UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The syntax of A′-dependencies in Bamileke Medumba Keupdjio, Hermann Sidoine 2020

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2020_may_keupdjio_hermann.pdf [ 4.73MB ]
Metadata
JSON: 24-1.0389713.json
JSON-LD: 24-1.0389713-ld.json
RDF/XML (Pretty): 24-1.0389713-rdf.xml
RDF/JSON: 24-1.0389713-rdf.json
Turtle: 24-1.0389713-turtle.txt
N-Triples: 24-1.0389713-rdf-ntriples.txt
Original Record: 24-1.0389713-source.json
Full Text
24-1.0389713-fulltext.txt
Citation
24-1.0389713.ris

Full Text

THE SYNTAX OF A′-DEPENDENCIES IN BAMILEKE MEDUMBA by  Hermann Sidoine Keupdjio  B.A., University of Yaoundé 1 (Cameroon), 2009 M.A., University of Yaoundé 1 (Cameroon), 2011  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Linguistics)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)   March 2020  © Hermann Sidoine Keupdjio, 2020  ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled: The Syntax of A'-dependencies in Bamileke Medumba  submitted by Hermann KEUPDJIO in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Linguistics  Examining Committee: Rose-Marie Déchaine (University of British Columbia) Supervisor  Hotze Rullmann (University of British Columbia) Supervisory Committee Member  Martina Wiltschko (University of British Columbia) Supervisory Committee Member Marcin Morzycki (University of British Columbia) University Examiner Ori Simchen (University of British Columbia)  University Examiner     Chris Collins (New York University)  External Examiner    Additional Supervisory Committee Members: Michael Rochemont (University of British Columbia; deceased) Supervisory Committee Member  iii  Abstract  In this dissertation, I investigate the syntax of A′-dependencies (wh-movement, focus movement, relativization and topicalization) in Bamileke Medumba, a Grassfields Bantu language spoken in the Western Region of Cameroon.     I first examine the in-situ/ex-situ partition with Medumba wh-/focus construals and propose an analysis in which the necessity of movement is driven by interpretation. This approach correctly predicts structural and semantic differences between in-situ and ex-situ wh-questions and foci in Medumba. Thus, they differ in Medumba with regard to: (i) exhaustivity –– in that in-situ wh-questions and foci are non-exhaustive whereas their ex-situ counterparts are exhaustive –– (ii) question-answer pairs –– in that the information-theoretic structure of the answer must match the information-theoretic structure of the question ––  and (iii) fragment answers –– in that fragment answers to in-situ wh-questions are not focus-marked whereas fragment answers to ex-situ wh-questions are focus-marked.   I also examine A′-agreement, analyzed as the reflex of Phasal-Agree. I show that A′-agreement is not only a crucial diagnostic for A′-movement, for Phasal-Agree and for the locality of movement (cyclic phase-by-phase movement (Biberauer and D’Alessandro 2006; Chomsky 2000, 2001; van Urk 2015; van Urk and Richards 2015)) but also a diagnostic for intermediate phases.   Finally, I examine resumption in Medumba A′-construals. Resumptive pronouns in Medumba surface both in island violation contexts (including apparent complement CPs analyzed as disguised adjunct clauses) as well as in contexts where there is absolutely no island violation (root clauses), where they alternate with gaps. I argue that resumptive structures are derived in iv  Medumba via the economy principle of Last Resort (Koopman and Sportiche 1986; Rizzi 1990; Chomsky 1991, 1998; Shlonsky 1992; Bobaljik 1995, Lasnik 1995; Ura 1996; Pesetsky 1997; Collins 2001; Bošcović 2011). To get a unified account of resumptive structures in Medumba, I propose that Last Resort is conditioned by syntactic and semantic constraints. Syntactic Last Resort derives resumptive pronouns in Medumba island violation contexts, to salvage A′-dependencies that would otherwise result in ungrammaticality. Semantic Last Resort is a condition on interpretation that derives resumption in configurations that would otherwise result in ambiguity.    v  Lay Summary  Every language has a way of asking a content question. However, strategies used in forming content questions vary from one language to another. In this dissertation, I look at the properties of Medumba content questions, including other related constructions such as focus constructions and relative clauses. The findings show that: (i) Medumba speakers, when asking a content question, can leave the question word in its original position or move it to the clause initial position. I propose that the use of either strategy depends on the interpretation of the question; (ii) in questions where the question word is in initial position, there is a change in the tonal melody of verbal heads and auxiliaries; (iii) in questions where the question word is in clause initial position, there is either a gap or a pronoun in the original position of the question word depending on the type of clause.     vi  Preface  This dissertation is an original intellectual product of the author, Hermann S. Keupdjio. vii  Table of Contents  Abstract ................................................................................................................................. iii Lay Summary ........................................................................................................................... v Preface ................................................................................................................................... vi Table of Contents .................................................................................................................. vii List of Tables ........................................................................................................................ xvi List of Figures ..................................................................................................................... xviii List of Abbreviations ............................................................................................................. xix Acknowledgements ............................................................................................................ xxiii Dedication ......................................................................................................................... xxvii Chapter 1: Introduction .................................................................................................................... 1 1.1 Why wh-movement? ............................................................................................................................. 1 1.2 Why wh-movement in Medumba? ....................................................................................................... 3 1.3 Medumba: the language and the people .............................................................................................. 6 1.4 Previous work on Medumba ................................................................................................................. 9 1.5 Preview of the dissertation ................................................................................................................. 11 1.5.1 Aʹ-movement and the in-situ/ex-situ partition in Medumba ......................................................... 11 1.5.2 Aʹ-movement and Aʹ-agreement in Medumba .............................................................................. 12 1.5.3 Aʹ-movement and the tense/aspect system in Medumba ............................................................. 14 viii  1.5.4 Aʹ-movement and resumption in Medumba .................................................................................. 14 Chapter 2: Aʹ-movement and the in-situ/ex-situ partition in Medumba .......................................... 16 2.1 The ex-situ/in-situ paradox: if movement is possible, why isn’t it obligatory? ................................... 16 2.1.1 The problem: Medumba uses both ex-situ and in-situ strategies .................................................. 17 2.1.2 A possible (but non-viable) solution: feature-driven movement ................................................... 18 2.1.2.1 Implementation 1: overt versus covert movement .............................................................. 18 2.1.2.2 Implementation 2: deletion of head versus tail .................................................................... 19 2.1.3 Another possible (but non-viable) solution: Agree ........................................................................ 21 2.1.3.1 Why Agree + EPP-features do not work ............................................................................... 22 2.1.3.2 Why Agree + Edge-features do not work .............................................................................. 23 2.1.3.3 Why Agree + labelling does not work ................................................................................... 25 2.1.4 Another possible (and viable) solution: interpretation-driven movement .................................... 25 2.1.4.1 Why interpretation-driven movement is conceptually necessary ........................................ 27 2.1.4.2 How interpretation-driven movement works ....................................................................... 27 2.1.5 Preview of chapter ......................................................................................................................... 29 2.2 Six diagnostics for Aʹ-movement ......................................................................................................... 29 2.2.1 Diagnostic 1: Aʹ-movement relates the head and tail of an Aʹ-chain ............................................. 29 2.2.1.1 Relation between the head and tail of an Aʹ-chain: the general picture .............................. 29 2.2.1.2 Relation between the head and tail of an Aʹ-chain: the picture in Medumba ..................... 30 2.2.2 Diagnostic 2: Aʹ-movement permits long-distance dependency .................................................... 34 2.2.2.1 Long-distance dependencies: the general picture ................................................................ 34 2.2.2.2 Long-distance dependencies: the picture in Medumba ....................................................... 34 2.2.3 Diagnostic 3: Aʹ-movement supports reconstruction ..................................................................... 39 2.2.3.1 Reconstruction: the general picture ..................................................................................... 39 2.2.3.2 Reconstruction: the picture in Medumba ............................................................................. 40 ix  2.2.4 Diagnostic 4: Aʹ-movement conditions Strong and Weak Crossover ............................................. 42 2.2.4.1 Strong & Weak Crossover: the general picture .................................................................... 42 2.2.4.2 Strong & Weak Crossover: the picture in Medumba ............................................................ 43 2.2.5 Diagnostic 5: Aʹ-movement is island-sensitive ............................................................................... 46 2.2.5.1 Island-sensitivity: the general picture ................................................................................... 46 2.2.5.2 Island-sensitivity: the picture in Medumba .......................................................................... 49 2.2.6 Diagnostic 6: Aʹ-movement conditions Aʹ-agreement ................................................................... 51 2.2.6.1 Aʹ-agreement: the general picture ....................................................................................... 51 2.2.6.2 Aʹ-agreement: the picture in Medumba ............................................................................... 52 2.3 Five arguments in favour of interpretation-driven movement ........................................................... 54 2.3.1 Argument 1: ex-situ wh/focus requires exhaustivity ..................................................................... 55 2.3.1.1 The mechanics of the derivation .......................................................................................... 55 2.3.1.2 Ex-situ wh-/focus blocks entailment relations ...................................................................... 61 2.3.1.3 Ex-situ wh-/focus blocks universal quantification ................................................................ 62 2.3.1.4 Ex-situ wh/focus blocks additive particles ............................................................................ 64 2.3.1.5 Ex-situ wh/focus blocks functional readings ......................................................................... 65 2.3.2 Argument 2: in-situ wh/focus does not require exhaustivity ......................................................... 67 2.3.2.1 The mechanics of the derivation ........................................................................................ 68 2.3.2.2 In-situ wh/focus allows entailment relations ....................................................................... 70 2.3.2.3 In-situ wh/focus allows universal quantification .................................................................. 71 2.3.2.4 In-situ wh/focus allows additive particles ............................................................................ 71 2.3.2.5 In-situ wh/focus allows functional readings ......................................................................... 72 2.3.3 Argument 3: Q/A pairs are conditioned by in-/ex-situ contrast ..................................................... 74 2.3.3.1 Ex-situ wh-questions are answered by ex-situ focus ............................................................ 74 2.3.3.2 In-situ wh-questions are answered by in-situ focus ............................................................. 75 2.3.4 Argument 4: fragment answers are conditioned by in-/ex-situ contrast ....................................... 76 x  2.3.4.1 Fragment answers to ex-situ wh require focus-marking ...................................................... 76 2.3.4.2 Fragment answers to in-situ wh- do not require focus-marking .......................................... 77 2.3.4.3 Implications for the syntax of fragment answers ................................................................. 78 2.3.5 Argument 5: subjects are (predictably) different ........................................................................... 81 2.3.5.1 Temporal auxes and resumption diagnose in-situ versus ex-situ subject ............................ 81 2.3.5.2 Deriving subject wh/focus .................................................................................................... 82 2.3.5.3 Subject wh/focus is always exhaustive ................................................................................. 86 2.3.5.4 Clefted subject wh/focus is non-exhaustive ......................................................................... 88 2.3.6 A loose end: the semantics of exhaustivity .................................................................................... 91 2.4 The broader landscape of Aʹ-movement ............................................................................................. 91 2.4.1 Wh-ex-situ doesn’t always “mean” the same thing ....................................................................... 92 2.4.1.1 Wh-ex-situ can be just inquisitive: English ............................................................................ 92 2.4.1.2 Wh-ex-situ can be inquisitive and exhaustive: Medumba .................................................... 92 2.4.2 Wh-in-situ doesn’t always “mean” the same thing ........................................................................ 93 2.4.2.1 Wh-in-situ can be just inquisitive: Medumba ....................................................................... 93 2.4.2.2 Wh-in-situ can be supplemented with reprise “a.k.a echo” particle: Medumba ................. 93 2.4.2.3 Wh-in-situ can be a reprise question: English ....................................................................... 96 2.4.3 Wh-questions always and only contrast in-situ versus ex-situ ....................................................... 97 2.4.3.1 Prolific inquisitive forms in French: seven ways of asking a question? ................................ 97 2.4.3.2 The French paradigm reduces to an in-situ/ex-situ partition ............................................... 98 2.4.4 A prediction about wh-interrogative and wh-relative in Medumba ............................................ 100 2.5 Conclusion ......................................................................................................................................... 106 Chapter 3: Aʹ-movement and Aʹ-agreement in Medumba ............................................................. 108 3.1 What is Aʹ-agreement? ..................................................................................................................... 108 3.1.1 The locus and form of Aʹ-agreement in Medumba ...................................................................... 109 xi  3.1.1.1 Detecting the basic form of the verb: yes/no question frame ........................................... 109 3.1.1.2 Detecting the reflex of Aʹ-agreement: HL tone overwrite with Aʹ-movement ................... 111 3.1.2 The proposal: Phasal-Agree derives Aʹ-agreement ...................................................................... 114 3.1.2.1 There are (at least) two phases: vP and CP ......................................................................... 114 3.1.2.2 Aʹ-agreement is Phasal-Agree ............................................................................................. 115 3.2 How phasal agree derives Aʹ-agreement in Medumba ..................................................................... 117 3.2.1 Phasal-Agree predicts subject/non-subject asymmetry ............................................................... 117 3.2.2 Phasal-movement and the absence of Superiority effects in Medumba ..................................... 121 3.2.3 The surface realization of Aʹ-agreement ...................................................................................... 125 3.2.3.1 The problem: iterated agreement is incompatible with agree ........................................... 