Karsten Koch, Jeff Muehlbauer, Dominique Quis, and Christine Ravinski. I have no idea where you all will end up, but it will make for a great story. Many thanks to Rachel Wojdak and Sugunya (Add) Ruangjaroon for looking out for me as a new student, telling me what to expect in graduate school, and sharing many parts of their lives. Thanks to Florence Woo for afternoons of quilting and walking. Thanks to Ryan Waldie for knowing that when “X” is not explained, nothing is explained. Thanks to the Thursday night dinner & reading group. Thanks to students inside and out of UBC linguistics: Oladiipo Ajiboye, Mario Chavez-Peon, Chin Seok Koon, Yunhee Chung, Kerim Demirci, Atsushi Fujimori, Masaru Kiyota, Diana Gibraiel, Yoko Ikegami, Paola Quintanar, Kristin Speth, Tanya Slavin, Ian Wilson and Noriko Yamane-Tanaka. I didn’t ever expect to meet someone like Qin Shujun. Her search for what is right and willingness to try anything rekindles my idealism, and she makes everything more fun. Plus she introduced me to dim sum and heroically proofed my thesis (remaining typos testify only to how big the job was). Thanks to my parents, Dave and Kathy Cook. I am grateful to have a dad who taught me to read, think, and be interested in the world around me from an early age. The longer I live the more I realize the many ways my mom has set an example for me. Thanks to my brothers and sisters for interrupting me with mail, phonecalls, and visits: Sam, Denver, Martha, Aselefech, Ty, Laura, Marshall, Asrat, Tesfu, Nettie, Ellie, Rosie, Tommy, and Joe. Their schedules and lives helped keep mine in perspective. My extended family has also been extremely kind: Gladys Scherwitz, Judy Cook, Bill and Sue Scherwitz, Joe and Nella Cook, Caroline Cook, and Cy and Judy Shuster all provided much-needed encouragement (both personal and intellectual) for my endeavors. I would also like to remember my grandfathers, Philip Cook and William Scherwitz, who strongly believed in the value of an education, but passed away long before my schooling ended. Thanks to Gerald and Suzanne Muehlbauer, Ned and Debbie Wicker, and Kathy Ritenour for their support. Last but not least, I am grateful to Jeff. He more than anyone knows what the last six years have been; thank you for always asking what the truth of a flower is.  xvi  Nature gives most of her evidence in answer to the questions we ask her. Here, as in the courts, the character of the evidence depends on the shape of the examination, and a good cross-examiner can do wonders. He will not indeed elicit falsehoods from an honest witness. But, in relation to the total truth in the witness’s mind, the structure of the examination is like a stencil. It determines how much of that total truth will appear and what pattern it will suggest. C.S. Lewis The Discarded Image  xvii  CHAPTER 1 INDEXICAL VERSUS ANAPHORIC CPS  1.1  Proposal: Indexical vs. anaphoric CPs  This thesis proposes that there is a fundamental division between sentence-types (CPs) which anchor to the speech act, and those which do not. I call the first type indexical CPs: these are sentences which are obligatorily interpreted deictically – with respect to the speaker, the speech time, and the speech place (Bühler 1934, Bar-Hillel 1954, Fillmore 1975, 1982, Ehlich 1982, Kaplan 1989, Schlenker 2003). (1)  CPindexical  I call the second type of CP anaphoric. Anaphoric CPs are interpreted with respect to some other element; just as anaphoric expressions are interpreted relative to an antecedent (Bühler 1934, Ross 1967, Langacker 1969, Fillmore 1975, Reinhart 1976), so with anaphoric CPs. (2)  XPantecedent … CPanaphoric  Starting with the idea that a proposition cannot be evaluated until it is given a context (Austin 1950, Barwise 1981), I model the difference between indexical and anaphoric clauses within a situation semantics framework as a difference in the value of the situation. With indexical clauses the proposition is evaluated with respect to the speech situation (s0). With anaphoric clauses, the situation is not specified – rather it is anaphorically given (s). (3)  a.  INDEXICAL CP  CP 2 s0 2 C XP 5  b.  ANAPHORIC CP  …  CP 2 s 2 C XP 5 1  The division between indexical and anaphoric CPs is motivated on the basis of the clause-typing system of Plains Cree, an Algonquian language spoken on the plains of western Canada and the United States. Plains Cree has an explicit clause-typing system whereby every clause is morpho-syntactically coded for its clause-typing status. For example, a clause can have two entirely different sets of inflectional morphology, depending on whether it is in the INDEPENDENT  or CONJUNCT order. In (4a) there is a first-person morpheme ni- preceding the  stem wâpam ‘x sees animate’; a grammatical function coding morpheme -â, and a third-person element -w. In (4b), the ni- has been replaced by the element ê-, and the two morphemes -â and w have been replaced by a single morpheme coding a first-person subject and a third-person object. (4)  Plains Cree (Algonquian) a.  niwâpamâw atim ni- wâpam -â -w atim 1- see.VTA -DIR-3 dog ‘I see a dog.’  INDEPENDENT  b.  ê-wâpamak atim ê- wâpam -ak atim C1-see.VTA -1>3 dog ‘… I see a dog.’  CONJUNCT  INDEPENDENT order clauses are restricted only to (a subset of) matrix clauses, and instantiate what I am calling indexical clauses. I will show that they are associated with a particular set of semantic properties deriving from their anchoring to the speech act: they are interpreted relative to the speech time, speech place, and the speaker. CONJUNCT clauses have a much wider distribution, which depends on a further subdivision determined by the affixes on the left and right edges of the clause. All CONJUNCT clauses can be embedded; those with the left-edge ê- morpheme may also occur in matrix clauses – but, as I argue, without the deictic properties of INDEPENDENT clauses.  CONJUNCT  clauses, as a  class, instantiate what I am calling anaphoric clauses: they are licensed either by a linguistic antecedent or by a shared context (cf. Fillmore 1975, Reinhart 2003, Kratzer 2007). Anaphoric CONJUNCT  clauses differ in how this licensing is achieved; some specifically require  subordination to an antecedent, while others do not.  2  1.2  Relation to previous work  The clause-typing distinction between indexical and anaphoric CPs is connected to – and therefore brings together – several distinct fields that have had significant previous research, including clause-typing (§1.1.1, 1.1.2), indexicality (§1.1.3), anaphora (§1.1.4), and Algonquian linguistics (§1.1.5).  1.2.1 Connection to the matrix/embedded distinction The indexical versus anaphoric distinction cross-cuts the familiar division between matrix and embedded clauses (Hockett 1958) that has been discussed for many Indo-European languages. For Plains Cree, this means that the INDEPENDENT/CONJUNCT clause-typing system does not directly map onto the matrix/embedded distinction (cf. §2.3). Cross-linguistically, I expect that clauses which are morpho-syntactically or syntactically typed as matrix clauses will subclassify into indexical and anaphoric clauses, as in table 1.1. Matrix Embedded ✔ ✖ Indexical ✔ ✔ Anaphoric Table 1.1. Indexicality vs. embedding In addition, the properties of anaphoric clauses – which can occur in both matrix and embedded contexts – have properties in common with clauses that participate in clause-chaining (Longacre 1983, Finer 1985, Stirling 1993, among many others), but does not have to stipulate these chains as a special kind of clause (cf. Givón 2001). Both of these issues are addressed in chapter 4.  1.2.2 Connection to illocutionary force The distinction between indexical and anaphoric clauses also cross-cuts illocutionary force (i.e., the distinction between declaratives, interrogatives, and/or imperatives; cf. Cheng 1991, 3  Chomsky 1995, Rizzi 1997, Portner 1999). This means that Plains Cree’s INDEPENDENT/CONJUNCT  clause-typing system will not map directly onto illocutionary force; for  example, both indexical and anaphoric clauses may in principle be (and in Plains Cree, are) declarative or interrogative.  1.2.3 Parallel between indexical CPs and indexical expressions Indexical expressions go back at least as far as Bühler (1934), and are defined for the purposes of this thesis as in (5), following Bühler (1934), Bar-Hillel (1954), Fillmore (1975), and Kaplan (1989), among many others. (5)  Indexical expressiondef: a linguistic element whose interpretation requires identification of the speaker, speech time and/or speech location  Typical English indexical expressions include pronominal I as well as spatial and temporal relation to the speech act (i.e., the here and now; cf. Bühler 1934, Bar-Hillel 1954). Since an indexical expression looks to the speech act for its interpretation, it is a particular kind of deictic (from Greek δεικ- ‘point out’) expression, as defined in (6) (cf. Fillmore 1975, Kaplan 1989, Green 1989, Nunberg 1993, Schlenker 2003). (6)  Deictic expressiondef:  a linguistic element whose interpretation requires pointing to some aspect of the context in which it is used  Claiming that a clause is indexical therefore means that it has the same pointing function to the speech act that any other indexical expression has. In this thesis, I argue that Plains Cree’s INDEPENDENT  order morpho-syntactically codes such a clause and does in fact have this pointing  function. Claiming that clause-typing codes a relation to the speech act relates in significant ways to the debate about the syntactic representation of ‘speech-act’ elements. This is most clearly seen in the tense literature, where reference to the speech time goes back at least as far as Paul 1886, and has been used in literally countless syntactic and semantic analyses since (in chronological order: Reichenbach 1947, Klein 1994, Kamp 1981, Stowell 1982, Kamp & Rohrer 1983; Enc 1987, Stowell 1995, Abusch 1998, Demirdache & Uribe-Etxebarria 2000, among 4  many others). In the current work, temporal deixis on the speech time is taken to be part of the general deixis on the speech situation (cf. Kratzer 2007). Similarly, there is much current work on the representation of a speaker or a speech-act-phrase; this is apparent particularly in the evidential literature (cf. Cinque 1994, Rivero 1994, Rizzi 1997, Speas & Tenny 2003, 2005), but also in work on speaker-oriented truth (Lasersohn 2005, Stephenson 2007), and in the linguistic structure of discourse (Banfield 1982, Smith 2003). In Plains Cree, all of these phenomena are associated with a particular clause-type, and thus the work reported on here – and the analysis pursued – is an attempt to show how these concepts might be linked. In particular, I model these relations within a situation semantics framework, where every proposition must be evaluated with respect to a situation (Austin 1950, Barwise 1981, Barwise & Perry 1983, Barwise & Etchemendy 1987, Kratzer 2007a). I argue that in an indexical clause this situation as the speech situation1. Following Kratzer (1989, 2007a), I define a situation s as a partial world, where a partial world is a domain for truth evaluation that does not necessarily contain truth values for all possible propositions. (7)  Situationdef:  a partial world  Within this framework, propositions are evaluated relative to situations. The speech situation s0 is simply a situation in which someone is speaking. When we think about what would be necessary for a situation to be called a speech situation, we would minimally need to include the individuals who are doing the speaking and hearing (i.e., the Speaker and Hearer); the time at which the speaking occurred (i.e., Speech Time), and the place at which the speaking occurred (i.e., Speech Location). The definition of a speech situation is given in (8). (8)  Speech situationdef: a situation minimally involving (i) the Speaker/Hearer; (ii) the Speech Time; and (iii) the Speech Location  A speech situation thus captures the relation between, for example, temporal effects and person effects.  1  See Kaplan 1989 for a treatment of indexicals in contexts, rather than situations.  5  1.2.4 Parallel between anaphoric CPs and anaphoric expressions Anaphoric expressions have also been an enormous research topic in linguistics. Working on pronominal forms in English, there are as many proposals about the relevant principles governing the licensing of anaphora as there are linguists, where licensing is defined as in (9). (9)  Licensingdef: an element α is licensed iff there is some element β able to serve as its context of interpretation  Fillmore (1975) provides discussion of how anaphora can be licensed both by an antecedent (antecedence licensing), and in the absence of a linguistic antecedent (shared context, termed symbolic licensing). There is general agreement that the latter case is not an entirely separate licensing mechanism from the first case (Heim & Kratzer 1998, Reinhart 2003); however, most of the work has tried to specify the conditions on the first mechanism, antecedence licensing. In particular, there have been disagreements about whether anaphora can be accounted for strictly by notions of hierarchy (e.g., the antecedent c-commands the anaphor; see Reinhart 1976, 1983, Kayne 1994), or whether both hierarchy and precedence (e.g., the antecedent precedes the anaphor) are relevant (Langacker 1967, Ross 1967, 1969, McCawley 1988, Carden 1986, Williams 1997). There is also disagreement as to whether anaphora have special properties (this seems to be the standard position), or whether they are an elsewhere case (as explicitly argued in McCawley 1970 and Williams 1997). This thesis contributes to the discussion in at least three ways. First, it extends the discussion of pronominal anaphora into the domain of clauses by claiming that there is a particular kind of clause that can host (both antecedence and symbolic) anaphoric relations. Second, it explicitly claims that anaphoric clauses are an elsewhere case: anaphoric clauses occur in contexts where an indexical clause fails to occur. Third, the data set considered here has the same licensing conditions as discussed by Carden (1986) and Williams (1997): i.e., c-command and precedence are split into separate conditions, with some anaphoric clauses needing only precedence, some needing only c-command, and others requiring both.  6  1.2.5 Connection to the INDEPENDENT/CONJUNCT contrast Within Algonquian linguistics, the distinction in clause-typing discussed in this thesis has been difficult to understand and analyze, despite a relatively long history of linguistic work on Plains Cree (Howse 1865, Lacombe 1874, Bloomfield 1928, Wolfart 1973, Dahlstrom 1991, Ogg 1991, Wolfart & Carroll 1996, Blain 1997, Long 1999, Hirose 1999) and related languages (in particular Reinholtz 1996, 1999, 2007 for Swampy Cree). It is hoped that the analysis developed here for Plains Cree’s INDEPENDENT/CONJUNCT distinction will serve as a basis for a more finegrained description and analysis of the function of clause-typing across the Algonquian language family. In particular, while the INDEPENDENT/CONJUNCT contrast in Plains Cree maps in a oneto-one fashion to the indexical/anaphoric contrast (language 1 in table 1.2), it is possible that in other languages, the INDEPENDENT order extends across both clause-types (language 2), or alternatively, the CONJUNCT order extends across both (language 3). Clause type Indexical  Language 1 (=Plains Cree) independent  Anaphoric  conjunct  Language 2 (= ??) independent  Language 3 (= ??) independent conjunct conjunct  independent conjunct Table 1.2. Mapping the indexical/anaphoric contrast onto INDEPENDENT and CONJUNCT Within Plains Cree, Blain’s (1997) thesis on wh-questions proposes a structural analysis of two kinds of CONJUNCT clauses within a restricted set of contexts, but the extent to which the analysis can be generalized across the language has not been addressed. Although her analysis is only relevant for specific parts of this thesis, the findings here are consistent with her claims. Long’s (1999) thesis on complement clauses in Plains & Swampy Cree similarly proposes a structural analysis of one kind of CONJUNCT clause in ‘complement’ contexts, and provides a number of diagnostics to structurally distinguish complement clauses from adjunct clauses. The work in chapter 5 builds on this analysis, adding more clause-types, more diagnostics, and proposing a third syntactic clausal relation: chains. Other previous work on Plains Cree has primarily focussed on the (large amounts of) morpho-phonology (cf. Lacombe 1874, Bloomfield 1925, Wolfart 1973), its historical relation to 7  other Algonquian languages (Goddard 1967, 1974, Pentland 1979, 1999), and morpho-syntax (Dahlstrom 1991, Déchaine 1999, 2001, 2002, 2003; Hirose 2000). This thesis has depended – at times heavily – on this previous work: without having it as a foundation, much of the current work could not have been done. At the same time, much of the data presented in this thesis – looking at the co-occurrence restrictions between the verbal complex and particles, the structural relations between clauses, and the interpretation of utterances – is novel, and many of the generalizations have not been previously discussed. Unless otherwise cited, all data comes from primary sources: either original elicitation fieldwork, or texts (i.e., transcribed recordings of narratives collected by others). This leads us to a discussion of methodology.  1.3  Methodology: Data collection and presentation  This research is based on two methods of data collection. The first method is the elicitation of introspective speaker judgments. The data reported in this thesis was collected on the basis of work with six different fluent Plains Cree speakers from Alberta and Saskatchewan2. Elicitation sessions included a variety of tasks, including (i) translation tasks, where the speaker is either presented with an English sentence and asked to translate into Plains Cree or vice versa; (ii) judgment tasks, where the speaker is presented with a Plains Cree utterance and asked to judge its well-formedness and or felicity in a context; (iii) utterance-in-context tasks, where the speaker is asked to provide an appropriate utterance in a constructed context; and (iv) analytic tasks, where the speaker provides a reason for the (un)grammaticality or (in)felicity of the utterance (Cook & Mühlbauer 2006). Elicitation work is vital to this thesis in two ways: it provides linguists with information on what are impossible utterances, and with explicit information about the (im)possible meanings that an utterance may have. At the same time, it should be noted that – as with all elicitation – different tasks sometimes lead to quite different results. In particular, a Plains Cree utterance may be translated in a particular way into English (for example, past tense), but when tested in defined contexts, turn out not to have the same distribution or felicity 2  Relevant biographical information on consultants is as follows: S1 was male, mid-50s, from Thunderchild, SK; S2 was female, mid-40s, born Lac La Biche, AB, raised in Edmonton, AB; S3 was female, early 50s, from Ermineskin, AB; S4 was male, mid-50s, from Louis Bull, AB; S5 was female, early 40s, from Little Pine, SK; S6 was female, mid-60s, from Cold Lake, AB.  8  conditions as the English translation. Likewise, speakers may offer what turn out to be crucial insights in analytic tasks, but it is vitally important to test these insights via other tasks. The second method of data collection involves the use of corpus material. In particular, I draw from published narratives of four Plains Cree speakers that were recorded, then transcribed and (minimally) edited by Wolfart & Ahenakew3 (Whitecalf 1993, Minde 1997, Kâ-Nipitêhtêw 1998, Ahenakew 2000). These speakers are referred to in the thesis by the initials of their names:  SW, EM, JKN,  and AA, respectively. The narratives range in length from approximately  one hour of speech (SW) to approximately two and a half hours of speech (each of EM, JKN, and AA).  Generalizations from corpora are vital in that they show the possible utterances in an  organically constructed context, and also indicate the robustness of a given phenomena. Following the premises of ‘cross-methodological validation’ (Carden & Dietrich 1982, Matthewson 2004), this thesis draws on both elicited and corpus data. Used together, these two methods of data collection can be used to cross-check the validity of the data set. For example, in elicitation contexts, speakers will sometimes rule utterances as impossible because the appropriate context has not been established; in such a case the corpora can (i) give evidence that the construction really does exist, and (ii) provide a ready-made context for that construction. Likewise, sometimes the construction being targeted will not appear in the corpora for accidental reasons (e.g., the appropriate context did not occur), but a fluent speaker can readily provide a judgment and context for the utterance when asked. In many cases, the two sources of data were combined, where pieces of the corpora were presented to the elicitation consultant and the consultant was asked about the meaning, or asked about possible permutations on the attested piece. This method was particularly useful for when judgments within a particular discourse context were needed, since the Plains Cree corpora provided a ready-made context without potential interference from a distinct framing language (in this case English). All Plains Cree data, whether from elicitation or textual sources, is presented in a (minimally) four-line format as follows (the lines enclosed in parentheses are given as relevant):  3  In essence, any transcription involves editorial decisions, even at the level of word breaks. Editorial work included marking of punctuation. False starts and hiccups were transcribed as they were heard.  9  (10)  (context-of-utterance) Plains Cree data in standard Roman orthography Morphemic breakdown Morpheme-by-morpheme gloss ‘Free English translation.’ (comment(s) by consultant about utterance)  In addition, text taken from textual sources is cited from the relevant text by speaker and paragraph number within the transcription4. Following the practices of Wolfart (e.g., 2000), textual data also includes the relevant contextual punctuation marking introduced by the editors (Ahenakew & Wolfart 1997, 1998; Wolfart & Ahenakew 1993, 2000) in the following way. Preceding the cited clause, ellipsis [...] indicates preceding linguistic material with no intervening period [.] or semicolon [;]. Any other intervening punctuation, including a comma [,], colons [:], and initial or ending quotations [“”] are marked. Following the cited clause, all punctuation is marked; if the punctuation is anything other than a period or semicolon, another ellipsis follows. Finally, it should be noted that because the translation is not always word-for-word, there are times where the punctuation of the English translation differs from the Plains Cree (for example, the Plains Cree clause may be in sentence initial position, but the English translation of that clause is in non-initial position). The internal morphological structure of stems is not usually given, since the relation between elements within the stem is different from the relation between elements external to the stem (Wolfart 1973). If necessary, stem-internal morpheme breaks are given within brackets [STEM-MORPHEMES]. A list of the abbreviations used in the morpheme-by-morpheme gloss is given in the front matter. As with any gloss, these are approximate and should not be taken as having any analytic or ‘real’ value. There are three symbols that may precede the Plains Cree line of an example: an asterisk [*], an exclamation point [!], or a pound sign [#]. The asterisk marks a string that was judged by one or more fluent speakers to be ill-formed – i.e., an impossible utterance. The exclamation point marks an utterance that may be well-formed, but cannot have the relevant interpretation. Such utterances are often judged ungrammatical if presented in the context of the relevant 4  This annotation makes it easy to distinguish between data from textual sources and elicitation data.  10  structure (e.g., coordination), but judged grammatical in the context of some other construction (e.g., temporal modification). Finally, the pound sign marks a string that is grammatical, but infelicitous in a particular discourse context.  