125 3.2.3.2 The solution: Resume Agree ............................................................................................... 127 3.3 A surprising result: Phasal-Agree does not predict root/non-root CP asymmetry ........................... 129 3.3.1 Hypothesis 1: Spec-C-to-Spec-C movement with non-root CPs ................................................... 134 3.3.2 Hypothesis 2: Apparent complement CPs as disguised vP-adjuncts ............................................ 137 3.3.3 Predictions of the vP-adjunct analysis of apparent complement CPs .......................................... 138 3.3.3.1 Apparent complement CPs behave like adjunct islands ..................................................... 138 3.3.3.2 CPs are never in complement position: the lack of embedded interrogatives ................... 139 3.3.3.3 CPs are never in argument position: the lack of subject and complement CPs .................. 145 3.3.3.4 Apparent complement CPs follow matrix VP-adjuncts ....................................................... 146 3.3.3.5 vP-adjuncts don’t trigger Aʹ-agreement, but VP-adjuncts do ............................................. 147 3.3.3.6 Apparent complement CPs strand under VP-gapping ........................................................ 150 3.3.3.7 Apparent complement CPs reconstruct .............................................................................. 151 3.4 Towards a cross-linguistic formal typology of Phasal-Agree ............................................................. 152 3.4.1 Type 1: Phasal-Agree at C ............................................................................................................. 155 3.4.1.1 Concordial agreement in Kilega .......................................................................................... 155 3.4.1.2 Anti-agreement in Bantu .................................................................................................... 159 xii  3.4.1.3 Wh-copying in German ....................................................................................................... 160 3.4.1.4 Stylistic inversion in French ................................................................................................ 162 3.4.2 Type 2: Phasal-Agree at v ............................................................................................................. 164 3.4.2.1 Downstep deletion in Kikuyu .............................................................................................. 165 3.4.2.2 Aʹ-particle in Duala .............................................................................................................. 167 3.4.3 Type 3: phrasal-agree at v and C .................................................................................................. 171 3.4.3.1 Chamorro Phasal-Agree at v and C ..................................................................................... 172 3.4.3.2 Medumba Phasal-Agree at v and C ..................................................................................... 175 3.4.4 Broader implications of the Phasal-Agree analysis ...................................................................... 180 3.5 Conclusion ......................................................................................................................................... 181 Chapter 4: Aʹ-movement and the tense-aspect system in Medumba ............................................ 184 4.1 The abundance of Medumba temporal auxiliaries ........................................................................... 184 4.1.1 Three observations about tense, Aʹ-movement and Aʹ-agreement ............................................. 184 4.1.1.1 Observation 1: Some temporal auxiliaries require Aʹ-agreement ...................................... 184 4.1.1.2 Observation 2: Some temporal auxiliaries show optional Aʹ-agreement ........................... 185 4.1.1.3 Observation 3: Some temporal auxiliaries are incompatible with Aʹ-agreement ............... 187 4.1.2 Five ingredients of the analysis .................................................................................................... 188 4.1.2.1 Ingredient 1: tense-marking versus tense-shifting ............................................................. 188 4.1.2.2 Ingredient 2: an exploded Tense projection ....................................................................... 188 4.1.2.3 Ingredient 3: N-prefix diagnoses in-situ heads ................................................................... 198 4.1.2.4 Ingredient 4: auxiliary allomorphy tracks agreement and head-movement ...................... 200 4.1.2.5 Ingredient 5: Phasal-Agree diagnoses intermediate phase between T and v ..................... 213 4.1.3 How this chapter will unfold ........................................................................................................ 213 4.2 A primer on temporal auxes: the convergence of syntax, semantics and morphology .................... 213 4.2.1 Zero-marked past tense ............................................................................................................... 214 xiii  4.2.1.1 Retrospective construals are compatible with zero-marked past tense ............................ 215 4.2.1.2 Four reasons why unmarked verbs are really zero-marked past tense .............................. 215 4.2.2 Concurrent construals are compatible with present tense tʃʷɛ̀ɛt́ ................................................ 222 4.2.3 Irrealis àʔ ...................................................................................................................................... 223 4.2.3.1 Irrealis àʔ is compatible with future interpretation ........................................................... 223 4.2.3.2 Irrealis àʔ is compatible with “future in the past” .............................................................. 223 4.2.3.3 Irrealis àʔ forms express degrees of certainty .................................................................... 224 4.2.3.4 Negative irrealis is suppletive: kʉ ́....................................................................................... 224 4.2.3.5 Irrealis àʔ is in Mood (not Tense) ....................................................................................... 225 4.2.4 What are Medumba temporal auxiliaries? .................................................................................. 226 4.2.4.1 Medumba temporal auxiliaries are not temporal adverbs ................................................. 226 4.2.4.2 Medumba temporal auxes are grammaticalized verbs ...................................................... 230 4.3 Why some temporal auxiliaries are compatible with Aʹ-agreement ................................................ 234 4.3.1 Why retrospective temporal-shifters are compatible with Aʹ-agreement ................................... 235 4.3.2 Why neutral tense-shifters permit Aʹ-agreement in retrospective contexts ............................... 237 4.3.3 Temporal auxes provides evidence for Phasal-Agree .................................................................. 241 4.3.3.1 Detecting intermediate phases: evidence from aux-stacking ............................................. 241 4.3.3.2 Detecting shifty phases: evidence from neutral tense-shifters and IPFV kʉ ́...................... 244 4.4 Why neutral tense-shifters prohibit Aʹ-agreement in prospective contexts ..................................... 246 4.4.1 Irrealis àʔ is phonologically defective .......................................................................................... 246 4.4.2 CP with MoodP complement as a defective phase ...................................................................... 246 4.5 Coming full circle ............................................................................................................................... 250 4.5.1 Broader implications for the analysis of zero-marked verb forms ............................................... 250 4.5.1.1 On the absence of a state/event contrast .......................................................................... 250 4.5.1.2 Distinguishing zero-marked past tense from zero-marked perfective aspect .................... 257 4.5.2 Broader implications for the analysis of dialect variation ............................................................ 258 xiv  4.5.3 Broader implications for the analysis of graded tense ................................................................. 260 4.5.3.1 Previous descriptions of Medumba: Nganmou 1991; Kouankem 2012 ............................. 260 4.5.3.2 Mucha 2016 ........................................................................................................................ 261 4.5.4 Broader implications for the analysis of Aʹ-movement and Aʹ-agreement .................................. 262 Chapter 5: Aʹ-movement and resumption in Medumba ................................................................ 264 5.1 What form does resumption take? ................................................................................................... 264 5.1.1 Medumba resumptive pronouns can be simplex ......................................................................... 264 5.1.2 Medumba resumptive pronouns can be complex ....................................................................... 267 5.1.3 Implications of the (simple versus complex) pronoun partition .................................................. 271 5.2 When does resumption occur? ......................................................................................................... 271 5.2.1 Resumption is optional in root clauses ........................................................................................ 272 5.2.2 Resumption is obligatory in island contexts ................................................................................. 273 5.2.3 Resumption is obligatory in complement (qua disguised adjunct) clauses .................................. 277 5.2.4 Implications of the (obligatory versus optional) resumption partition ........................................ 279 5.3 Why does resumption occur? ........................................................................................................... 287 5.3.1 The syntactic basis of obligatory resumption: economy of derivation ........................................ 287 5.3.2 The semantic basis of optional resumption: economy of interpretation ..................................... 299 5.3.3 Implications of the economy (of derivation versus interpretation) partition .............................. 305 5.4 Question/answer sequences and the gap/resumption partition in Medumba ................................ 305 5.4.1 Non-plural wh-XPs ........................................................................................................................ 306 5.4.1.1 Individual variable denotation possible with non-plural wh-XPs ....................................... 308 5.4.1.2 Natural function denotation unavailable with non-plural wh-XPs ..................................... 309 5.4.1.3 Pair list denotation unavailable with non-plural wh-XP ..................................................... 311 5.4.2 Plural wh-XPs ................................................................................................................................ 312 5.4.2.1 Individual variable denotation unavailable with plural wh-XPs .......................................... 313 xv  5.4.2.2 Natural function denotation unavailable with plural wh-XPs ............................................. 315 5.4.2.3 Pair-list denotation possible with bare wh-XP .................................................................... 316 5.5 Conclusion ......................................................................................................................................... 318 Chapter 6: Conclusion ................................................................................................................... 320 6.1 Summary and contributions .............................................................................................................. 320 6.2 Aʹ-movement and the in-situ/ex-situ partition in Medumba ............................................................ 321 6.2.1 Implications for future research ................................................................................................... 322 6.2.1.1 The semantics of exhaustivity marking: “Max” applies only to individuals ........................ 323 6.2.1.2 Question/Answer sequences and fragment answers ......................................................... 323 6.3 Aʹ-movement and Aʹ-agreement in Medumba ................................................................................. 324 6.3.1 Implications for future research ................................................................................................... 326 6.4 Aʹ-movement and the tense/aspect system in Medumba ................................................................ 327 6.4.1 Implications for future research ................................................................................................... 327 6.5 Aʹ-movement and resumption in Medumba .................................................................................... 328 6.5.1 Implications and future research ................................................................................................. 329 Bibliography ........................................................................................................................ 330 Appendices .......................................................................................................................... 346 Appendix A: kí-topicalization as a species of A′-movement ………………………………… 346 Appendix B: Medumba stem allomorphy ……………………………………………………. 360 xvi  List of Tables Table 1. 1: The internal structure of ʉ́-words (wh-words) in Medumba ......................................... 5 Table 3. 1: Verb tone classes and their surface tonal allomorphs in Medumba .......................... 110 Table 3. 2: The locus of A′-agreement in Medumba root-clauses ............................................... 118 Table 3. 3: The locus of A′-agreement in Medumba non-root clauses ........................................ 130 Table 3. 4: Phasal-Agree and clause typing in Medumba ........................................................... 133 Table 3. 5: Selectional properties of (some) embedding verbs in English and Medumba .......... 141 Table 3. 6: Formal typology of the locus of A′-agreement ......................................................... 154 Table 3. 7: Form of A′-agreement cross-linguistically ................................................................ 154 Table 3. 8: Type 1 Phasal-Agree ................................................................................................. 155 Table 3. 9: Type 2 Phasal-Agree ................................................................................................. 165 Table 3. 10: Type 3 Phasal-Agree ............................................................................................... 171 Table 4. 1: Auxiliary allomorphy in Medumba ........................................................................... 201 Table 4. 2: Diagnosing past tense with the construal of event verbs in Medumba ..................... 216 Table 4. 3: Grammaticalization of temporal auxiliaries in Medumba ......................................... 230 Table 4. 4: Lexical entries of aux usages and main verb usage in Medumba ............................. 234 Table 4. 5: Diagnosing past tense with the construal of change-of-state verbs in Medumba ..... 252 Table 4. 6: Summary of the construal of unmarked verbs in Medumba ..................................... 257 Table 4. 7: Diagnosing past tense and perfective marking in Medumba and Siamou ................ 258 Table 5. 1: The simplex pronouns paradigm in Medumba .......................................................... 265 Table 5. 2: 1PL denotation of complement complex pronouns in Medumba ............................. 268 Table 5. 3: Question and answer sequences in Medumba ........................................................... 306 Table 5. 4: Non-plural wh-XPs and the denotation of their answers in Medumba ..................... 307 xvii  Table 5. 5: Plural wh-XPs and the denotation of their answers in Medumba ............................. 312 Table 5. 6: Summary of wh-XPs and the denotation their answers in Medumba ....................... 318 Table 6. 1: The broader landscape of wh-questions: strategies and the question semantics ....... 323 Table 6. 2: Form of A′-agreement cross-linguistically ................................................................ 326 Table A 1: kí-movement as A′-movement ................................................................................... 352 Table B 1: Verb tone classes and their surface tonal allomorphs in Medumba .......................... 360 Table B 2: Number on Medumba count nouns ........................................................................... 366 Table B 3: Number on Medumba non-count (abstract & mass) nouns ....................................... 367 Table B 4: Pluralization of Medumba inherently singular non-count nouns .............................. 368 Table B 5: Pluralization of Medumba inherently plural mass nouns .......................................... 368   xviii  List of Figures  Figure 1. 1: Geographical location of the Medumba people ........................................................... 7 Figure 1. 2: Bamileke Medumba, the language ............................................................................... 8   xix  List of Abbreviations  -Q:  Non-interrogative 1:  1st person 2:  2nd person 3:  3rd person ACC:  Accusative ADJ:  Adjective AGR:  Agreement APL:  Associative plural ASP:  Aspect ASSC:  Associative AUX:  Auxiliary C:  Complementizer CL:  Class  D:  Determiner DEM:  Demonstrative DIM:  Diminutive   ECP:  Empty Category Principle EF:  Edge feature EM:  External merge EPP:  Extended Projection Principle EXH.:  Exhaustive xx  FOC:  Focus marker FUT:   Future FV:  Final vowel H:  High tone HAB:  Habitual HL:  High low falling tone HMN:  Human IM:  Internal merge INCL:  Inclusive INF.:  Infinitive IPFV:  Imperfective L:  Low tone LF:  Logical form LH:  Low high rising tone  LIT.:  Literal LNK:  Linker LOC:  Locative N-:  Homorganic nasal prefix N:  Noun NEG:  Negation NMLZ:  Nominalizer NOM:  Nominative OBJ:  Object xxi  OM:  Object marker P:  Phrase p:  Proposition PERS:  Person PF:  Phonetic form PFV:  Perfective PIC:  Phase Impenetrability Condition PL:  Plural POSS:  Possessive PRED:  Predicate PREP:  Preposition PRN:  Pronoun PROG:  Progressive PROX:  Proximate  PRS:  Present PRT:  Particle PST:  Past Q:  Interrogative QY/N:  Yes/no question particle REL:  Relative SA:  Subject agreement SG:   Singular SM:  Subject marker xxii  SPEC:  Specifier SPKR:  Speaker SUBJ:  Subject SUP:  Supremum SVC:  Serial verb construction t:  Trace TNS:  Tense TOP:  Topic marker UNMKD: Unmarked V:  Verb WH:  Wh-word xxiii  Acknowledgements “bʰúù ⁿʧʰʉ́ʔ kʉ́ʉ̀ʔ mʉ́ kfʉ́ʉ̀t bʰúʔ” “One hand cannot tie up a packet/it takes more than one hand to tie up a packet” This Medumba saying summarizes my journey through this process of dissertation-writing. It is usually said “It takes the village to raise a child”, and indeed it took the whole village to raise the linguist I am becoming. It is in this spirit of the village that I am very grateful to the whole village, the amazing people without whom this journey would not have started and without whom this dissertation would not exist or exist in the form it is. I just want to say mʉ́ làbtə̀ə́, ‘thank you’.   First and foremost, I wish to express my sincere gratitude to my supervisory committee. You believed in me since the first day I got to UBC and your guidance, advice, encouragement and support did not cease during these rough years of graduate school. You kept pushing me up towards the goal. Your comments, feedback and many insightful questions throughout the different stages of the writing process of this dissertation not only helped clarifying my ideas but also helped me gain some amount of confidence in the directions I was taking.   Rose-Marie Déchaine, thank you for getting me interested in tone. It was through lengthy hours of meeting with lengthy handouts that I realized that although very complex in Medumba, I should not be afraid using tones for syntactic diagnostics. Your expressions “on y arrivera” or “ça s’envient” or “ça se précise” kept fuelling me up when I felt like I should give up investigating tonal allomorphy in Medumba. You were patient listening to my confusing ideas and helped me see the big picture of what I was doing when I was laser-focused on a particular data set.    Hotze Rullman, thank you for your patience going with me through the various lengthy handouts for each chapter of this dissertation and for your comments and remarks through the xxiv  various drafts of each chapter of this dissertation. Your kind words of encouragement were the ammunition I needed to keep soldiering.     Martina Wiltschko, thank you for jumping on board in my supervisory committee after Michael Rochemont’s sudden passing away even though you already had a lot on your plate. I am very grateful for your advice and financial support throughout my years in graduate school.  I am also very grateful to the late Michael Rochemont for his comments and guidance at the early stages of this dissertation. It is very sad and very unfortunate that he would never see his favourite chapter (Ch. 2) of this dissertation and that I would never share with him the various implications of tonal allomorphy in Medumba.        To my cohort Adriana Osa G., Emily Sadlier-Brown, Oksana Tkachman (cohort in honour) Erin Guntly, thank you for your support and kind words throughout these years of graduate school. I am particularly thankful to Adriana Osa G., my twin sister from another mother, for cheering me up. You mastered my timetable more than me, you kept reminding me of the different appointments, meetings and deadlines. In all, gracias!    I am also thankful to my fellow graduate students at UBC, especially to the dissertators Andrei Anghelescu, James Crippen, Joash Gambarage and Natalie Weber for your comments and feedback during the meetings. I am particularly grateful to Dzeiwsh (James Crippen) for proofreading this dissertation.  I am particularly very grateful to Ncǎkò’, Katie Franich for proofreading this dissertation.  I wish to express my sincere gratitude to the person who inspired and introduced me into syntax, my M.A. thesis supervisor at University of Yaoundé 1, Prof. Edmond Biloa. I am also grateful to my former professors at the University of Yaoundé 1, namely Prof. Philip Ngessimo Mutaka, Prof. Cledor Nseme, Dr. Emmanuel Makasso and the late Prof. Pius Tamanji, for your xxv  devotion in shaping my dream of becoming a linguist at the Department of African Languages and Linguistics of the University of Yaoundé 1.   I am also thankful to the organizers of the African Linguistics School (ALS), and more particularly to Prof. Enoch Aboh, Prof. Chris Collins, Prof. Claire Halpert and Prof. Hilda Koopman for your advice when I was in the process of applying for graduate school.   To all my friends and colleagues in Cameroon, Dr. Constantine Kouankem, Alain Yonkeu Mbiami, Elodie Ngo Ngue, Rebecca Bedim and Guy Merlin Tchagoua, thank you for your encouragement and being part of this journey. To many of you whose names are not listed here, you know who you are; to all, thank you.  To Ian and Barbara Rokeby, thank you for your prayers and for helping me in my integration process in Canada. The skiing and boating experience was amazing.     To Achille and Mireille Fossi, thank you for your words of encouragement, your support and for welcoming me in Canada.  To my family, thank you for believing in my dreams, for your support and encouragement since I left home and throughout these years I have been away and have not seen you. To my parents, I love you and I miss you every day. Mom, I promise I will come back home to see you after graduation. To my elder brother and his wife Basile and Nazareth Nana; to Mr. and Mrs. Tchana, thank you for everything. To papa Henri Toukam, thank you for always encouraging me. To my lovely nieces and nephews, thank you for encouraging your ‘tonton’. I kept going partly because you kept asking me when I will be done with the dissertation.  To my elder brother Henri Innoçant (in memoriam), you are particularly missed.    My dearest, lovely and very supportive wife, Christelle Keupdjio Toukam, ‘ma Madouce’, words are not enough to express my gratitude for putting up with me throughout these years, for xxvi  cheering me up, for keeping telling me it is possible, I can do it. I finally finally did it ‘lovy lovy’. To Nolann Myronn Keupdjio Toukam, my little ‘Boubounet’, thank you for being the light that brightens my way through the final stages of this dissertation.         xxvii  Dedication  À mes parents et Pour toi, Madouce.                   1  Chapter 1: Introduction  1.1 Why wh-movement? Wh-movement has been a subject of inquiry in modern generative linguistic theory since Chomsky’s 1977 “On wh movement”. What has been fascinating about the phenomenon of displacement in natural languages is that some items can be pronounced in one position but interpreted in other positions. In the English example in (1) and the Medumba example in (2), although the wh-XP who or á wʉ́ is pronounced in clause-initial position, it is understood that the wh-XP is introduced as the complement of the verb betray/ⁿ-sʷɛ́ɛ̀n.   Thus, wh-questions instantiate a non-local dependency, where the wh-XP is associated with a position from which, by hypothesis, it originates.     (1) Who did Watat betray? (2)  á      wʉ́      Wàtɛ̀ɛ́t      nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n          á          FOC   who     Watat.H    AGR.AUX.T2 N-AGR.sell    C.Q.H                                           ‘Who did Watat betray? One reason why it is important to investigate wh-movement in natural languages lies in the fact that it has led to the identification of configurations that serve as diagnostics for non-local dependencies, which in turn raises the question of how to model these non-local dependencies, and what the underlying mechanism is. The key diagnostic properties of wh-movement are in (3)1. (3) General characteristics of wh-movement (Chomsky 1977:86, (49))  (a) Wh-movement leaves a gap.  1 I only list here the diagnostics given in Chomsky’s 1977. See chapter 2 for the complete list of the diagnostic properties of wh-movement. 2   (b) Where there is a bridge, there is an apparent violation of Subjacency  (c) Wh-movement observes the Complex NP Constraint (CNPC).  (d) Wh-movement observes the Wh-Island Constraint. These properties all pertain to the relation between the moved element and its extraction site. The cluster of properties in (3) follows from the assumption that wh-movement moves an XP, which in turn leaves behind at the extraction site a category with phonological null content called a gap (3a). Wh-movement in some contexts can circumvent island violations (3b). This implies that wh-movement can proceed in a stepwise fashion in successive cycles through all the intermediate CPs that separate the launching site from the ultimate landing site. Finally, wh-movement is sensitive to islands (3c&d).   The Minimalist Program (Chomsky 1995 and subsequent work) explores the idea that human language may be a ‘‘perfect system’’ optimally designed to meet certain interface conditions imposed by other cognitive systems that the language faculty interacts with. In such a system, the question of how to develop a general theory of movement (reduced to internal Merge) — and more particularly how to analyze wh-movement — remains a puzzle. Specifically, investigating wh-movement within the Minimalist framework raises the question of how the properties of wh-movement can be explained in terms of principles of the interface systems and what principles of efficient computation underlie the derivation of non-local dependencies.   Some questions that arise when investigating wh-movement in natural languages are: (i) What is moved? (ii) Where does it move to? (iii) How local is the movement operation? and (iv) What is left behind after movement? To this list is added one of the biggest questions –– which still remains an unresolved puzzle ––what forces movement and why is movement required at all?   3  1.2 Why wh-movement in Medumba? Wh-movement in Medumba seems to be a perfect test case for theories of wh-movement because wh-question formation in Medumba involves both the in-situ (i.e. leaving the wh-XP in place) and the ex-situ (i.e. moving the wh-XP to the clause-initial position) strategies. The examples from (4) to (7) provide a synopsis of the major properties of wh-questions in Medumba. The clauses (4&5) contain one main verb each and are called root clauses. (6&7) are non-root clauses embedded within another clause.  (4) Root-clause wh-in-situ        Wàtɛ̀t   nɔ́ʔ             sʷɛ̀n     á      wʉ́         á        Watat   AUX.T2     sell      FOC   who        C.Q.H             T.H    V.L                                          Lit.: Watat betrayed who? (5) Root-clause wh-ex-situ        a. á      wʉ́     Wàtɛ̀t      nɔ́ɔ̀ʔ                  ⁿ-sʷɛ́ɛ̀n         ___   á         b. á      wʉ́     Wàtɛ̀t      nɔ́ɔ̀ʔ                   ⁿ-sʷɛ́ɛ̀n         í         á             FOC   who    Watat     AGR.AUX.T2     N-AGR.sell    3SG.H        C.Q.H                                          T.HL                     V.HL                      ‘Who did Watat betray (him/her)? (6) Non-root clauses wh-in-situ        Nùmí       nɔ́ʔ           tʃúp   ᵐbʉ̀   Wàtɛ̀t   nɔ́ʔ            sʷɛ̀n     á      wʉ́         á        Numi      AUX.T2    say    C.L   Watat   AUX.T2x   sell      FOC   who      C.Q.H                T.H      V.H                T.H                V.L                       Lit.: Numi said that Watat betrayed who? (7) Non-root clauses wh-ex-situ       á    wʉ́    Nùmí  nɔ́ɔ̀ʔ                ⁿ-tʃúp  ᵐbʉ̀  Wàtɛ̀ɛ́t    nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n         í    á        FOC who Numi  AGR.AUX.T2  N-say    C.L  Watat.H  AGR.AUX.T2 N-AGR.sell   3SG.H  C.Q.H                         T.HL                      V.H              T.HL            V.HL                 ‘Who did Numi say that Watat betrayed [him/her]? The color coding used in the examples highlights the properties of Medumba wh-questions which constitute the focus of this dissertation as follows: 4   (i) Red: a wh-XP in Medumba can stay in its canonical position –– whether in a root clause (4) or in a non-root clause (6) –– or can move to the clause-initial position as in (5) and (7); this property is discussed in chapter 2. When a wh-XP moves, if it occurs in a root clause, then its extraction site is left empty (a gap) as in (5a) or filled by a pronoun, usually called a resumptive pronoun in this context as in (5b). However, if movement occurs in a non-root clause, then the extraction site of the moved wh-XP is always filled by a resumptive pronoun as shown in (7). This property is discussed in chapter 5.  (ii) Blue: in a sentence where there is movement of the wh-XP, there is a change in the tonal melody of verbal heads and temporal auxiliaries. (4) and (6) show that in sentences where the wh-XP is in its canonical position, the temporal auxiliary nɔ́ʔ surfaces with a high tone (H) and the verb sʷɛ̀n ‘sell’ with a low tone (L). In contexts where there is movement of the wh-XP to the clause-initial position, the auxiliary nɔ́ʔ and the verb sʷɛ̀n ‘sell’ surface with an HL tone melody (see 5&7). I develop an analysis which treats this perturbation of tone melody as an instance of A′-agreement. In non-root clauses, the verb (boldface) that introduces the embedded clause (matrix verb) never shows A′-agreement as shown in (7) where tʃúp ‘say’ keeps its H-tone. The phenomenon of A′-agreement is investigated in chapter 3 and 4.  It is to be noted that the above properties are attested not only in the context of content questions (i.e. wh-movement in the narrow sense), but are also found with focus movement, relativization and topic-movement. These are all instances of A′-movement. Using less Anglo-centric terminology, wh-words in Medumba could be called ʉ́-words. An ʉ́-word in Medumba contrasts a w-form used only for human, and a k-form used elsewhere as shown in table 1.1.   5  Base Animacy Feature Form Gloss Examples                    DP [w- + WH]  Human w-ʉ́ HMN-WH who (8a)                    DP [k- + WH]    Unmarked  k-ʉ́ UNMKD-WH what (8b) PP  [ⁿʣʉ́-   DP]         [nùúm-DP] [P  ⁿʤʉ́-[DP k-ʉ́ ]]     way       UNMKD-WH how (9a) [P nùúm-[DP  k-ʉ́ ]]     PREP         UNMKD-WH          why (9b) ADV  Other s-ʉ́ TEMP-WH when (10a) já LOC.WH where (10b) Table 1. 1: The internal structure of ʉ́-words (wh-words) in Medumba The forms given in table 1.1 are illustrated in the following examples. For simplicity, I will be glossing wh-XPs in the remainder of the dissertation as just wh.  (8) Wh-DP  a. Wàtɛ̀t   nɔ́ʔ             kɛ̀           á      w-ʉ́                á             Watat   AUX.T2    choose      FOC   HMN-WH        C.Q.H                   T.H     V.L                                              Lit.: Watat chose who?  b. Wàtɛ̀t   nɔ́ʔ             kɛ̀          á      k-ʉ́                     á             Watat   AUX.T2    choose      FOC   UNMKD-WH        C.Q.H                   T.H     V.L                                             Lit.: Watat chose what? (9) Wh-PP   a. Wàtɛ̀t   nɔ́ʔ             kɛ̀ɛ́      Nùmí    á       ⁿʤʰʉ́-k-ʉ́                    á                Watat   AUX.T2    choose  Numi     FOC   WAY-UNMKD-WH        C.Q.H                     T.H              V.LH                                                  Lit.: Watat chose Numi how?  b. Wàtɛ̀t   nɔ́ʔ             kɛ̀ɛ́      Nùmí     á       nùúm-k-ʉ́           á                Watat   AUX.T2    choose   Numi     FOC   PREP-UNMKD-WH        C.Q.H                      T.H              V.LH                                                 Lit.: Watat chose Numi why?   6  (10) Wh-Adv  a. Wàtɛ̀t   nɔ́ʔ             kɛ̀ɛ́      Nùmí    á      s-ʉ́                 á                Watat   AUX.T2    choose  Numi     FOC   TEM-WH        C.Q.H                     T.H              V.LH                                                 Lit.: Watat chose Numi when?  b. Wàtɛ̀t   nɔ́ʔ             kɛ̀ɛ́      Nùmí      á      já       á                Watat   AUX.T2     choose  Numi     FOC   LOC.WH        C.Q.H                     T.H              V.LH                                                 Lit.: Watat chose Numi where? 1.3 Medumba: the language and the people Medumba (Mə̀dʲʉ̂ᵐbɑ̀; Mə̀ɟʝʉ̂ᵐbɑ̀) is a Grassfields Bantu language spoken in Cameroon. Medumba speakers originate from the Ndé division of the West Region of Cameroon (figure 1), with their main settlements in Bangangté, Bakong, Bangoulap, Bahouoc, and Tonga. However, there is also an important number of Medumba speakers settled in Bazou (in the neighborhoods of Plateau, Carrière, Coteau, Tergal, Femtchouet, and Comfort) and in the neighboring villages; namely, Bafetba, Bamaha, Nsiteun, Kouba and Bassamba (Keupdjio 2011).  7   Figure 1. 1: Geographical location of the Medumba people  Medumba is a major language of the Bamileke cluster and is part of its Eastern Group (figure 2). The Bamileke cluster is a cluster of eleven languages including Fe’fe’, Ghomala, Kwa’, Medumba, Nda’nda’ (Eastern Bamileke); Mengaka, Ngombale, Ngiemboon, Ngomba, Nwe, Yemba (Western Bamileke). The Bamileke cluster — along with Ngemba, Nkambe and Nun — 8  is part of the Eastern Grassfields subgroup which, together with the Ring, the Momo languages and the Southwest Grassfields languages, constitute the Grassfields grouping (Watters 2003).   Figure 1. 2: Bamileke Medumba, the language   9   Medumba people are located in an area where sacred kingship played a pivotal role in government, justice, and diplomacy (Feldman-Savelsberg 1995; Warnier 2015). The modern history of the Bamileke area, which was a German colony placed under French trusteeship after World War I by the League of Nations in 1919, is closely associated with the nationalist movement2 of the ‘Union des populations du Cameroun’ (UPC), which developed primarily in the coastal hinterland (Basaá) and the western highlands (Bamileke). From 1956 to the late 1960s, this area of Cameroon experienced a period of unrest (Meredith 2014), which resulted in a brutal repression3 by the colonial army and the Cameroonian neocolonial army. This episode of the history of the Bamileke people –– which to date is still an untold history –– continues to shape the Bamileke political culture and has an impact on language identity (Bandia 1993) and the linguistic landscape (Ndjio 2009).  