1.4  Plains Cree terminology  Plains Cree, like other Algonquian languages, has three inflectional classes of words: verbs, which take one set of inflectional morphology, nouns, which take another set of inflectional morphology, and particles, which cannot be inflected (Wolfart 1973). Particles are a syntactically and semantically heterogeneous class which I will not deal with here (but see Ogg 1991 for discussion). Nominal stems may be inflected for possession and plurality (Lacombe 1874, Hockett 1966, Wolfart 1973, Dahlstrom 1991). Modifiers may attach to the left of the stem. A simplified template for nouns is given in (11). (11)  Template for nominal stems [  POSS  [ MOD [STEM] PL/OBV ] ]  Depending on the context in which they occur, nouns fall into one of three referential categories: inanimate, animate, or obviative (a subclass of animate), but nouns are not inherently specified as to their category (see arguments in Wolfart 1973, Mühlbauer 2008). Verbal complexes consist of a stem, which almost always has internal structure (Wolfart 1973, Hirose 2000, Déchaine 2003), including a root, a possible medial, and at least one final. (12)  Template for verbal stems [STEM root – (medial) – final ]  The finals are inflected for animacy and arguably code argument structure (i.e., the introduction of argument positions and the assigning of grammatical function) (cf. Hirose 1999, Déchaine 2003). To the left of the stem is the pre-verb domain, which hosts, among other things modifiers and tense/aspect/modality markers (Edwards 1954, Wolfart 1973, Cook 2004). The left and right edges of the verbal complex external to the stem have person/number marking and, in the case of CONJUNCT clauses, a closed class of left-edge morphemes hosting an 11  ablaut process called initial change (IC, cf. Wolfart 1973); these latter will be of central concern to the thesis. As we saw at the beginning of the chapter, there are two agreement paradigms5. These are called orders: there are the INDEPENDENT and CONJUNCT orders, represented by the templates in (13) and (14) respectively. (The * in the template indicates the possibility of iteration; parenthesis indicates that the element is only sometimes present.) (13)  template for INDEPENDENT order  [VERBAL COMPLEX PERSON [ (PRE-VERB *) [STEM ] ] PERSON (NUMBER) ] (14)  template for CONJUNCT order  [VERBAL COMPLEX ê-/kâ-/IC [ (PRE-VERB *) [STEM ] ] PERSON (NUMBER) ] Orders may subclassify for modes; in Plains Cree, at least for the data set I have, the CONJUNCT order is the only one to have any modes (cf. Wolfart 1973, who documents three modes for the INDEPENDENT  order in older forms of Plains Cree). These modes include (following Wolfart  1973) a primary division between simple CONJUNCT, and the changed CONJUNCT, and a further division depending on the suffixation of -i, yielding the subjunctive CONJUNCT, the and the iterative CONJUNCT. MODE  FORM SIMPLE simple (ka-)nipât subjunctive nipâci CHANGED changed ê-nipât kâ-nipât nêpât iterative nêpâci Table 1.3. The CONJUNCT modes For the data set I am working with, the simple CONJUNCT is almost universally prefixed with the irrealis preverb ka-. Both the changed CONJUNCT formed by ablaut (nêpât) and the iterative CONJUNCT  (nêpâci) are essentially absent from the data (i.e., no consultant recognized or  produced the forms, and they were only attested a handful of times in the corpora). 5  There is a third paradigm – the IMPERATIVE order. The imperative order cannot host most agreement, any of the elements on the far left edge, or most of the preverbs. I will not discuss it further in this thesis.  12  For a more in-depth discussion of Plains Cree’s grammar, see Wolfart (1973, 1996); Wolfart & Carroll (1981); and Dahlstrom (1991). For the purposes of this thesis, other Algonquian-specific terms will be introduced as necessary.  1.5  Layout of the thesis  There are six chapters following the introduction. Chapter two presents a series of arguments that the sentence-level verbal complexes I am looking at in Plains Cree form a uniform syntactic class, which within the Principles & Parameters framework of Chomsky (1981) and Chomsky & Lasnik (1993) is called a CP. I further argue that in CONJUNCT clauses the left-edge elements head the CP (i.e., they are complementizers), while in INDEPENDENT clauses the left edge elements are in spec, CP. The next two chapters address the main proposal of the thesis, that the fundamental distinction in Plains Cree’s clause-typing system is between indexical and anaphoric clauses. Chapter three is concerned with structural and semantic contexts for indexical clauses, instantiated by Plains Cree INDEPENDENT order. Structurally, I show that indexical clauses (a) are subject to anti c-command (they cannot be c-commanded); and therefore (b) require all dependent elements to be resolved locally (within the clause). Semantically, I show that indexical clauses have indexical temporal and referential properties. Chapter four is concerned with what happens when the structural and semantic context is such that an indexical clause cannot occur. Here I claim that we get anaphoric clauses, instantiated by Plains Cree’s CONJUNCT order. Syntactically, I extend Williams’ (1997) analysis of anaphora, and show that anaphoric clauses are licensed by either a precedence condition or a c-command condition. Semantically, I show that the value of temporal/spatial/referential dependent elements within an anaphoric clause are determined by an antecedent. Finally, I use Fillmore’s (1975) contextual licensing principles of anaphora to derive the distribution of matrix anaphoric clauses. The next two chapters develop more specific syntactic and semantic analyses of anaphoric CONJUNCT clauses.  13  Chapter five argues that anaphoric clauses fall into three syntactic classes, defined by their relation to another clause: (i)  chained clauses, which are governed solely by precedence and do not form a constituent with any other clause;  (ii)  adjunct clauses, which are governed by c-command and form a constituent of another clause; and  (iii)  mediated argument clauses, which are licensed by an argument-position (subject or object) and adjoined within the clause.  Chapter six argues that anaphoric clauses also fall into three semantic classes, which cross-cut the syntactic classification of chapter 5, but map onto the form of the complementizer: (i)  the complementizer kâ- introduces presupposed clauses, where the truth of the proposition being presented is assumed within the discourse;  (ii)  the  (null) complementizer introduces a-veridical clauses, where the truth of the proposition being presented is unevaluated within the discourse; and  (iii)  the complementizer ê- introduces an unspecified clause which does not carry any inherent semantic value.  Chapter seven concludes by summarizing the main findings of the thesis, and pointing out possible directions for further research.  14  CHAPTER 2 MAPPING INDEXICAL AND ANAPHORIC CPS ONTO PLAINS CREE’S MORPHO-SYNTAX  2.1  Proposal: A one-to-one mapping in Plains Cree  I claim that there are two clause-types in Plains Cree: indexical clauses and anaphoric clauses. Indexical clauses have an indexical speech situation (s0) in spec, CP, as in (1a), and anaphoric clauses have an anaphoric situation (s) in the same position (1b). (1)  a. INDEXICAL CP  CP 2 s0 2 C XP 5  b. ANAPHORIC CP  CP 2 s 2 C XP 5  Syntactically and semantically, there is a one-to-one relation between the element in spec, CP (indexical vs. anaphoric), and the kind of dependencies which a CP may have. Indexical clauses code a proposition that is evaluated with respect to the speech situation (see chapter 3). Anaphoric clauses code a proposition that is evaluated to an anaphoric situation (see chapter 4). Morpho-syntactically, the distinction between indexical CPs and anaphoric CPs could logically have one of three patterns. The first possibility is that there is no morpho-syntactic distinction between different kinds of CPs; the distinction between them is contextually determined. The second possibility is that there is a morpho-syntactic differentiation between different kinds of CPs, but the distinction is cued to factors other than the indexical vs. anaphoric property. The third possibility is that there is a morpho-syntactic differentiation between different kinds of CPs which is specifically cued to the distinction between indexical and anaphoric  15  clauses. I claim that this is the pattern we see in Plains Cree (Algonquian): the two clause types are morpho-syntactically realized by two different clausal paradigms, called orders in the Algonquianist literature, and there is a direct mapping between the morpho-syntax, syntax, and semantics of indexical and anaphoric clauses. For Plains Cree, I show that the INDEPENDENT order instantiates indexical clauses. It is characterized by proclitics on the left edge of the clause (1st-person ni- in (2a); 2nd-person ki- in (2b)), and by a particular set of right edge agreement (e.g., the speech-act-participant suffix -n in (2a-b), and the third person suffix -w in (2c)). (2)  Indexical clause = INDEPENDENT order a.  nimîcison ni-mîciso -n 1-eat.VAI-SAP ‘I’ve eaten.’  b.  kimîcison ki-mîciso -n 2-eat.VAI -SAP ‘You’ve eaten.’  c.  mîcisow mîciso-w eat.VAI-3 ‘S/he’s eaten.’  For Plains Cree, I also show that the CONJUNCT order, of which there are several varieties, morpho-syntactically instantiates anaphoric clauses. What CONJUNCT clauses share with each other is a distinct set of right-edge morphology (for example, 3rd-person -t in (3a-c)).  16  (3)  Anaphoric clauses = CONJUNCT order (citation forms) a.  ê-wâpamât ê- wâpam -â -t C1-see.VTA-DIR-3 ‘…s/he sees him/her’  b.  kâ-wâpamât kâ- wâpam -â -t C2-see.VTA -DIR-3 ‘…when s/he saw him/her’  c.  wiyâpamâci1 IC- wâpam -â -t -i IC- see.VTA-DIR-3-PL ‘…whenever s/he saw him/her’  On the left edge, the data in (3) illustrates a number of different clause-typing elements, all of which are associated with an ablaut process known in Algonquian linguistics as initial change (IC); these include the proclitics ê- and kâ-, (ablauted from i- and kî-, respectively) and stem infixation (-iy-). In the absence of initial change, CONJUNCT clauses are termed ‘simple CONJUNCT’; they have an irrealis element (ka- or -i)2. These again share the right-edge agreement. (4)  Anaphoric clauses = CONJUNCT order (citation forms) a.  ka-wâpamât ka- wâpam -â -t IRR-see.VTA -DIR-3 ‘…him/her to see him/her’  b.  wâpamâci wâpam -â -t -i see.VTA-DIR-3-SUBJ ‘…if/when s/he see him/her’  1  Historically, ablaut of the stem could also take place without the subjunctive suffix –i (see, e.g., Wolfart 1973:46, who notes that it seemed to be disappearing in favor of the ê- proclitic). I have not found any examples of this kind in any of the corpora I have worked with, and none of the speakers I work with recognize or use these forms; even with the subjunctive marker, ablaut is now extremely restricted, and I have little to say about them. 2 Historically, a simple CONJUNCT clause like (4a) was reported to be possible without ka- (Wolfart 1973:46). However, I have not worked with a speaker who controls difference between a simple conjunct clause with vs. without the ka- proclitic. Writing 35 years ago, Wolfart (1973:45) comments that forms with ka- (or its alternant (ki)ta-) was by far more common than forms without it; since then, the completely bare verbal complex seems to have (all but?) disappeared.  17  The irrealis markers are different from the clause-typing elements in that they have a wider distribution: ka- can occur in both INDEPENDENT and CONJUNCT clauses, and -i can co-occur with the clause-typing elements ê-, kâ-, and internal change (IC). I will treat clauses like in (4) as having a null complementizer for reasons that will become clear later in the chapter. The mapping between the two clause types and the two orders in Plains Cree is thus represented as in (5).3 (5)  TWO TYPES OF CLAUSES  INDEXICAL CP (= INDEPENDENT) Structure:  CP 2 s0 2 niC XP ki5 -  ANAPHORIC CP (= CONJUNCT) CP 2 s 2 C XP kâ- 5 êIC  There are two parts to my claim which must be defended: (i)  that what I’m talking about are CPs (rather than some smaller structure like an IP or a VP); and  (ii)  that there are two kinds of CPs (i.e., indexical and anaphoric).  In the current chapter, I take up the claim that the two clause-types both have the properties of CPs. To the extent that this characterization is accurate, indexical and anaphoric CPs differ not in the amount of structure they have, but in the s vs. s0 contrast in spec, CP. This chapter lays the groundwork for the later chapters, where I address the second claim, analyzing the syntax and semantics of the two clause types in detail.  3  On this account, the ni-, ki-, and - proclitics are all in the same syntactic position as the s0 constant, in effect meaning there are three ways to spell out s0. If this is an accurate representation, we raise the interesting question of having many forms mapping to the same meaning. Since I take the the s0 constant to be a characteristic property of the INDEPENENT paradigm, rather than a property of a particular morpheme, the answer to the question is not crucial to the analysis. Alternatively, we could say that s0 is external to CP altogether (Lisa Matthewson, p.c.). Again, nothing hinges on this decision.  18  First, I lay out the evidence that both INDEPENDENT and CONJUNCT verbal complexes in Plains Cree are full CPs, and that the left-edge elements (pronominal proclitics and clause-typing proclitics introduced by initial change) in particular are hosted in CP. This is important because, in principle, both the pronominal proclitics and initial change could target any level of the clause, including VP, IP and CP, and this in fact seems to be a place where there is variation across the Algonquian family. For example, it has been argued that the pronominal proclitics in Blackfoot are hosted in the IP domain (Ritter & Wiltschko 2005, 2007), and similarly, the process of initial change in Ojibwa is associated with tense (James 1982, Blain 1999, Mathieu 2008). If the pronominal proclitics and clause-typing proclitics were in a lower position, we would not expect them to correlate with clause-typing distinctions. Second, I show that the pronominal proclitics ni- ‘1st person’ and ki- ‘2nd person’ differ from the clause-typing proclitics ê-, kâ- and IC in a number of respects. I model this difference as a difference in whether the element is introduced in spec, CP (i.e., for the pronominal proclitics) or head the CP-projection (i.e., the clause-typing elements).  2.2  Diagnostics for CPs in Plains Cree  When we compare multiple elements in a class, we need a set of criteria which makes each element a member of that class. Here I am comparing members of the class of CP – that is, things which are CPs (i.e., maximal clauses). Thus, in this section, I provide the set of criteria used in determining that Plains Cree verbal complexes are CPs. First, the ordering properties of the pronominal and clause-typing proclitics are consistent with having the highest position in the clause. Second, the distributional and interpretational properties of Plains Cree’s peripheral agreement are consistent with CPs, but not IPs or VPs (cf. Déchaine 2001, 2002). Third, the sensitivity of verbal complexes to the matrix/embedded distinction is consistent with CPs, but not IPs or VPs. Finally, the complementary distribution of pronominal proclitics as opposed to clause-typing proclitics supports the claim that the pronominal proclitics are hosted by the same layer of the clause as the clause-typing proclitics.  19  Although no one of these criteria is conclusive evidence about the nature of Plains Cree verbal complexes, if they are taken together they present a coherent argument for the current analysis.  2.2.1 Ordering properties Aspect (temporal structure of the predicate), tense (temporal anchoring of the event), and modality ((ir)reality of the proposition) are taken to be in VP and IP domains of the clause (Pollock 1989, Hornstein 1990, Cinque 1999, Givón 2001, Demirdache & Uribe-Etxebarria 2002, among many others). Cook (2003, 2004) has argued that the linear ordering of elements in Plains Cree’s preverb domain reflects hierarchical structure (cf. Baker 1986, Kayne 1994, Cinque 1999): if element α precedes element β, then element α dominates element β. This position leads us to expect that within Plains Cree’s verbal complexes, elements associated with the CP-layer of the clause must precede elements associated with the IP-layer of the clause. This is borne out in the data: the pronominal proclitics and clause-typing elements precede all tense, aspect, and modality preverbs. This is schematized in (6a) for INDEPENDENT clauses and (6b) for CONJUNCT clauses (cf. also Edwards 1954). (6)  a.  [ PERSON] [modality/temp.] [ aspect] [ STEM ] [ ni-, ki- ] [ka-, kî-] [ati-, mêkwâ-, wî- ] [ … ]  INDEPENDENT  b.  [ CLAUSE-TYPING ] [modality/temp.] [ aspect] [ STEM ] [ ê-, kâ-, IC ] [ka-, kî-] [ati-, mêkwâ-, wî- ] [ … ]  CONJUNCT  A representative pair of data showing the position of the temporal shifting preverb kî-4 relative to the pronominal proclitic ni- and clause-typing proclitic ê- is given in (7).  4  See chapters 3 and 7 for discussion of kî- as a temporal shifting device.  20  (7)  a.  nikî-wâpamâw ana nâpêw ni-kî- wâpam -â -w ana nâpêw 1-PREV-see.VTA-DIR-3 DEM.AN man ‘I saw that man.’  INDEPENDENT  b.  êwakw ânima pêyak kisêyiniw ê-kî-nakiskawak,… êwakw anima pêyak kisêyiniw ê- kî- nakiskaw -ak TOPIC DEM.INAN one old.man C1-PREV-meet.VTA -1>3 ‘I met a certain old man about that, ...’ (JKN 1.1)  CONJUNCT  Likewise, (8) demonstrates that aspectual elements such as inceptive ati- always follow pronominal proclitics like ni- and clause-typing proclitics like ê-. (8)  a.  nitati-kinosin ni(t)- ati- kinosi -n 1- INCEP-tall.VAI-SAP ‘I am getting taller.’  INDEPENDENT  b.  ..., êkos êkwa, ê-ati-tipiskâk êkwa, ... êkosi êkwa ê- ati- tipiskâ -k êkwa TOPIC and C1-INCEP-night.VII-0 and ‘And so, when it was getting to be night, …’ (AA 2.2)  CONJUNCT  The temporal anchoring preverb kî- precedes the aspectual preverb ati-; this is consistent with kîoccupying a higher position in the clause. (9)  kî- precedes atieight hours nikî-ati-nôcihtânân, ... eight hours ni- kî- ati- nôcihtâ -nân eight hours 1- PREV-INCEP-pursue.VAI -1PL ‘It had taken us eight hours to go, …’ (AA 3.4)  The ordering of the preverbal elements with respect to the pronominal/clause-typing elements is consistent with the claim that the latter elements are in the highest position of all. Although the ordering does not tell us what that position is, the ordering is consistent with the claim I am making that the elements are in CP.  21  (10)  CP 3IP [person] 3 VP [c-typing] kî3 ati5  The irrealis preverb ka- interacts with the left edge in a more complex way: it is internal to the pronominal proclitics as in (11), and in complementary distribution with the CONJUNCT proclitics. (11)  a.  ..., “â, êkota nika-pôsipayihon,” ... â êkota ni- ka- pôsipayiho -n INTERJ there 1- IRR-jump.VAI -SAP ‘..., “Well, I will jump on that,” … ’ (AA 8.3)  b. * ê-ka-pôsipayihoyân ê- ka- pôsipayiho -yân C1-IRR- jump.VAI -1 --- (intended: ‘I will jump.’)  INDEPENDENT  CONJUNCT  However, unlike either the pronominal proclitics or the clause-typing proclitics, ka- can occur in both matrix INDEPENDENT and embedded CONJUNCT clauses. (12)  Irrealis ka- across different clause-types a.  ..., “â, êkota nika-pôsipayihon,” ... â êkota ni- ka- pôsipayiho -n INTERJ there 1- IRR-jump.VAI -SAP ‘..., “Well, I will jump on that,” … ’ (AA 8.3)  INDEPENDENT  b.  nikî-kwêcimâw Nettie ka-pê-itohtêt ni- kîkwêcim -â -w N ka- pê- itohtê -t 1- PREV-ask.VTA -DIR-3 N IRR-come-go.VAI-3 ‘I asked Nettie to come.’  CONJUNCT  The data in (12) shows that ka- crosscuts both the matrix/embedded distinction and Plains Cree’s INDEPENDENT/CONJUNCT  (13)  distinction. In addition, ka- precedes kî-, as in (13).  ..., nika-kî-itwân êwakw anima, ... ni- ka- kî- itwâ -n êwakw anima 1- IRR-PREV-say.VAI-SAP TOPIC DEM.INAN ‘..., I can say that, …’ (AA 2.1)  22  I take the distribution of ka- to be significant. In particular, even though ka- interacts with clause-typing elements, it is crucially not restricted to one of the clause-types (i.e., INDEPENDENT or CONJUNCT). Therefore, it must be sitting in a lower position than both the pronominal proclitics (with which it co-occurs) and the clause-typing proclitics (which are restricted to CONJUNCT order).  Taking its ordering with respect to kî-, its complementarity with the clause-typing elements ê- and ka-, and its irrealis meaning into account, I model ka- as a finiteness complementizer (Cfin; Rizzi 1997). This position is distinct from and lower than force complementizers (Cforce).5 (14)  CPforce 3CPfin [person] 3 IP [c-typing] ka3 VP kî3 ati5  The co-occurrence restrictions between ka- and the clause-typing elements can be seen as an instance of local head-to-head interaction (e.g., only one complementizer may be overt at a time). Summarizing, we see that the position of the pronominal proclitics and clause-typing elements is consistent with them being in the CP-layer of the clause. I now look at some evidence that they are in fact in this layer of the clause.  2.2.2 Peripheral agreement diagnoses CPs Algonquian languages are famous for the abundance of agreement they exhibit. Consider the INDEPENDENT  clause in (15). From left to right, we see a pronominal proclitic ni-, the root wâp, a  valency marker -am that is codes the animacy of the internal argument, a valency marker -â, a third person suffix -w, and a plural marker -ak.  5  I have not given evidence that ka- must be a finiteness complementizer, and in fact nothing in the following argumentation depends on it being such.  23  (15)  niwâpamâwak ni- [wâp- am] -â -w-ak 1- [light- eye.AN]see.VTA -DIR -3 -PL ‘I see them.’  INDEPENDENT  The CONJUNCT clause in (16) has a left-edge clause-typing element ê-; it shows some of the same agreement on the right edge, including identical valency markers, but it has person (-t ‘3rd') and number (-ik ‘pl’) agreement that is different from the INDEPENDENT clause. (16)  ê-wâpamâcik ê- [wâp -am] -â -t -k C1- [light -eye.AN]see.VTA -DIR -3 -PL ‘…I see them.’  