1.4 Previous work on Medumba Medumba is relatively well described for a Grassfields Bamileke Bantu language.  This is due to the work done in the 1960s and 1970s by Jan Voorhoeve on (i) morpheme structure constraints (Voorhoeve 1965); (ii) personal pronouns (Voorhoeve 1967); (iii) noun classes (Voorhoeve 1967 and 1969); (iv) tone of nouns (Voorhoeve 1971); (v) traditional Bamileke narratives (Voorhoeve 1976); (vi) general linguistic description of the Bangangte dialect (Voorhoeve 1977). This work laid the foundations for research conducted by L. Hyman, on the closely related Bamileke language Fe’fe’ (Hyman1970).   2 The Nationalist movement advocated for freedom and the right of the people of Cameroon to self determination  3 There is a locality in the Ndé division called ⁿtɔ́ɔ̀-lʉ́ which literally means ‘blood-covered market’. My parents told me when I was growing up that the locality got this name in the aftermath of a massive repression, a massacre by the colonial army where many were reported dead. The use of the compound ⁿtɔ́ ‘market’ and lʉ́ ‘blood’ depicts the gruesomeness of the atrocities.     10   Research on Medumba was re-invigorated in the early 2010s by a research group at Boston University (lead by C. O’Connor), at the University of British Columbia (lead by R.-M. Déchaine) and at the university of Delaware by K. Franich. Some of these publications include: Danis et al. 2012, on a prosodic account of downstep and contour formation in Medumba; Franich et al. 2012, on tonal merger in Medumba nouns; Goldman et al. 2014, on interspeaker variation in noun class realization in Medumba; Franich 2016a on internal and contextual cues to tone perception in Medumba; Franich 2016b, on the perception of tonal contours in Medumba; Franich 2017, on evidence for metrical prominence asymmetries in Medumba; and Franich 2018, on tonal and morphophonological effects on the location of perceptual centers (p-centers).  Also relevant are the notable scholarly contribution of Medumba speaker-linguists listed here in chronological order: (i) Tondji 1979 D.E.S (MA) thesis (University of Yaoundé) on phonetic variation in Medumba (ii) Nganmou 1991, doctoral dissertation (University of Yaoundé) on tense and aspect; (iii) Wandji 1993, MA thesis (University of Yaoundé 1) on the Medumba noun phrase; (iv) Njike 2010, MA thesis (University of Yaoundé 1)  on the syntax of adverbs; (v) Kouankem 2011, PhD dissertation (University of Yaoundé 1) on DP syntax; (vi) Keupdjio 2011, MA thesis (University of Yaoundé 1) on wh-movement and clause structure; (vii) Kouankem 2013, an article on DP concord.   Current work on Medumba is part of a more general effort towards describing, analyzing and documenting the languages of Africa, in the face of rising levels of language endangerment (Kandybowicz and Torrence 2017) and language death. According to Opala 2002 and Okol 2014, Cameroon — along with Nigeria, Sudan, and Ethiopia — is reported to have one of the highest language mortality rates in Africa. Given the wide range of research topics on Medumba, it is almost impossible to do a summary of previous work done on the language. In appendix B, I describe one aspect of Medumba which has not been previously investigated systematically, 11  namely stem allomorphy, which is a core diagnostic for A′-agreement. Unless cited otherwise, the data in this dissertation are based on my introspection as a Medumba speaker-linguist. 1.5 Preview of the dissertation This dissertation is organized as follows:  o A′-movement and the in-situ/ex-situ partition (Ch. 2);  o A′-movement and A′-agreement (Ch. 3);  o A′-movement and the tense/aspect system (Ch. 4);  o A′-movement and resumption (Ch. 5); and  o Prospects for future research (Ch.6). These chapters are self-contained, so it is possible to read them independently. In the following subsections, I give a preview for each of the chapters, the main theoretical problem and my proposal to solve it.  1.5.1 A′-movement and the in-situ/ex-situ partition in Medumba Chapter 2 examines the in-situ/ex-situ partition in Medumba with regard to wh- and focus construals. The main problem addressed in this chapter is illustrated in (11) whereby a wh-XP or a focus XP can be construed in-situ (11a) or can be moved to the clause-initial position (11b). (11) a. (i)  Wàtɛ̀t   nɔ́ʔ             kɛ̀      á      wʉ́           á  [in-situ wh-question]             (ii) Wàtɛ̀t   nɔ́ʔ             kɛ̀      á      Nùmí   [in-situ focus]             Watat  AUX.T2      choose       FOC   DP           C.Q.H                        T.H        V.L                                                        Lit.: Watat chose who? /Numi.  a. (i)  á      wʉ́         Wàtɛ̀t      nɔ́ɔ̀ʔ                  ᵑ-kɛ́ɛ̀    ___    á [ex-situ wh-question]                   (ii) á      Nùmí       Wàtɛ̀t      nɔ́ɔ̀ʔ                  ᵑ-kɛ́ɛ̀    ___    lá [ex-situ focus]                       FOC    DP           Watat     AGR.AUX.T2    N-AGR.sell     C                                                T.HL                        V.HL                               ‘Who did Watat choose? / NumiFOC Watat chose’ 12  Given (11), there arises the question of how in-situ and ex-situ wh-/focus construals are derived in Medumba. More specifically, why do wh-/focus construals involve displacement of the wh-XP or focus XP in some contexts and no displacement in others? In other words, if movement is possible in Medumba, why isn’t it obligatory? To these questions is also added the question of whether in-situ and ex-situ wh-questions and foci are derived by the same mechanism in the language or by two different mechanisms; and what implications this has for current theories of A′-movement.   To answer these questions I adopt the hypothesis that the necessity of movement is driven by interpretation. On this view, movement in Medumba wh-questions and focus construals is driven by interpretation. I propose that there is a covert exhaustive operator at C which marks any wh-/focus XP within its vicinity as exhaustive. This proposal aims at an account of the properties of in-situ and ex-situ wh-/focus construals in Medumba. As such, ex-situ wh- /focus XPs are exhaustive while their in-situ counterparts are predictably non-exhaustive. 1.5.2 A′-movement and A′-agreement in Medumba Chapter 3 examines A′-movement and A′-agreement. The latter is realized in Medumba as an HL tone melody that overwrites the lexical tone of verbal heads, as well as that of temporal and aspectual auxiliaries. As shown in (12), when there is A′-movement form a non-root clauses, there is A′-agreement in the form of an HL overwrite tone melody with the embedded CP on V and T. With the matrix CP, A′-agreement is only on the matrix T and not on the matrix V as in (12a). Absence of A′-agreement when there is movement leads to ungrammaticality (see (12b)). (12) a. á    wʉ́  Nùmí  nɔ́ɔ̀ʔ             ⁿ-tʃúp ᵐbʉ̀ Wàtɛ̀ɛ́t   nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n      í    á                  FOC WH  Numi AGR.AUX.T2 N-say  C.L Watat.H AGR.AUX.T2 N-AGR.sell 3SG.H  C.Q.H                                  T.HL                     V.H               T.HL              V.HL                          ‘Who did Numi say that Watat betrayed [him/her]?’  13   b. *á     wʉ́  Nùmí  nɔ́ʔ         tʃúp  ᵐbʉ̀  Wàtɛ̀ɛ́t    nɔ́ʔ          sʷɛ̀n      í    á                      FOC WH  Numi  AUX.T2  say    C.L  Watat.H  AUX.T2   sell         3SG.H      C.Q.H                                      T.H            V.H            T.H          V.L                             [Who did Numi say that Watat betrayed [him/her]?]    The data in (12) raise several questions, among which are the following:  (i) What is the formal mechanism for deriving A′-agreement in Medumba? In other words, what determines when A′-agreement surfaces on a head or not? And why those heads? (ii) What accounts for the absence of A′-agreement with matrix V in non-root clauses?  (iii) Is A′-agreement an idiosyncratic property of Medumba? If not, are there parallels in the literature to phenomena in other languages that bear similarity to the form of A′-agreement found in Medumba? And to what degree and with what implications for both existing analyses of other forms of A′-agreement and the analysis proposed for A′-agreement in Medumba?  A′-agreement (depending on its form) has previously been analyzed and referred to in the literature as wh-agreement, wh-copying, extraction morphology or complementizer agreement (see a.o. Kinyalolo 1991; Carstens 2005; Chung 1994; Reintges, LeSourd and Chung 2006; Hedinger 2008; Lochbihler and Mathieu 2010; Felser 2004). I propose in chapter 3 that all instances of A′-agreement reflect the activity of the same underlying formal mechanism, namely Phasal-Agree. A phase-bound operation (OP) between a probe (P) and a goal (G), where P is a phase-head and G an A′-bound XP; applies in such a way that the reflex of OP is either on P or on the complement of P. In other words, A′-agreement is the reflex of cyclic phase-by-phase movement, with agreement surfacing within the phase domain each time an A′-bound XP reaches a phase edge. With regard to the absence of A′-agreement with a matrix V in extraction from a 14  non-root CP, I propose that apparent complement CPs are disguised adjunct clauses in Medumba. Thus, movement from apparent complement CPs does not go through the matrix VP edge.   1.5.3 A′-movement and the tense/aspect system in Medumba Chapter 4 focuses on A′-agreement and the tense/aspect system in Medumba. More specifically, I look at the distribution of A′-agreement in aux-stacking contexts. As shown in (13), in aux-stacking contexts, only some auxes get A′-agreement. In (13b) there is A′-agreement with Aux.T2, Aux.T3 but not with Aux.β.  (13) a. Nùmí nɔ́ʔ  fə̀  ⁿ-ʧə́k  ᵑ-kɛ́ ↓ⁿdʒʷɛ́n      Numi AUX.T2 AUX.T3 N-AUX.β N-fry chips    T.H  T.L       T.H  V.H         ‘Numi fried the chips (long ago in the morning of the day before)’  b. á kʉ́ Nùmí nɔ́ɔ̀ʔ  ᵐ-fə̀ə́  ⁿ-tʃə́k  ᵑ-kɛ́ɛ̀     á         FOC WH Numi AGR.AUX.T2 AGR.AUX.T3 N-AUX.β N -AGR.fry C.Q.H     T.HL  T.LH       V.H              V.HL           ‘What did Numi fry (long ago in the morning of the day before)’   The question that arises from this data is what conditions the locus of A′-agreement in Medumba? That is, why does A′-agreement not surface on every auxiliary, and why is the locus of A′-agreement only on the auxiliaries it is on? To answer these questions, I use the mechanism that was proposed in chapter 3; that is, Phasal-Agree derives A′-movement in aux-stacking contexts. More precisely, I propose that there is an intermediate phase between vP and CP. Thus, absence of A′-agreement with Aux.β. is because Aux.β.  is not a phase. 1.5.4 A′-movement and resumption in Medumba Chapter 5 examines resumption in A′-construals. The problems addressed in this chapter are illustrated in (14) where a resumptive pronoun is obligatory in island violation contexts as in (14a-ii) and optional in contexts in which there is no island violation, specifically in root clauses where resumption alternates with a gap, as in (14b).  15  (14) a. Extraction out of an adjunct island (i) *á    wʉ́    Nùᵑgɛ̀   nɔ́ɔ̀ʔ           nɛ́ɛ̀n   ⁿ tɔ́n      [káà      Wàtɛ̀ɛ́t     tʃɔ́ɔ̀ʔdə̀ ___]        à?  (ii)  á    wʉ́    Nùᵑgɛ̀   nɔ́ɔ̀ʔ           nɛ́ɛ̀n   ⁿ tɔ́n       káà      Wàtɛ̀ɛ́t     tʃɔ́ɔ̀ʔdì       í          á?       FOC  WH   Nuga    AGR.AUX.T2 N-go  market  before  Watat.H   AGR.greet  3SG.H   C       ‘*who did Nuga go to the market before Watat greeted [him/her]?’  b. Root clause extraction  (i)  á      wʉ́     Wàtɛ̀t      nɔ́ɔ̀ʔ                  ⁿ-sʷɛ́ɛ̀n         ___    á (ii) á      wʉ́     Wàtɛ̀t      nɔ́ɔ̀ʔ                   ⁿ-sʷɛ́ɛ̀n          í        á       FOC   WH     Watat      AGR.AUX.T2     N-AGR.sell    3SG.H            C.Q.H                                   T.HL              V.HL                  ‘Who did Watat betray (him/her)?  The data in (14) raise the following questions: (i) Which mechanism regulates resumption in island contexts? (ii) Which mechanism regulates resumption in non-root clauses?  The overarching question addressed in this chapter is the theory of resumption in Medumba and the implications it has for the syntax and semantics of A′-movement.  I propose that the economy principle of Last Resort drives resumption in Medumba (see a.o. Koopman and Sportiche 1986; Rizzi 1990; Chomsky 1991, 1998; Shlonsky 1992; Bobaljik 1995, Lasnik 1995; Ura 1996; Pesetsky 1997; Collins 2001; Bošcović 2011). In proposing a unified account of resumptive strategies in Medumba, this chapter argues that Last Resort conditions can be syntactically or semantically conditioned. Syntactic Last Resort derives resumptive pronouns in island violation contexts to salvage A′-dependencies that would otherwise result in ungrammaticality. Semantic Last Resort is a condition on interpretation that derives resumption in configurations that would otherwise result in ambiguity.   In the next chapter, I focus on A′-movement and the in-situ/ex-situ partition in Medumba.   16  Chapter 2: A′-movement and the in-situ/ex-situ partition in Medumba  2.1 The ex-situ/in-situ paradox: if movement is possible, why isn’t it obligatory? In many languages, forming a wh-question4  consists either of moving the wh-word to the clause left-peripheral position (ex-situ) or leaving the wh-word in place at its base-generated position (in-situ). Generative theories of wh-question formation have mostly focused on languages that exhibit only one of the facets of wh-question formation, and more specifically on languages that use the ex-situ (movement) strategy. Less attention has been given to languages that use both the in-situ and the ex-situ strategies. However, there are some puzzles with regard to why (i) wh-question formation in some languages involves movement of the wh-phrase but in others it does not and (ii) why movement is even required to start with. For instance, while approximately 240 languages found in the World Atlas of Language Structures (WALS) show obligatory movement of wh-words to the clause-initial position, in approximately 540 languages wh-words are not “obligatorily” clause-initial (Dryer 2008, Déprez et al. 2013). Though it is unclear what is meant by “not obligatorily” initial in that it could either mean non-displacement of wh-words at all or the possibility of having both displacement and non-displacement of wh-words, it is clear that for languages that allow both the in-situ and the ex-situ strategy, things are a bit fuzzy. The examples in below illustrate wh-questions in English-type systems where the wh-word surfaces in clause-initial position (ex-situ) (1) and in Mandarin-type systems where the wh-word surfaces in its base-generated position (in-situ) (2).  4 Wh-questions are also known as ‘content questions’ or ‘information seeking questions.’ They are characterized by a dedicated wh-word and require an answer that replaces or otherwise responds to this wh-word.  17  (1) Who did you see ___?        English (2) Qiaofeng mai-le  shenme ne      Mandarin         Qiaofeng buy-ASP what  QWH         ‘What did Qiaofeng buy?’     [Cheng 1991: 11, p. 30] 2.1.1 The problem: Medumba uses both ex-situ and in-situ strategies  In Medumba, wh-questions and focus exhibit an in-situ/ex-situ partition in that wh-XPs and focused XPs are either construed in-situ in their base-generated position or are moved to the clause left-peripheral position (ex-situ). In addition, whether in-situ or ex-situ, wh-phrases and focused XPs are always5 preceded by the invariable high tone focus particle á in Medumba. For wh-questions, there is also a variable (H/L)6 Q-particle a that surfaces in final-position. These are given (3) and (4) for wh-questions and focus respectively. (3)  a. In-situ wh-question     Wàtɛ́t nɔ́ʔ  sʷɛ̀n á wʉ́ á?         Watat AUX.T2 sell FOC WH C.Q.H    T.H  V.L            Lit.: ‘Watat betray who?’  b. Ex-situ wh-question   á wʉ́  Wàtɛ́t nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n á?     FOC WH Watat AGR.AUX.T2 N-AGR.sell C.Q.H    T.HL      V.HL             ‘Who did Watat betray?’  (4)  a. In-situ focus   Wàtɛ́t nɔ́ʔ  sʷɛ̀n á Nùᵑgɛ̀    Watat AUX.T2 sell FOC Nuga      T.H  V.L   ‘Watat betrayed NugaFOC’  5 In some dialects, subject wh-/focus XPs can surface without the focus particle.  6 The Q-particle [a] surfaces as H when the preceding syllable carries an H-tone and L when it carries an L-tone  18   b. Ex-situ focus   á Nùᵑgɛ̀ Wàtɛ́t nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n lá   FOC Nuga Watat AGR.AUX.T2 N-AGR.sell C.-Q         T.HL       V.HL      ‘NugaFOC Watat betrayed’  The preceding data sets show that there are at least three types of systems7 to keep in mind while investigating wh-questions as listed below. (5)  a. Lg 1: wh-ex-situ languages  b. Lg 2: wh-in-situ languages  c. Lg 3: in-situ and ex-situ-wh languages   2.1.2 A possible (but non-viable) solution: feature-driven movement  Wh-questions are generally viewed as involving movement of the wh-XP to the clause left-peripheral position. Movement is usually motivated the presence of some wh/Q-feature at C that drives movement of the wh-phrase from its base position to Spec-C. This can be implemented in two ways. Implementation 1 consists of whether the moved wh-XP checks the relevant feature overtly or covertly (Chomsky 1977, 1986, 1998; Huang 1982; Aoun and Li 1993; Cheng 2009). As for implementation 2, it consists of whether the head or tail of the chain is deleted (Chomsky 1995; Bobaljik 2002; Bošković and Nunes 2007; Nunes 2011, 2017).  2.1.2.1 Implementation 1: overt versus covert movement  Implementation 1 of feature-driven movement involves movement of the wh-phrase to the left periphery. For languages where wh-question formation involves fronting of the wh-XP, the surface configuration is derived by overt movement of the wh-XP at S-structure to check  the wh/Q-feature at C whereas in languages where the wh-phrase surfaces in-situ in wh-question formation, the  7 I abstract away from multiple wh-questions here (see chapter 3, section 3.2.2 for details). 19  surface configuration is derived by covert movement of the wh-phrase at LF (Huang 1982, Aoun and Li 1993, Pesetsky 2000, Cheng 2009).  (6)  a. Overt movement at S-structure   b. Covert movement at LF            CP       CP            wo    wo        Spec                         C’              Spec                         C’         whi              wo          wo                           CWH/Q                     TP         CWH/Q                       TP                        Attract   6                           Attract      6                                                           ti      wh    Overt movement        Covert movement  For Medumba-type languages, where both wh-ex-situ and wh-in-situ are possible, it is unresolved how this analysis would apply. We must first determine under which conditions overt and covert movement have applied, and secondly, why they need to be applied at all in a single system. 2.1.2.2 Implementation 2: deletion of head versus tail  Implementation 2 of feature-driven movement involves the copy theory of movement (Chomsky 1995). It stipulates that movement in overt syntax creates a chain with two or more copies which are then interpreted at the PF and LF interfaces (see also Bobaljik 2002, Bošković and Nunes 2007; Nunes 2011, 2017). With this implementation, wh-ex-situ and wh-in-situ are derived by the same movement operation. Their surface realization depends on which copy the two interfaces interpret. If the head of the chain (i.e. the upper copy) is interpreted by both interfaces, it creates the typical movement operation in which movement is visible as the upper copy of the moved XP is pronounced and is also the one interpreted by LF. This derives wh-questions in wh-ex-situ languages. However, the tail of the chain (i.e. the lower copy) can be pronounced by PF while LF 20  interprets the upper copy; in such a derivation, the lower copy looks as if it has not been moved. This derives wh-questions in wh-in-situ languages.  (7) a. Copy + deletion of the tail of the chain b. Copy + deletion of the head of the chain            CP       CP            wo    wo        Spec                         C’            wh-XP                         C’       wh-XP          wo          wo                           CWH/Q                     TP         CWH/Q                       TP                                       6                                    6                                                      wh-XP             wh-XP               As for Medumba-type languages, where both wh-fronting and wh-in-situ are possible, this leaves unresolved exactly the nature of the mechanism that determines whether the lower copy or the upper copy is deleted.     Under a feature-driven account, wh-questions are derived by movement of the wh-XP regardless of their surface configurations. This predicts no difference –– be it structural or semantic –– between in-situ and ex-situ wh-questions. This approach faces two problems:   (i) The first problem is theory internal and arises from Spec-to-Spec movement in long-distance wh-extraction. If movement of the wh-XP to Spec-C is triggered by the need to check some wh/Q-feature at C, there arises a question as to what triggers movement through the intermediate CPs in long-distance movement given that in those cases only the highest C bears the wh-/Q-features as shown in (8). (8)  [ wh CWH/Q [   wh C  wh ]  Several proposals have been made to regulate movement through intermediate CPs.  21  • Collins (1997) proposes that movement through intermediate CPs is triggered by uninterpretable, non-interrogative wh-features in intermediate C-heads.  • McCloskey (2002) working with the phase-based derivation, proposes that spurious features on intermediate C-heads are triggers for movement of a wh-XP to the phase-edge.  • Felser (2004) similarly argues that movement through intermediate CPs are triggered by pseudo-interrogative or other ‘peripheral’ (force, focus, or similar) features in intervening phase heads.   While it is clear from the different analyses that the motivation of movement through the intermediate CPs is an outstanding question for theoretical syntacticians, things become unclear when phases are invoked. All of these proposals focus only on the intermediate CP phase and none address the question of how movement proceeds through intermediate vP phases.  (ii) The second problem faced by the feature-driven movement approach to wh-questions come from languages like Medumba where in-situ and ex-situ wh-questions are not equivalent to each other. In fact, feature-driven movement cannot account for languages like Medumba where there is a structural and semantic difference between in-situ and ex-situ wh-questions.  In sum, although feature-driven movement remains a powerful tool in deriving wh- questions in wh-ex-situ and wh-in-situ languages, the structural and semantic differences between in-situ and ex-situ questions in Medumba are challenges for this approach. 2.1.3 Another possible (but non-viable) solution: Agree  With the advent of the operation Agree (Chomsky 2000, 2001, 2007, 2008 and others), feature checking does not necessarily require movement. It is taken care of by Agree, a feature-checking operation between a probe and a goal. The operation Agree takes place only if the probe and the goal both bear uninterpretable features, which make them active. After the operation Agree applies, 22  the uninterpretable features of the probe and the goal are checked, making them inactive or invisible for further Agree operations. This is schematized in (9) with α serving as a probe and β as a goal. (9)   a. Feature matching phase    b. Agree phase               ...          ...              3         3  α-F          ...         α+F                    ...            3       3          β-F    ...      β+F          ...   With regard to wh-movement, Agree predicts a configuration where the features of the probe C and the goal wh-XP are checked at a distance without any movement required as illustrated in (10). (10)       CP                    3        Spec             CP                      3                    C[WH]          TP                                6                    Agree          wh-XP[WH]  Agree analyses predict no feature-driven movement. In fact, they predict no movement at all and therefore predicts everything in-situ. Therefore, the existence of ex-situ wh-questions becomes a problem and Agree analyses must invoke additional mechanisms. It is usually assumed that content questions in wh-ex-situ languages are derived in this system by an Agree operation followed by Move. However, there are several puzzles with regard to Move, especially the mechanisms that force it and why it applies at all. Attempts to solve this problem include the use of EPP-features, edge features and the satisfaction of labeling (Chomsky 2007, 2013).          2.1.3.1 Why Agree + EPP-features do not work  It has been proposed that Move only takes place if the head carrying the probe feature is also endowed with an EPP-feature. That is, while the feature checking operation is satisfied under 23  Agree, the presence of an EPP-feature at C forces movement of the wh-phrase to Spec-C as illustrated in (11).  (11)        CP                          wo              Spec                         CP                                  wo                                CEPP                       TP                                                             6                                 Move                            wh            In wh-ex-situ languages, the head C has an EPP feature that triggers movement of the wh-XP to Spec C in addition to the wh/Q-features that are checked under Agree. In contrast, in wh-in-situ languages the head C lacks the EPP feature. For Medumba-type languages which allow both wh-in-situ and wh-ex-situ, is it the case that C has an EPP-feature for ex-situ construals and no EPP-feature for in-situ construals? If so, why and what determines the presence or absence of the EPP-feature?  (12) EPP and wh-ex-situ versus wh-in-situ languages  Lg 1:  [ wh CEPP   wh ] wh-ex-situ only   Lg 2:  [  C   wh ] wh-in-situ only                    ✘  Lg 3: ?a. [ wh CEPP   wh ]     ?b [  C   wh ]                     ✘ 2.1.3.2 Why Agree + Edge-features do not work  Chomsky (2007, 2008) analyzes features that favor Move as edge features. Edge features (EFs) are the properties of lexical items that enable them to be merged and to enter a computational system 24  (Chomsky 2008:139). In such an analysis, every lexical item has an edge feature. Chomsky argues that EFs permit free merge to the edge and Move satisfies EFs of phase heads.  (13) The edge feature EF of a phase head P can seek a goal in the complement of P, which it can raise to SPEC-P (Chomsky 2007: 24).  While A′-movement is Internal Merge driven by EFs of the phase head, A-movement is Internal Merge contingent on probe by uninterpretable inflectional features (Chomsky 2007: 24). This is illustrated below as applied to wh-movement. (14)      CP            3        Wh-XP       CP                     3                    CEF          TP                              3                            T               vP                                        3          <Wh-XP>      vP     3               vEF     VP      5      <Wh-XP> It appears that Move is conditioned by EFs and is to be a free operation that may or may not be applied. One can therefore conclude that in wh-fronting languages, EFs permits Move to the edge whereas in wh-in-situ languages, EFs do not permit Move to the edge. However, it remains unclear under which conditions EFs permit or do not permit movement to the edge, especially for Medumba-type languages.    (15) EFs and ex-situ versus in-situ languages    Lg 1:  [ wh CEF   wh ] wh-ex-situ   Lg 2:  [  CEF   wh ] wh-in-situ 25   Lg 3: ?a. [ wh CEF   wh ]     ?b. [  CEF   wh ] 2.1.3.3 Why Agree + labelling does not work  In Chomsky’s 2013 labeling theory (Problems of Projection), the necessity of movement to intermediate positions is derived by the failure of labeling if one of sister non-terminal projections does not move. He assumes that movement is free (untriggered). In such an analysis, the only way for non-terminal projections that are sisters to be labeled is if there is an agreement relation between them that can project. For wh-interrogatives, such an agreement is not available, until its features match those of C, so that [[wh-XP] [C+wh [ TP]]] can project a common label, namely wh-interrogative. This is probably satisfied in wh-in-situ languages by covert movement, but this is not clear in Chomsky’s description. The labeling theory tries to derive movement (especially movement to intermediate CPs) in wh-questions without recourse to features. However, it still faces problems with Medumba-type languages. In fact, if the necessity of movement (be it overt or covert) is to avoid failure of labeling, then this would incorrectly predict wh-in-situ to be ruled out in Medumba if there is no movement of some sort. But I show in the next section that wh-in-situ in Medumba does not seem to undergo any movement of any sort. 2.1.4 Another possible (and viable) solution: interpretation-driven movement    To derive wh-questions in Medumba, I propose an analysis where movement is driven by interpretation.  (16) Movement is driven by interpretation In this analysis, movement in wh-questions in Medumba is not motivated by the desire to check wh/Q-features at C, but instead is driven by interpretation. This correctly predicts that there must 26  be structural and semantic differences between in-situ and ex-situ wh-/focus in Medumba. I show that in-situ and ex-situ wh-/ focus differ in Medumba:  o Ex-situ wh- and focus require exhaustivity o In-situ wh- and focus don’t require exhaustivity o Question/answer pairs condition the in-situ/ex-situ partition  o Fragment answers condition the in-situ/ex-situ partition o Subject wh-/focus are (predictably) different Also, interpretation-driven movement captures the properties of Agree that distinguish it from Move: while Agree is legislated by formal features, Move is driven by interpretation.   This analysis is similar to the free (untriggered) approach to movement advocated in Chomsky (2008, 2013) and Safir (2010, 2018), paraphrased as follows: (17) The Chomsky-Safir conjecture: A wh-phrase has to move somewhere else the interpretation will fail.  However, the interpretation-driven movement analysis proposed here differs from the Chomsky-Safir conjecture in one crucial way: interpretation-driven movement permits — but does not require — movement; in contrast, the Chomsky-Safir conjecture requires movement. Simply put, the hypothesis that movement is driven by interpretation predicts that if the wh-XP stays in-situ the sentence gets one interpretation and if the wh-XP moves, the resulting sentence gets a different interpretation. So, movement in some sense is free in Medumba; not in the sense that the interpretation will fail in the absence of movement (as would be the case according to the Chomsky-Safir Conjecture) but free in the sense that the absence of movement does not lead to ungrammaticality. 27  2.1.4.1 Why interpretation-driven movement is conceptually necessary Languages like Medumba, in which both the ex-situ and the in-situ strategies are used in forming wh-questions and focus pose a challenge  for theories in which (i) ex-situ wh-questions are derived either by movement or by copy and deletion of the tail of the chain;  and (ii) in-situ wh-questions are derived by covert movement or by copy and deletion of the tail of the chain. The conceptual problem is that economy and the minimization of the cost of a derivation are core concepts of the Minimalist Program (Chomsky 1995 and subsequent work). As such, it is problematic within such a framework to allow in the same language, processes, such as the possibility of having overt movement alongside covert movement, or processes such as the deletion of the head versus deletion of the tail of the chain as concurrent processes aiming at the same result. That is, using two different mechanisms to achieve the same result. Also, given the interpretative difference between in-situ and ex-situ wh-questions and focus in Medumba, it is hard to conceptually motivate why configurations derived by the same process of movement (be it overt or covert) are interpreted differently and why that should exist in a system that promotes economy of derivations. Interpretation-driven movement is advantageous in that it derives ex-situ wh-questions and focus without invoking additional mechanisms for in-situ configurations. Put simply, in-situ wh-questions and focus are derived for free within this approach. Also, with interpretation-driven movement, the question of what triggers movement through intermediate CPs, and what is the nature of those features is no longer at issue as movement is not feature-driven but free.   2.1.4.2 How interpretation-driven movement works  As stated above, interpretation-driven movement permits but does not require movement. That is when a wh-/focus-XP is merged, it can freely move to the CP edge or remain in-situ depending on which interpretation the resulting derivation conveys. In Medumba, ex-situ wh-questions and ex-28  situ foci are interpreted as exhaustive whereas their in-situ counterparts are non-exhaustive. I propose that ex-situ wh-/focus XPs in Medumba need to move within the vicinity of a covert exhaustification operator (øExh.) at C in order to be interpreted as exhaustive (see also Chierchia, Fox and Spector 2008). By vicinity, I mean being within the same phase as the covert exhaustification operator. Given that the targeted phase here is C, it means XPs at Spec-C or XPs within its complement, more specifically elements at Spec-T, would be affected. (18)  CP  wo     Wh-/foc-XP      CP       wo      C øExh.          TP             wo    Wh-/foc-XP    TP             6          …	As for in-situ wh-/focus XPs, they remain in-situ and are interpreted as non-exhaustive. Consider for instance the derivation of an object wh-question within a phase-based framework (Chomsky 2000, 2001). When the first vP phase is built, if the object Wh-XP remains in its first merge position (the V complement position), it is sent to transfer and can no longer participate further in the merge operation. Thus, when all items are merged from the numeration, the resulting derivation is interpreted as a non-exhaustive wh-question in Medumba. In contrast, if the first vP phase is built and the object Wh-XP moves and adjoins to vP, then it will remain visible for subsequent merge operations when VP (complement of the v phase-head) is sent to transfer. Thus, when the CP phase is built, the Wh-XP can be remerged from the vP phase-edge to the CP phase-edge. The resulting derivation is interpreted in Medumba as an exhaustive wh-question (see section 2.3 for the step-by-step derivation of ex-situ and in-situ wh-questions). 29  2.1.5 Preview of chapter  The chapter is organized as follows:  • (six) diagnostics for A′-movement (§2.2);  • (five) arguments in favour of interpretation-driven movement (§2.3);  • broader landscape of A′-movement (§2.4). 2.2  Six diagnostics for A′-movement  This section is an overview of the different general diagnostic properties of A′-movement and thus, can be skipped by readers with sufficient background knowledge. A′-movement typically exhibits the following properties: o A′-movement is associated with gapping or resumption (§2.2.1); o A′-movement permits long-distance dependencies (§2.2.2); o A′-movement supports reconstruction (§2.2.3); o A′-movement conditions Strong and Weak Crossover (§2.2.4); o A′-movement is island-sensitive (§2.2.5); o A′-movement conditions A′-agreement (§2.2.6). 2.2.1 Diagnostic 1: A′-movement relates the head and tail of an A′-chain 2.2.1.1 Relation between the head and tail of an A′-chain: the general picture The relation between the head and the tail of A′-chains can vary both within and across languages. In some languages, the tail of the chain is left unpronounced (a gap) as shown in (19) for English.     (19) Who did Mary see ___? In other languages, the tail of the chain is spelled out as a resumptive pronoun or alternates between a gap and a resumptive pronoun. This is shown in (20) for Lebanese Arabic where the tail of the 30  chain is spelled out as a resumptive pronoun and in (21) for Hebrew where A′-movement leaves behind a gap or a resumptive pronoun. (20) S-Sabe      yalli      raɦ təʃɦaT-o  Zeena mən l-madrase harab   the-boy      that       FUT.kick.3SF-him Z.  from the-school ran-away.3SM  ‘The boy that Zeina will kick [him] out of school ran away.’      [Lebanese Arabic, Aoun and Choueiri 2000: 15b, p. 10] (21) a. raʔiti   ʔet ha-yeled she-/ʔasher rina ʔohevet ʔoto       saw-I  ACC the-boy that   Rina loves  him        ‘I saw the boy that Rina loves [him].’  b. raʔiti   ʔet ha-yeled she-/ʔasher rina ʔohevet ___      saw-I  ACC the-boy that   Rina loves          ‘I saw the boy that Rina loves.’      [Hebrew, Borer 1984: 1a, 1d, p. 220] 2.2.1.2 Relation between the head and tail of an A′-chain: the picture in Medumba  In Medumba, the relation between the head and tail of an A′-chain patterns as follows:  Generalization 1: SUBJECT extraction always leaves a resumptive pronoun. Generalization 2: With extraction from root CPs, if the head of the A′-chain is an OBJECT, then the tail of the A′-chain can be a gap or a resumptive pronoun (i.e. with wh-/foc-movement as well as relativization). Generalization 3: With extraction from non-root CPs, the tail of the A′-chain is always a resumption pronoun.  As stated in the above generalizations, A′-movement of an XP from a root clause in Medumba can leave a gap or a resumptive pronoun at the extraction site. This is illustrated below 31  for A′-movement from the subject position. (22-24) show that subject extraction is always associated with a resumptive pronoun8.  (22) Subject wh-question a. Gap á wʉ́  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n Nùᵑgɛ̀  à?  FOC WH AGR.AUX.T2 N-sell  Nuga  C.Q.L T.HL                 ‘Who betrayed Nuga?’  b. Resumption á wʉ́ à nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n Nùᵑgɛ̀  à?  FOC WH 3SG.L AGR.AUX.T2 N-sell  Nuga  C.Q.L T.HL                 ‘Who [he] betrayed Nuga?’  (23) Subject Focus  a. Gap   á Wàtɛ̀ɛ́t     nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n Nùᵑgɛ̀   FOC Watat.H AGR.AUX.T2 N-sell  Nuga   T.HL                ‘Watat betrayed Nuga?’   b. Resumption á má-ⁿdʒùm à nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n Nùᵑgɛ̀  lá  FOC SG-male 3SG.L AGR.AUX.T2 N-sell  Nuga  C.-Q T.HL                ‘The boyFOC [he] betrayed Nuga’     8 I treat the subject wh- and the subject focus examples (23a) and (24a) respectively as being in-situ (see section 2.3.5 for discussion). 32  (24) Subject relativization  a. Gap *má-ⁿdʒùúm zə̀ ___ nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n Nùᵑgɛ̀  lá …  SG-male.H C.CL1  AGR. AUX.T2 N-sell  Nuga  C.-Q T.HL              [The boy that betrayed Nuga…]’  b. Resumption má-ⁿdʒùúm zə̀ à nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n Nùᵑgɛ̀  lá …   SG-male.H C.CL1 3SG.L AGR.AUX.T2 N-sell  Nuga  C.-Q T.HL             [The boy that [he] betrayed Nuga…]  Unlike A′-movement of the subject which is conditioned by the type of A′-extraction, A′-movement of the object alternates between a gap and a resumptive pronoun. (25) illustrates wh-movement of object wh-XP á wʉ́ ‘who’; (26) shows focus movement of the DP-object á má-ⁿdʒùm ‘boyFOC’ and (27) shows relativization of the object-DP má-ⁿdʒùm ‘boy’. In all these examples, when the XP moves from the complement-of-V position, it leaves behind a gap or a resumptive pronoun (in this case the animate H-tone pronoun í). (25) Object wh-question a. Gap  á wʉ́  Wàtɛ́t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n ___  á?  FOC WH Watat  AGR.AUX.T2 N-AGR.sell  C.Q.H T.HL       V.HL       ‘Who did Watat betray?’ a. Resumption  á wʉ́  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n í  á?  FOC WH Watat  AGR.AUX.T2 N-AGR.sell 3SG.H C.Q.H T.HL      V.HL        ‘Who did Watat betray [him/her]?’  33  (26) Object Focus  a. Gap  á má-ⁿdʒùm Wàtɛ́t nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n ___ lá FOC SG-male Watat AGR.AUX.T2 N-AGR.sell  C.-Q          T.HL       V.HL     ‘The boyFOC Watat betrayed’  b. Resumption á má-ⁿdʒùm Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n í lá  FOC SG-male Watat  AGR.AUX.T2 N-AGR.sell 3SG.H C.-Q T.HL      V.HL      ‘The boyFOC Watat betrayed [him]’ (27) Object relativization  a. Gap má-ⁿdʒùúm zə̀ Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n ___ lá …  SG-male.H C.CL1 Watat  AGR.AUX.T2 N-AGR.sell  C.-Q T.HL       V.HL             ‘The boy that Watat betrayed…’ b. Resumption má-ndʒùúm zə̀       Wàtɛ̀t nɔ́ɔ̀ʔ             ⁿ-sʷɛ́ɛ̀n  í lá …  SG-male.H REL.CL1    Watat AGR.AUX.T2 N-AGR.sell 3SG.H C.-Q T.HL      V.HL      ‘The boy that Watat betrayed [him] …’   In sum, in Medumba root clauses, A′-moved constituents leave a gap or a resumptive pronoun at extraction site. This gap/resumption alternation has semantic correlates. The gap configurations are interpreted as specific or non-specific whereas the resumptive configurations are interpreted as specific. This is further discussed in chapter 5 in which I link this split in interpretation to the general de dicto and de re distinction. More specifically, the gap is ambiguous between the de dicto and de re interpretation in Medumba whereas resumption forces only the de re interpretation.  34  2.2.2 Diagnostic 2: A′-movement permits long-distance dependency  2.2.2.1 Long-distance dependencies: the general picture  A hallmark property of A′-movement is that the distance between the head and the tail of the A′-chains can be arbitrarily long. That is, A′-movement is unbounded so that the moved XP can cross-over several CP boundaries as shown in (28). (28) [ CP who [ does Lucy think] [ CP wh [ that Susan said] [ CP wh [ that Mary saw ___ ]]]]? In (28), the wh-XP who crosses the lower CP ([that Mary saw __ ]), the intermediate CP introduced by the verb say and lands at the spec-position of the highest CP introduced by the verb think.  2.2.2.2 Long-distance dependencies: the picture in Medumba  Long distance (non-root clause) A′-extraction in Medumba always leaves a resumptive pronoun at the extraction site. The examples in (29), (30) and (31) illustrate respectively wh-movement, focus movement and relativization from an embedded subject position. These examples show that if there is A′-movement from an embedded subject position in Medumba, the tail of the A′-chain can only be associated with a resumptive pronoun (the (b) examples) and not a gap (the (a) examples).    (29) Wh-movement a. Gap *á     wʉ́   Nùmí      nɔ́ɔ̀ʔ               ⁿ-tʃúp    ᵐbʉ̀   ___  nɔ́ɔ̀ʔ                ⁿ-sʷɛ́ɛ̀n     Nùᵑgɛ̀      à   FOC WH   Numi     AGR.AUX.T2  N-say    C.L              AGR.AUX.T2  N-sell        Nuga       C.Q.L             T.HL                           T.HL    [Who did Numi say that betrayed Nuga?] b. Resumption á     wʉ́   Nùmí      nɔ́ɔ̀ʔ                 ⁿ-tʃúp    ᵐbʉ̀   á         nɔ́ɔ̀ʔ                 ⁿ-sʷɛ́ɛ̀n    Nùᵑgɛ̀      à FOC WH   Numi     AGR.AUX.T2    N-say     C.L  3SG.H   AGR.AUX.T2    N-sell      Nuga C.Q.L                     T.HL                                T.HL                           ‘Who did Numi say that [he] betrayed Nuga? 35  (30) Focus movement a. Gap *á mɛ́n     Nùmí    nɔ́ɔ̀ʔ                ⁿ-tʃúp    ᵐbʉ̀   ___    nɔ́ɔ̀ʔ                ⁿ-sʷɛ́ɛ̀n     Nùᵑgɛ̀  lá  FOC child   Numi AGR. AUX.T2  N-say     C.L               AGR.AUX.T2  N-sell        Nuga   C.-Q     T.HL             T.HL          [The childFOC Numi said that betrayed Nuga.’] b. Resumption á       mɛ́n   Nùmí    nɔ́ɔ̀ʔ               ⁿ-tʃúp    ᵐbʉ̀    á          nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n      Nùᵑgɛ̀  lá  FOC  child  Numi   AGR.AUX.T2  N-say    C.L    3SG.H   AGR.AUX.T2  N-sell        Nuga   C.-Q             T.HL                                               T.HL                  ‘The childFOC Numi said that [he] betrayed Nuga.’ (31) Relativization a. Gap *mɛ́n zə̀      Nùmí   nɔ́ɔ̀ʔ                ⁿ-tʃúp   ᵐbʉ̀  ___   nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n      Nùᵑgɛ̀        lá …   child C.CL1 Numi  AGR.AUX.T2  N-say   C.L            AGR.AUX.T2  N-sell        Nuga   C.-Q                T.HL                  T.HL               [The child that Numi said betrayed Nuga …’] b. Resumption mɛ́n   zə̀      Nùmí  nɔ́ɔ̀ʔ                ⁿ-tʃúp   ᵐbʉ̀    á         nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n      Nùᵑgɛ̀        lá …  child C.CL1 Numi  AGR.AUX.T2  N-say   C.L    3SG.H  AGR.AUX.T2  N-sell        Nuga        C.-Q           T.HL                 T.HL              ‘The child that Numi said [he/she] betrayed Nuga …’ The same pattern is found with object extractions from an embedded clause as given below for wh-movement (32), focus movement (33) and relativization (34). Again, it is illicit if the tail of the A′-chain is associated with a gap (the (a) examples) and licit if and only if the tail of the A′-chain is associated with a resumptive pronoun (the (b) examples).       36  (32) Wh-movement a. Gap *á     wʉ́   Nùmí       nɔ́ɔ̀ʔ                 ⁿ-tʃúp   ᵐbʉ̀  Wàtɛ̀ɛ́t    nɔ́ɔ̀ʔ                 ⁿ-sʷɛ́ɛ̀n      ___   á    FOC WH  Numi      AGR.AUX.T2     N-say   C.L  Watat.H   AGR.AUX.T2  N-AGR.sell             C.Q.H              T.HL                 T.HL             V.HL   [Who did Numi say that Watat betrayed?] b. Resumption á     wʉ́   Nùmí       nɔ́ɔ̀ʔ             ⁿ-tʃúp   ᵐbʉ̀   Wàtɛ̀ɛ́t      nɔ́ɔ̀ʔ              ⁿ-sʷɛ́ɛ̀n í á FOC WH   Numi     AGR.AUX.T2 N-say   C.L   Watat.H   AGR.AUX.T2  N-AGR.sell 3SG.H  C.Q.H                        T.HL              T.HL          HL            ‘Who did Numi say that Watat betrayed [him/her]?’ (33) Focus movement a. Gap *á       mɛ́n      Nùmí    nɔ́ɔ̀ʔ               ⁿ-tʃúp   ᵐbʉ̀   Wàtɛ̀ɛ́t     nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n.    ___    lá     FOC   child    Numí   AGR.AUX.T2  N-say    C.L   Watat.H  AGR.AUX.T2  N-AGR.sell       C.-Q    T.HL                     T.HL  V.HL  [The childFOC Numi said that Watat betrayed] b. Resumption  á       mɛ́n      Nùmí     nɔ́ɔ̀ʔ              ⁿ-tʃúp   ᵐbʉ̀   Wàtɛ̀ɛ́t    nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n        í       lá  FOC   child     Numí   AGR.AUX.T2  N-say    C.L  Watat.H  AGR.AUX.T2  N-AGR.sell  3SG.H C.-Q T.HL                        T.HL             V.HL     ‘The childFOC Numi said that Watat betrayed [him/her]’ (34) Relativization a. Gap  *mɛ́n zə̀        Nùmí nɔ́ɔ̀ʔ                ⁿ-tʃúp   ᵐbʉ̀   Wàtɛ̀ɛ́t     nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n        ___  lá …    child C.CL1  Numi AGR.AUX.T2 N-say   C.L    Watat.H  AGR.AUX.T2  N-AGR.sell    C.-Q    T.HL                  T.HL      [The child that Numi said Watat betrayed …]    37  b. Resumption mɛ́n zə̀        Nùmí nɔ́ɔ̀ʔ               ⁿ-tʃúp   ᵐbʉ̀  Wàtɛ̀ɛ́t   nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n       í          lá …  child C.CL1  Numi  AGR.AUX.T2 N-say   C.L  Watat.H  AGR.AUX.T2 N-AGR.sell  3SG.H  C.-Q T.HL                       T.HL           V.HL       ‘The child that Numi said Watat betrayed [him/her] …’  A′-movement is always to the left-most edge of the clause in Medumba. As a result, partial wh-movement or movement to the edge of the embedded clause is not possible in Medumba under the interrogative reading. That is, Medumba lacks embedded questions (see chapter 3 for details and analysis). More precisely such configurations have only the root clause interpretation and are interpreted as direct quotes. Moreover, in these configurations –– unlike extraction from a non-root clause where resumption is obligatory –– the tail of the A′-chain can be associated with a gap or with a resumptive pronoun. That is, resumption is optional. The fact that resumption is optional in apparent partial wh-movement in Medumba confirms that they are actually not embedded clauses.  (35) and (36) confirm that resumption is optional in Medumba construals involving apparent partial wh-movement. (35) Subject extraction a. Wh-movement Nùmí      nɔ́ʔ        bɛ́ttə́    ᵐbʉ̀    á     wʉ́    (à)      nɔ́ɔ̀ʔ                ⁿ-sʷɛ́ɛ̀n      Nùᵑgɛ̀ à Numi     AUX.T2   ask      C.L   FOC  WH   3SG.L   AGR.AUX.T2  N-sell        Nuga C.Q.L                                                T.HL                       # ‘Numi asked who betrayed Nuga  = Numi asked: “who (he) betrayed Nuga?” b. Focus movement Nùmí      nɔ́ʔ           tʃúp    ᵐbʉ̀   á mɛ́n (á)        nɔ́ɔ̀ʔ                ⁿ-sʷɛ́ɛ̀n      Nùᵑgɛ̀  lá  Numi    AUX.T2    say     C.L   FOC child    3SG.H   AGR.AUX.T2  N-sell        Nuga   C-Q                              T.HL               # ‘Numi said that the childFOC betrayed Nuga’ = Numi said: “the childFOC (he) betrayed Nuga”  38  c. Relativization Nùmí nɔ́ʔ          tʃúp   ᵐbʉ̀    mɛ́n  zə̀       á         nɔ́ɔ̀ʔ              ⁿ-sʷɛ́ɛ̀n     Nùᵑgɛ̀            lá …  Numi AUX.T2    say     C.L    child C.CL1 3SG.H  AGR.AUX.T2  N-sell       Nuga   C.-Q                      T.HL                # ‘Numi said that the child that betrayed Nuga …’ = Numi said: “the child that [he] betrayed Nuga …” (36) Object extraction a. Wh-movement Nùmí       nɔ́ʔ-bɛ́ttə́   ᵐbʉ̀   á     wʉ́   Wàtɛ̀ɛ́t      nɔ́ɔ̀ʔ                 ⁿ-sʷɛ́ɛ̀n        (í)    á Numi      Aux-ask    C.L   FOC WH   Watat.H   AGR.AUX.T2   N-AGR.sell   3SG.H      C.Q.H                               T.HL      V.HL            # ‘Numi asked who Watat betrayed’ = Nuga asked: “who did Watat betrayed (him/her)” b. Focus movement Nùmí   nɔ́ʔ          tʃúp    ᵐbʉ̀    á      mɛ́n    Wàtɛ̀ɛ́t      nɔ́ɔ̀ʔ                ⁿ-sʷɛ́ɛ̀n         (í)         lá  Numí   AUX.T2    say     C.L  FOC   child   Watat.H   AGR.AUX.T2   N-AGR.sell    3SG.H   C-Q                     T.HL                 V.HL     # ‘Numi said that the childFOC Watat betrayed [him/her]’ = Numi said “the childFOC Watat betrayed (him/her)” c. Relativization Nùmí nɔ́ʔ         tʃúp  ᵐbʉ̀  mɛ́n      zə̀       Wàtɛ̀ɛ́t     nɔ́ɔ̀ʔ               ⁿ-sʷɛ́ɛ̀n        í      lá … Numi AUX.T2  say   C.L   child    C.CL1  Watat.H  AGR.AUX.T2   N-AGR.sell.  3SG.H C.-Q                           T.HL                   V.HL       # ‘Numi said that the child that Watat betrayed …’ = Numi said: “the child that Watat betrayed (him/her)…”  To summarize, the tail of A′-chains is always spelled out as a resumptive pronoun in Medumba non-root clauses and, A′-movement is always to the leftmost edge of the clause. Movement to the edge of an apparent embedded clause has root clause interpretation, involves optional resumption and is interpreted as a direct quote (see Collins and Branigan 1997). 39  2.2.3 Diagnostic 3: A′-movement supports reconstruction 2.2.3.1 Reconstruction: the general picture  Reconstruction refers to a context in which in an A′-chains, with regard to binding principles, the head of the chain is interpreted in its original position as if movement has not taken place (Chomsky 1976, Huang 1993). The examples below illustrate reconstruction with regard to Principle A of Binding Theory (Chomsky 1981) and show that Principle A holds despite the fact that the reflexive is not c-commanded.  (37) Principle A: An anaphor9 must be bound in its Domain (Sportiche et al. 2014:168)   (38a) is a well-formed sentence even though the reflexive himself is not c-commanded by its antecedent John. (38b) is ambiguous and shows that either the matrix subject DP John or the embedded subject DP Bill can be the antecedent of the reflexive himself.  The ambiguity of (38b) is resolved when the antecedent matrix DP and the antecedent embedded DP differ in gender feature. This is confirmed in (38c) where the matrix subject DP John is the antecedent of the reflexive and in (38d) where the embedded subject DP Mary is the antecedent of that reflexive. (38) a. Which pictures of himself did John like which pictures of himself?   b. Which pictures of himself did John think Bill saw which pictures of himself?   c. Which pictures of himself did John think Mary saw which pictures of himself?   d. Which pictures of herself did John think Mary saw which pictures of herself?           [Huang 1993: 3, p. 103]  9 Anaphors includes reflexives and reciprocals 40  Taken together, the examples in (38) establish that the head of an A′-chain which contains the reflexive is interpreted as if movement has not occurred; that is the reflexive reconstructs to its original position. 2.2.3.2 Reconstruction: the picture in Medumba  A′-movement reconstructs in Medumba. The examples in (39), (40) and (41) illustrate subject wh-movement, subject focus movement and subject relativization respectively. The (a) examples provide the baseline sentences which establish that Principle A holds in Medumba. The reflexive tʰú-vʉ́dl-í ‘himself/herself’ is co-indexed with and is c-commanded within its local domain by the antecedent DP Nùᵑgɛ̀, making biding possible.    The (b) examples show that Principle A still holds even though the subject reflexive tʰú-vʉ́dl-í (himself/herself) has been moved and now a resumptive pronoun surfaces at the extraction site.  (39) Subject wh-movement a. Nùᵑgɛ̀j   tʃúp  ᵐbʉ̀ sə́vʰə̀ə́   tʰú-vʉ́dl-íj  vʰʉ̀ʉ́     sí        Nuga   say   C.L   picture   head-body-3SG fall.H down            ‘Nuga said that the picture of himself fell down’ b. á       jíìt            sə́vʰə̀ə́   tʰú-vʉ́dl-íj          Nùᵑgɛ̀j   tʃúùp     ᵐbʉ̀   áj        vʰʉ̀ʉ́     sí        á       FOC  SG.which   picture   head-body-3SG  Nuga   AGR.say  C.L  3SG.H  fall.H    down  C.Q.H                 v.HL       ‘Which picture of himself/herself did Nuga say that [it] fell down?’ (40) Subject focus movement   a. Nùᵑgɛ̀j   tʃúp     ᵐbʉ̀   á       sə́vʰə̀ə́    tʰú-vʉ́dl-íj         nɔ́ɔ̀ʔ               ᵐ-vʰʉ́ʉ̀    sí             Nuga   say       C.L    FOC   picture   head-body-3SG   AGR.AUX.T2   N-fall     down                           HL       ‘Nuga said that [picture of himself/herself]FOC fell down’ b. á       sə́vʰə̀ə́    tʰú-vʉ́dl-íj       Nùᵑgɛ̀j   tʃúùp     ᵐbʉ̀     áj         nɔ́ɔ̀ʔ              ᵐ-vʰʉ́ʉ̀   sí     lá       FOC   picture   head-body-3SG   Nuga   AGR.say   C.L    3SG.H   AGR.AUX.T2   N-fall   down  C.