CONJUNCT  The amount and kinds of agreement that these clauses exhibit provide evidence that these verbal complexes are structurally quite big. If a clause is composed of multiple domains (cf. Pollock 1989, Rizzi 1997, Cinque 1999, among others), then agreement may logically occur in any of these domains. For the purposes of this discussion, I will assume that a clause has at least the layers of VP (i.e., the predicate domain where theta-marking agreement occurs); IP (the inflectional domain where grammatical function agreement occurs); and CP (the clause-typing domain where discourse function agreement occurs). (17)  CP 5 IP 5 VP 5    discourse function agreement    grammatical function agreement    theta-marking agreement  Agreement that remains constant across distinct clause-typing environments (e.g., matrix vs. embedded clauses or declarative vs. interrogative clauses) is not a good candidate for CPagreement, and I will not discuss it here (see Déchaine 2001, 2002 for discussion). However, for several reasons, the peripheral person and number agreement which alternates in Plains Cree’s INDEPENDENT  and CONJUNCT order provide evidence that the verbal complexes under analysis  are full CPs, rather than a reduced constituent (e.g., an IP or a VP):  24  (i)  the agreement does not map directly onto either theta-roles or grammatical-function;  (ii)  the agreement does not interact with finiteness; and  (iii) the agreement correlates with clausal embeddedness. First, person and number agreement are insensitive to grammatical function (i.e., subject vs. object vs. indirect object). In (18), the third person -w suffix of the INDEPENDENT order can act as a subject (18a), an object (18b), or an indirect object (18c). (18)  a.  mâtow mâto -w cry.VAI-3 ‘s/he is crying/cries’  b.  ninakiskawâw ni- nakiskaw -â -w 1- meet.VTA -DIR-3 ‘I met him/her.’  c.  niwihtamawâw ni- wihtam -aw -â -w 1- tell.VTA-BEN-DIR-3 ‘I tell this to him/her.’  Similarly, in (19) the third person -t suffix of the CONJUNCT order can mark subjects (19a), objects (19b), or indirect objects (19c). (19)  a.  ê-mâtot ê- mâto -t C1-cry.VAI-3 ‘s/he is crying’  b.  ê-mâmitonêyihtamikot ê- mâmitonêyihtam -iko -t C1-trouble.VTA -INV -3 ‘it (inan.) troubles him/her’  c.  ê-itôtâkot ê-itôtê-aw-iko-t C1-do.VTA-BEN-INV-3 ‘it did this to him/her’  25  The plural markers which occur on the far right edge of the verbal complex may also be associated with either the subject or object position6. For example, -ak in the INDEPENDENT order may be associated with a subject (20a) or an object (20b). (20)  a.  wâpamêwak wâpam -ê -w -ak see.VTA-DIR-3 -PL = ‘They see him/her.’ ≠ ‘S/he sees them.’  SUBJECT  b.  niwâpamâwak ni- wâpam -â -w -ak 1- see.VTA-DIR-3 -PL ≠ ‘We see him/her.’ = ‘I see them.’  OBJECT  The -k plural suffix of the CONJUNCT order may also be associated with a subject (21a) or an object (21b). (21)  a.  ê-wâpamâcik ê- wâpam -â -t -k C1-see.VTA-DIR-3 -PL = ‘ … they see him/her.’ ≠ ‘…s/he sees them.’  SUBJECT  b.  ê-wâpamakik ê- wâpam -ak -k C1-see.VTA-1>3-PL ≠ ‘…we see him/her.’ = ‘…I see them.’  OBJECT  Finally, the INDEPENDENT order pronominal proclitics ni- ‘1st’ and ki- ‘2nd’ may be associated with either a subject or an object argument. In (22), ni- and ki- are associated with the subject of a classically unergative predicate pâhpi- ‘laugh.VAI’.  6  The determination of which argument the plural is associated with is complex and depends on multiple factors. In general, they are associated with an object if a speech act participant is the subject, and associated with subject elsewhere (including cases where 3PL acts on 3PL). The point here is that it does not mark only one type of argument.  26  (22)  a.  nipâhpin ni- pâhpi -n 1- laugh.VAI-SAP ‘I laugh.’  b.  kipâhpin7 ki- pâhpi -n 2- laugh.VAI-SAP ‘You laugh.’  In (23), ni- and ki- are associated with the subject of a classically unaccusative predicate nêstosi‘tired.VAI’. (23)  a.  ninêstosin ni- nêstosi -n 1- tired.VAI-SAP ‘I am tired.’  b.  kinêstosin ki- nêstosi -n 2- tired.VAI-SAP ‘You are tired.’  In (24), ni- and ki- are associated with the subject of a transitive verb miskam- ‘find.VTA’; notice that the theme-sign immediately following the stem is the direct form -â. (24)  a.  nimiskamâw ni- miskam -â -w 1- find.VTA-DIR-3 ‘I found him/her.’  b.  kimiskamâw ki- miskam -â -w 2- find.VTA -DIR-3 ‘You found him/her.’  In (25), however, ni- and ki- are associated with the object of the transitive verb; here the subject/object association has been reversed by the use of the inverse marker (Dahlstrom 1991, Déchaine and Reinholtz 1997, 2008).  7  Note that the second-person forms are pragmatically very odd, since the contexts where a statement about the addressee can be made felicitously are extremely restricted (cf. Ross 1970; Rutherford 1970 for discussion of this problem in English). These forms are given to demonstrate their formal grammaticality.  27  (25)  a.  nimiskamik ni- miskam -ik -w 1- find.VTA-INV-3 ‘S/he found me.’  b.  kimiskamik ki- miskam -ik -w 2- find.VTA-INV-3 ‘S/he found you.’  The above data shows that the peripheral elements in Plains Cree verbal complexes is independent of the subject or object position. Plains Cree does have elements which are associated with subjects (e.g., the different subject marker -yi (Mühlbauer 2007) and the inverse marker -ik (Déchaine & Reinholtz 1997, 2008)), and elements associated with objects (direct theme signs, Déchaine & Reinholtz 2008). The insensitivity of peripheral person marking to these positions must therefore mean that it is not in the IP domain (i.e., subject domain) or the VP-domain (i.e., object domain). Combining this evidence with the ordering facts discussed above, I conclude that these peripheral elements must be external to IP. Another fact about the peripheral person marking in Plains Cree which suggests it is external to the IP-domain is that there is no correlation between agreement and finiteness of clauses. The irrealis clauses which are used in dependent clauses for commands, wishes, etc. (i.e., the contexts where English has non-finite clauses) show exactly the same kind of agreement as other kinds of embedded clauses. The (non-)contrast is shown in (26), where (26a) shows a realis clause, introduced by a factive predicate and (26b) shows an irrealis clause (with the irrealis ka-) introduced by a predicate of desire (ninitawêyimâw). The right-edge agreement (3rd-person -t) stays constant. (26)  a.  ê-wankisiyân Clare ê-nipât ê- wankisi -yân C ê- nipâ -t C1-forget.VAI-1 C C1-sleep.VAI-3 ‘I forgot Clare was sleeping.’  FINITE  b.  ninitawêyimâw Jeff ka-nipât ni- nitawêyim -â -w J ka- nipâ -t 1- want.VTA -DIR-3 J IRR-sleep.VAI-3 ‘I want Jeff to sleep.’  NON-FINITE  This is different from some other Algonquian languages, such as Blackfoot, where person  28  marking changes in some irrealis conditions (Frantz 1991, Ritter & Wiltschko 2007). For example, in (27a) the 2nd-person pronominal proclitic ki- occurs on the left edge of the embedded finite clause; in (27b), there is no pronominal proclitic, and the agreement is on the right edge of the embedded non-finite clause. This is one factor that has led Ritter & Wiltschko to posit that person marking is hosted in the IP domain. (27)  Blackfoot person marking changes relative to finiteness. a.  nitsíkohtaahsí’taki kikáó’toohsi nit- ik- ohtyaahs -i’taki k- ikáá- o’too -hs -yi 1- very-source-good -feel.VAI 2- PERF- arrive.VAI-conj-conj ‘I’m glad that you have arrived.’ (Frantz 1991:112)  FINITE  b.  ikkamáyo’kainoainiki, nitáakahkayi ikkam -á -yo’kaa -inoainiki nit- yáak- wa:hkayi if -dur-sleep.VAI -2PL.SUBJ 1- FUT- go^home ‘If you (pl) are sleeping, I’ll go home.’ (Frantz 1991:113)  NON-FINITE  The final distinguishing characteristic of Plains Cree’s peripheral agreement which is consistent with placing it in the CP layer is its sensitivity to the matrix/embedded distinction. A verbal complex with the INDEPENDENT agreement -w in (28a) cannot occur in an embedded clause (notice here that the INDEPENDENT clause has no pronominal proclitic, but still is restricted to matrix clauses). (28)  Only CONJUNCT agreement allowed in embedded clauses a.  b.  * ninitawêyimâw nîcêwâkan (ka)-mîcisow ni- nitawêyim -â -w ni- wîcêwâkan mîciso-w 1- want.vta -DIR-3 1- friend eat.VAI-3 --- (intended: ‘I want my friend to eat.’) ninitawêyimâw nîcêwâkan ka-mîcisot ni- nitawêyim -â -w ni- wîcêwâkan ka- mîciso -t 1- want.VTA -DIR-3 1- friend IRR-eat.VAI-3 ‘I want my friend to eat.’  It is also impossible for the pronominal proclitics (ni- ‘1st-person’ and ki- ‘2nd-person’) to occur in an embedded clause, regardless of the right-edge agreement. In (29), the first-person nicannot occur in an embedded clause with right-edge INDEPENDENT agreement (29a), nor can it  29  occur in an embedded clause with right-edge CONJUNCT agreement (29b). The embedded clause must have CONJUUNCT agreement without the pronominal proclitic (29c). (29)  Person proclitics are impossible in embedded clauses a.  * Sam nitawêyihtam ni-(ka)-nikamon S nitawêyihtam -w ni-ka- nikamo -n S want.VTI -3 1-IRR-sing.VAI -SAP ---  INDEPENDENT  b.  * Sam nitawêyihtam ni-(ka)-nikamoyân S nitawêyihtam -w ni-ka- nikamo -yân S want.VTI -3 1-IRR-sing.VAI -1 ---  CONJUNCT  This again differs from Blackfoot, where the pronominal proclitics occur in both matrix (30a) and embedded (30b) clauses. (30)  a.  kitáakahkayi kit- áak- ahkayi 2- FUT- go^home ‘You’re going home.’ (Frantz 1991:15)  MATRIX  b.  nitsíkohtaahsí’taki kikáó’toohsi nit- ik- ohtyaahs -i’taki k- ikáá- o’too -hs -yi 1- very-source-good -feel.VAI 2- PERF- arrive.VAI-conj-conj ‘I’m glad that you have arrived.’ (Frantz 1991:112)  EMBEDDED  In summary, then, the peripheral agreement in Plains Cree is associated with the CP domain: it is sensitive to the matrix/embedded distinction, a characteristic property of CP-level elements; and it is insensitive characteristic properties of IP-level elements such as the subject/object distinction and finiteness. In both of these respects, there is variation across the Algonquian language family, meaning that it is important to establish the position of agreement for each individual language.  30  2.2.3 Clause-typing diagnoses CPs If the CP-layer associates the proposition to a larger structure (Rizzi 1997), this implies that elements whose presence conditions the distribution and interpretation of a clause relative to its larger context must be in the CP-layer of the clause. If the element can only occur in matrix clauses, or only in embedded clauses, that element invokes a CP structure. For example, the complementizer that in English is taken to be in C: it specifies the clause as a complement clause and can only be found in embedded structures (Rosenbaum 1967, Emonds 1976). (31)  a. *  that I’m tired. [CP [C that [IP I’m tired ] ] ]  b.  I told my brother that I’m tired. I told my brother [CP [C that [IP I’m tired ] ] ]  In Plains Cree, the form of the verbal complex is dependent on the matrix/embedded distinction. As we have already seen, INDEPENDENT clauses cannot be embedded. (32)  INDEPENDENT  a.  b.  clauses are sensitive to the matrix/embedded distinction  ninêstosin ni-nêstosi-n 1-tired.VAI-SAP ‘I’m tired.’ * nikî-wîhtamawâw nisîmis ninêstosin ni- kîwîhtamaw -â -w ni- sîmis ni-nêstosi -n 1- PREV-tell.VTA -DIR-3 1- sibling 1- tired.VAI-SAP --- (intended: ‘I told my younger brother that I’m tired.’)  MATRIX  EMBEDDED  One could argue that some overt complementizer or subordinator is needed to subordinate INDEPENDENT  clauses (e.g., if one tried to posit matrix clauses as IPs). However, adding an overt  subordinating particle like osâm ‘because’ does not help, as (33) shows.  31  (33)  osâm ‘because’ can only introduce INDEPENDENT clauses a.  b.  * nawac ê-kî-cihkêyihtahkik, osâm kî-sâkihitowak INDEPENDENT nawac ê- kî- miyawâtam -k -k, osâm kî- sâkih -ito -w -ak more C1-PREV-happy.VTI -0 -PL because PREV-love.VTA-REFL-3 -PL --..., nawac ê-kî-miyawâtahkik, osâm ê-kî-sâkihitocik, … CONJUNCT nawac ê- kî- miyawâtah -k -k osâm ê- kî- sâkih -ito -t -k more C1-PREV-happy.VTI -0 -PL because C1-PREV-love.VTA-REFL-3-PL ‘..., they had been happier even when they were poor, because they used to love one another, …’ (EM 6)  In CONJUNCT clauses, the presence and particular form of clause-typing element determines the distribution of the verbal complex as a whole with respect to matrix and embedded environments. The element kâ- and internal change both restrict the verbal complex to embedded clauses. (34)  kâ-clauses must be embedded a.  b.  (35)  * atim kâ-mêkwâ-nipât atim kâ- mêkwâ- nipâ -t dog C2- MIDST- sleep.VAI-3 --nikî-atoskân atim kâ-mêkwâ-nipât ni-kî- atoskâ -n atim kâ- mêkwâ- nipâ -t 1-PREV-work.VAI-SAP dog C2- MIDST- sleep.VAI-3 ‘I worked while the dog slept.’  IC-clauses  a.  b.  MATRIX  EMBEDDED  must be embedded  * kiyîsîhtâci pêyak wâskahikan iy-kîsîhtâ -t -i pêyak wâskahikan IC-finish.VAI-3-SUBJ one house ---  MATRIX  ..., kiyîsîhtâci pêyak wâskahikan, kotakihk ê-itohtêt; EMBEDDED iy-kîsîhtâ -t -i pêyak wâskahikan kotak -ihk ê- itohtê -t IC-finish.VAI-3-SUBJ one house other -LOC C1-go.VAI -3 ‘..., and when he had finished one house, he went to the next;’ (AA 1.9)  32  Substituting the clause-typing proclitic ê- for kâ- or internal change correlates with a change in the distribution of the CONJUNCT verbal complex; with ê-, the verbal complex may occur in both matrix and embedded environments. (36)  ê-clauses allow both matrix and embedded clauses a.  atim ê-nipât atim ê- nipâ -t dog C1- sleep.VAI -3 ‘...the dog is sleeping.’  MATRIX  b.  nikî-wâpahtên atim ê-nipât ni- wâpahtê-n atim ê- nipâ -t 1- see.VTI -SAP dog C1- sleep.VAI -3 ‘I saw that the dog was sleeping.’  EMBEDDED  Finally, if the left-edge process of initial change is absent (so-called simple CONJUNCT), the verbal complex is also restricted to embedded clauses. A simple CONJUNCT clause can only be introduced by higher predicates, as in (37). (37)  Simple CONJUNCT clauses must be embedded a.  b.  (38)  * (ka-)mîcisot ka- mîciso -t IRR-eat.VAI-3 --ninitawêyimâw nîcêwâkan ka-mîcsot ni- nitawêyim -â -w ni- wîcêwâkan ka- mîciso -t 1- want.VTA -DIR-3 1- friend IRR-eat.VAI -3 ‘I want my friend to eat.’  MATRIX  EMBEDDED  Subjunctive CONJUNCT clauses must be embedded a.  b.  * wâpamaki Jeff wâpam -ak -i J see.VTA-1>3-SUBJ J --wâpamaki Jeff, nika-wîhtamawâw kâ-itwêyan wâpam -ak -i J ni- ka- wîhtamaw -â -w kâ- itwê -yan see.VTA-1>3-SUBJ J 1- IRR-tell.VTA -DIR-3 C2- thus.say.VAI-2 ‘Should I see Jeff, I’ll tell him what you said.’  MATRIX  EMBEDDED  33  In CONJUNCT clauses, then, it is the left-edge element that determines the distribution of the verbal complex. The table summarizing the distribution is given below. Matrix Embedded ✖ ✔ kâ✖ ✔ IC✔ ✔ ê ✖ ✔ Table 2.1. Left-edge clause-typing proclitic determines distribution of clause Since it is the choice of clause-typing element that correlates with distribution in matrix vs. embedded environments, these are the elements that look much like that in English – their distributional effect is consistent with putting them in C. (39)  [CP [C ê- / kâ- / IC [IP … ] ] ]  2.2.4 Pronominal proclitics are complementary with clause-typing In the last section we saw that the left-edge proclitics in the CONJUNCT order have a clause-typing function. If we compare these proclitics to the pronominal proclitics, we see additional evidence that pronominal proclitics are in CP: they are in complementary distribution with the clausetyping elements (including ê-, kâ-, and internal change) (cf. Wolfart 1973, Blain 1997). This is illustrated in (16-17), with the first-person marker ni- and the clause-typing element ê-. We have already seen that ni- and ê- can both occur in matrix clauses, However, as (40) shows, ni- and êcannot co-occur with right-edge INDEPENDENT order agreement -n.  34  (40)  a.  * ni(t)-ê-kî-mâton nit- ê- kî- mâto -n 1- C1-PREV-cry.VAI-SAP --- (intended: ‘I cried.’)  b.  * ê-ni-kî-mâton ê- ni- kî- mâto -n C1-1- PREV-cry.VAI-SAP --- (intended: ‘I cried.’)  And as (41) shows, ni- and ê- cannot co-occur with right-edge CONJUNCT order agreement -yân. (41)  a.  * ni(t)-ê-kî-mâtoyân nit-ê-kî-mâtoyân nit-ê-kî-mâto-yân --- (intended: ‘…I was crying’)  b.  * ê-ni-kî-mâtoyân ê-ni-kî-mâto-yân C1-1-PREV-cry.VAI-1 --- (intended: ‘…I was crying’)  There is no semantic reason why a first-person marker should be incompatible with a complementizer, since many languages permit this (e.g., English: John told me that I was going to win.). Thus, we can conclude that the incompatibility of the person prefixes and complementizers is a syntactic problem: their complementarity arises from the fact that they are both within the same layer of the clause (i.e., the CP).  2.2.5 Interim summary: Verbal complexes are CPs I have presented a four-part argument that verbal complexes in Plains Cree are CPs, with the pronominal proclitics and clause-typing proclitics specifically hosted in the CP-layer of the clause: (i)  pronominal proclitics and clause-typing proclitics precede all tense/aspect and modality preverbs;  (ii)  peripheral agreement has does not have any properties associated with IP (e.g., subjecthood, sensitivity to finiteness);  35  (iii)  the distribution of the clause in matrix vs. embedded contexts is determined by the presence of pronominal proclitics and/or choice of clause-typing proclitics;  (iv)  pronominal proclitics and clause-typing proclitics are in complementary distribution.  In the next section I distinguish between the different verbal complexes, showing that INDEPENDENT CONJUNCT  clauses have an overtly filled specifier position (spec, CP), that CHANGED  clauses have an overt complementizer (i.e., C), and that SIMPLE CONJUNCT clauses  have a covert complementizer (C).  2.3  Diagnosing C vs. spec, CP  In this section, I give evidence to support the claim that the pronominal forms (ni- ‘1st’ and ki‘2nd’) are in spec, CP while the clause-typing proclitics are in C. The diagnostics I use to determine whether pronominal proclitics are heads or not are: (a) whether they select for a complement; (b) whether substitution of elements within the same class changes the distribution of the clause; and (c) the (non-)significance of a covert element. In each case, the pronominal proclitics diverge from the clause-typing proclitics; the former are consistent with specifiers, the latter with heads. DIAGNOSTIC  PRONOMINAL PROCLITICS (=SPEC)  CLAUSE-TYPING PROCLITICS (=HEAD)  ✖ ✔ select for complement? ✖ ✔ substitution determines distribution? ✖ ✔ covertness significant? Table 2.2. Diagnostics for specifiers vs. heads I discuss each of these diagnostics in turn in the following subsections.  36  2.3.1 Selection of complement Turning our attention first to the pronominal proclitics, we see that they are not specific to clauses. Rather, they are analogous to the paradigm for possessors in nominals (Wolfart 1973, Ahenakew 1987, Dahlstrom 1991), and in fact their use in the verbal is argued to be a historical extension from the nominal domain (Goddard 1967). In (42a), we see the pronominal prefix attaching to a verbal stem and associated with one of the arguments of the predicate; in (42b), this same prefix attaches to a nominal stem and marks the possessor: (42)  a.  nicihkêyihtên ni- cihkêyihtê -n 1- happy.VTI -SAP ‘I’m happy.’  b.  nimaskisin ni- maskisin 1- shoe ‘my shoe’  The full paradigms are given in Table 2.3. For the first and second person, both nominals and clauses have a left edge ni- or ki-; and share the same right-edge plural marking, including 1pl. excl. -nân; 1-2.pl -naw; and 2.pl -wâw (note that the verbal version of the latter has an extended form –nâwâw). In the third person, the verbal complex lacks the prefix o-, (a fact I will return to in §3.2.3.2) but again the verbal and nominal paradigms have the same right-edge marking in the 3.obv form (-yiwa).  PERSON CATEGORY 1.sg. 2.sg. 1.pl. excl. 1.pl. incl. 2.pl 3.sg. 3.pl 3.obv  INDEPENDENT ORDER ninipân kinipân ninipânân kinipânaw kinipânâwâw nipâw nipawak nipâyiwa  NOMINAL POSSESSION nimis kimis nimisinân kimisnaw kimisiwâw omisa omisiwâwa omisiyiwa  Table 2.3. Person marking in INDEPENDENT and possession paradigms 37  Pronominal proclitics do not select for a particular kind of XP: they are neutral with respect to the distinction between DPs and CPs. (43)  a.  CP 2 ni- 2 ki- C 4  b.  DP 2 ni- 2 ki- D 4  Syntactically, this is a classic difference between heads and specifiers: heads must select for a complement, but a specifier does not. In the case of complementizer C heads, the head selects for a clausal constituent. However ni- and ki- are not selecting for a clausal constituent – the presence of ni- or ki- does not identify the constituent it attaches to as a clause. This is evidence that they are hosted in spec, CP, rather than C. While pronominal proclitics are found in both verbal and nominal contexts, the clausetyping proclitics occur only in verbal contexts. For example, in (44), the proclitic ê- may not attach to a nominal stem, but it may attach to a verbal stem (identifiable by the -t agreement). (44)  a.  b.  * ê-minôs ê- minôs C1-cat -ê-minôsit ê- minôs –i -t C1-cat -EPEN-3 ‘S/he is a cat.’  Thus, the clause-typing proclitics do identify a clause. Notice that categorial properties of the verbal stem are identifiable by the right-edge agreement, and recall that we saw reason to suppose that this agreement is very high in the clause, above the IP-layer of the clause. Following Déchaine (2001), I therefore take this agreement to occupy the phrase which is selected by the complementizer, as in (45): the whole verbal predicate mâto- ‘cry.VAI’ rises from its lower position to sit in the spec of the Agr Phrase, giving rise to the discontinuity of the two elements.  38  (45)  [CP [C ê- [XPverb mâtoyân ] ] ] CP 3 3 AgrP C 3 êXPi 3 IP 4 Agr 5 mâto -yân … ti …  Since nominal stems lack this verbal agreement, the clause-typing proclitic cannot select for an appropriate complement, yielding ungrammaticality, as with minôs ‘cat’. (46)  [CP [C ê- [XPnoun minôs ] ] ]  Adding the appropriate selectional material (i.e., clausal agreement) satisfies selection. Notice that when the nominal stem is framed by the clause agreement, we get a predicate reading: ‘he is a shoe’; this is a common strategy in Plains Cree. (47)  ê-minôsit ê- minôs -i -t C1-cat -EPEN-3 ‘…s/he is a cat.’  (48)  [CP [C ê- [AgrP minôsit ] ] ] CP 3 3 AgrP C 3 êXPi 3 4 Agr IP minôs -t 5 … ti …  With respect to selection, the clause-typing proclitics exhibit behaviour that is quite distinct from the behaviour of the pronominal proclitics: the former behave like heads, the latter do not.  39  2.3.2 Substitution (does not) determine distribution Specifier positions also differ from heads in that interchanging the form of the former should not necessarily change the clause’s distribution relative to external linguistic structure, while interchanging the form of the latter (i.e., the complementizer) should. For example, there are several wh-words that may move to spec, CP in English (Ross 1967, Huang 1982, Richards 1997). (49)  a.  What did you do yesterday?  [CP whati [C did [IP you do ti yesterday ] ] ]  b.  Why did you do that?  [CP whyi  c.  Where did you find it?  [CP wherei [C did [IP you find it ti] ] ]  [C did [IP you do it ti] ] ]  Substituting what for who will change the question that is being asked, but it does not change the fact that the clause is a wh-clause with wh-syntax. However, the element in C is invariant across this movement: it is always an auxiliary moved from I. We cannot substitute a different complementizer, such as that, or while. (50)  a.  * What that you did yesterday?  [CP whati [C that [IP you did ti yesterday] ] ]  b.  * What while you did yesterday?  [CP whati [C while [IP you did ti yesterday] ] ]  If we apply this logic to Plains Cree clauses, we get a split between pronominal proclitics and clause-typing proclitics. Clause-typing proclitics look like English complementizers: substitution changes the distribution of the clause. For example, the ê- proclitic obligatorily occurs in clauses associated with an object position. Absence of the ê- proclitic (51b) or replacement with a different proclitic (51c) result in a clause that cannot be interpreted as an object clause. (51)  a.  Jeff ê-wanikiskisit ê-mîcisot J ê- wanikiskisi -t ê- mîciso -t J C1- forget.VAI -3 C1-eat.VAI-3 ‘Jeff forgot that he had eaten.’  [CP [C ê- [XP mîcisot ] ] ]  40  b.  ! Jeff ê-wanikiskisit mîcisoci J ê- wanikiskisi -t mîciso -t -i J C1-forget.VAI -3 eat.VAI-3 -SUBJ --- (intended: ‘Jeff forgot that he was eating.’)  [CP [C  [XP mîcisoci ] ] ]  c.  ! Jeff ê-wanikiskisit kâ-mîcisot J ê- wanikiskisi -t kâ-mîciso -t J C1-forget.VAI -3 C2-eat.VAI -3 --- (intended: ‘Jeff forgot that he was eating.’)  [CP [C kâ- [XP mîcisot ] ] ]  Pronominal proclitics, on the other hand, do not distinguish between clause-types. For example, substituting the pronominal form does not change the CP’s inability to be embedded. In (52), the clause hosting the proclitic ni- cannot be embedded (52b). (52)  a.  b.  niwâpamik ni- wâpam -ik -w 1- see.VTA-INV-3 ‘He saw me.’ * Jeff niwîhtamâk niwâpamik J ni- wîhtam -aw -ik ni- wâpam -ik J 1- tell.VTA -BEN-INV 1- see.VTA-INV ---  MATRIX  EMBEDDED  In (53), ki- has been substituted for ni-, and the clause is still unable to be embedded (53b). (53)  a.  b.  kiwâpamik ki- wâpam -ik 2- see.VTA-INV ‘He saw you.’ * Jeff niwîhtamâk kiwâpamik J ni- wîhtam -aw -ik ki- wâpam -ik J 1- tell.VTA-BEN-INV 2- see.VTA -INV ---  MATRIX  EMBEDDED  41  Finally, in (54) there is no proclitic at all, and the clause is again unable to be embedded (54b). (54)  a.  b.  nimâma wâpamik ni- mâma wâpam -ik 1- mom see.VTA-INV ‘My mother saw him/her.’  MATRIX  * Jeff niwîhtamâk nimama wâpamik J ni- wîhtam -aw -ik ni- mâma wâpam -ik J 1- tell.VTA-BEN-INV 1- mom see.VTA-INV ---  EMBEDDED  I therefore posit a structure in which the pronominal proclitics are in spec, CP, rather than C: substitution of the pronominal proclitic (i.e., ni- vs. ki- vs. -) does not change the distribution of the clause. (55)  CP 2 ni2 kiC 5 -  The behaviour of the clauses with a - proclitic leads us to the next point: the significance of non-overtness.  2.3.3 The significance of non-overtness Just as substitution of different forms has different consequences for specs vs. heads, so the absence of a form has consequences. For heads, the absence of a form means either that the head is gone (resulting in less structure), or that there is a null head (which should affect the form and function of the phrase it projects to (56)). (56)  a.  XP 5  b.  CP 2 2  XP 5  42  The structure in (56b) more accurately captures what we see with the clause-typing proclitics. In (57a) we see a clause wâpahtam Jeff ‘Jeff saw it’ introducing a dependent proposition with the clause-typing proclitic ê-. If this clause-typing proclitic is absent, the utterance becomes ungrammatical.  (57)  Clauses associated with object position require ê- proclitic a.  b.  wâpahtam Jeff ê-kî-mispohk wâpahtam -w J ê- kî- mispon -k see.VTI -3 J C1-PREV-snow.VII-0 ‘Jeff saw it had snowed.’ * wâpahtam Jeff kî-mispohk wâpahtam -w J kî- mispon -k see.VTI -3 J PREV-snow.VII-0 ---  Likewise, clauses without an overt clause-typing proclitic (used, for example, in some conditionals) become ungrammatical if an overt clause-typing proclitic is added. This is demonstrated in (58). (58)  Antecedents of conditionals require ê- proclitic a.  kspî nîcêwâkan sipwêhtêci wâpahki, nika-kaskêyihtên8 kspî ni- wîcêwâkan sipwêhtê -t -i wâpah -k -i ni-ka- kaskêyihtê -n if/when 1- friend leave.VAI-3-SUBJ dawn.VII-0 -SUBJ 1- IRR-lonely.VTI -SAP ‘If my friend leaves tomorrow, I will be lonely.’  b.  * kspî nîcêwâkan ê-sipwêhtêci wâpahki, nika-kaskêyihtên kspî ni-wîcêwâkan ê- sipwêhtê -t -i wâpah -k -i ni- ka- kaskêyihtê -n if/when 1- friend C1-leave.VAI-3-SUBJ dawn.VII-0-SUBJ 1- IRR-lonely.VTI -SAP ---  This bi-directional implication between the presence of a phonologically overt clause-typing proclitic versus the functional and distributional properties of the clause is behaviour that is accounted for by representing them as C heads.  8  The kspî element in these examples is regularly used in these construction by one of the consultants I worked with. It is not clear to me if this is a morpho-phonological permutation of kîspin ‘if’, a morpho-phonological permutation of êkospî ‘then’, or an entirely different particle. I have thus left this particle in its surface form.  43  There is an additional complication with simple CONJUNCT clauses in that the preverb kaand suffix -i have a different distribution. For example, the -i clause cannot be substituted for the ka- clause in (59). (59)  a.  b.  nikwêcimâw Jeff ka-nikamot ni-kwêcim -â -w J ka- nikamo -t 1-ask.VTA -DIR-3 J IRR-sing.VAI-3 ‘I asked Jeff to sing.’ * nikwêcimâw Jeff nikamoci ni-kwêcim -â -w J nikamo -t -i 1-ask.VTA -DIR-3 J sing.VAI-3-SUBJ ‘I asked Jeff to sing.’  However, I do not treat them as complementizers on par with the clause-typing proclitics because they do not have the distributional restrictions that the true clause-typing proclitics have. As we saw earlier in the chapter, the irrealis preverb ka- can occur in matrix and embedded clauses, and in both INDEPENDENT and CONJUNCT clauses. The relevant data is repeated in (60): in (60a), ka- is occurring in an INDEPENDENT clause to indicate a future event, and in (60b) it is in a CONJUNCT clause to indicate a clausal relation akin to the Indo-European subjunctive (see chapter 6 for details). (60)  Interpretations of irrealis ka- across Plains Cree’s orders a.  b.  ..., “â, êkota nika-pôsipayihon,” ... â êkota ni- ka- pôsipayiho -n INTERJ there 1- IRR-jump.VAI -SAP ‘..., “Well, I will jump on that,” … ’ (AA 8.3)  INDEPENDENT  nikî-kwêcimâw Nettie ka-pê-itohtêt ni- kîkwêcim -â -w N ka- pê- itohtê -t 1- PREV-ask.VTA -DIR-3 N IRR-come-go.VAI-3 ‘I asked Nettie to come.’  CONJUNCT  future  irrealis  Likewise, the suffix -i, which is glossed as a subjunctive marker in simple CONJUNCT clauses (61a), also appears in CONJUNCT clauses that have clause-typing proclitics. In the latter cases it indicates plurality of inanimate referents, as in (61b), or plurality of realis events (61c) (see Mühlbauer 2008 for discussion).  44  (61)  Interpretations of –i across Plains Cree’s CONJUNCT modes a.  b.  c.  miywâsiki … miywâsi -k –i nice.VII -0 -SUBJ ‘if it/they are nice’  SIMPLE SUBJUNCTIVE  ê-miywâsiki maskisina ê- miywâsi -k -i maskisin -a C1-nice.VII -0-PL shoe -PL ‘...the shoes are nice.’  Ê-CONJUNCT  êkwa piyê-takohtêtwâwi êkwa iy- pê- takohtê -twâw -i and IC-COME-arrive.VAI-3PL -PL ‘And when the men would come home  IC-CONJUNCT  non-singular unrealized event  non-singular inan. referent  non-singular realized event  mâna nâpêwak kî-kîsowihkasowak, … mâna nâpêw -ak kîkîsowihkaso -w -ak usually man -PL PREV-warm.self.VAI-3 -PL they used to warm themselves …’ (EM 50) This means that, for ka- and -i, the choice of clause-type affects the interpretation of the proclitic, rather than the proclitic affecting the choice of clause-type. I take the data above to signify that ka- or the -i are in a position immediately below the clause-typing domain; concretely, I call this Cfiniteness, following Rizzi (1997), and call the clausetyping domain where the pronominal proclitics and clause-typing proclitics sit Cforce. (62)  a.  CPforce 2 2CPfiniteness C 2  2 ka5 -i  Let us now turn to the specifier position. If a specifier position is phonologically null, the projection as a whole does not change, and we expect that the function and distribution of the constituent also will not change.  45  (63)  CP 2  2 C For the pronominal forms, we observe this latter pattern. While the pronominal proclitics  are obligatory if a speech act participant (1st or 2nd person) is one of the participants in the event, there is no pronominal proclitic at all if no speech act participant is an event participant. Thus (64) shows two examples that differ only in the presence/absence of a phonologically overt pronominal proclitic: if there is an overt pronominal proclitic, it gets interpreted (in this example) as a subject (64a); if there is none, then the subject is obligatorily unspecified (thus denoted by the passive translation to English in 64b) (cf. Déchaine & Reinholtz 1999 on unspecified subject constructions). (64)  a.  niwâpamâw ni-wâpam -â -w 1- see.VTA-DIR-3 ‘I see him/her  b.  wâpamâw wâpam -â -w see.VTA-DIR-3 ‘s/he was seen.’  Other morpho-syntactic configurations of the INDEPENDENT order, which involve only third persons, obligatorily lack an overt pronominal proclitic, as in (65). (65)  a.  wâpamêw wâpam -ê -w see.VTA-DIR-3 ‘s/he saw him/her.’  b. *  niwâpamêw ni- wâpam -ê -w 1- see.VTA-DIR-3 ---  A phonologically null – or absent – pronominal proclitic does not change the external syntax of the verbal complex. The form wâpamêw ‘s/he sees him/her’, which lacks a pronominal proclitic,  46  cannot suddenly appear in an embedded clause. In this sense, the pronominal proclitics behave like elements in a specifier, rather than a head, position.  2.3.4 Interim Summary The preceding pages have shown several ways in which the pronominal proclitics in the INDEPENDENT order  and the clause-typing proclitics in the CONJUNCT order differ; I have argued  that these differences correspond to the split between heads and specifiers. This means that the mapping between Plains Cree’s morpho-syntax and the distinction between indexical versus anaphoric clauses is quite transparent. Given that the INDEPENDENT order corresponds to indexical clauses (the topic of chapter 3), indexical clauses in Plains Cree may host an element in spec, CP. (66)  INDEXICAL CLAUSE  CP 2 s0 2 C XP 5     Plains Cree INDEPENDENT CP 2 ni2 kiC XP 5 wâpamâw  And given that the CONJUNCT order corresponds to anaphoric clauses (the topic of chapter 4), anaphoric clauses in Plains Cree may host a complementizer in C that corresponds with the anaphoric situation.  47  (67)  ANAPHORIC CLAUSE  CP 2 s 2 C XP 5     Plains Cree CONJUNCT CP 2 s 2 C XP ê5 kâwâpamât IC-  -  2.4  The indexical/anaphoric distinction ≠ matrix/embedded distinction  The final point I want to make in this chapter is that, although Plains Cree’s clause-typing split interacts with the matrix/embedded distinction in many ways, it does not pick out the matrix/embedded distinction. In this section, I briefly look at two elements in Plains Cree that do distinguish matrix vs. embedded contexts: negation, and the interrogative marker cî. The relevant point is that the distribution of both of these elements is not determined by the morpho-syntactic INDEPENDENT/CONJUNCT distinction  in Plains Cree, but rather by the syntactic matrix/embedded  distinction.  2.4.1 Negation distinguishes matrix and embedded clauses Plains Cree has two negative elements: êkâ and namôya9 (Lacombe 1874, Wolfart 1973, Dahlstrom 1991, Déchaine & Wolfart 1998, 2000). Unlike English negation, the form of negation in Plains Cree is sensitive to the distinction between matrix and embedded clauses (Wolfart 1973, Reinholtz & Wolfart 1996, Reinholtz 1999, Déchaine & Wolfart 1998, 2000; Déchaine & Wiltschko 2002, 2006).  9  This negator actually has multiple morpho-phonological forms, including nama, ma, môya, and môy. In general, the môy(a) forms are most commonly found with clausal negation (as opposed to constituent negation), but more work is needed to understand the interaction of form with function and distribution.  48  The namôya form occurs in matrix environments. For example, the verbal complexes under negation in (68a-b) have the same form (CONJUNCT), but differ as to whether they are embedded. The môy form of negation cannot be used in embedded clauses. (68)  a.  b.  môy ê-kiskêyimak môy ê- kiskêyim -ak NEG C1-know.VTA-1>3 ‘I didn’t know him.’ * nitâyimêyihtên môy ê-kiskêyimak ni(t)- âyimêyihtê -n môy ê- kiskêyim -ak 1- consider.difficult.VTI-SAP NEG C1-know.VTA-1>3 --- (intended: ‘It was hard that I didn’t know him.’)  MATRIX  EMBEDDED  comment: in this sentence, êkây feels better In (69), the matrix negator môya occurs with both INDEPENDENT clauses (69a) and CONJUNCT (69b). This means that the form of negation does not map onto a particular morpho-syntactic form in Plains Cree. (69)  a.  môy ninêstosin môy ni- nêstosi -n NEG 1- tired.VAI-SAP ‘I’m not tired.’  INDEPENDENT  b.  …, namôy ê-môhcwêyimakik, … namôya ê- môhcwêyim -ak -k NEG C1-consider.VTA-1>3-PL ‘…, I do not consider them stupid, …’ (JKN 1.3)  CONJUNCT  The êkâya form of negation occurs in embedded clauses or in clauses that have the irrealis marker ka-.10 Without ka-, the negator êkâya is prohibited from matrix clauses, whether they be INDEPENDENT (70a)  or CONJUNCT (70b).  10  This is a long-standing puzzle in Plains Cree syntax and semantics: why do these environments pattern together? A third environment where êkâya negation is used is in imperatives, which, like clauses with ka-, have an irrealis flavor; this suggests that, whatever the answer, the puzzle is not specific to the morpheme ka-. Thus, on the one hand, êkâya’s distribution is syntactically conditioned (by the matrix/embedded split), and on the other hand it is semantically conditioned (by the realis/irrealis split) (Déchaine & Wolfart 1998, 2000).  49  (70)  êkâ negation cannot occur in matrix clauses a.  * êkâ nikiskêyimâw êkâ ni-kiskêyim-â-w NEG 1-know.VTA-DIR-3 --- (intended: ‘I don’t know him/her.’)  b. * êkâ ê-kiskêyimak êkâ ê- kiskêyim -ak NEG C1-know.VTA-1>3 --- (intended: ‘I don’t know him/her.’)  INDEPENDENT  CONJUNCT  Since INDEPENDENT clauses are never allowed in embedded contexts, they are unsurprisingly bad here too (71a); in a CONJUNCT embedded clause, êkâ negation is fine (71b). (71)  êkâ negation occurs in embedded clauses a.  b.  * nitâyimêyihtên êkâ nikiskêyimâw ni(t)- âyimêyihtê -n êkâ ni- kiskêyim -â -w 1consider.difficult.VTI-SAP NEG C1- know.VTA -DIR-3 ‘It was hard because I didn’t know him.’  INDEPENDENT  nitâyimêyihtên êkâ e-kiskêyimak CONJUNCT ni(t)- âyimêyihtê -n êkâ ê- kiskêyim -ak 1consider.difficult.VTI-SAP NEG C1-know.VTA-1>3 ‘It was hard because I didn’t know him.’ (AA 2.1, presented in elicitation)  Proof that êkâ negation is not selecting for CONJUNCT clauses can be found when we look at clauses with the modal ka- (cf. Lacombe 1874, Déchaine & Wolfart 1998). In (72a-b), we see examples of êkâ co-occurring with an INDEPENDENT clause hosting ka-. (72)  INDEPENDENT +  ka- + êkâ negation  a.  êkâ ka-kimiwan êkâ ka- kimiwan NEG IRR-rain.VII ‘It better not rain!!’  b.  êkâ nika-mîcison êkâ ni- ka- mîciso -n NEG 1- IRR-eat.VAI-SAP ‘I won’t eat (right now).’  50  Summing up, the distribution of negation in Plains Cree is sensitive to (although not entirely determined by) the matrix/embedded distinction. Relevant to the current discussion is the fact that the distribution of negation is not sensitive to the distinction between Plains Cree’s INDEPENDENT  and CONJUNCT orders.  2.4.2 Interrogative marking distinguishes matrix and embedded clauses A second element that is sensitive to the matrix/embedded distinction in Plains Cree is the interrogative marker cî, which can only occur in matrix clauses. This is not surprising, given that interrogative force is a kind of illocutionary force (Searle 1965, Austin 1950), which in turn is thought to be a function of the CP-domain (Cheng 1991, Chomsky 1995, Portner 1999). Again, the point I want to make here is that cî picks out matrix clauses, which are a heterogeneous class in terms of Plains Cree’s morpho-syntax. Both INDEPENDENT and CONJUNCT verbal complexes host Plains Cree cî, as shown in (73): in (73a), cî follows an INDEPENDENT verbal complex, and in (73b), it follows an ê-CONJUNCT verbal complex. (73)  a.  kimîcison cî ki-mîciso-n cî 2-eat.VTA-SAP Q ‘Have you eaten?’  INDEPENDENT  b.  ê-nêstosiyan cî ê-nêstosi-yan cî C1-tired.VAI-2 Q ‘…are you tired?’  CONJUNCT  However, cî may not be embedded under a higher predicate. Thus, in (74a), cî is in second position and has scope over the clause it follows – the matrix clause kiwâpamâw ‘you saw her’. In (74b), which was an attempt to form an embedded interrogative, cî is ungrammatical.  51  (74)  a.  kiwâpamâw cî Rose-Marie ê-kî-pâhpit ki- wâpam -â -w cî RM ê- kî- pâhpi -t 2- see.VTA-DIR-3 Q RM C1-PREV-laugh.VAI-3 ‘Did you see that Rose-Marie laughed?’  b. * Rose-Marie môy niwâpamâw [ê-sipwêhtêt cî] RM môya ni- wâpam -â -w ê- sipwêhtê -t cî RM NEG 1- see.VTA -DIR-3 C1-leave.VAI-3 Q --- (intended: ‘I didn’t see if/whether Rose-Marie left.’)  MATRIX  EMBEDDDED  Similarly, in (75) we observe that kîspin ‘if’ is used to introduce an indirect yes/no question (75a), and that it is ungrammatical to replace kîspin ‘if’ with cî ‘Q’ (75b). (75)  a.  nikwêcimâw Rose-Marie kîspin ê-wî-itohtêt ni- kwêcim -â -w RM kîspin ê- wî- itohtê -t 1- ask.VTA-DIR-3 RM if C1-INT-go.VAI-3 ‘I asked Rose-Marie if/whether she was coming.’  b. * nikwêcimâw Rose-Marie cî ê-wî-itohtêt ni- kwêcim -â -w RM cî ê- wî- itohtê -t 1- ask.VTA -DIR-3 RM Q C1-INT-go.VAI-3 --In summary, then, cî picks out matrix clauses, but it does not pick out INDEPENDENT clauses. Together, negation and the interrogative cî provide evidence that the morpho-syntactic division between INDEPENDENT and CONJUNCT in Plains Cree does not correspond to the matrix/embedded distinction.  2.5  Summary  This chapter has been concerned with how the indexical vs. anaphoric division in clauses maps onto Plains Cree’s morpho-syntax. I first argued that the left-edge pronominal proclitics and clause-typing proclitics are hosted in the CP-domain in Plains Cree, a place where Plains Cree differs from at least some other Algonquian languages, and that these were thus candidates for cuing the indexical/anaphoric division, which I claim is hosted in spec, CP.  52  (76)  a. INDEXICAL CP  CP 2 s0 2 C XP 5  b. ANAPHORIC CP  CP 2 s 2 C XP 5  I then argued that the pronominal proclitics in the INDEPENDENT order are in spec, CP, while the clause-typing proclitics in the CONJUNCT order are complementizers. (77)  a. INDEXICAL CP  CP 2 ni2 ki- C XP 5  b. ANAPHORIC CP  CP 2 s 2 êXP kâ5 IC-  This results in a one-to-one mapping between the indexical vs. anaphoric clauses on the one hand, and Plains Cree’s clause-typing morpho-syntax on the other. I now turn to the external syntax and the semantics of each of these clauses.  53  CHAPTER 3 INDEXICAL CLAUSES: PLAINS CREE’S INDEPENDENT ORDER  3.1  Proposal: The syntax and semantics of indexical clauses  In chapter 2, we looked at the internal structure of indexical clauses, and I argued that they have an indexical speech situation variable in spec, CP.  (1)  Internal structure of an indexical clause CP 2 s0 2 ni- C XP ki5 wâpamâw In Plains Cree, an indexical clause is instantiated by the INDEPENDENT order. In the  summary given in table 3.1, we see that INDEPENDENT clauses are characterized by left-edge 1st and 2nd person marking, and by a unique set of right-edge person marking.  PERSON CATEOGRY 1.sg. 2.sg. 1.pl. excl. 1.pl. incl. 2.pl 3.sg. 3.pl 3.obv  INDEPENDENT ORDER  ninipân kinipân ninipânân kinipânaw kinipânâwâw nipâw nipawak nipâyiwa  Table 3.1. Summary of the INDEPENDENT order paradigm in Plains Cree  54  In this chapter I turn to indexical clauses’ external properties; i.e., how an indexical clause relates to clause-external linguistic material. I make claims about the reflexes of indexicality in these clauses’ structural, semantic, and discourse properties. Structurally, I take there to be two ways in which clauses can be related: the hierarchical notion of c-command as defined in (2), and the linear notion of precedence as defined in (3). (2)  C-commanddef: A constituent α c-commands β iff β is dominated by the lowest node of a major category that dominates α.  (3)  Precedencedef: A constituent α precedes β iff constituent α is linearly ordered before β within a given domain.  Given these conditions1, I claim that indexical clauses are subject to anti-c-command and antiprecedence. Thus they are prohibited in configurations like (4) where the indexical clause is dominated by another clause. (4)  *  CP 5 CPIND 5  They are also prohibited in configurations like (5), where the indexical clause is non-initial within the domain (indicated by the square brackets). (5)  *  [ CP ... CPIND ... ]  In §3.2, I show that these conditions on indexical clauses derive the distribution of Plains Cree’s INDEPENDENT  order clauses. The anti-c-command condition derives the fact that Plains Cree’s  INDEPENDENT  clauses are always matrix clauses. The anti-precedence condition derives the fact  that variables introduced in Plains Cree’s INDEPENDENT order clauses must have a clause-internal antecedent – i.e., even a non-c-commanding antecedent is ruled out. In other words, indexical clauses “must be free.”  1  There is much disagreement in the literature about whether c-command and precedence are two separate conditions (Ross 1967b, Carden 1986, Williams 1997) or whether one can be derived from the other (cf. Reinhart 1983, perhaps laid out most explicitly in Kayne 1994). As we will see in this chapter and chapter 4, Plains Cree exhibits patterns that are best captured by positing both conditions.  55  Semantically, indexical clauses have a privileged relation to the speech act (cf. Banfield 1982): they are indexed to it (cf. Bühler 1934, Bar-Hillel 1954, Kaplan 1989 on indexical expressions). Indexicality is a subset of deixis that picks out the speaker, the speech time and/or the speech location.2 We therefore expect that an indexical clause will have the following particular deictic properties: (i)  referentially, they are anchored to the speaker;  (ii)  temporally, they are anchored to the speech time; and  (iii)  spatially, they anchored to the speech place.  In §3.3, I show that these properties account for the restricted interpretation of Plains Cree’s INDEPENDENT  order clauses: they are evaluated relative to speech time, relative to the speaker,  and relative to the speech location. Modelling this within a situation semantics framework, where every proposition must be evaluated with respect to a situation (Austin 1950, Barwise 1981, Barwise & Perry 1983, Kratzer 1989, 2007), I argue that in an indexical clause this situation is the speech situation. As discussed in chapter 1, a situation s is a partial world; the speech situation s0 is simply a situation in which someone is speaking. The speech situation minimally must include the individual who is doing the speaking (i.e., the speaker I); and the temporal/spatial location of the speaking (i.e., speech time now and speech place here). Therefore, if the truth of a proposition expressed by an indexical clause is evaluated relative to the speech situation, this logically entails that the clause be evaluated relative to both the individual (speaker) and temporal/spatial location to be coded. In §3.3 I look at the referential, temporal, and spatial anchoring properties of indexical clauses to empirically motivate the semantic claims about them.  3.2  The structural context of indexical clauses  In this section I discuss the structural contexts of indexical clauses, focussing on the implications of the claim that indexical clauses are subject to anti-c-command and anti-precedence. We expect 2  The confluence of speaker, speech time, and speech place is called the origo in some treatments (e.g., Bühler 1934; Garrett 2001).  56  that the exponent of indexical clauses in Plains Cree, INDEPENDENT clauses, will be excluded from all embedded contexts (§3.2.1). However, being a matrix clause is not enough to ensure an indexical clause. The implication goes only one way: indexical clauses must be matrix clauses, but there can be matrix clauses which are not indexical. In order for a matrix clause to be an indexical clause it must also satisfy anti-precedence: it cannot be preceded by another clause within its domain. This means, for example, that an indexical clause cannot enter into cross-clausal dependencies. Thus, in a language that morpho-syntactically marks indexical clauses (such as Plains Cree’s INDEPENDENT  order), we expect that all and only indexical matrix clauses will lack cross-clausal  dependencies (§3.2.2).  