-Q            v.HL           T.HL        ‘[Picture of himself/herself]FOC Nuga said that [it] fell down’ 41  (41) Subject relativization a. Nùᵑgɛ̀j   tʃúp     ᵐbʉ̀ sə́vʰə̀ə́   tʰú-vʉ́dl-íj  vʰʉ̀ʉ́     sí       Nuga   say       C.L   picture   head-body-3SG fall.H down           ‘Nuga said that the picture of himself fell down’  b. sə́vʰə̀ə́    tʰú-vʉ́dl-íj          sə̀        Nùᵑgɛ̀   tʃúùp     ᵐbʉ̀     áj          vʰʉ̀ʉ́       sí       lá …      picture   head-body-3SG   C.CL5  Nuga   AGR.say  C.L   3SG.H    fall.H    down   C.-Q              v.HL   ‘The picture of himself/herself that Nuga said that [it] fell down …’  The same pattern is found with A′-movement from the object position. As shown in the following examples, Principle A continues to hold even though the object reflexive has been extracted and the sentence is interpreted as if movement has not occurred.  (42) Object wh-movement a. Nùᵑgɛ̀j     kɛ̀  á jíìt         sə́vʰə̀ə́ tʰú-vʉ́dl=íj          á      Nuga     choose FOC SG.which    picture head-body=3SG    C.Q.H      ‘Nuga chose which picture of himself/herself?’ b. á jíìt         sə́vʰə̀ə́ tʰú-vʉ́dl-íj     Nùᵑgɛ̀j     kɛ́ɛ̀  ___  á     FOC SG.which    picture head-body-3SG    Nuga     AGR.choose  C.Q.H              v.HL    ‘Which picture of himself/herself did Nuga choose?’ (43) Object focus movement a. Nùᵑgɛ̀j     kɛ̀  á sə́vʰə̀ə́ tʰú-vʉ́dl=íj        Nuga     choose FOC picture head-body=3SG                      ‘Nuga chose [picture of himself/herself]FOC b. á sə́vʰə̀ə́ tʰú-vʉ́dl-íj     Nùᵑgɛ̀j     kɛ́ɛ̀  ___   lá      FOC picture head-body-3SG    Nuga     AGR.choose   C.-Q             v.HL      ‘Picture of [himself/herself]FOC Nuga chose’ (44) Object relativization a. Nùᵑgɛ̀j     kɛ̀ɛ́  sə́vʰə̀ə́ tʰú-vʉ́dl=íj        Nuga     choose.H picture head-body=3SG                      ‘Nuga chose [picture of himself/herself]FOC  42  b. sə́vʰə̀ə́ tʰú-vʉ́dl-íj     sə̀     Nùᵑgɛ̀j      kɛ́ɛ̀ ___ lá vʰʉ̀ʉ́ sí     picture head-body-3SG    C.CL5    Nuga     AGR.choose C.-Q fall.H down             v.HL    ‘The picture of himself/herself that Nuga chose fell down’  The fact that principle A continues to hold after the reflexive has been extracted confirms that A′-movement reconstructs in Medumba. 2.2.4 Diagnostic 4: A′-movement conditions Strong and Weak Crossover 2.2.4.1 Strong & Weak Crossover: the general picture  In A′-movement contexts, the trace of the moved element is subject to Strong Crossover (Postal 1971, 1993, Warsow 1972, Chomsky 1981, Safir 1986, 1996, McCloskey 2006). (45) Strong Crossover   The trace of movement to an A′-position may not be anaphorically linked with a c- commanding pronoun (McCloskey 2006: 101)  Strong Crossover gives rise to a Condition C effect.  (46) Condition C: Non-pronominal DPs and wh-traces must not be bound by an element in an A-position (McCloskey 2006: 101).      In the example below, the trace of the moved wh-XP in the embedded subject position is c-commanded and co-indexed with the pronoun she, giving rise to a Condition C violation.   (47) *Who did shei claim [ti had arrived earliest]?         [McCloskey 2006: 23, p.101]  Weak Crossover arises when A′-movement crosses over a c-commanding phrase containing a pronoun it binds. This is illustrated in the following examples. (48) *Whoi does hisi mother like ___? (49) *Which boyi did you say hisi mother like ___? 43  In the above examples, even though there is no Condition C violation, the bound reading of the pronoun is excluded. These are known as Weak Crossover configurations because the judgment about the unavailability of the bound reading of the pronoun is less robust than in Strong Crossover configurations (Wasow 1972; Lasnik and Stowell 1991; Moulton 2013). 2.2.4.2 Strong & Weak Crossover: the picture in Medumba  A′-movement in Medumba conditions Strong and Weak Crossover. With regard to Strong Crossover, (50) is the baseline sentence which establishes that Condition C holds in Medumba. That is, an R-expression such as Nùᵑgà must be free, and so, cannot be co-indexed with a c-commanding antecedent such as bʰúʔtʰú ‘some/the idiots’.  (50) *bhúʔthúi tʃúp mbʉ̀ mʉ̀ àʔ ʒʷí  Nùᵑgɛ̀i  idiot  say  C.L 1SG.L IRR kill  Nuga   [*the idioti said that I would kill Nugai] Even after A′-movement has taken place, which creates a Strong Crossover environment, Condition C continues to hold as shown in (51a) for subject wh-movement, (51b) for subject focus movement and (51c) for subject relativization. (51) Subject extraction a. Wh-movement   *á wʉ́i bhúʔthúi tʃúùp  mbʉ̀ ài àʔ ʒʷí Wàtɛ̀t  à   FOC WH idiot  AGR.say  C.L 3SG.L IRR kill Watat  C.Q.L      V.HL   [*whoi did the idioti say that [hei] would kill Watat] b. Focus movement  *á Nùᵑgɛ̀i bhúʔthúi tʃúùp  mbʉ̀ ài àʔ ʒʷí Wàtɛ̀t lá   FOC Nuga idiot  AGR.say  C.L 3SG.L IRR kill Watat C.-Q      V.HL   [*NugaFOCi the idioti said that [hei] would kill Watat]  44  c. Relative clause  *Nùᵑgɛ̀i zə̀ bhúʔthúi    tʃúùp mbʉ̀ ài àʔ ʒʷí  Wàtɛ̀t lá     Nuga  C.CL1 idiot     AGR.say  C.L 3SG.L IRR kill Watat  C.-Q           V.HL   [*Nugai that the idioti said that I would kill Watat] The same pattern obtains when the object is extracted as illustrated below in (52a) for object wh-movement, in (52b) for object focus movement, and in (52c) for object relativization. (52) Object extraction a. Wh-movement   *á wʉ́i bhúʔthúi tʃúùp  mbʉ̀ mʉ̀ àʔ-ʒʷí ___ á     FOC WH idiot  AGR.say  C.L 1SG.L IRR-kill C.Q.H      V.HL   [*whoi did the idioti said that I would kill] b. Focus movement  *á Nùᵑgɛ̀i bhúʔthúi tʃúùp  mbʉ̀ mʉ̀ àʔ-ʒʷí  ___ lá     FOC Nuga idiot  AGR.say  C.L 1SG.L IRR-kill  C.-Q       V.HL  [*NugaFOCi the idioti said that I would kill] c. Relative clause  *Nùᵑgɛ̀i zə̀ bhúʔthúi tʃúùp  mbʉ̀ mʉ̀ àʔ-ʒʷí ___ lá …     Nuga  C.CL1 idiot  AGR.say  C 1SG.L IRR-kill C.-Q       V.HL   [*Nugai that the idioti said that I would kill … ]  Weak crossover arises when A′-movement crosses over a c-commanding phrase containing a pronoun it binds.  In the following examples, the A′-extracted XPs cross over a possessive DP that contain a pronoun. As the extracted phrase binds the pronoun, this results in ungrammaticality as illustrated in (53a) for wh-movement, in (53b) for focus movement, and in (53c) for relativization.   45  (53) a. wh-movement  *á wʉ́j ʃʷín-íj   àʔ-tʃɔ̀ʔdə̀ ___ à  FOC WH friend-3SG.POSS.C1 IRR-greet  C.Q.L   [Lit.: Whoi will hisi friend greet?] b. Focus movement  *á Nùᵑgɛ̀j ʃʷín-íj   àʔ-tʃɔ̀ʔdə̀ ___ lá  FOC NUGA friend-3SG.POSS.C1 IRR-greet  C.-Q  [Lit.: Nugai  hisi friend will greet.] c. Relativization  *mɛ́nj zə̀ ʃʷín-íj   àʔ-tʃɔ̀ʔdə̀ ___ lá …   child C.CL1 friend-3SG.POSS.C1 IRR-greet  C.-Q   [Lit.: the childi  that hisi friend will greet … ] Weak Crossover environments are licit in Medumba when the tail of the chain is spelled out as a resumptive pronoun.  (54) a. wh-movement  á wʉ́j ʃʷín-íj   àʔ-tʃɔ̀ʔdì íj á   FOC WH friend-3SG.POSS.C1 IRR-greet 3SG.H C.Q.H   Lit.: Whoi will hisi friend greet [him]? b. Focus movement  á Nùᵑgɛ̀j ʃʷín-íj   àʔ-tʃɔ̀ʔdì íj lá   FOC WH friend-3SG.POSS.C1 IRR-greet 3SG.H C.-Q  Lit.: Nugai  hisi friend will greet [himi] c. Relative clause  mɛ́nj zə̀ ʃʷín-íj   àʔ-tʃɔ̀ʔdì íj lá  child C.CL1 friend-3SG.POSS.C1 IRR-greet 3SG.H C.-Q  Lit.: the childi  that hisi friend will greet [himi] 46  2.2.5 Diagnostic 5: A′-movement is island-sensitive 2.2.5.1 Island-sensitivity: the general picture Islands are configurations that render otherwise legitimate syntactic dependencies illicit (Boeckx 2007). They are considered to be a standard diagnostic property of A′-movement. The notion of islandhood originates from Ross (1967) and includes complex DPs, adjoined clauses, coordinate structures, ‘left branches’, sentential subjects, and embedded interrogative clauses. Chomsky (1973, 1986) further investigates these domains and proposes that they are constrained by the principle of subjacency10. Concretely, subjacency bars movement from crossing two bounding nodes such as the top clausal S and NP (modern IP/TP and DP); these are treated as barriers in Government and Binding Theory (Chomsky 1986, Haegeman 1994) and as phases in Minimalism (Chomsky 2000, 2001). Although the literature distinguishes between strong or absolute islands and weak or selective islands (Boeckx 2007, Szabolcsi 2006), there is no clear-cut distinction between the two. Strong islands prohibit extraction of any kind and include complex DPs (relative clauses and complement clauses), adjunct clauses and coordinate structures as illustrated in the examples below adapted from Szabolci 2006: 482-483). (55) Complex DP (with relative clause)  a. *Which kid must you call [the teacher who punished ___ ]?  b. *What size shoes did you call [the man who wears ___ ]?  c. *Where must you call [the teacher who put the book ___ ]?  d. *How did you call [the man who behaved ___ ]?   10 A condition that bars movement from crossing in one step, two or more bounding nodes. 47  (56) Complex DP (with complement clause)  a. *Which man did you hear [the rumor that my dog bit ___ ]?  b. *What size shoes did you hear [the rumor that I wear ___ ]?  c. *Where did you hear [the rumor that I put the book ___ ]?  d. *How did you hear [the rumor that I behaved ___ ]? (57) Adjunct island  a. *About which topic did you leave [because Mary talked ___ ]?  b. *Which topic did you leave [because Mary talked about ___ ]?  c. *How did you leave [because Mary behaved ___ ]? (58) Coordinate structure  a. *Which man did you invite [Mary and ___ ]?  b. *Which man did you invite [ ___ and Mary]? With regard to weak islands –– which include embedded interrogative clauses –– extraction of arguments is permitted, but extraction of adjuncts is prohibited (see a.o. Huang 1982; Lasnik and Saito 1984, 1992; Chomsky 1986; Szabolci 2006; Boeckx 2007). As illustrated in (59), they show an argument/adjunct asymmetry, in that extraction is licit from an argument position, but not from an adjunct position.  (59) Wh-island  a. *About which topic did John ask [who was talking ___]?  b. *?Which topic did John ask [who was talking about ___]?  c. *How did John ask [who behaved ___]? 48  The examples below illustrate cases where an argument can be extracted (the (a) examples), an adjunct such as how, why and when cannot be extracted (the (b) examples) (See Szabolcsi 2006 for discussion).  (60) a. Which problem did John ask [how to phrase ___ ]?  b. *How did John ask [which problem to phrase ___ ]? (61) a. Which man did John ask [whether to fire ___ ]?  b.i *Why did John ask [whether to fire him ___ ]?  b.ii ??When did John ask [whether to fire him ___ ]?  Nevertheless, as pointed out by Ross (1967) there are at least three contexts where islands can be circumvented (see also Boeckx 2007). The first includes configurations where the tail of the A′-chain is spelled out as a resumptive pronoun (62). (62) a. *Which woman did John laugh [after Bill kissed __]  b. Which woman did John laugh [after Bill kissed her] [Boeckx 2007: 5-6, p. 155] The second context which circumvents islandhood involves pied-piping where the moved constituent takes along the whole island. In this case, movement is not strictly speaking out of an island as it is the entire island that moves. (63) a. *Whose did you buy [whose book]?  b. [Whose book] did you buy [whose book]?  [Boeckx 2007: 7-8, p. 155] The third context where islandhood is circumvented involves sluicing11. Ross (1969) noted that configurations in which islands are part of the ellipsis site are acceptable; see (64).  11 Sluicing is a type of elliptical construction in which the sentential portion of a content question is deleted, leaving only the remnant Wh-XP (Ross 1969, Merchant 2001). 49  (64) a. *That he will hire someone is possible, but I will not divulge who that he will hire is          possible.   b. That he will hire someone is possible, but I will not divulge who <that he will hire ___  is possible>        [Ross 1969: 73, p. 277)] 2.2.5.2 Island-sensitivity: the picture in Medumba Extraction out of an island is prohibited in Medumba if the tail of the A′-chain is a gap. (65) shows that a wh-XP12 cannot be extracted out of the embedded interrogative CP […ᵐbʉ́ʉ̀ Wàtɛ̀ɛ́t nɔ́ɔ̀ʔ ⁿ-tʃɔ́ɔ̀ʔdə́ __ ] ‘…whether Watat greeted __’. (66) shows that if a wh-XP is extracted out of the adjunct CP […káà Wàtɛ̀ɛ́t tʃɔ́ɔ̀ʔdə̀ __ ] ‘before Watat greeted’, the resulting structure is also ungrammatical.  Similarly, if the wh-XP is extracted out of a complex DP (in this case a relative clause) [mɛ́n zə̀ à nɔ́ɔ̀ʔ ⁿ-tʃɔ́ɔ̀ʔdə́ __] ‘the child who greeted __’ the result is ill-form as shown in (67). The coordinate structure constraint stipulates that extraction out of coordinated XPs is prohibited. (68a) and (68b) show that movement of a wh-XP out of either branch of a coordinated structure is illicit in Medumba.     (65) Wh-island *á wʉ́ Nùᵑgɛ̀  nɔ́ɔ̀ʔ   ᵐ-bɛ́ttə́  [ᵐbʉ́ʉ̀  FOC WH Nuga  AGR.AUX.T2 N-ask     C.HL         T.HL     Wàtɛ̀ɛ́t nɔ́ɔ̀ʔ           ⁿ-tʃɔ́ɔ̀ʔdə́  ___  ]   á     Watat.H AGR.AUX.T2 N-AGR.greet   C.Q.H        HL       V.HL     ‘*Who did Nuga ask whether Watat greeted ___?    12 Only wh-movement is illustrated here but the same pattern holds for focus movement and relativization 50  (66) Adjunct Island *á    wʉ́    Nùᵑgɛ̀   nɔ́ɔ̀ʔ      nɛ́ɛ̀n    ⁿ tɔ́n      [káà      Wàtɛ̀ɛ́t tʃɔ́ɔ̀ʔdə̀ ___ ] à?   FOC WH   Nuga    AGR.AUX.T2 N-go   market   before  Watat.H AGR.greet  C.Q.L          T.HL      V.HL       ‘*who did Nuga go to the market before Watat greeted?’ (67) Complex DP (Relative clause) *á     wʉ́   Nùᵑgɛ̀   nɔ́ɔ̀ʔ               ⁿ-dʒʉ́n  [mɛ́n   zə̀       à          nɔ́ɔ̀ʔ                 ⁿ-tʃɔ́ɔ̀ʔdə́ ___  ] á?   FOC  WH   Nuga   AGR.AUX.T2  N-see    child  C.CL1 3SG.L   AGR.AUX.T2   N-AGR.greet    C.Q.H         T.HL        T.HL      V.HL      ‘*who did Nuga see the child that greeted? (68) a. The coordinate structure constraint 1 *á wʉ́ Nùᵑgɛ̀  nɔ́ɔ̀ʔ  ⁿ-tʃɔ́ɔ̀ʔdə́ [Wàtɛ̀t   búù   ___]  à?    FOC WH Nuga  AGR.AUX.T2 N-AGR.greet  Watat  PL             C.Q.L     T.HL       V.HL    *who did Nuga greet Watat and –––? b. The coordinate structure constraint 2 *á wʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ⁿ-tʃɔ́ɔ̀ʔdə́     [ ___ búù Wàtɛ̀t]  à?    FOC WH Nuga AGR.AUX.T2 N-AGR.greet  PL Watat  C.Q.L    T.HL       V.HL     *who did Nuga greet ––– and Watat?   In Medumba A′-chains, crossing of an island boundary is prohibited if the tail of the A′-chain is a gap.  However, if the tail of the A′-chain is a resumptive pronoun, extraction out of an island is permitted. This is illustrated in (69) where the same island domains that were deemed illicit with a gap are now licit with a resumptive pronoun.      (69) a. wh-island á wʉ́ Nùᵑgɛ̀  nɔ́ɔ̀ʔ  ᵐ-bɛ́ttə́  [ᵐbʉ́ʉ̀  FOC WH Nuga  AGR.AUX.T2 N-ask     C.HL     T.HL     Wàtɛ̀ɛ́t nɔ́ɔ̀ʔ           ⁿ-tʃɔ́ɔ̀ʔd=í  á     Watat.H AGR.AUX.T2 N-AGR.greet=3SG.H C.Q.H        T.HL      V.HL     ‘*Who did Nuga ask if Watat greeted [him/her]?’ 51   b. Adjunct island á      wʉ́    Nùᵑgɛ̀  nɔ́ɔ̀ʔ     nɛ́ɛ̀n   ⁿ tɔ́n     káà      Wàtɛ̀ɛ́t tʃɔ́ɔ̀ʔdì=í         á?  FOC  WH   Nuga   AGR.AUX.T2 N-go   market before  Watat.H AGR.greet=3SG.H   C.Q.H         T.HL      V.HL    ‘*Who did Nuga go to the market before Watat greeted [him/her]?’  c. Complex DP (Relative clause) á    wʉ́   Nùᵑgɛ̀  nɔ́ɔ̀ʔ               ⁿ-dʒʉ́n  mɛ́n  zə̀        à         nɔ́ɔ̀ʔ              ⁿ-tʃɔ́ɔ̀ʔd=í                 á?  FOC WH  Nuga  AGR.AUX.T2 N-see  child  C.CL1  3SG.L AGR.AUX.T2  N-AGR.greet=3SG.H C.Q.H     T.HL                  T.HL           V.HL      ‘*Who did Nuga see the child that greeted [him/her]?’  d. The coordinate structure constraint 1 á wʉ́ Nùᵑgɛ̀  nɔ́ɔ̀ʔ  ⁿ-tʃɔ́ɔ̀ʔdə́ Wàtɛ̀t búù   jí  á?  FOC WH Nuga  AGR.AUX.T2 N-AGR.greet Watat  PL 3SG.H  C.Q.H     T.HL       V.HL    *Who did Nuga greet Watat and [him/her]?  e. The coordinate structure constraint 2 á wʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ⁿ-tʃɔ́ɔ̀ʔd [=í  búù Wàtɛ̀t]  à?  FOC WH Nuga AGR.AUX.T2 N-AGR.greet=3SG.H PL Watat  C.Q.L    T.HL       V.HL *Who did Nuga greet [him/her] and Watat?  To sum up, A′-movement across an island boundary is prohibited in Medumba if the tail of the A′-chain is a gap. But islands can be circumvented if the tail of the chain is a resumptive pronoun. 2.2.6 Diagnostic 6: A′-movement conditions A′-agreement 2.2.6.1 A′-agreement: the general picture A′-agreement is the reflex of Agree (see chapter 4 for discussion) which surfaces within the V-domain or C-domain when there is A′-extraction. In Kilega for instance, A′-agreement takes the form of concordial agreement. The morpheme that surfaces within the verbal complex when there is wh-movement agrees in φ-features with the moved wh-XP. A′-agreement only happens when 52  the wh-XP has moved; this is confirmed by the fact that in-situ contexts do not exhibit A′-agreement. This is illustrated in (70).  (70) Kilega A′-agreement   a. bá-bo  bí-kulu     b-á-kás-íl-é  mwámí  bi-kí mu-mw-ílo           2-that 2-woman 2SM-V-give-PFV-FV     1chief   8-what 18-3-village              lit: those women gave what the chief in the village  [Kinyalolo 1991: 13a, p. 21] b. bi-kí  bí-á-kás-íl-é                 bá-bo bí-kulu     mw-ámí    ___ mu-mw-ílo      8-what 8.AGR-V-give-PFV-FV 2-that 2-woman 1-chief  18-3-village    ‘what did those women give the chief in the village?’  [Kinyalolo 1991: 13b, p. 21] 2.2.6.2 A′-agreement: the picture in Medumba  A′-agreement is realized in Medumba as an HL tonal melody that overwrites the lexical tone of main verbs, temporal auxiliaries as well as aspectual auxiliaries. With root CPs, A′-agreement shows a subject/object asymmetry: with subject extraction, agreement is only with T; but with object extraction agreement is with V and T. (71) shows that when the subject undergoes A′-movement, the HL melody surfaces only on the temporal auxiliary; this is illustrated in (71a.i) for wh-movement, (71b.i) for focus movement, and in (71c.i) for relativization. Failure of A′-agreement in these contexts leads to ungrammaticality (see the examples in (71a.ii), (71b.ii) and (71c.ii)). (71) Subject extractions  a.i wh-movement á wʉ́ nɔ́ɔ̀ʔ  ⁿ-dʒʉ́n  Nùᵑgɛ̀  à?  FOC WH AGR.AUX.T2 N-see  Nuga  C.Q.L T.HL  V.H               ‘Who saw Nuga?’  a.ii *á wʉ́ nɔ́ʔ  jʉ́n Nùᵑgɛ̀  à?    FOC WH AUX.T2 see Nuga  C.Q.L T.H  V.H                ‘Who saw Nuga?’ 53   b.i focus movement á Wàtɛ̀ɛ́t     nɔ́ɔ̀ʔ  ⁿ-dʒʉ́n  Nùᵑgɛ̀   FOC Watat.H AGR.AUX.T2 N-see  Nuga   T.HL  V.H               ‘Watat saw Nuga’   b.ii *á Wàtɛ̀ɛ́t     nɔ́ʔ  jʉ́n Nùᵑgɛ̀     FOC Watat.H AUX.T2 see Nuga   T.H  V.H               ‘Watat saw Nuga’   c.i Relativization  má-ⁿdʒùúm zə̀ à nɔ́ɔ̀ʔ  ⁿ-dʒʉ́n  Nùᵑgɛ̀  lá …  SG-male.H C.CL1 3SG.L AGR.AUX.T2 N-see  Nuga  C.