3.2.1 Indexical clauses must be matrix clauses In this section I show that indexical INDEPENDENT clauses cannot be introduced by higher predicates or by subordinating particles. Further, elements which can be independently argued to be restricted to embedded clauses will be ungrammatical with INDEPENDENT clauses; for example, I show that the embedded negator êkâya cannot be used with an INDEPENDENT clause. English clauses are unspecified with respect to the indexical/non-indexical distinction; there is no morpho-syntactic marking to distinguish them. The form of an English matrix clause can be morpho-syntactically identical to its embedded counterpart. (6)  a.  I’m happy.  b.  I told her I’m happy.  In Plains Cree, however, these two contexts are morpho-syntactically distinguished: an indexical INDEPENDENT CONJUNCT  clause can occur in a matrix context (7a), but is replaced by a non-indexical  clause in the corresponding embedded context (7b).  57  (7)  a.  nicihkêyihtên ni- cihkêyihtê -n 1- happy.VTI-SAP ‘I’m happy.’  INDEPENDENT  b.  niwîhtamawâw ê-cihkêyihtamân ni- wîhtamaw -â -w ê- cihkêyihtam -ân 1- tell.VTA -DIR-3 C1-happy.VTI -1 ‘I told him/her I’m happy.’  CONJUNCT  3.2.1.1 Embedding predicates do not introduce indexical clauses Many verbs in Plains Cree may introduce an embedded clause, but indexical clauses (Plains Cree’s INDEPENDENT order) are impossible in an embedded position; another clause type (Plains Cree’s CONJUNCT order) must be used. (8)  CP1 6 matrix CPi 5 ✔ CONJUNCT ✖ INDEPENDENT  The examples below illustrate. In (9) the embedded clause is a simple CONJUNCT clause (9a); an indexical INDEPENDENT clause is ungrammatical (9b). (9)  a.  b.  ninitawêyimâw nîcêwâkan ka-mîcisot ni- nitawêyim -â -w ni- wîcêwâkan ka- mîciso -t 1- want.VTA -DIR-3 1- friend IRR-eat.VAI-3 ‘I want my friend to eat. * ninitawêyimâw nîcêwâkan mîcisow ni- nitawêyim -â -w ni- wîcêwâkan mîciso-w 1- want.VTA -DIR-3 1- friend eat.VAI-3 --- (intended: ‘I want my friend to eat.’)  CONJUNCT  INDEPENDENT  Likewise, in (10) we observe an embedded ê-conjunct clause (10a); again the indexical INDEPENDENT  counterpart is ungrammatical.  58  (10)  a.  nikî-wâpahtên ê-kimiwahk ni- kî- wâpahtê -n ê- kimiwan -k 1- PREV-see.VTI -SAP C1-rain.VII -0 ‘I saw that it was raining.’  b.  * nikî-wâpahtên kimiwan ni-kî- wâpahtê -n kimiwan 1-PREV-see.VTI -SAP rain.VII --- (intended: ‘I saw that it was raining.’)  CONJUNCT  INDEPENDENT  Similarly, predicative particles (e.g., piko ‘be.necessary.that’) cannot introduce an INDEPENDENT clause (cf. Wolfart 1973, Ahenakew 1987). Rather, they always introduce a CONJUNCT clause. This is illustrated in (11), where both simple CONJUNCT clauses (prefixed with the irrealis marker ka-) and changed CONJUNCT clauses (prefixed with the complementizer ê-) are grammatical (11aa’), but INDEPENDENT clauses are not (11b). (11)  a.  piko ka-wâpamak ana nâpêw piko ka- wâpam -ak ana nâpêw be.necessary IRR-see.VTA-1>3 DEM.AN man ‘I have to see that man.’  SIMPLE CONJ  a’.  piko ê-wâpamak ana nâpêw piko ê- wâpam -ak ana nâpêw be.necessary C1-see.VTA -1>3 DEM.AN man ‘I have to see that man.’  Ê- CONJUNCT  b.  * piko niwâpamâw ana nâpêw piko ni- wâpam -â -w ana nâpêw be.necessary 1- see.VTA-DIR-3 DEM.AN man --  INDEPENDENT  In summary, Plains Cree INDEPENDENT clauses cannot be embedded. This is a way in which Plains Cree’s INDEPENDENT clauses are more restricted than clauses that appear in matrix contexts in English: the latter can occur in embedded contexts without any change in the morpho-syntax, whereas the indexical INDEPENDENT clauses cannot.  59  3.2.1.2 Subordinating particles do not introduce indexical c\lauses In addition to embedded clauses, there are a number of subordinators which introduce different kinds of adjoined dependent clauses. These subordinators are uninflected particles which sit external to and precede the verbal complex; they act as restrictors on the complementizer of the clause they introduce, specifying the type of embedded clause. Syntactically, I posit that these particles are complementizers. Since indexical clauses are by hypothesis subject to anti-c-command, we expect that Plains Cree’s INDEPENDENT clauses should never occur with these subordinators. This expectation is fulfilled by the data as exemplified in the following table. The distribution of each subordinator according to clause-type is given for four different speakers (AA, EM, JK, SW).3 None of the subordinators introduce an indexical INDEPENDENT clause for any of the speakers, while all of them may introduce an anaphoric CONJUNCT clause (the numbers give the number of attested examples for each speaker). Subordinator  indexical:  anaphoric:  INDEPENDENT  CONJUNCT  JK SW AA EM JK SW AA ✖ ✖ -osâm ‘reason’ --✔ (5) -✖ ✖ ✖ ✖ ✔ (12) ✔ (4) ✔ (47) iyikohk ‘as far as’ ✖ ✖ ✖ ✖ ✔ (4) kiyâm ‘although’ ✔ (6) ✔ (5) ✖ ✖ ✖ ✖ pâmwayês ‘before’ ✔ (2) ✔ (2) ✔ (3) ✖ ✖ ✖ ✖ ✔ (7) mayaw ‘as soon as’ ✔ (3) ✔ (4) Table 3.2. Distribution of subordinators by clause-type in Plains Cree  EM ✔ (24) ✔ (46) ✔ (5) ✔ (10) ✔ (2)  The distributional facts of INDEPENDENT clauses in the context of subordinators are quite striking: they simply do not occur. By contrast, every subordinator introduces some form of an anaphoric CONJUNCT clause4. This data was confirmed in elicitation sessions, where speakers rejected utterances where an indexical INDEPENDENT clause had been substituted for the anaphoric CONJUNCT clause: minimal pairs are given in (12-13). For example, concessive clauses are introduced by kiyâm ‘although’, and always appear in the kâ-conjunct (12a), usually followed by the particle âta ‘even’. An indexical INDEPENDENT clause is ungrammatical (12b). 3  There are a limited number of subordinators that occur with any regularity. The table is intended to be an exhaustive list of those which occur across multiple speakers. 4 See chapter 6 for details.  60  (12)  kiyâm ‘although’ only in CONJUNCT kiyâm âta kâ-pipok, âhci piko mân ê-kî-yîkinikêt nikâwînân. kiyâm âta kâ-pipon -k although even C2-winter.VII-0  a.  CONJUNCT  âhci piko mâna ê- kî- yîkinikê -t ni-kâwî -nân still usually C1-PREV-milk.VAI-3 1- mother-1.PL ‘Even during the winter our mother would still milk the cows.’ (EM 17) b.  * kiyâm âta pipon, âhci piko mân ê-kî-yîkinikêt nikâwînân. INDEPENDENT kiyâm âta pipon âhci piko mâna ê- kî- yîkinikê -t ni-kâwî -nân although even winter.VII still usually C1-PREV-milk.VAI-3 1- mother-1.PL ---  Degree clauses are introduced by the element iyikohk ‘so’. As a degree marker,5 iyikohk ‘so’ always introduces an anaphoric CONJUNCT clause (13a); the corresponding indexical INDEPENDENT  (13)  clause is ungrammatical (13b).  iyikohk ‘so’ only in CONJUNCT a.  b.  “… êkotowahk mân ê-kî-mîciyân, iyikohk ê-nôhtêhkatêyân,” … CONJUNCT êkotowahk mâna ê- kî- mîci -yân iyikohk ê- nôhtêhkatê -yân that.kind usually C1-PREV-eat.VTI-1 DEG C1-hungry.VAI-1 ‘ “…I was so hungry that I would eat that kind,” …’ (EM 71) (alt. trans. “I used to eat the kind because I was so hungry.”) * êkotowahk mân ê-kî-mîciyân, iyikohk nikî-nohtêhkatân6 êkotowahk mâna ê-kî-mîci-yân iyikohk ni-kî-nôhtêhkat-â-n that.kind usually C1-PREV-eat.VTI-1 DEG 1-PREV-hungry-VAI-SAP --- (intended: ‘I was so hungry that I would eat that kind.’)  INDEPENDENT  5  Note that, like many particles, iyikohk occurs in a number of varied contexts, with a number of interpretations. While some of these contexts do allow INDEPENDENT clauses, these contexts do not have the dependence of the degree clauses given above. See chapter 6 for further discussion. 6 In this example, I have presented the INDEPENDENT order clause with the temporal sequencer kî-, since kî- is necessary to get a time disjoint from utterance time (cf. §7.1), which is what we have in the preceding ê-kî-mîciyân clause. The INDEPENDENT is also bad if the kî- is omitted: (i) * …êkotowahk mân ê-kî-mîciyân, iyikohk ninohtêhkatân êkotowahk mâna ê- kî- mîci -yân iyikohk ni- nôhtêhkatâ -n that.kind usually C1-PREV-eat.VTI-1 DEG 1- hungry.VAI-SAP --- (intended: ‘I was so hungry that I would eat that kind.’)  61  3.2.1.3 Embedded negation does not modify indexical clauses The last embedded context presented here is one specific to Cree – the interaction of clause-type with negation. As we saw in chapter 2, Plains Cree has two forms of negation: namôya and êkaya. These two forms are sensitive to the matrix/embedded distinction. namôya occurs in unembedded contexts. The êkâya form of negation occurs only in embedded clauses (cf. Déchaine & Wiltschko 1998, 2006)7 ; the relevant contrast is shown in (14). (14)  a.  b.  nitâyimêyihtên êkâ e-kiskêyimak ni(t)- âyimêyihtê -n êkâ ê- kiskêyim -ak 1consider.difficult.VTI-SAP NEG C1-know.VTA-1>3 ‘It was hard because I didn’t know him.’ (AA 2.1, presented in elicitation) * nitâyimêyihtên môy ê-kiskêyimak ni(t)- âyimêyihtê -n môy ê- kiskêyim -ak 1- consider.difficult.VTI-SAP NEG C1-know.VTA-1>3 --- (intended: ‘It was hard because I didn’t know him.’)  Thus we expect that it will not be possible to negate an indexical INDEPENDENT clauses with êkâ. This is correct, as shown in (15): replacing môy negation with êkâ negation yields ungrammaticality. (15)  a.  môy ninohtêhkatân môy ni- nohtêhkatâ -n NEG 1- hungry.VAI-SAP ‘I’m not hungry.’  b.  * êkâ ninohtêhkatân êkâ ni- nohtêhkatâ-n NEG 1- hungry.VAI-SAP --  Thus, negation provides further evidence for the indexical analysis: we see that the only type of negation available for INDEPENDENT clauses is the negation that is restricted to unembedded environments.  7  Unless the irrealis ka- preverb is present; see chapter 2.  62  3.2.1.4 Summary: Indexical clauses cannot be subordinated We have seen three independent pieces of evidence that indexical INDEPENDENT clauses are subject to anti-c-command: they cannot be introduced by a higher predicate, they cannot be introduced by a subordinating particle, and they cannot be negated by the êkâyâ negator. In the next section I turn to the anti-precedence condition on indexical clauses and show how this condition accounts for the lack of cross-clausal dependencies in indexical INDEPENDENT clauses.  3.2.2 Indexical clauses exclude cross-clausal dependencies The purpose of this section is to show that cross-clausal dependencies, such as the binding of a variable by a clause-external variable, are excluded from indexical clauses. First, I examine a class of variables known in the Algonquianist literature as relative roots (§3.2.2.1), and show that the antecedence relation is affected by clause-type, an observation which, regardless of whether the current analysis is correct or not, offers an important insight into the grammar of Algonquian languages. Second, I examine temporal and locative proforms and show that unless the proform has a morphologically marked deictic component, they are ungrammatical in indexical INDEPENDENT clauses (§3.2.2.2). Third, I discuss how reference to argument expressions is restricted in indexical INDEPENDENT clauses (§3.2.2.3)  3.2.2.1 Relative roots Relative roots are a class of proforms (locative, manner, temporal, etc.) found across all Algonquian languages (Bloomfield 1962; Wolfart 1973; Valentine 2001; Rhodes 1976). They are termed roots because they may be found in the root position of a stem (even though they may also be found in places where they are not in a root position). They are relative because they do not have an independent interpretation, but rather are interpreted relative to the an antecedent which is obligatory for the utterance in which they occur to be well-formed (Bloomfield 1962, Wolfart 1973). More generally, they are variables that are quite unspecified as to their features: 63  their specific function is determined in part by its position in the clause (there are at least three possible positions), and in part by nature of its antecedent. For each of the two relative root variables that I look at here, there are at least three kinds of antecedents. Although they have been widely discussed in the literature (cf. Bloomfield 1928, 1946, 1962; Wolfart 1973, Rhodes 1976, 2003, Pentland 1979, Dahlstrom 1991, Bruening 2001, among others), the principles that determine (im)possible antecedence relations remain very poorly understood. In Plains Cree, the inventory of these antecedent-dependent elements is a closed class and includes: it/isi ‘thus’8; oht/ohci ‘originating.from’ isko- ‘to.such.an.extent’, and tahto- ‘so.many’ (Wolfart 1973:66). In this thesis I have chosen to look at the two relative roots which are found across all of the speakers I have worked with: the relative root of manner: it/is(i) ‘thus’ and the relative root of origin: oht/ohc(i) ‘originating from’. Both relative roots may be found in a variety of positions within the clause. In the following examples, I have bolded the relative root and underlined its antecedent (the element without which the utterance would be ungrammatical). First, they may occur in a root position: in (16a), isi- ‘thus’ is the root and Jane is the antecedent; in (16b), ohc- ‘origin’ is in the root position and Calgary is the antecedent. (16)  ROOT POSITION  a.  Jane isiyîhkâsow J isiyîhkâso -w J THUS.be.called.VAI-3 ‘Her name is Jane.’  b.  Calgary nitohcîn C nit- ohcî -n C 1- ORIG.VAI-SAP ‘I am from Calgary.’  8  These glosses are meant only to give a rough idea of their meaning; as will become clear, their semantics are underspecified. The addition of -i causes a palatalization of both relative roots: it  isi; and oht  ohci (cf. Piggott 1971, Wolfart 1973). Due to morpho-phonological processes which lead to the deletion of –i (for example, vowel hiatus), many times the surface form will be palatalized but not have -i. Finally, ohc(i) alternates with ôh- based on factors that are as yet undescribed in the literature. The alternations do not seem to have any direct correlation to the syntactic and semantic generalizations presented here (although I have never seen the ôh- form with an ê-CONJUNCT verbal complex), so I will not be concerned further about which form shows up.  64  Second, relative root variables may occur in a preverbal position, as in (17)9. (17)  PREVERBAL POSITION  a.  mâka kahkiyaw pâh-pîtos kitis-âyânânaw. mâka kahkiyaw pâh- pîtos kit- is- âyâ -nânaw but all RED- different 2- THUS-be.VAI-2.PL ‘but we are all different.’ (EM 19)  b.  mistahi mân âya, tôhtôsâpoy nikî-ohci-pimâcihikonân êkwa aya, ... mistahi mâna aya tôhtôsâpoy ni- kî- ohci- pimâcih -iko -nân êkwa aya a.lot usually CONN milk 1- PREV-ORIG-sustain.VTA-INV-1PL and CONN ‘She used to have lots of milk on which to sustain us, ...’ (EM 16)  Finally, a relative root variable can be an adposition. With verbs of motion, isi indicates motion towards goal (e.g., towards waskahikanihk ‘the house.LOC’ in 18a) and ohci indicates motion from the origin (e.g., away from waskahikanihk ‘the house.LOC’ in 18b) (cf. Edwards 1954). (18)  ADPOSITION – VERB OF MOTION  a.  nipimohtân wâskahikanihk isi ni- pimohtâ -n wâskahikan -ihk isi 1- walk.VAI -SAP house -LOC THUS ‘I’m walking towards the house.’  b.  nipimohtân wâskahikanihk ohci ni- pimohtâ -n wâskahikan -ihk ohci 1- walk.VAI-SAP house -LOC ORIG ‘I’m walking from the house.’  If the verb is not a verb of motion, the adpositional relative root indicates manner for isi, as in (19a), and instrumental for ohci, as in (19b).  9  I take the clausal material occurring external to the verbal complex to be part of the CP constituting the verbal complex; e.g., in (17) I take the adverbial pâh-pîtos ‘different’ to be a modifier of the verb ayâ- ‘be’. I do not know of any good analysis of the mechanisms driving some clausal elements to be external to the verbal complex, and others to be internal to it (although see Dahlstrom 1995, Mühlbauer 2003, and Déchaine 2007 for a more detailed description of the issue). On a very broad view, the issue seems to be one of non-concatenative morpho-syntax (cf. non-concatenative morpho-phonology in Semitic; Arad 2000).  65  (19)  ADPOSITION – OTHER VERBS  a.  ..., âta ê-kî-kiskêyihtahkik âh-âyîtaw isi maskihkiy, ... âta ê- kî- kiskêyihtam -k -k âh- âyîtaw isi maskihkiy even C1-PREV-know.VTI -0 -PL RED-side THUS medicine ‘..., although they used to know both sides of medicine, ...’ (AA 10.1) (Lit: ‘... they knew both sides of medicine that way.’)  b.  môhkomân ohci ê-wî-mansamân môhkomân ohci ê- wî- mansam -ân knife ORIG C1-INT-cut.VTI -1 ‘I am going to cut it with a knife.’  Notice that in all these examples the underlined antecedent precedes the bolded relative root it binds: for example the antecedent môhkomân ‘knife’ must precede the adposition ohci ‘with’. This is a context where there is an fixed ordering between two elements in Plains Cree10 (Wolfart 1973, see also Rhodes 2003 for Ojibwa). As we will see, however, this pattern is part of a more general principle about the relation that must hold between a dependent element and its antecedent (cf. chapter 4). Now that we have seen the different positions where a relative root position may be introduced, I turn to the different possible antecedents. I focus on relative roots in the preverbal position because it is this position that (i) shows the most variation in possible antecedents, but (ii) has antecedents that are both clause-internal and clause-external, allowing us to test the claim about indexical INDEPENDENT clauses. 3.2.2.1.1 Relative roots with predicate modifier antecedents One type of antecedent that preverbal relative roots may be anaphoric on is a predicate modifier (i.e., an adverbial or oblique argument). Syntactically, predicate modifiers are usually assumed to be introduced quite low in the clause, either in the vP, or the functional domain (AspP or TP). 10  There are some speakers for whom the ordering some examples is not fixed. In particular, when the relative root is stem-internal, it does not require that the antecedent precede the stem, as in (i) volunteered by a consultant. (i) nitisiyîhkâson Clare ni(t)- isiyîhkâso -n C 1- THUS.be.called.VAI-SAP C ‘My name is Clare.’ I take this to be a separate grammar, where the stem is now opaque – it has ‘word-level’ properties in the sense of DiSciullo & Williams (1987), and thus the relative root is not available for syntactic operations. See also Hirose (2000) for discussion of variation with respect to the syntactic visibility of stems in Plains Cree.  66  Plains Cree’s predicate modifiers are consistent with this claim: they are linearly internal to elements in information-structure positions (i.e., topic/focus), negation, and quantifiers (cf. Dahlstrom 1995, Mühlbauer 2003). For concreteness, I place them as modifiers to the vP. In (20), I give a proposed structure, where the relative root variable (RR.vbl) is associated with a predicate modifier (indicated by the coindexation). (20)  CP 2 TP / AspP SUBORD. 2 ASP/QUANT vP 2 vP PRED.MODi 5 RR.vbliNEG  Since predicate modifiers are within the same CP as the relative root variable they are associated with, we expect them to be possible antecedents for both INDEPENDENT and CONJUNCT clauses. This is accurate for both relative roots, as shown in table 3.3. Antecedent  indexical INDEPENDENT anaphoric CONJUNCT JK SW AA EM JK SW AA EM ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ Predicate isi✔ ✖ ✔ ✔ ✔ ✔ ✔ ✔ modifier ohciTable 3.3 Predicate modifier antecedents occur in both clause-types The preverbal isi- can have lexical manner adverbs as antecedents, as in (21a-b). Both indexical INDEPENDENT  and anaphoric CONJUNCT clauses allow manner adverbs such as kwayask ‘proper’  to be antecedents to isi-. (21)  a.  ..., pik ôma ka-mâmawôhkamâtoyahk, piko ôma ka-mâmawôhkama-ito-yahk necessary DEM.INAN IRR-work.VTA -RECIP-21pl  CONJUNCT  kwayask ka-kakwê-isi-pimâtisiyahk, ... kwayask ka-kakwê-isi-pimâtisi-yahk proper IRR-try-THUS-live.VAI -21 ‘..., we must work together to try to lead a good life, ...’ (EM 37)  67  b.  ..., matwân cî kwayask nika-kî-isi-tâhkôtên … INDEPENDENT matwân cî kwayask ni-ka- kî- isi- tâhkôtê -n EVID Q proper 1- IRR-PREV-THUS-discuss.VTI-SAP ‘..., I wonder if I will be able to discuss it with proper faithfulness, …’ (JKN 6.1)  The other relative root, ohci, introduces instrumental adjuncts11. In (22a), the topic element êwako ‘that’ is the antecedent for ohci and represents the means of washing the floor. In (22b), the deictic element êkoni is the antecedent for ohci and indicates the means of blessing the addressee. The former is an anaphoric CONJUNCT clause; the latter an indexical INDEPENDENT clause. (22)  a.  ..., pihko ê-siswêwêpinahkik êkwa pihko ê- siswêwêpinam -k -k êkwa ash C1-sprinkle.VTI -0 -PL and ‘Some I even saw sprinkle ashes about and  CONJUNCT  êwako ê-ohci-wâpiskahahkik aya, … êwako ê- ohci- wâpiskaham -k -k aya TOPIC C1- ORIG-wash.VTI -0 -PL CONN use that to wash the floor-boards …’ (EM 82) (lit: ‘...and wash the floor-boards with that ...’) b.  “hâw, êkoni ôhi, k-ôh-sawêyimitin nîst ôma, … INDEPENDENT hâw êkoni ôhi ki- oh- sawêyim -iti -n nîsta ôma indeed DEIC.TOPIC DEM.INAN 2- ORIG-bless.VTA-1>2-SAP 1.EMPH DEM.INAN ‘ “Indeed, with these I myself will bless you, … ’ (JKN 7.2)  11  The preverbal ohci- can also introduce directional adjuncts, just like the adpositional ohci. Directional adjuncts also being predicate modifiers, they can occur with either indexical (INDEPENDENT) or anaphoric (CONJUNCT) clauses, as shown in (i). Notice that with the CONJUNCT example, there is a demonstrative intervening between the locative element ôtê ‘there’ and the verbal complex; this is indicative of a cleft construction (cf. Blain 1997) and significantly, is absent in the INDEPENDENT example. (i) a. ..., ôtê k-ôh-osâpamikowâw. ôtê ki- oh- osâpam -iko -wâw there 2- ORIG-watch.jealously.VTA-INV -2PL ... , that they are [jealously] watching you from over there, ... (JKN 3.17) b. ôtê ana ê-ohci-kitâpamiht, ... ôtê ana ê- ohci- kitâpam -ih -t there DEM.AN C1-ORIG-watch.over.VTA-USC-3 he is watched over from there, ... (JKN 4.9)  68  3.2.2.1.2 Relative roots with CP-modifier antecedents CP-modifiers can also act as antecedents to a preverbal relative root variable, including the deictic topic marker êkosi and negation (both namôya and êkâya). The topic marker is part of information structure, and on independent grounds, negation is a CP-modifier (Déchaine & Wiltschko 2002). (23)  CP 3 NEGi TP / AspP TOPICi 2 ASP vP 2 vP PRED.MOD 5 vbli-  Since negation and topic markers are CP-modifiers, again we expect that they are possible antecedents in both INDEPENDENT and CONJUNCT clauses. This is borne out in all four corpora, as shown in table 3.4. Antecedent Predicate modifier CPmodifier  indexical INDEPENDENT anaphoric CONJUNCT JK SW AA EM JK SW AA EM ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ isi- (oblique) ✔ ✖ ✔ ✔ ✔ ✔ ✔ ✔ ohci- (oblique) ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ isi- (topic) ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ohci- (negation) Table 3.4. CP-modifier antecedents are possible in both clause-types  The topic-marker êkosi ‘that way’ is an antecedent for isi (24); consistent with being a CPmodifier, it occurs in a clause-initial position (Baker 1985, Cinque 1999; Cook et. al 2003).12  12  êkosi ends in the segmental sequence si-; following Wolfart (1973), I analyze this element as having a bi-partite structure êkw ‘deixis’ and isi ‘thus’.  69  (24)  a.  êwakw ânim [CP êkos ê-kî-isi-pimâcihocik kayâs ayisiyiniwak ]. CONJUNCT êwakw anima êkosi ê- kî- isi- pimâciho -t -k kayâs ayisiyiniw -ak topic DEM.INAN DEIC.so C1-PREV-THUS-live.VAI -3 -PL long.ago person -PL ‘That is how the people made a living long ago.’ (EM 12)  b.  …, [CP êkosi mîna mâna nikî-isi-mâmitonêyihtên ]. êkosi mîna mâna ni- kî- isi- mâmitonêyihtê -n DEIC.so also usually 1- PREV-THUS-think.about.VTI-SAP ‘…, and that is how I used to think about it.’ (EM 29)  INDEPENDENT  Negation is an antecedent for ohci in both anaphoric CONJUNCT and indexical INDEPENDENT clauses13; here it is suppletive with the temporal preverb kî-14 and has a past orientation. In (25a), êkâ ‘NEG’ antecedes ohci- in a CONJUNCT clause; in (25b) môy negation antecedes ôh- in an INDEPENDENT clause. (25)  a.  …, êkâ ê-ohci-nisitawêyimakik ayisiyiniwak. êkâ ê -ohci- nisitawêyim -ak -k ayisiyiniw -ak NEG C1-ORIG- know.VTA -1>3-PL person -PL ‘…, because I did not know people.’ (EM 8)  CONJUNCT  b.  mâka, niya wiya môy nôh-pakwâtên anima ... mâka niya wiya môya ni-ôh- pakwâtê -n anima but 1 EMPH NEG 1-ORIG-hate.VTI-SAP DEM.INAN ‘But I did not mind [it]...’ (EM 8)  INDEPENDENT  3.2.2.1.3 Relative roots with cross-clausal antecedents Like other kinds of variables, relative roots can have an antecedent in another clause, creating a cross-clausal dependency15. These include clause-external wh-words (Blain 1997; Cook 2003,  13  In this discussion, I claim that NEG is an antecdent to ohci- in the sense that (i) the presence of NEG is sufficient to license ohci-; and (ii) the presence of NEG yields a particular interpretation of ohci-. An alternative hypothesis is that ohci- under negation is a negative polarity item. I think that the NPI analysis is not mutually exclusive to the discussion here – e.g., as an NPI, ohci- would be licensed by negation, which is consistent with the discussion here. The main point I am trying to make here is that ohci- must always be licensed by something. There is still the question of how ohci gets the interpretation it does – i.e., how is the ‘past’ interpretation related to the directional and instrumental readings. For this question, I think a consideration of the very abstract ORIGIN concept would be useful, although I do not have time and space to consider it here. Finally, if ohci- is analyzed as an NPI, it would have to be a strong NPI in that negation is the only context that triggers it (e.g., ohci- cannot be triggered in interrogatives, relative clauses, etc., at least in Plains Cree). Since the NEG…ohci pattern is robust across the Algonquian family, it might be worthwhile to look at NPI contexts across different languages to see if there is variation on this account. Thanks to A. Dahlstrom (p.c.) for bringing this to my attention, and to L. Matthewson (p.c.) for discussion. 14 See §7.4 for a discussion on the role of negation and modality in the interpretation of kî-. 15 These cross-clausal dependencies lead Bruening (2001) to posit a relative root phrase.  70  2004) as in (26a), clause-external non-wh antecedents (Wolfart 1973) as in (26b), and discourse antecedents (Bloomfield 1928, 1946) as in (26c). (26)  a.  XP 5 WHI CP 5 vbli  b.  …XPi … CP 5 vbli-  c.  CPi 5 CP 5 vbli  Now, if indexical clauses have an anti-c-command condition, there will be no higher clause to host the antecedent. We therefore expect that cross-clausal antecedents will only be possible with anaphoric CONJUNCT clauses; indexical INDEPENDENT clauses should be ungrammatical. This is correct, as summarized in table 3.5. Antecedent Predicate modifier CP-modifier  INDEPENDENT  JK ✔ ✔ ✔ ✔ ✖ ✖ ✖ ✖ ✖  SW ✔ ✖ ✔ ✔ ✖ ✖ ✖ ✖ ✖  AA ✔ ✔ ✔ ✔ ✖ ✖ ✖ ✖ ✖  CONJUNCT  EM ✔ ✔ ✔ ✔ ✖ ✖ ✖ ✖ ✖  JK ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔  SW ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔  AA ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔  isi- (oblique) ohci- (oblique) isi- (deictic) ohci- (negation) Cross-clausal isi- (manner) whohci- (reason) Cross-clausal isi- (manner) non whohci- (reason) Superordinate isiclause Table 3.5. No cross-clausal antecedents for indexical INDEPENDENT clauses  EM ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔  Let us look at each case in turn. Starting first with the wh-antecedents, we observe that the relative root isi- may be bound by the manner wh-word tânisi ‘how’ (27a). Similarly, the relative root ohci- may be bound by the reason wh-word tânêhkî ‘why’ in (27b). 71  (27)  a.  [CP [pred tânisii ]j [DP proi [CP Opi ... isii ... ] tj ] tânisi ê-isi-pimipayik, ... tânisi ê-isi-pimipayi-k Q.manner C1-THUS-go.VII-0 ‘how it happened’ (SW16)  b.  [CP [pred tânêhkîi ]j [DP proi [CP Opi … ôhi … ] tj ] − tânêhk âwa, k-ôh-ihtakot ôta? tânêhkî awa kâ- ôh- ihtako -t ôta Q.RAT DEM.AN C2-ORIG-exist.VAI-3 here ‘why does this one here exist?’ (JK3:13)  Relative root variables may also be bound by non-wh antecedents. In (28) we have a biclausal structure: the êkos ânima sequence is a kind of nominal predication structure (Déchaine 1997, Blain 1999) roughly equivalent to ‘the way is this’. The anaphoric CONJUNCT clause modifies the subject anima ‘this’. Here the relative root variable isi- in the embedded modifying clause has the deictic manner element êkosi ‘this way’ in the higher clause as its antecedent. (28)  [CP [pred êkosii ]j [DP animai [CP Opi … isii … ] tj ] ..., êkos ânim ê-isi-kitâpamât; (EM38) êkosi anima ê- isi- kitâpam -â -t TOP DEM.INAN C1-THUS-look.VTA-DIR-3 ..., that is the way she looks upon them;  Similarly, the relative root variable ohci- can have a cross-clausal antecedent like the êwakôhci (from êwakw ‘that’ + ohci ‘originate’) in (29b), which is again arguably acting as the subject of a higher nominal predication structure (Déchaine 1997, Blain 1999). Such antecedents are fine with an anaphoric CONJUNCT clause, but not with indexical INDEPENDENT clauses.  72  (29)  [CP [pred êwakôhci ]j [DP proi [CP Opi … ohi … ] tj ] a.  ê-nôhtê-osêhkêmit, êwakôhci kôh-âtoskêt Jeff ê- nôhtê-osêhkêmi -t êwakw ohci kâ- ôh- âtoskê -t J C1-WANT-have.car.VAI-3 TOPIC ORIG C2-ORIG-work.VAI-3 J ‘He wants a car, that is why Jeff is working.’  b.  * ê-nôhtê-osêhkêmit, êwakôhci ôhc-atoskêw Jeff ê- nôhtê-osêhkêmi -t êwakw ohci ôhc- âtoskê -w J C1-WANT-have.car.VAI-3 TOPIC ORIG ORIG-work.VAI-3 J --- (intended: ‘He wants a car, that’s why Jeff is working.’)  CONJUNCT  INDEPENDENT  The final kind of cross-clausal antecedent is specific to the relative root variable isi-. This antecedent is not a word- or phrase-level constituent, but rather the a preceding (set of) clause(s) (cf. Bloomfield 1928). For example, in (30), the narrator is describing of the things they had to do, and she then says that, through those actions they were able to avoid starvation. Thus, all of the things described in the initial clauses serve as an antecedent to the manner variable isi- in the purpose clause. (30)  piko mitoni --~ tâpitawi pikw ê-~ ê-wî-kakwê-tôtamâhk kîkway, piko mitoni tâpitawi piko ê- wî- kakwê-tôtam -ân -k kîkway QUANT much truly QUANT C1-INT-TRYdo.VTI -1 -PL thing ‘we very much had to try and do things at all time k-êsi-pihkohtamâsohk ka-mîcihk. ka- isi- pihkohtamâso -hk ka- mîci -hk IRR-THUS-manage.VAI -IMP IRR-eat.VTI-IMP in order to manage to have something to eat.’ (AA 9.1)  Once again, since the antecedent is external to the clause, the behaviour of isi- exemplified in (30) is unattested with indexical INDEPENDENT clauses. To sum up, relative roots show that indexical INDEPENDENT clauses exclude cross-clausal anaphoric relations that are possible in anaphoric CONJUNCT clauses. This is important because it is consistent with our expectation that dependency relations must be resolved locally (i.e., clauseinternally) with indexical clauses.  73  3.2.2.2 Spatio-temporal variables must be bound in indexical clause Plains Cree has dedicated spatial and temporal proform variables, including ita ‘where’; itê ‘where’ and ispî ‘when’; these occur on the far left edge of the clause, and must be bound by an antecedent. The anti-c-command and anti-precedence conditions predict that they will be excluded from indexical clauses. This is correct: as shown in table 3.6, they are unattested. Variable  indexical INDEPENDENT anaphoric CONJUNCT JK SW AA EM JK SW AA EM ✖ ✖ ✖ ✖ ✔ (5) itê ‘where’ ✔ (3) ✔ (7) ✔ (8) ✖ ✖ ✖ ✖ ita ‘where’ ✔ (46) ✔ (8) ✔ (9) ✔ (27) ✖ ✖ ✖ ✔ (4) ispî ‘when’ --✔ (7) ✔ (9) Table 3.6. Distribution of spatio-temporal proform variables by clause-type Elicitation data confirms that locative itê and ita, and temporal ispî are incompatible with indexical INDEPENDENT clauses. (31)  locative proform variables are bad in INDEPENDENT a.  * itê itohtêwak itê itohtê -w -ak LOC go.VAI-3 -PL --- (intended: ‘They are going there/somewhere.’)  b. * ita itohtêwak itê itohtê -w -ak LOC go.VAI-3 -PL --- (intended: ‘They are going somewhere.’)  (32)  INDEPENDENT  INDEPENDENT  temporal proform variable is bad in INDEPENDENT a.  ispî kâ-pîhtikwêt, … ispî kâ- pîhtikwê -t TEMP C2-go.in.VAI -3 ‘when she went in, …’  b. * ispî pîhtikwêw ispî pîhtikwê -w TEMP go.in.VAI -3 --- (intended: ‘When/then she went in.’)  CONJUNCT  INDEPENDENT  74  However, if these variables are bound by the deictic element êkw- (Wolfart 1973), they become perfectly acceptable, as demonstrated by the locative proforms with êkw- in (33) and the minimal pair of temporal proforms in (34). (33)  (34)  Locative proforms bound by êkw- are good in INDEPENDENT a.  itohtêwak êkotê itohtê -w -ak êkotê go.VAI-3 -PL there ‘They went over there.’  b.  itohtêwak êkota itohtê -w -ak êkotê go.VAI-3 -PL there ‘They went there.’  Temporal proforms require êkw- in INDEPENDENT a.  b.  * ispî kimiwan êkotê kâ-itohtêyâhk ispî kimiwan êkotê kâ- itohtê -yân -k time rain.VII there C2- go.VAI -1 -PL --êkospî kimiwan êkotê kâ-itohtêyâhk êkospî kimiwan êkotê kâ- itohtê -yân -k then rain.VII there C2-go.VAI -1 -PL ‘At that time it was raining, when we went there.’  Syntactically, êkw- acts as an antecedent to the variable, precluding the necessity of a crossclausal antecedent. Semantically, recall from chapter 1 that deictic expressions point to their referent (Fillmore 1975, Kaplan 1989, a.o.); thus the presence of êkw- fixes the reference of the spatio/temporal proform in a given context. In fact, êkw- is a general-purpose deictic antecedent. For example, we saw earlier that the deictic topic marker êkosi ‘that way’ was one of the possible antecedents for the isi variable in indexical clauses. The relevant example is repeated in (35).  75  (35)  ..., êkos ânim ê-isi-kitâpamât; êkosi anima ê- isi- kitâpam -â -t TOP DEM.INAN C1-THUS-look.VTA-DIR-3 ‘..., that is the way she looks upon them;’ (EM 38)  It is also used for deictic referents, in combination with the referential ani (cf. Reinholtz & Wolfart 2001). For example, in (36) êkoni is associated with nâpêwa ‘man’, and is used with either a demonstrative gesture (deixis), or to refer back to man previously talked about (anaphora). (36)  ê-wâpamât iskwêw êkoni nâpêwa ê- wâpam -â -t iskwêw êkoni nâpêw -a C1-see.VTA-DIR-3 woman that.one man -OBV ‘The woman saw that man.’  The morphological bi-partite structure of êkw- affixed elements (cf. Wolfart 1973:38) has syntactic and semantic significance as well: the initial êkw- morpheme acts as an antecedent to a variable introduced by the second unit of the demonstrative. (37)  3 DEMP CP 3 5 DEM XP niwâpamâw êkw2 pro 2 LOC ita/itê TEMP ispî ARG ani MANNER isi  3.2.3 Pronominal proclitics are indexical I have so far shown that indexical clauses have a particular set of structural properties: (i) they cannot be embedded; and (ii) dependent elements must have their dependency resolved clauseinternally or be deictic. I have claimed that these properties reflect the syntax of indexicality, here implemented as anti-c-command and anti-precedence conditions. A third outcome of the 76  indexical analysis is that indexical clauses should not have anaphoric pronominal forms. In particular, the pronominal proclitics in Plains Cree INDEPENDENT clauses should have a deictic/indexical dependency, rather than an anaphoric one. Turning to the INDEPENDENT mode paradigm in table 3.7, this includes 1st-person ni- and 2nd-person ki- . In Plains Cree, unlike many other Algonquian languages, it does not include 3rd-person o-  PERSON CATEOGRY 1.sg. 2.sg. 1.pl. excl. 1.pl. incl. 2.pl 3.sg. 3.pl 3.obv  INDEPENDENT MODE ninipân kinipân ninipânân kinipânaw kinipânâwâw nipâw nipawak nipâyiwa  NOMINAL POSSESSION nimis kimis nimisinân kimisnaw kimisiwâw omisa omisiwâwa omisiyiwa  Table 3.7. Person-marking in INDEPENDENT clauses vs. possessed nominals If these pronominal proclitics are indexical, they should have a more restricted behaviour than general pronominals: (a) indexical pronominals cannot be bound16, and (b) indexical pronominals cannot lack referential features.  3.2.3.1 Indexical proclitics cannot be bound 1st- and 2nd- person pronominal forms are anaphoric variables; rather, they are a sub case of deixis: they directly point to the speech act participants (speaker and hearer). Thus, in possession, niminôsim ‘my cat’ is the cat of the speaker (38a); in the clause ninêstosin ‘I am tired’ it is the speaker who is tired (38b).  16  As we will see, English forms like I and you can be bound in some contexts. Although they are often considered prototypical indexical forms, I would argue following Heim (1991) that the binding facts mean I and you cannot be dedicated indexicals; i.e., they only have an indexical function in some contexts.  77  (38)  a.  niminôsim ni- minôs -im 1- cat -DISJ ‘my cat’  b.  ninêstosin ni-nêstosi -n 1- tired.VAI-SAP ‘I am tired.’  In Plains Cree, the pronominal forms are dedicated indexicals, where pronominal forms in English are not, as can be seen by looking at the contexts in which first- and second person may function as variables (cf. Heim 1991, Partee 1989, Kratzer 1998, Rullmann 2003, 2004, among others). For example, for at least some speakers of English, the second ‘I’ in (39) is a boundvariable; in ellipsis contexts it can be bound by the higher subject (e.g., John). (39)  Only I got a question I understood; John didn’t. = (i) John didn’t get a question John understood. Bound-variable = (ii) John didn’t get a question I understood. Indexical (adapted from Rullmann 2004)  In Plains Cree however, the bound variable reading of ‘I’ and ‘you’ must be represented by anaphoric CONJUNCT agreement. Ellipsis is done as in (40), with the contrastive conjunction mâka ‘but’, matrix negation môya, and the subject Jeff17: (40)  niya niwâpamâw wacask, mâka môya Jeff. niya ni-wâpam-â-w wacask mâka môya J 1.EMPH 1see.VTA-DIR-3 muskrat but NEG J ‘I saw a muskrat but Jeff didn’t (see a muskrat).’  In order to get a bound-variable reading of 1st-person, the CONJUNCT form in (41a) is used; when this utterance was presented to the consultant, the bound-variable reading was volunteered, and the consultant strongly dispreferred the non-bound-variable reading. When the INDEPENDENT form is substituted, the utterance is ruled ungrammatical – in other words, contexts that allow a bound variable reading prohibit indexical clauses.  17  Ellipsis structures have not, to my knowledge, been discussed in the Plains Cree literature previously.  78  (41)  a. = ≠? b.  niya niwâpamâw atim kâ-nitonak, mâka môya Jeff. CONJUNCT niya ni- wâpam -â -w atim kâ- niton -ak mâka môya J 1.EMPH 1- see.VTA -DIR-3 dog C2- look.for.VTA-1>3 but NEG J ‘I saw the dog I was looking for, but Jeff didn’t see the dog he was looking for.’ ‘I saw the dog I was looking for, but Jeff didn’t see the dog I was looking for.’  * niya niwâpamâw atim ninitonâw, mâka môya Jeff. INDEPENDENT niya ni- wâpam -â -w atim ni- niton -â -w mâka môya J 1.EMPH 1- see.VTA -DIR-3 dog 1- look.for.VTA-DIR-3 but NEG J ---  Likewise, the counterpart examples discussed by Lakoff (1968) are obligatorily translated into the CONJUNCT order. For example, in (42), the dreamer is dreaming that s/he is someone else, and as that other person, kisses the dreamer. In such a context, the English I in ‘I kissed myself’ is not indexical. The indexical clause in Plains Cree is ruled ungrammatical. (42)  context: Speaker is describing a dream in which s/he was someone else (Bridget Bardot), and as that other person, kisses the dreamer a.  CONJUNCT  ê-pakwatamân awâs-tipskaw Bridget Bardot êsa niya êkwa ê-ocêmisoyân ê- pakwâtam -ân awas-tipiskâw BB êsa niya êkwa ê- ocêmiso -yân C1-dream.VTI -1 last-night BB EVID 1.PRO and C1-kiss.VAI -1 ‘I dreamt I was Bridget Bardot, and I kissed myself.’ b.  INDEPENDENT  !  ê-pakwâtamân awas-tipskaw Bridget Bardot êsa niya ekwa nitocêmison ê- pakwâtam -ân awas-tipiskâw BB êsa niya êkwa ê- ocêmiso -yân C1-dream.VTI -1 last-night BB EVID 1.PRO and C1-kiss.VAI -1 --comment: ‘I’m not familiar with this [nitocêmison] form.’  Thus, the pronominal proclitics in Plains Cree have a fixed reference: they do not introduce a bound-variable dependency.  3.2.3.2 The absence of third-person proclitics Unlike 1st-and 2nd-person pronominals, which are deictic on the the speech act, third-person pronominals such as her/him/she/he/it must be assigned reference by an antecedent or by some  79  corresponding deixis (i.e., pointing) (cf. Postal 1969, Ritter 1995, Heim & Kratzer 1998, Déchaine & Wiltschko 2002). They are also only accidentally part of the speech situation – there is nothing in the features of a third person that necessarily links them to a speech act. In possession, the dependence of third-person pronominals can be seen by the infelicity of uttering ominôsima ‘his/her cat’ without specifying the antecedent of o- (e.g., Fred in (43)). (43)  #(Fred) wâpamêw ominôsima wâpam-ê-w F o- minôs -im -a see.VTA-DIR-3 F 3-cat -DISJ-OBV ‘Fred saw his cat’  However, notice that the o- is necessary to show the relation between Fred and minôs (44). The prefix o- thus codes reference, but does not identify the referent. (44) * Fred wâpamêw minôsima F wâpam -ê -w minôs -im -a F see.VTA-DIR-3 cat -DISJ-OBV --This means that the 3rd-person prefix o- should not be present in an indexical clause, where it would be located in spec, CP. This is correct. (45)  a.  Fred nêstosiw F nêstosi -w F tired.VAI-3 ‘Fred is tired.’  b.  * Fred onêstosiw F o- nêstosi -w F 3- tired.VAI-3 ---  Notice that this argument does not say that there are no third-person arguments in INDEPENDENT clauses (notice that 45a is completely grammatical), or even that there is no agreement for third person arguments (45a has the third-person subject agreement suffix -w). Rather, the point is  80  that there is no third-person pronominal form sitting in spec, CP. Since this is the position occupied by the indexical speech situation, this is the position that is of relevance here18. (46)  CP 2 ni2 ki5 * o-  This raises the question of how third persons in general (i.e., apart from the pronominal proclitic) behave in indexical clauses, which I turn to next.  3.2.3.3 Referents are deictically anchored in indexical clauses We have so far considered the properties of first- and second-person referents, and I have shown that they are always indexical in indexical clauses. I have not so far considered the properties of 3rd-person referents. In particular, while 1st and 2nd person referents, who I take to be speech act participants (cf. Fillmore 1975, Benviniste 1950) and thus licensed by the s0 speech situation, 3rd person referents have no indexical properties since they are not speech act participants. The current analysis therefore predicts a different structure for third-person referents in indexical vs. non-indexical clauses, with accompanying distributional and interpretation differences. There are two difficulties with testing this prediction: first, there is in general a lack of criteria that could be used to distinguish different kinds of referents; second, reference-tracking in Plains Cree specifically is not at all well-understood. What I have to say here will be tentative in nature; this is a huge topic for further research. Minimally we see that the forms used to refer to third-person referents differ between indexical INDEPENDENT clauses and anaphoric CONJUNCT ones: in the former we have the suffix w, and the latter we have -t (3rd.animate) and -k (3.inanimate). Of course until we have an idea  18  This predicts that Algonquian languages which do have the third-person prefix (e.g., Ojibwa, Blackfoot) would either lack the indexical/anaphoric distinction described here for Plains Cree, or that the third-person prefix would on independent grounds be in some other position (e.g., in IP).  81  about the semantics of these agreement markers, this does not provide evidence of what the difference is: but at least these facts are consistent with my claim. Let us suppose that referents can be defined over some unit of discourse (i.e., they are the topic of that unit) (cf. Longacre 1979, Fox 1987a, b, Smith 2003 for the correspondence between reference-tracking and topicality in English and other languages; see also the discussion of reference tracking in centering theory: Brennan, Friedman & Pollard 1987, Grosz, Joshi & Weinstein 1995, etc.). We could then say that INDEPENDENT clauses only pick out referents which are topics; if the referent they refer to is not a topic, an overt nominal will be necessary. There is some preliminary evidence that this is on the right track, although much more work would need to be done to work out this analysis in detail. I discuss three cases from the corpora that I have found: (i)  cases where the indexical INDEPENDENT clause uses an overt nominal even when the nominal was also present in the previous CONJUNCT clause;  (ii)  cases where an indexical INDEPENDENT clause requires re-introduction of a referent (via an overt nominal) in a subsequent clause; and  (iii)  cases where an indexical INDEPENDENT clause has no overt nominals, but refers to the main characters of a story.  I also consider the properties of the disjoint-subject marker -yi, and show that it has only occurs in two restricted environments in indexical INDEPENDENT clauses, both of which are consistent with the analysis of INDEPENDENT clauses. While the data presented here is not conclusive, and by no means a full account of overt nominals, it may serve as a step towards understanding the structure of discourse in Plains Cree.  3.2.3.3.1 Distribution of overt nominals in a discourse The first piece of evidence suggesting that indexical INDEPENDENT clauses do not contain anaphoric reference to referents comes from data like (47). Here we find a CONJUNCT order clause accompanied by the overt nominal awa nâpêsis ‘little boy’; this is followed by an indexical INDEPENDENT clause, which is also accompanied by the same overt noun phrase (awa nâpêsis ‘the little boy’).  82  (47)  overt nominal + CONJUNCT …, awa nâpêsis êkwa awa ê-nihtâwikit, awa nâpêsis êkwa awa ê- nihtâwiki -t DEM.AN boy and DEM.AN C1-born.VAI -3 ‘…, and when the little boy was born, overt nominal + INDEPENDENT nipamihâw mân âwa nâpêsis ê-~ ê-~, … ni- pamih -â -w mâna awa nâpêsis 1- care.VTA-DIR-3 usually DEM.AN boy I would look after him too.’ (AA 5.5)  Based on observations made about English and theories of anaphora, we would expect the main clause to be fine without this referent (cf. Ross 1967, Langacker 1969, Reinhart 1976, 1983, and the accompanying English translation). Based on the analysis of indexical clauses, however, we expect this kind of ‘repetition’, since the reference to a third-person should be defined within the clause (i.e., by the overt nominal). Another suggestive piece of evidence has to do with what happens when an overt nominal is introduced by an indexical INDEPENDENT clause. Here the following INDEPENDENT clauses can also refer to awa nâpêsis ‘the little boy’, but only as long as each indexical INDEPENDENT  clause successively refers to him. Thus, in (48), all three indexical INDEPENDENT  clauses refer to him. (48)  ..., nipamihâw mân âwa nâpêsis ê-~ ê-~, kâ-sipwêhtêcik mân ôki, ni- pamih -â -w mâna awa nâpêsis kâ-sipwêhtê -t -k mâna ôki 1- care.VTA-DIR-3 usually DEM.AN boy C2-leave.VAI -3-PL usually DEM.AN ‘..., I would look after him too, when they [his parents] went out, niya mâna nikanawêyimâw êkwa ê-pamihak. niya mâna ni- kanawêyim -â -w êkwa ê- pamih -ak 1.EMPH usually 1- care.VTA -DIR-3 and C1-care.VTA-1>3 I kept him and looked after him. êkosi piyis aci-misikicisiw, ... êkosi piyis aci- misikicisi -w TOPIC finally INCEP-big.VAI -3 So at last he was getting quite big, …’ (AA 5.5-6)  83  However, when the speaker then switches to clauses referring only to herself, and then mentions the boy again in the INDEPENDENT, an overt nominal is again used. The following example picks up immediately after the last example left off. The INDEPENDENT clauses in the first line (bolded) refer only to the speaker; the INDEPENDENT clause at-ôhpikiw ‘he was growing up’ in the second line (bolded) refers to the little boy again, and has the overt nominal phrase nâpêsis awa ‘the little boy’ (bolded and underlined). (49)  …, êkwa êkos êkwa nikî-pônatoskân êkwa nîtê kâwi nikî-isi-kîwân. êkwa êkosi êkwa ni-kî- pônatoskê -n êkwa nîtê kâwi ni- kî- isi- kîwê -n and TOPIC and 1- PREV-stop.work.VAI-SAP and there again 1- PREV-thus-gohome.VAI-SAP ‘…, and then I quit my job and went back home over there. kîtahtawê kâ-pîhtamân aya (-- at-~ at-ôhpikiw awa nâpês-~, nâpêsis awa, ... kêtahtawê kâ- pêhtam -ân aya ati- ôhpiki -w awa nâpêsis awa suddenly C2- hear.VTI-1 CONN INCEP-grow.VAI-3 DEM.AN boy DEM.AN Later I heard (the little boy was growing up, …’ (AA 5.6)  Here, it appears that once reference has been shifted away from the boy, it cannot be picked up again with an indexical INDEPENDENT order clause; an overt nominal phrase is used to reestablish the referent. The same pattern happens again in (50) (taken from later in the same corpus). Here an indexical INDEPENDENT clause is accompanied by an overt nominal phrase an îskwêw ‘that woman’, and the next indexical INDEPEPENDENT clause also refers to her. In the third and fourth lines, we have indexical clauses which refer only to the speaker19, and when the next indexical clause refers again to the woman, the overt nominal phrase is re-used. (50)  …, êkwa nikî-~ kî-âcimâw an îskwêw an âyi, ê-asiwasot ôtê Battleford, êkwa kîâcim -â -w ana iskwêw ana ayi ê- asiwaso -t ôtê B and PREV-tell.VTA-DIR-3 DEM.AN woman DEM.AN CONN C1-be.locked.up.VAI-3 there B “…, and then that woman was said to be locked up over there at Battleford, êtikwê kî-~ kî-kîskwêyêyihtam, êtikw ânim ê-kî-pâskiswât onâpêma. êtikwê kî- kîskwêyêyihtam -w êtikwe anima ê- kî- pâskisw -â -t o- nâpêm -a DUBIT PREV-crazy.VTI -3 DUBIT DEM.INAN C1-PREV-kill.VTA -DIR-3 3-husband-OBV she must have gone mad, I suppose, upon killing her husband.  19  I do not count the clauses in the direct quote, since those are part of a separate discourse (cf. Banfield 1982).  84  ê-wî-~ ê-~ nikî-itohtân ê-nitawi-nitonawak, ni-kî- itohtâ -n ê- nitawi- nitonaw -ak 1-PREV-go.VAI-SAP C1-golook.for.VTA-1>3 I went there to go and look for her, «mwâc, môy ôta ayâw, môy ôhci-pimâtisiw,» nikî-itikawin, namwâc môya ôta ayâ -w môy ohci- pimâtisi -w ni-kî- it -ikawi -n NEG NEG here be.VAI-3 NEG PREV-live.VAI -3 1- PREV-say.VTA-USC -SAP and I was told “No, she is not here, she has died,” êkosi môy nôh-wâpamâw an îskwêw. êkosi môy n- ôh- wâpam -â -w ana iskwêw topic NEG 1-PREV-see.VTA-DIR-3 DEM.AN woman and so I never did get to see that woman.” (AA 5.7) The final set of data I talk about has to do with topics of the story. Mühlbauer (2007) argues that when we look at the distribution of nominals in a Plains Cree discourse, we find that there are two different kinds of nominals, introduced at two different stages of the discourse. In the initial stage of a discourse, speakers introduce a set of referents, and identify the relation of each referent to the speaker, via a kin-term, some intermediate individual, or by shared space/time. In the second stage of a discourse, the speaker talks about the events surrounding these referents (cf. also Janzen 2004 for similar patterns in American Sign Language). Referents that are introduced in the intial stage I will refer to as topic referents. For example, in the following piece of narrative, taken from an earlier point of the same narrative as the past two pieces of data) the narrator (Alice Ahenakew) starts with an intransitive verb in the INDEPENDENT order, then introduces her relation to the two ‘main characters’ môniyâwak ‘white people’ via the transitive verb ê-kî-atoskawakik ‘I worked for them.’ (51)  êkwa ôtê mîna mâna nikî-atoskân ôtê isi, êkwa ôtê mîna mâna ni- kî- atoskê -n ôtê isi and here also usually 1- PREV-work.VAI-SAP here DIR ‘And then I also used to work over in this direction, môniyâwak ê-kî-atoskawakik. môniyâw -ak ê- kî- atosk-aw -ak -ik white -PL C1-PREV-work-BEN-1>3-PL I used to work for White people.’ (AA 5.5)  These two main characters are then identified independently of any events in the story, as given in (52). 85  (52)  pêyak ana, aya, Irish an[a] îskwêw, êkwa ana nâpêw German, ... pêyak ana aya Irish ana iskwêw êkwa ana nâpêw G one DEM.AN CONN Irish DEM.AN woman and DEM.AN man G ‘The one, the woman was Irish, and the man was German,’ (AA 5.5)  And then the narrator proceeds to tell the story of working for these two people, and how, eventually, some time after she stopped working for them, the woman ends up killing her husband. Several other referents come into this story at different points: e.g., the couple’s children (a boy and two girls), several other nouns are also used; however the story centers around these two individuals. These I call the topic referents of the story. In fact, other than the instance we saw above with the awa nâpêsis ‘little boy’, all other instances of INDEPENDENT clauses in this narrative that lack overt nominals for all their arguments refer to one of these two referents. There are three such cases, distributed throughout the story. The first one refers to the woman and has no nominal phrases outside the verbal marking (53). (There is no overt nominal in the previous clause either.) (53)  INDEPENDENT  refers to text-level woman  …, nitaw-ôpêpîmiw êkwa ... nitaw- opêpîmi -w êkwa gohave.baby.VAI-3 and …, she went to have her baby, … (AA 5.5) The second case is a transitive verb with the man as the subject and the woman as the object. Here there is a possessive form wîwa ‘his wife’ referring to the woman as the man’s wife. While the demonstrative ana ‘that.AN’ refers to the man, the demonstrative without an accompanying noun does not have enough information to identify the referent. The man has not been brought up since line 2. (54)  INDEPENDENT  refers to text-level man  ..., êkotê êkwa itohtahêw ana wîwa; êkotê êkwa itohtah -ê -w ana w- îw -a there and take.VTA-DIR-3 DEM.AN 3- wife-OBV ‘…, then that man took his wife there;’ (AA 5.5) Finally, the last line of the story contains just a bare INDEPENDENT clause. The verb is transitive, and is the punch line of the story: the woman introduced at the beginning of the story has killed 86  the man introduced at the beginning of the story (notice, for example, the emphatic flavour of the English with the affirmative did. (55)  INDEPENDENT  refers to both text-level individuals  …, kî-nipahêw. kînipah -ê -w PREV-kill.VTA-DIR-3 …, and she did kill him.’ (AA 5.6) To sum up then – INDEPENDENT clauses referring to third persons seem to have a specific set of properties with them. First, they will repeat overt nominals even when the relevant referent was introduced in the previous anaphoric clause. Second, a referent introduced in an indexical clause is reintroduced after an intervening indexical clause. Third, other indexical INDEPENDENT clauses lacking overt nominals correspond to referents that are the main ‘topics’ of the story. These observations provide an opening into much further research.  3.2.3.3.2 Restrictions on switch-reference in INDEPENDENT clauses Plains Cree has a switch-reference marker -yi, which marks a subject disjoint from an argument of some other predicate (different subject (DS); Dahlstrom 1991, Mühlbauer 2007, in prep). Because -yi requires that arguments across two predicates be evaluated, the current analysis predicts that -yi in indexical INDEPENDENT clauses will be either ill-formed or have a deictic function. The latter prediction is borne out and actually predicts a pattern has not generally been noticed in the language. Thus, teaching grammars such as Ahenakew (1987) and Hunter, Karpinski, & Mulder (2001) show, as part of the standard paradigm, independent forms with the -yi suffix (usually termed obviative agreement), but it is not coincidental that all of their examples use a CONJUNCT form of the verb. In fact, a look at Wolfart (1973) shows that, in running speech, -yi is only attested in some of the expected forms in the INDEPENDENT order, (Wolfart 1973:41; Ahenakew 1987 also gives text counts which show that -yi is very restricted in the indexical INDEPENDENT clauses). Table 3.8 replicates Wolfart’s findings.  87  Syntactic context  Expected form  Attested in INDEPENDENT  order? ✔ Intransitive DS STEM-yiwa ✖ Transitive DS > 1 ni-STEM-ikoyiwa ✖ DS > 2 ki-STEM-ikoyiwa ✔ DS > 3OBV STEM-êyiwa Table 3.8. (Un)attested co-occurrence of -yi in INDEPENDENT order (Wolfart 1973) Even when we limit ourselves to the forms which do occur in running speech, which I will discuss below, it is difficult to reproduce these forms in elicitation contexts for reasons that will become clear below. Consider the following pair, both of which were presented in elicitation. With the anaphoric CONJUNCT form, the consultant found the sentence infelicitous without context, but when asked how it would be interpreted, (e.g., if accidentally overheard) could translate it (56a). By contrast, consultants do not even recognize the form in (56b) - it is uninterpretable. (56)  Context: presentation of different subject marking in CONJUNCT and INDEPENDENT forms a.  # nâpêwa ê-nikamoyit nâpêw -a ê- nikamo -yi -t man -OBV C1-sing.VAI-DS-3 ‘Someone’s guy is singing.’  CONJUNCT  comment: who are you talking about? b. !  anihi nâpêwa nikamoyiwa anihi nâpêw -a nikamo -yi -w -a dem man -OBV sing.vai-DS -3 -OBV --  INDEPENDENT  comment: I’ve never heard that before. The inability of consultants to recognize the latter form in elicitation contexts highlights the difference between indexical clauses, and anaphoric clauses, which even when unembedded, can be interpreted with respect to some previous antecedent. When we look at running speech, we do see INDEPENDENT clauses with the -yi suffix on them – but only in two specific contexts (Mühlbauer 2007, in prep). The first is when the subject of the verb is possessed, as in (57).  88  (57)  Clare omâmâwa ka-nikamoyiwa C o-mâmâ -a ka- nikamo -yi -w -a C 3-mother-obv IRR-sing.VAI-DS -3 -OBV ‘Clare’s mother will sing.’  INDEPENDENT  In this case, there is internal structure of the DP subject Clare omâmâwa ‘Clare’s mother’; in particular, there are two referents: Clare and omâmâwa ‘her mother’. The different subject marks disjunction between the subject of the verb omâmâwa ‘her mother’ and the subject of the possession construction Clare as represented in (58); crucially, there is no cross-clausal antecedence relation between the two subjects. (58)  -yi: (Subjomâmâwa ≠ SubjClare ) XP 3 DPi IP 2 2 DP2 DP1 2 5 5 -yi XP Clare omâmâwa  This context accounts for almost all -yi marked INDEPENDENT clauses, and it should be underlined that this is the only context I know of where a consultant has accepted a -yi marked INDEPENDENT  clause in an elicitation context.  The second context where -yi can occur is when it marks what again look like text-level referents. For example, in (59), the speaker is telling a joke about a dead prairie-chicken found by a woman going to church. The prairie chicken is introduced as an obviative referent relative to the woman, and is the referent on which the joke hangs. At the point of the story where this clause is uttered, there are only two possible referents in the discourse: nôcikwêsiw ‘old woman’ and pihêwa ‘(obviative) prairie chicken’. Further, previous to this clause, there have been no switching of subjects between clauses. In this situation, then, we have an utterance like (59).20  20  When these kinds of examples are presented to speakers in elicitation contexts, they are accepted as fine, but when the speakers are asked to reproduce them, the context seems to evaporate and the different subject marking disappears.  89  (59)  context: middle of story about woman and prairie chicken ..., mâk êtikwê miyâkosiyiwa, ... mâka êtikwê miyâkosi -yi -w -a but EVID stink. VAI -DS -3 -OBV ‘..., but it [the prairie chicken] must have been smelling already, ...’ (AA7.1)  In this case we can say that the -yi is not anaphorically dependent on the previous clause, but rather it is deictically pointing to the (only) obviative referent in the story. Thus the subordinate subject marker -yi, has deictic behaviour in INDEPENDENT clauses (cf. its anaphoric behaviour in CONJUNCT clauses, discussed in chapter 4).  3.3  The semantics of indexical clauses: Indexicality  We now turn from the structural conditions on indexical INDEPENDENT clauses to the implications of the clause being anchored to the speech act (via the situation variable in spec, CP); in particular focussing on the temporal implications (i.e., that indexical INDEPENDENT clauses are evaluated with respect to speech time) and referential implications (i.e., that indexical INDEPENDENT  clauses are evaluated with respect to the speaker).  First, I show that indexical clauses have a privileged temporal relationship to the speech time (§3.3.1). Second, I show that indexical clauses have a privileged referential relationship to the speaker (§3.3.2).  3.3.1 Temporal deixis: Relating reference time to speech time In this section, I consider the temporal properties of indexical clauses. Within a Reichenbachian framework (cf. Paul 1886, Reichenbach 1947, Hinrichs 1986, Enc 1987, Hornstein 1990, Kamp & Reyle 1993, Klein 1994, Kratzer 1998, Demirdache & Uribe-Etxebarria 2002, among others), there are traditionally three times: the speech time (aka utterance time), the reference time (aka topic time), and the event time (aka situation time). These are defined in (60) and related to the corresponding linguistic structures.  90  (60) a. b.  Speech Time (T0) Reference Time (Tref)  c.  Event time (Tsit)  the time of speaking the time the sentence makes a claim about the time of the event or situation  speech act propositional structure predicate structure  From these times, a large number of relations can be made to model different tense/aspectual distinctions (see, for example Klein 1994). The relation between these times can be sequencing in nature, i.e., x precedes y, x follows y. It may also be a relation of inclusion: x includes y or conversely, y includes x21. Following recent work in the tense/aspectual literature, I model the relations between these times as a [± coincidence] relation, where [- coincidence] captures the sequencing relation, and [+ coincidence] captures the inclusion relation (cf. Demirdache & Uribe-Etxebarria 199x, 200x; Ritter & Wiltschko 2005, 2007)22. Tense relations are generally taken to be relations anchored to the speech time, and aspectual relations those anchored to the situation time. In this thesis, since we are interested in clause-typing, I am primarily concerned with the former type of relation. More recently, many analyses (e.g., Gennari 2003, among others) also make reference to an evaluation time, as in (61). (61)  Evaluation Time (Teval)def: the time with respect to which the truth of the sentence (i.e., proposition) is evaluated  The status of this fourth time with respect to the others is often vague. In this thesis, I integrate the evaluation time into the other times in a specific way, such that, for any clause, there are only 21  In order to relate multiple times, multiple relations are often necessary. For example, the specification of an English past perfect specifies a precedence relation between the reference time and the utterance time, and a precedence relation between the reference time and the situation time (cf. Klein 1994). (i) - COIN (Tref, T0) - COIN (Tref, Tsit) Because these relations can be established (at least semi-)independently, we expect that they could be established via multiple parts of the grammar: e.g., there is no need to assume that a single morpheme would give both the relation between Tref and Tsit and the relation between Tref and T0. 22 The [± coincidence] feature is in fact an over-simplified analysis, since the precedence relation could also in theory be reversed, where the topic time must precedent the situation time, yielding a future tense (Kamp & Rohrer 1983). To distinguish the future from the past, we therefore need some additional specification. While there is on independent grounds substantial agreement that the future needs a modal component (Jespersen 1924, Comrie 1985, Hornstein 1990, Abusch 1998, Copley 2002, Matthewson 2006, among others), the modal component is again not sufficient to derive the temporal properties of the future, and the ordering component of the future must be specified as [ FOLLOW]. Thus an alternative analysis fully compatible with the data here is to specify the relations as [OVERLAP] and [ PRECEDENCE].  91  three relevant times. Specifically, I model the relation often characterized as a relation between reference time and speech time as a relation between reference time and evaluation time. (62)  Tense relations: + COIN (Tref, Teval) -  COIN  (Tref, Teval)  What the indexical/non-indexical split in clauses does is give the value of the evaluation time Teval. Within the situation semantics framework, all of these times will be derived from situations. Here I focus on the ‘times’ aspect, and use the ‘times’ notation. But if indexical INDEPENDENT  clauses are evaluated with respect to a speech situation (s0), then the time they will  be evaluated with respect to (i.e., the evaluation time) is the speech time T0. (63)  s0  T0  By transitivity, in indexical clauses the tense relation will always be between the reference time and the speech time. (64)  Indexical: Evaluation time is speech time (Teval = T0) + COIN (Tref, T0) - COIN (Tref, T0) In anaphoric CONJUNCT clauses, by contrast, the proposition is evaluated with respect to  some (unspecified) situation; thus it will be evaluated with respect to some (unspecified) time (in chapter 4, I argue that this time is established according to general principles of anaphora). In anaphoric CONJUNCT clauses, the tense relations will therefore be as in (65). (65)  Anaphoric: Evaluation time is anaphorically give (Teval = T) + COIN (Tref, T) - COIN (Tref, T) The definition in (64) captures the fixed reference to speech time that indexical clauses  have; in §3.2.2.4 I use the distinction between (64) and (65) to capture the different interpretations of indexical (Plains Cree INDEPENDENT) and anaphoric (Plains Cree CONJUNCT) clauses. 92  The second part of the claim is that a bare indexical clause in Plains Cree has a [+ coincidence] value;23 in order to get a [- coincidence] value, the temporal shifting preverb kîmust be added to the verbal complex. Putting everything together, the temporal value of a bare INDEPENDENT  (66)  order clause in Plains Cree is thus as in (66).  + COIN (Tref, T0) Thus, while bare INDEPENDENT clauses do not map directly onto either of the English  past/present tense distinctions in terms of distribution, they are more like a present tense in that the reference time coincides with the speech time. This captures the generalization that Plains Cree’s INDEPENDENT clauses always have ‘present relevance’, a notion that is also recurrent in the literature for other languages (cf. Huddleston 1969, Lakoff 1970, Bennett & Partee 1972, Comrie 1976, Dowty 1979, Klein 1992, 1994, among others). There are at least two possible interpretations that the [+ COIN (Tref, T0)] relation is compatible with, depending on what the relation between Tevent and Tref is. First, it is compatible with a ‘present tense’ interpretation, where Tevent also coincides with Tref. (67)  present: ‘I am happy.’ + COIN (Tref, T0) + COIN (Tref, Tevent)  Second, the coincidence relation between Tref and Teval could be conjoined with a noncoincidence relation between Tevent and Tref. This is approximately the model for the English present perfect (Klein 1994): the reference time must be in the posttime (i.e., after) the situation time (i.e., expressed by the predicate), and must also include the utterance time. (68)  present perfect ‘I have eaten three apples.’ + COIN (Tref, T0) - COIN (Tref, Tevent)  Since Tevent is associated with the predicate domain and Tref with the propositional domain, and these are both below the clause-typing domain, I will not be concerned with the representation of 23  I leave as an open question whether the unmarked value of [+ coincidence] should be treated as an inherent (i.e., universal) value of indexical clauses, or whether this is specific to Plains Cree’s indexical clause.  93  aspectual value (cf. Klein 1992, Kratzer 1998, Bohnemeyer & Swift 2004). I show that while there is variation between a present and present perfect interpretation (which is indirectly related to aspectual value), there is always a [+ COIN] relation between the reference time Tref and the speech time T0. If the predicate is stative, the state holds at speech time; if the predicate is an activity (i.e. a-telic), the event or resultant state holds at speech time; if the predicate is an accomplishment (i.e., telic), then the resultant state holds at speech time (i.e., like a present perfect). This is summarized in table 3.9. ASPECT CLASS INTERPRETATION Stative predicates present Activity predicates present Telic predicates present perfect Table 3.9. Interpretation of INDEPENDENT clauses by aspectual class  3.3.1.1 Contrasting temporal interpretations of indexical and non-indexical clauses The temporal difference between indexical and non-indexical clauses is represented in (69): while the [+coincidence] relation is indexically given as the speech time in INDEPENDENT order clauses, it is unspecified in anaphoric CONJUNCT clauses, and must be given by context (specifically, it must have an antecedent). (69)  a.  + COIN (Tref, T0)  INDEPENDENT  b.  + COIN (Tref, T)  CONJUNCT  Probably the most striking illustration of the anaphoric nature of non-indexical CONJUNCT clauses is that they may be embedded with respect to another clause, in which case they are dependent on the higher clause for temporal interpretation (cf. Chapter 4). In (70) the crying is interpreted relative to when Kim spoke to me, not relative to speech time. (The baby could still be crying, but the sentence says nothing about that.)  94  (70)  Kim nikî-itik bêbîsis ê-mâtot K ni- kî- it -ikw bêbi -sis ê- mâto -t K 1- PREV-tell.VTA-INV baby-DIM C1-cry.VAI-3 ‘Kim had told me that the baby was crying = + COIN (crying, T; T = TELL)  The sentence in (70) tells us that the temporal interpretation of embedded anaphoric (CONJUNCT) clauses can be set with respect to some time that is not the speech time (i.e., it is not fixed). However, even in unembedded contexts, the temporal interpretation of CONJUNCT clauses is not fixed. In (71), both utterances are translated exactly the same way, but they are nevertheless temporally distinguished: the CONJUNCT form can be used to report something you heard earlier, or something cotemporaneous with speech time; the indexical INDEPENDENT has only the cotemporaneous interpretation. (71)  a.  ê-pêhtawak nisîmis wayawihtamihk ê-mêtawêt ê- pêhtaw -ak ni- sîmis wayawihtam -ihk ê- mêtawê -t C1-hear.VTA-1>3 1- sibling outside -LOC C1-play.VAI -3 ‘…I heard my little brother playing outside.’  CONJUNCT  comment: This one could mean any time; it could mean before, or it could mean I’m hearing him now b.  nipêhtawâw nisîmis wayawihtamihk ê-mêtawêt ni- pêhtaw -â -w ni- sîmis wayawihtam-ihk ê- mêtawê -t C1-hear.VTA-DIR-3 1- sibling outside -LOC C1-play.VAI-3 ‘I heard my little brother playing outside.’  INDEPENDENT  comment: …like I’m on the phone with you, and he’s making noise, and I’m telling you about it Second, in places where INDEPENDENT clauses must be marked with kî- (e.g., states that hold at some time other than utterance time), CONJUNCT clauses can be bare. The temporal value of the indexical clause is interpreted with respect to speech time, so an overt [- coincidence] element like kî- is needed (72a, 73a). The temporal value of the anaphoric CONJUNCT clause (72b, (72b) is interpreted with respect to the time established by the previous clause: the times of the events coincide so kî- is not needed.  95  (72)  (73)  a.  kâ-wâpamak Jeff, kî-nôhtêhkatêw kâ-wâpam -ak J kî- nôhtêhkatê -w C2-see.VTA-1>3 J PREV-hungry.VAI -3 ‘When I saw Jeff, he was hungry.’  INDEPENDENT  b.  kâ-wâpamak Jeff, ê-nôhtêhkatêt kâ-wâpam -ak J ê- nôhtêhkatê -t C2-see.VTA-1>3 J C1-hungry.VAI -3 ‘When I saw Jeff, he was hungry.’  CONJUUNCT  a.  kâ-wâpamak Jeff, kî-âhkosiwpayiw kâ- wâpam -ak J kî- âhkosiwpayi-w C2-see.VTA -1>3 J PREV-sick.VAI -3 ‘When I saw Jeff he got sick (suddenly).’  INDEPENDENT  b.  kâ-kî-wâpamak Jeff, ê-âhkosiwpayit kâ- kî- wâpam -ak J ê- âhkosiwpayi -t C2-PREV-see.VTA -1>3 J C1-sick.VAI -3 ‘When I saw Jeff, he got sick (suddenly).’  CONJUNCT  24  Third, in sequences of clauses, if kî- is added to an indexical INDEPENDENT clause, it fixes the relation relative to the speech time, as in (74a, 75a); however if kî- is added to an anaphoric CONJUNCT  clause (74b, 75b), it shifts the event time with respect to the preceding clause, rather  than with respect to speech time. See chapter 4 for more details. (74)  a.  ê-pê-kîwêt Jeff nikî-mîcisonân. ê- pê- kîwê -t J ni- kî- mîciso -nân C1-come-go.home.VAI-3 J 1- PREV-eat.VAI-1PL ‘When Jeff came, then we ate.’ = come home < eat  b.  =  INDEPENDENT  ê-pê-kîwêt Jeff, ê-kî-mîcisoyâhk CONJUNCT ê- pê- kîwê -t J ê- kîmîciso -yân -k C1-come-go.home.VAI-3 J C1-PREV-eat.vai -1 -PL ‘…Jeff came home, we had eaten.’ (we = speaker & someone else) eat < come home  24  Stative predicates in unembedded clauses still strongly prefer to be interpreted as holding at utterance time, but as the example shows, this is only a preference when it comes to CONJUNCT clauses.  96  (75)  a.  kâ-wâpamak Jeff, kî-ahkosiwpâyiw kâ- wâpam -ak J kî- ahkosiwpayi -w C2- see.VTA-1>3 J PREV-sick.VAI -3 ‘When I saw Jeff he got sick (suddenly) (at that time).’  INDEPENDENT  b.  kâ-wâpamak Jeff, tontoni ê-kî-ahkosiwpayit kâ- wâpam -ak J ê- kî- ahkosiwpayi -t C2-see.VTA -1>3 J C1-PREV-sick.VAI -3 ‘When I saw Jeff, he had gotten really very sick.’  CONJUNCT  Finally, the asymmetry between INDEPENDENT and CONJUNCT clauses can be seen in structures where the temporal adverbial tâkosih ‘yesterday’ is in final position (74a). Indexical INDEPENDENT clauses  (76)  are bad, but non-indexical CONJUNCT clauses are fine.  a.  ?* kacis pêyakwâw nimîcison otâkosihk kacis pêyakwâw ni- mîciso -n otâkosihk only once 1- eat.VAI-SAP1 yesterday intended: I only ate once yesterday  b.  kacis pêyâkwâw ê-mîcisoyân otâkosihk kacis pêyakwâw ê- mîciso -yân otâkosihk only once C1-eat.VAI -1 yesterday ‘I only ate once yesterday.’  INDEPENDENT  CONJUNCT  As we will see in the following sections, there is a general prohibition on final past time adverbials with indexical clauses lacking the temporal shifter kî-, across all aspectual classes; this is expected if there is a fixed coincidence relation between reference time and speech time. However, there is no such prohibition on non-indexical CONJUNCT clauses; this is expected if the coincidence relation is between reference time and some unspecified time (i.e., that might be in the past). In this last section I have exemplified the difference between indexical INDEPENDENT clauses and non-indexical CONJUNCT clauses. In the next section, I focus on the temporal interpretations available for indexical clauses in stative and (atelic and telic) eventive predicates. These sections demonstrate the [+coincidence] relation between the reference time Tref and speech time T0. Because the relation is always evaluated relative to speech time T0, these sections also provide additional indirect evidence that the evaluation time for indexical clauses is always T0.  97  3.3.1.2 Indexical clauses present statives that hold at T0 Stative predicates are interpreted as holding at speech time in indexical (INDEPENDENT) clauses, as evidenced by the following characteristics: (i)  they cannot be translated with a true past ‘used to (be)’ constructions;  (ii)  they cannot be used in explicitly past contexts;  (iii)  they are obligatory if the time span referred to (either by context or by a modifier) includes the utterance time; and  (iv)  they cannot co-occur with final past-time adverbials unless there is an overt temporal shifter.  First, if Plains Cree was simply vague or ambiguous with respect to temporal relations, we would expect that it could be mapped onto either tense form in English25. However, when translations which unambiguously identify the predicate as not holding at speech time, such as the periphrastic ‘used to be’ (termed English true past constructions in Lakoff 1970), they are systematically rejected. The preverb kî-, which is often called a ‘past tense’ (see chapter 7 for discussion), is used if the predicate does not hold at speech time. (77)  Permanent stative predicates must have kî- for unambiguous past translation a.  Jeff kinosiw J kinosi -w J tall.VAI-3 = ‘Jeff is tall’ ≠ ‘Jeff used to be tall.’  b. ≠ =  Jeff kî-kinosiw J kî- kinosi -w J PREV-tall.VAI-3 ‘Jeff is tall.’ ‘Jeff used to be tall.’  25  For example, the Plains Cree diminutive -sis is ambiguous with respect to indicating size or quality (i). Notice that both English translations, which are non-ambiguous, are valid. (i) acimosis atimw -sis dog -DIM = ‘small dog (e.g., Pomeranian)’ = ‘puppy (e.g., baby German Shepherd)’  98  (78)  Temporary stative predicates must have kî- for unambiguous past translation a.  nôhtêhkatêw Tomio nôhtêhkatê -w T hungry.VAI-3 T = ‘Tomio is hungry.’ ≠ ‘Tomio was/used to be hungry.’  b.  kî-nôhtêhkatêw Tomio kînôhtêhkatê -w T PREV- hungry.VAI-3 T ≠ ‘Tomio is hungry.’ = ‘Tomio was/used to be hungry.’  Adding overt past contexts which disambiguate the interpretations (criteria (ii)) confirms these translations. For example, when talking about a person who’s no longer living, a stative predicate must be marked with kî- (cf. Wolfart 1990, 2000:170). The absence of kî- codes that the dead person can still have this state attributed to them, as nicely summarized by one consultant. (79)  context: talking about a person who’s no longer living b. # nôhkomipân kahkiyaw kiskêyihtam n- ôhkom -pân kahkiyaw kiskêyihtam -w 1-grandmother-former all know.VTI -3 ‘My grandmother (no longer here) knows everything.’ comment: OK if you believe you still have some communication with grandmother…spiritual [communication], she’s dead, but you still have a relationship with her – for example, a dream where she speaks to you b.  nôhkomipân kahkiyaw kî-kiskêyihtam n- ôhkom -pân kahkiyaw kî- kiskêyihtam -w 1- grandmother-former all PREV-know.VTI -3 ‘My grandmother (no longer here) knew everything.’  99  (80)  context: talking about a friend who recently died a. #  John kinosiw J kinosi -w J tall.VAI-3 ‘John is tall.’  b.  John kî-kinosiw J kî- kinosi -w J PREV-tall.VAI-3 ‘John was tall.’  Likewise, in (81), the state being talked about (living in a particular house) held for some previous time span that does not include the time of speech. Again the bare indexical clause cannot be used. (81)  context: walking by a house that speaker used to live in. Speaker points it out a. #  niwîkin ôta ôma wâskahikan nêwaskiy ni- wîki -n ôta ôma wâskahikan nêwo askiy 1- live.VAI-SAP here DEM.INAN house four year ‘I’ve lived in this house for four years.’  b.  nikî-wîkin ôta ôma wâskahikan nêwiskiy ni- kî- wîki -n ôta ôma wâskahikan nêwo askiy 1- PREV-live.VAI-SAP here DEM.INAN house four year ‘I lived at this house for four years.’  The past context for the INDEPENDENT clause in (82) is disambiguated by the initial clause: the speaker is talking about a state that held many years in the past. Again, the clause must be marked with kî-. (82)  context: old person talking about when they were young a.  b.  * kâ-(kî)-oskinîkiskwêwiyân nikatawasisin kâ- kî- oskinîkiskwêwi -yân ni- katawasisi -n C2-PREV-young.woman.VAI-1 1- beautiful.VAI-SAP --kâ-(kî)-oskinîkiskwêwiyân, nikî-katawasisin kâ- kî- oskinîkiskwêwi -yân ni- kî- katawasisi -n C2-PREV-young.woman.VAI-1 1- PREV-beautiful.VAI-SAP ‘When I was a young woman, I was beautiful.’  100  Conversely, if the state does hold at present, a indexical (INDEPENDENT) clause must be bare (not marked with kî-). In (83) the speaker is talking about a person recently met. In order for the state to be interpretable as holding at utterance time, the unmarked clause must be used. (If the clause is marked with kî-, then the state must hold at a time previous to speech time). (83)  context: talking about tall guy named Bernie after meeting him at a party a.  iyikohk kinosiw Bernie iyikohk kinosi -w B so tall.VAI-3 B ‘Bernie is/was really tall.’  b.  # iyikohk kî-kinosiw Bernie iyikohk kîkinosi -w B so PREV-tall.VAI-3 B --comment: *laughter* how did he get short? is it because he’s older and shrunk?  Likewise, the contrast in (84) shows that with the adverbial phrase ispî otâkosihk ‘since yesterday’, only the bare form is good. (84)  a.  nicihkêyihtên sspî otâkosihk ni- cîhkêyihtê -n ispî otâkosin -k 1- happy.VTI -SAP TIME be.evening.VII-0 ‘I’ve been happy since yesterday.’  b. * nikî-cihkêyihtên sspî otâkosihk ni- cîhkêyihtê -n ispî otâkosin -k 1- happy.VTI -SAP TIME be.evening.VII-0 --- (intended: ‘I’ve been happy since yesterday.’) A final piece of evidence that predicates in an indexical clause must hold at speech time is their co-occurrence restrictions with past time adverbials. Distant past-time adverbials like kayâs ‘long ago’ cannot co-occur at all. (85) * kayâs nimiyosin kayâs ni- miyosi -n long.ago. 1- pretty.VAI-SAP --- (intended: ‘A long time ago I was pretty.’)  101  More recent past time adverbials can co-occur with unmarked statives. If an unmarked INDEPENDENT clause  is contrasted with a kî-marked INDEPENDENT, the difference seems to be that  in the latter case, the state of happiness no longer holds. (86)  a.  otâkosihk nicihkêyihtên otâkosin -k ni- cihkêyihtê -n be.evening-0 1- happy.VTI -SAP ‘I was happy yesterday.  b.  otâkosihk nkî-cihkêyihtên otâkosin -k ni- kî- cihkêyihtê -n be.evening.VII-0 1- PREV happy.VTI -SAP ‘I was happy yesterday.’ comment: there’s sort of an implied BUT… she’s not happy now  However, there is a restriction on this co-occurrence: the adverbial must be in initial position. In final position, they are bad, as in (87). (87)  a.  ninitawêyihtên mîcimâpoy ni- nitawêyihtê -n mîcimâpoy 1- want.VTI -SAP soup = ‘I want soup.’ ≠ ‘I wanted soup.’  b. * ninitawêyihtên mîcimâpoy otâkosihk ni- nitawêyihtê -n mîcimâpoy otâkosin -k 1- want.VTI -SAP soup be.evening.VII-0 --At this point, I am not sure what accounts for the distribution of the adverbials, although one possible line of analysis to pursue is that initial adverbials relate to the event time, while final adverbials relate to reference time (cf. Currie 1995 on time adverbials in Salish; also Klein 1992 on English adverbials with the perfect).  3.3.1.3 Indexical clauses present activities that coincide with T0 Like stative predicates, unmarked activity predicates are interpreted as present in indexical clauses, with either a habitual or imperfective reading, as shown in (88). 102  (88)  Activity predicates must be marked with kî- for unambiguous past reading a.  nikisipêkinên wîyâkana ni- kisipêkinê -n wîyâkan -a 1- wash.VTI -SAP dish -PL = ‘I am washing dishes (right now)’ = ‘I wash dishes in general’  b.  nikî-kisipêkinên wîyâkana ni-kî-kisipêkinê-n wîyâkan-a 1-PREV-wash.VTI-SAP dish-PL = ‘I had washed the dishes’  Activity predicates in bare indexical clauses act like stative predicates with respect to periphrastic past constructions; speakers reject periphrastic ‘used to’ past translations as in (89), even with the habitual element mâna ‘usually/at times’. (89)  a.  nimâton ni- mâto -n 1- cry.VAI-SAP = ‘I am crying / I cry.’ ≠ ‘I used to cry.’  b.  nimâton mâna ni- mâto -n mâna 1- cry.VAI-SAP usually = ‘I cry at times.’ ≠ ‘I used to cry (at times).’  Activity predicates also behave like stative predicates in that speakers reject the bare indexical clause in past contexts. Notice that in many cases (e.g., (90)), speakers will volunteer an English past translation, but reject the past context. (90)  nimîcison ni- mîciso -n 1- eat.VAI-SAP = ‘I’m eating right now.’ = ‘I eat.’ (Like say you lost weight, and someone asks ‘aren’t you eating anymore?’) ≠ ‘I ate’ (consultant allows translation ‘I ate’ but says it is bad in a past context, e.g., talking about the big meal you had earlier in the day)  One possible explanation for this translation/interpretation discrepancy is that it stems from the convergence of tense (past) and aspect (completive) in English past tense constructions: in these  103  contexts, the speaker is attending to the fact that event time can precede and be completed by reference time, rather than attending to the relation between reference time and speech time. At any rate, when a past relation is established (e.g., by using a context), the speaker systematically rejects the bare indexical clause forms. Also like stative predicates, when the context is set up such that the reference time includes the speech time, an indexical clause is obligatorily bare. This includes present perfect contexts, as in (91). (91)  context: walking across British Columbia, reach a friend’s house after three weeks, but still have a long ways to go a.  mistahi nipimohtân mîhcêt kîsikâk mistahi ni- pimôhtê -n mîhcêt kîsikâ -k much 1- walk.VAI-SAP many be.day.VII-0 ‘I’ve walked a lot/many miles in many days.’  b. # mistahi nikî-pimohtân mistahi ni- kîpimohtê -n much 1- PREV-walk.VAI -SAP ‘I had walked a lot/ many miles’ comment: I wouldn’t guess that you’re going to walk more In terms of past time adverbials, unmarked activities can be modified by tâkosihk yesterday, but like with statives, the adverbial must be in initial position. (92)  a.  otâkosihk, mistahi nimâton otâkosin -k mistahi ni- mâto -n be.evening.VII-0 much 1- cry.VAI-SAP ‘I cried yesterday.’  b. ??  mistahi nimâton otâkosihk mistahi ni- mâto -n otâkosin -k much 1- cry.VAI-SAP be.evening.VII-0 --- (intended: ‘I cried yesterday.’)  104  Also like statives, the utterance is much worse if the temporal adverbial refers to a distant past. In (93), the consultant fixed the unmarked form by adding the temporal sequencer kî-, and a locative demonstrative.26 (93)  a. ??  kayâs kimiwan kâyâs kimiwan long.ago rain.VII ---  b.  kayâs kî-kimiwan ôta kayâs kîkimiwan ôta long.ago PREV-rain.VII here ‘It rained here a long time ago.’  3.3.1.4 Indexical clauses present telic predicates whose result state coincides with T0 Unmarked telic predicates can only be interpreted as completed in bare indexical clauses. In terms of the coincidence relations, the result holds at speech time. Concurrent with their completion status, unmarked telic predicates can only translate into the English temporal system as a past tense. In (94) English present translations are rejected; however, like stative and activity predicates, the periphrastic past ‘used to’ is also rejected.  26  The addition of the locative in tandem with kî- is suggestive of the well-known link between spatial and temporal deixis. In particular, while unmarked predicates are always taken to be spatially centered around the speaker (consistent with their indexical status), the marked predicate is clearly not; thus the overt proximal locative ôta ‘here’ specifies that although the event is temporally distant it is still spatially coincident with the speech situation.  105  (94)  Telic predicates can only have completed translation a.  Wâpastim pîkonam wâpamowina27 W pîkonam -w wâpam -win -a W break.VTI-3 see.VTA-NOM-OBV ‘W. broke a mirror.’ ≠ ‘W. is breaking a mirror.’ ≠ ‘W. breaks mirrors.’ ≠ ‘W. used to break mirrors.’ comment: The mirror is broken.  b.  Jeff paskiswêw atimwa J paskisw -ê -w atimw -a J shoot.VTA-DIR-3 dog -OBV ‘Jeff shot the dog.’  Incompleted events are marked by preverbs or adverbials; in (93) the addition of mêkwâ corresponds with a progressive. (95)  a.  atim nipahêw minôsa atimw nipah -ê -w minôs -a dog kill.VTA-DIR-3 cat -OBV ‘A/some dog killed a cat.’  b.  atim mêkwâ-nipahêw minôsa atimw mêkwâ- nipah -ê -w minôs -a dog IMPkill.VTA-DIR-3 cat -OBV ‘A dog is killing a cat right now.’ comment: Maybe they’re looking out the window and see it happening.  27  It appears that the aspectual properties may not be hard-coded in these predicates; the other interpretations are made available by adding modifiers: mâna ‘usually’ to get the habitual, mêkwac- ‘midst’ to get an imperfective: (i) a. Wâpastim pîkonam wâpamwona mâna W pîkonam -w wâpam -win -a mâna W break.VTI-3 see.VTA-NOM-PL usually ‘Wâpastim breaks mirrors.’ (Lit.: ‘Wâpastim has broken a mirror repeatedly.’?) b.  Wâpastim mêkwâc-pîkonam wâpamwona W mêkwâc pîkonam -w wâpam -win -a W now break.VTI-3 see.VTA-NOM-PL ‘W. is breaking mirrors right now.' This does not impact the analysis of clause-typing, which only requires an evaluation time of T0 (compatible with both (ia) and (iib)), but it suggests further work is need on aspectual classes in Plains Cree.  106  The present perfect contexts help to tease apart the aspectual interpretation (i.e., the result holds at T0) from the temporal interpretation (i.e., Tref precedes T0). If the former is the relevant interpretation of the clause, then we predict that present perfect contexts will be fine with bare indexical INDEPENDENT clauses, whereas if the latter analysis is more accurate, then we would expect some additional marking to be necessary in these contexts (just as mêkwâ, which is used for imperfectives, is necessary in present contexts). In present perfect contexts, the bare indexical clause is used. For example, the ‘news’ context of the present perfect (Comrie 1976, Fenn 1987, Klein 1994) necessitates an unmarked INDEPENDENT  clause. Notice that trying to use a kî- marked clause elicits the kind of response  we expect where reference time does not coincide with the speech time: the present relevance of the father’s fall and leg-breaking is no longer apparent. (96)  context: a child’s father has just broken his leg; child runs to tell the news and get help a.  b.  nipâpa wîsakisin, pîkonam oskât, pê-wîcihinân ni- pâpa wîsakisin -w pîkonam -w o- skât pê- wîcihi -nân 1- papa fall.VAI -3 break.VTI -3 3- leg come-help.VTA-1.PL ‘My dad got hurt, he broke his leg, come and help us!’ # nipâpa kî-wîsaksin, kî-pîkonam oskât, pê-wîcihinân ni- pâpa kî- wîsakisin-w kî- pîkonam -w o- skât pê- wîcihi -nân 1- papa PREV-fall.VAI-3 PREV-break.VTI-3 3- leg come-help.VTA-1.PL --comment: …wouldn’t say it that way; sounds like the son is heartless/ungrateful  (97)  context: I found a hat that belonged to you several months ago, but didn’t tell you until just now (months later). I want to pretend I just found it a.  kitastotin nimiskên ki(t)- astotin ni- miskê -n 2- hat 1- find.VTI-SAP ‘I found your hat.’ (Last spring) comment: He will think you just found it.  107  b. #  kitastotin nikî-miskên ki(t)- astotin ni- kî- miskê -n 2- hat 1- PREV-find.VTI-SAP ‘I had found your hat.’ comment: if you wanted to pretend that you just found it, you wouldn’t use the kî-  Using the adverbial phrase anohc piko ‘just now’ also obligatorily requires a bare indexical clause, as in (98). (98)  context: on a walk; snake has just slithered across path and into bushes a.  b.  anohc piko niwâpamâw kinêpik anohc piko ni- wâpam -â -w kinêpikw today QUANT 1- see.VTA-DIR-3 snake ‘I saw a snake just now.’ * anohc piko nikî-wâpamâw kinêpik anohc piko ni- kî- wâpam -â -w kinêpikw today QUANT 1- PREV-see.VTA-DIR-3 snake --comment: They’re sort of cancelling each other out–one is now, the other is before  Finally, a bare predicate is used for a ‘perfect of result’ context, as shown in (99) and (100). (99)  context: explaining to someone why you can’t see a.  niwanîhtân niskîsikôhkâna ni- wanîhtâ -n ni- skîsikwihkân -a 1- lose.VAI -SAP 1- glasses -PL ‘I lost my glasses.’  b. ?# nikî-wanîhtân niskîsikôhkâna ni- kî- wanîhtâ -n ni- skîsikwihkân-a 1- PREV-lose.VAI-SAP 1- glasses -PL ‘I lost my glasses.’  108  (100) context: we are waiting for Tom to come so we can start dancing A:  Tom ôta ayâw T ôta ayâ -w T here be.VAI-3 ‘Is Tom here?’  B:  pê-takosin pê- takosin -w DIR-arrive.VAI-3 ‘(Yes) he’s arrived.’ comment: this one is implying that he’s still here, that we’re ready to dance  What all of these contexts have in common is that they are contexts where we want to model that the result holds at speech time. Thus, just like stative and activity predicates, telic predicates in indexical clauses have a fixed temporal relation to the speech time. And in all of these contexts, bare indexical INDEPENDENT clauses are felicitous, indicating that we are tracking the coincidence of the reference time with speech time (T0). Bare indexical clauses with telic predicates can co-occur with otahkosihk ‘yesterday’, as in (101). Sometimes examples with past time adverbials are augmented by another clause with a stative predicate. In these examples, the telic predicate is overtly related to some non-telic predicate that holds at utterance time.28 (101) a.  b.  otâkosihk nîcêwâkan sipwêhtew, êkwa anohc nikaskêyihtên. otâkosin -k ni- wîcêwâkan sipwêhtê -w êkwa anohc ni- kaskêyihtê -n be.evening.VII-0 1- friend leave.VAI-3 and today 1- lonely.VTI -SAP ‘Yesterday my friend left, and now I’m lonely.’ nikisîpêkinên ôma wîyâkan, kanâtan ni- kisîpêkinê -n ôma wîyâkan kanâtan 1- wash.VTI-SAP DEM.INAN dish clean.VII ‘I washed this dish, it’s clean.’  The data in (101) is consistent with the data we saw earlier, where the past time adverbial was acceptable as long as it was in initial position.  28  This parallels observations made by Lakoff (1970) regarding ‘present relevance’ in the English tense system.  109  3.3.1.5 Interim summary Given the times (situation time, reference time, evaluation time, and speech time) and relations [±coincidence] that seem to be relevant to linguistic structure, we would expect indexical clauses to always make reference to the speech time. More specifically within a situation semantics analysis, the speech time is the temporal component of the speech situation by which the proposition in an indexical clause is evaluated. Consistent with our expectations, in this section we have seen data to show that there is a [+coincidence] relation between reference time and speech time in indexical clauses. Crucially, as we saw in §3.3.1.1, this interpretation is different from non-indexical clauses, where speech time is not necessarily taken into account. In the next section I argue that we can make the same claims for referents that we do for times, to derive the evidential force of indexical INDEPENDENT clauses.  3.3.2 Referential deixis: The role of the speaker in indexical clauses In this section, I show that indexical INDEPENDENT clauses always make reference to a speaker, a property which has not been previously discussed in the Algonquianist literature. If we think about which referents are necessary for a speech act, we see that first-person is crucial. Without a speaker, there is no speech act, and as long as there is a speaker, a speech act can occur29. Within a speech act, then, only first person is a referential constant; all other referents can come or go. If an indexical clause contains reference to a speech act (or, more formally, the speech situation), and every speech act contains a speaker, then that logically