-Q T.HL   V.H          [The boy that saw Nuga…]’  c.ii *má-ⁿdʒùúm zə̀ à nɔ́ʔ  jʉ́n Nùᵑgɛ̀  lá …    SG-male.H C.CL1 3SG.L AUX.T2 see Nuga  C.-Q T.H  V.H             [The boy that saw Nuga…]’ Object extraction is illustrated in (72) where the HL melody surfaces both on the verb and the temporal auxiliary. This is true for wh-movement (72a.i), focus movement (72b.i) and relativization (72c.i).  Failure of A′-agreement with object-extraction leads to ungrammaticality (cf. (72a.ii), (72b.ii) and (72c.ii)). (72) Object extractions  a.i Wh-movement  á wʉ́  Wàtɛt  nɔ́ɔ̀ʔ  ⁿ-dʒʉ́ʉ̀n  á?  FOC WH Watat  AGR.AUX.T2 N-AGR.see C.Q.H T.HL       V.HL        ‘Who did Watat see?’   a.ii *á wʉ́  Wàtɛ́t  nɔ́ʔ  jʉ́n  á?     FOC WH Watat  AUX.T2 see C.Q.H T.H        V.H        ‘Who did Watat see?’ 54   b.i Focus movement  á Nùᵑgɛ̀ Wàtɛ́t nɔ́ɔ̀ʔ  ⁿ-dʒʉ́ʉ̀n lá FOC Nuga Watat AGR.AUX.T2 N-AGR.see C.-Q         T.HL       V.HL      ‘NugaFOC Watat saw’   b.ii *á Nùᵑgɛ̀ Wàtɛ́t nɔ́ʔ  jʉ́n lá    FOC Nuga Watat AUX.T2 see C.-Q         T.H    V.H      ‘NugaFOC Watat saw’ c.i Relativization   má-ⁿdʒùúm zə̀ Wàtɛ́t  nɔ́ɔ̀ʔ  ⁿ-dʒʉ́ʉ̀n lá …  SG-male.H C.CL1 Watat  AGR.AUX.T2 N-AGR.see C.-Q T.HL       V.HL              ‘The boy that Watat saw…’  c.ii *má-ⁿdʒùúm zə̀ Wàtɛ́t  nɔ́ʔ  jʉ́n lá …  SG-male.H C.CL1 Watat  AUX.T2 see C.-Q T.H   V.H              ‘The boy that Watat saw…’ To summarize, it appears from what precedes that:  o A′-movement is associated with a gap or a resumptive pronoun;  o A′-movement permits long-distance dependencies;  o A′-movement supports reconstruction; o A′-movement conditions Strong and Weak Crossover;  o A′-movement is island-sensitive;  o A′-movement conditions A′-agreement.  2.3 Five arguments in favour of interpretation-driven movement In this section, I present the different arguments in favour of interpretation-driven movement.  These arguments support the idea that there is structural and semantic difference between in-situ and ex-situ construals as predicted by interpretation-driven movement. The arguments are: 55  o ex-situ wh/focus requires exhaustivity (§2.3.1); o in-situ wh/focus doesn’t require exhaustivity (§2.3.2); o question/answer pairs are conditioned by the in-/ex-situ contrast (§2.3.3); o fragment answers are conditioned by the in-/ex-situ contrast (§2.3.4); o subjects are (predictably) different (§2.3.5). 2.3.1 Argument 1: ex-situ wh/focus requires exhaustivity The diagnostics presented in this section establish that ex-situ wh-questions and foci are exhaustive in Medumba. These diagnostics are: (i)  the entailment test; (ii)  the use of universal quantifiers; (iii) the use of additive particles;  (iv) and the unavailability of function denotations (natural functions and random functions also called pair list reading).  The first three diagnostics follow Kiss’s (1998, 2010) analysis of exhaustivity in Hungarian and the last one is a new diagnostic that I have developed for Medumba. In the following subsections, I first show the mechanics of the derivation of ex-situ wh-questions and foci. As predicted by interpretation-driven movement, movement is driven by interpretation and involves the presence of a covert exhaustive operator at C which marks any constituent it associates with as exhaustive. Then I show the diagnostics that confirms the exhaustiveness of ex-situ wh-questions and foci.   2.3.1.1 The mechanics of the derivation In (73), I show a step-by-step derivation of an ex-situ object wh-question in Medumba. The ingredients for this consist of: 56  o A numeration: a set of lexical items tokens needed for the derivation and which must be exhausted by the end of the derivation o Merge: an operation which selects a pair of syntactic objects and combine them into new syntactic objects. It consists of either selecting and combining syntactic objects directly from the numeration (first merge or external merge) or by selecting and combining syntactic objects contain within others (internal merge or movement).      (73) Step-by-step derivation of an ex-situ (exhaustive) wh-question Numeration: {áFOC; kʉ́WH; NùmíN; nɔ́ɔ̀ʔT2; ᵑ-kɛ́ɛ̀V; áC.Q; vø; Dø; CøExh; Tø1} a. vP phase I. Merge <V; Wh-XP> The wh-XP is merged as complement of V [VP [V ⁿʤʉ́ʉ̀n ] [Wh-XP á wʉ́]   II. Merge <DP; vP> and <v; VP>  The phase head v merges with its complement VP and the subject DP is merged at Spec-vP [vP [DP [Nùmí ] ] [vP [v ∅ ] [VP [V ⁿʤʉ́ʉ̀n] [Wh-XP á wʉ́] 57   III. Move Wh-XP and adjoin to vP  The wh-XP moves and adjoins to vP at the phase edge so that it remains visible for subsequent merge. The complement of the phase (VP) is sent to transfer. [vP [Wh-XP á wʉ́] [vP [DP [Nùmí ] ] [vP [v ∅ ] [VP [V ⁿʤʉ́ʉ̀n] [<Wh-XP>]  b. CP phase I. Merge <T2; vP> [TP2 [T2 nɔ́ɔ̀ʔ] [vP [Wh-XP á wʉ́] [vP [DP [Nùmí ] ] [vP [v ∅ ] [VP [V ⁿʤʉ́ʉ̀n] [<Wh-XP>] 58   II. Merge <T1; TP2> T1 merges with TP2 and the subject DP moves to Spec-T1 [TP1 [DP [Nùmí ]] [TP1 [T1 ø ] [TP2 [T2 nɔ́ɔ̀ʔ] [vP [Wh-XP á wʉ́] [vP [<DP>] [vP [v ∅ ] [VP [V ⁿʤʉ́ʉ̀n] [<Wh-XP>]  III. Merge <CøExh; TP1> C1 (øExh) merges with TP1 and the wh-XP moves to Spec-C1. TP is sent to transfer. [CP1 [Wh-XP á wʉ́] [CP1 [C1 øExh] [TP1 [DP [Nùmí ]] [TP1 [T1 ø ] [TP2 [T2 nɔ́ɔ̀ʔ] [vP [<Wh-XP>] [vP [<DP>] [vP [v ∅ ] [VP [V ⁿʤʉ́ʉ̀n] [<Wh-XP>] 59   IV. Merge <CQ; CP1> C2 hosting the Q-operator merges with CP1 [CP2 [C2 á] [CP1 [Wh-XP á wʉ́] [CP1 [C1 øExh] [TP1 [DP [Nùmí ]] [TP1 [T1 ø ] [TP2 [T2 nɔ́ɔ̀ʔ] [vP [<Wh-XP>] [vP [<DP>] [vP [v ∅ ] [VP [V ⁿʤʉ́ʉ̀n] [<Wh-XP>]  V. Move CP1 to Spec-C2 CP1 moves to Spec-C2 for linearization and the Q-particle is stranded in final position 60        This analysis predicts that overt exhaustification marking should be incompatible with ex-situ construals. This is indeed the case as an overt exhaustive particle such as ⁿdɔ̀ɔ́ʔ ‘only’ is illicit with ex-situ wh- and focus construals in Medumba as shown in (74a) and (74b) respectively13. (74) a. *[á ⁿdɔ̀ɔ́ʔ kʉ́]  Nùmí  nɔ́ɔ̀ʔ  ᵑ-kɛ́ɛ̀  á?     FOC only WH Numi  AGR.AUX.T2 N-AGR.fry C.Q.H    [?what only did Numi fry]  b. *[á ⁿdɔ̀ɔ́ʔ ⁿʤʷɛ́n]  Nùmí  nɔ́ɔ̀ʔ  ᵑ-kɛ́ɛ̀  lá     FOC only chips  Numi  AGR.AUX.T2 N-AGR.fry C.-Q    [?only chips Numi fried] The overt exhaustive operator and the wh-XP form a single constituent in Medumba. When this constituent is merged with V in the complement of V position, it does not need to move within the vicinity of the covert exhaustive operator at C to be interpreted as exhaustive.  Instead, it is interpreted as exhaustive in place. In fact, interpretation-driven movement blocks movement if the resulting sentence will not get a different interpretation. Interpretation-driven movement favours  13 I predict the overt exhaustification operator to be compatible with in-situ focus. Details are given in §2.3.2.5)  61  the most economical derivation. Thus, the non-movement operation is preferable unless the movement operation results in an interpretation that differs from the non-movement one. 2.3.1.2 Ex-situ wh-/focus blocks entailment relations  Entailment is a relationship between statements where one statement is true when it logically follows from one or more other statements, so that in every situation where p is true then q is also true. For instance, if the proposition in (75) is true, the proposition in (76) is true as well. (75) Susan and Tom ate carrots (76) Susan ate carrots In the above examples, the statement Susan and Tom ate carrots entails the statement Susan ate carrots. Exhaustivity marking perturbs the entailment relation. For example, the statement in (77) does not entail (78). (77) Only Susan and Tom ate carrots (78) Only Susan ate carrots In (77), the only people who ate carrots are Susan and Tom (but Susan and Tom may eat other things besides carrots). The entailment relation between (77) and (78) fails to hold because the exclusive particle ‘only’ in (78) presents Susan as the singleton member of the set, but this does not logically follow from (77).   In Medumba, the entailment relation is perturbed in ex-situ contexts. Consider for instance the ex-situ question in (79a) and the corresponding ex-situ answer in (79b). (79) a. Q: á kʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  Wàtɛ̀t à?    FOC WH Nuga AGR.AUX.T2 C-AGR.give Watat C.Q.L        T.HL      V.HL     ‘What did Nuga give to Watat?’  62    b. A1: á bɔ̀ búù ʧʉ̀ʔ à nɔ́ɔ̀ʔ          ᵐ-fáà         jí     lá    FOC bag 3PL hat 3SG.L AGR.AUX.T2 N-AGR.give   3SG.H C.-Q         T.HL              V.HL    ‘He gave [the bag and the hat]FOC to him’ In (79), the answer in (b) where the coordinated DP [á bɔ̀ búù ʧʉ̀ʔ] ‘[the bag and the hat]FOC’ is ex-situ does not entail (c) and (d). In fact, if (79b) is true in Medumba, then (79c-d) are false.  c. A2: #á bɔ̀ à nɔ́ɔ̀ʔ            ᵐ-fáà  jí lá    FOC bag 3SG.L AGR.AUX.T2   N-AGR.give 3SG.H C.-Q       T.HL               V.HL    ‘He gave [the bag]FOC to him’  d. A3: #á ʧʉ̀ʔ à nɔ́ɔ̀ʔ           ᵐ-fáà             jí lá    FOC hat 3SG.L AGR.AUX.T2  N-AGR.give 3SG.H C.-Q       T.HL               V.HL    ‘He gave [the hat]FOC to him’ The fact that the entailment relations are blocked in ex-situ contexts in Medumba can be explained if the ex-situ construals are assumed to be exhaustive. (79b) presents the bag and the hat as the only things that Nuga gave Watat. However, (79c&c) present either the bag or the hat as the only thing that Nuga gave Watat and this does not logically follow (79b).  2.3.1.3 Ex-situ wh-/focus blocks universal quantification A universal quantifier entails that the statement within its scope is true for everything or for every instance of a specific thing. For instance, in a situation with three children such as Lucy, Susan and Pete, if (80) is true: (80) Every child bought a hat then it is necessarily true that:  (i) Lucy bought a hat   (ii) Susan bought a hat   (iii) Pete bought a hat 63  This contrasts with exhaustivity marking in that exhaustivity entails that the statement about the entity which it marks is not true for other entities. In other words, it entails that other entities are excluded. For example, if (81) is true: (81) Only Lucy and Susan bought a hat then it is not true that:  (82) Pete bought a hat  In (81), other individuals are excluded from the set of people who bought a hat contra (80) in which everyone is included.    In Medumba, the construal of ex-situ focus is interpreted as exhaustive. The sentence in (83) for instance is true only in a context where Nuga gave the hat to Watat and nothing else. In other words, the hat exhausts the list of items given to Watat by Nuga. That is, other items are excluded from the list. (83) á ʧʉ̀ʔ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  Wàtɛ̀t lá  FOC hat Nuga AGR.AUX.T2 N-AGR.give Watat C.-Q      T.HL      V.HL   ‘The hatFOC (is the thing) Nuga gave to Watat’ Universal quantifiers, however, are predicted to be excluded with ex-situ focus in Medumba because, unlike exhaustivity which entails exclusivity, universal quantifiers do not. And so, are incompatible with exhaustivity. This explains the ungrammaticality of the Medumba sentence in (84). (84) *á ⁿʤɔ̀ɔ́ŋ ʧʉ̀ʔ fɛ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  Wàtɛ̀t lá    FOC every hat all Nuga AGR.AUX.T2 N-AGR.give Watat C.-Q        T.HL      V.HL   [Nuga gave [every hat]FOC to Watat] 64  2.3.1.4 Ex-situ wh/focus blocks additive particles The use of additive particles is another diagnostic of exhaustivity. I consider entities under the scope of the additive particle ᵐbà ‘even, also, too’ as well as the use of additive particles in follow-ups (additive follow-ups). In general, the use of an additive particles presupposes that the statement about the entity under its scope is true for other entities as well. For instance, if (85) is true: (85) Lucy also bought a hat Then it is true that there are other individuals who bought a hat. This contrasts with exhaustivity marking which entails exclusivity. In Medumba, the additive particle ᵐbà ‘even, also, too’ is incompatible with ex-situ foci. (86a) shows the baseline sentence and (86b) an ungrammatical sentence involving ex-situ focus with ᵐbà (even). (86) a. Nùᵑgɛ̀ nɔ́ʔ  fá ᵐbà ʧʉ̀ʔ  Wàtɛ̀t       Nuga AUX.T2 give even hat Watat      T.H      V.H       Lit.: Nuga gave even the hat to Watat  b. á ᵐbà ʧʉ̀ʔ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  Wàtɛ̀t lá      FOC even hat Nuga AGR.AUX.T2 N-AGR.give Watat C.-Q       T.HL      V.HL       [Nuga gave [even the hat]FOC to Watat]  As for additive follow-ups, they are also infelicitous if they follow an ex-situ construal  (87) a. á ʧʉ̀ʔ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  Wàtɛ̀t lá      FOC hat Nuga AGR.AUX.T2 N-AGR.give Watat C.-Q      T.HL      V.HL        ‘Nuga gave the hatFOC to Watat’  b. #ŋ́ŋ̀!  ᵐbà bɔ̀        yes  even bag       ‘yes! He also gave him the bag’  (86a) and (86b) above cannot be both true at the same time.  65  2.3.1.5  Ex-situ wh/focus blocks functional readings I show that exhaustivity marking is only compatible with individual-denoting expressions in Medumba. This has implications on what constituent can move ex-situ in Medumba and what the answer to ex-situ wh-questions can denote. In English, a wh-question such as which girl did each boy greet can be answered by a natural function (89a) or a random function also known as a pair-list reading (89b). (88) which girl did each boy greet?  (89) a. His sister  b. Tom greeted Susan; Sam greeted Lucy … In Medumba, the natural function denotation and the pair list reading are infelicitous as an answer to an ex-situ wh-question as shown in (90) and (91) respectively. (90) a. Q: á     jíìt       ᵑgùn    ᵐbà   jíìt       ʃùm  ló nɔ́ɔ̀ʔ                 ⁿ-tʃɔ́ɔ̀ʔdə́       á                FOC which   girl      even which   boy    ?  AGR.AUX.T2  N-AGR.greet    C.Q.H            T.HL                  V.HL  Lit: which girl did each boy greet? b. A1: #á ᵑgù-máɰáp  FOC girl-mother.3POSS  ‘His sister’  (91) a. Q: á     jíìt       ᵑgùn    ᵐbà   jíìt       ʃùm  ló nɔ́ɔ̀ʔ                ⁿ-tʃɔ́ɔ̀ʔdə́         á                FOC which   girl      even which   boy    ?  AGR.AUX.T2   N-AGR.greet    C.Q.H            T.HL                   V.HL  Lit: which girl did each boy greet?  b. A: #á Màrjà Pítà nɔ́ɔ̀ʔ                 ⁿ-tʃɔ́ɔ̀ʔdə́       lá, á Nùmí Nùᵑgɛ̀             FOC Mary Peter AGR.AUX.T2   N-AGR.greet    C.-Q FOC Numi Nuga         nɔ́ɔ̀ʔ                 ⁿ-tʃɔ́ɔ̀ʔdə́       lá         AGR.AUX.T2   N-AGR.greet    C.-Q         T.HL                   V.HL   ‘Peter greeted MaryFOC, Nuga greeted NumiFOC’ 66   The unavailability of the functional reading (be it a natural function or a random function (pair list)) with ex-situ construals is arguably due to exhaustivity. One way of understanding this is that exhaustivity (in some sense maximality) applies to entities that denote individuals but not to functions because functions denote sets. Accordingly, expressions denoting functions are ruled out with ex-situ construals. This converges with Szabolcsi (1997) who argues that the denotational semantic properties of some elements may determine their syntactic/scopal configurations. If exhaustivity applies only to individuals and not to functions, then this predicts that all event-modifying adjuncts (i.e. where, when, how and why wh-questions (the (a) examples) and their corresponding foci (the (b)) examples) will be prohibited from ex-situ contexts. This prediction is confirmed. See (92-95). (92) Locative  a. *á já Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ⁿ-ʤʉ́ʉ̀n Wàtɛ̀t à        FOC WH Nuga AGR.AUX.T2    N-AGR.see Watat C.Q.L     T.HL      V.HL         [where did Nuga saw Watat?]  b. *á ↓ⁿtɔ́n Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ⁿ-ʤʉ́ʉ̀n Wàtɛ̀t lá        FOC market Nuga AGR.AUX.T2    N-AGR.see Watat C.-Q     T.HL       V.HL       [Nuga saw Watat at the marketFOC] (93) Temporal  a. *á sʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ⁿ-ʤʉ́ʉ̀n Wàtɛ̀t à        FOC WH  Nuga AGR.AUX.T2    N-AGR.see Watat C.Q.L     T.HL      V.HL      [when did Nuga saw Watat?]  b. *á ᵑgɣúʔ-mùʔ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ⁿ-ʤʉ́ʉ̀n Wàtɛ̀t lá       FOC year-other  Nuga AGR.AUX.T2    N-AGR.see Watat C.-Q      T.HL      V.HL        [Nuga saw Watat [last year]FOC]  67  (94) Manner  *á ⁿʤʰʉ́-kʉ́  Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n  Wàtɛ̀t à    FOC manner-WH  Nuga AGR.AUX.T2    N-AGR.sell Watat C.Q.L      T.HL     V.HL    [how did Nuga betrayed Watat?] (95) Rationale  a. *á nùúm-kʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n  Wàtɛ̀t à       FOC for-WH  Nuga AGR.Aux N-AGR.sell Watat C.Q.L      T.HL       V.HL       [why did Nuga betrayed Watat?]  b. *á ⁿdɛ̀ɛ́n-ᵐbʉ́   á          bɔ̀-n-í           Nùᵑgɛ̀ nɔ́ɔ̀ʔ            ⁿ-sʷɛ́ɛ̀n       Wàtɛ̀t   lá      FOC know-that   3SG.H     hate-N-3SG.H     Nuga AGR.AUX.T2   N-AGR.sell  Watat  C.-Q         T.HL  V.HL       [[because s/he hate him/her]FOC, Nuga betrayed Watat] 2.3.2 Argument 2: in-situ wh/focus does not require exhaustivity   Wh-/focus XPs can be ex-situ in Medumba, in which case they are interpreted as exhaustive. In this section, I show that in the absence of movement in Medumba, in-situ wh-/focus are interpreted in place. In fact, no extra mechanism is needed in order to derive in-situ wh-/focus construals. They are base generated in-situ and are interpreted as non-exhaustive.  I first show that wh-/focus can be construed in-situ in Medumba. This is illustrated in (96) for object wh-question and in (97) object focus.  (96) In-situ object wh-question   Nùᵑgɛ̀ nɔ́ʔ  jʉ́n á wʉ́ á   Nuga AUX.T2 see FOC WH C.Q.H   T.H  V.H        ‘Nuga saw who?’    68  (97) In-situ object focus  Nùᵑgɛ̀ nɔ́ʔ  jʉ́n á Wàtɛ̀t   Nuga AUX.T2 see FOC Watat   T.H  V.H   ‘Nuga saw WatatFOC’  In the following subsections, I first show the mechanics of the derivation of in-situ focus and then the diagnostics that confirm the non-exhaustiveness of in-situ focus in Medumba. 2.3.2.1 The mechanics of the derivation In this section, I show the step-by-step derivation of an in-situ object wh-question in Medumba. (98) Step-by-step derivation of an in-situ (non-exhaustive) wh-question  Numeration: {áFOC; kʉ́WH; NùmíN; nɔ́ɔ̀ʔT2; ᵑ-kɛ́ɛ̀V; áC.Q; vø; Dø; Tø} a. vP phase I. Merge <V; Wh-XP> [VP [V jʉ́n ] [Wh-XP^ á wʉ́]  II. Merge <DP; vP> and <v; VP>  [vP [DP^ [Nùmí ] ] [vP [v ∅ ] [VP [V jʉ́n ] [Wh-XP^ á wʉ́]   69  b. CP phase I. Merge <T2; vP> [TP2 [T2 nɔ́ʔ] [vP [DP^ [Nùmí ] ] [vP [v ∅ ] [VP [V jʉ́n ] [Wh-XP^ á wʉ́]  II. Merge <T1; TP2> [TP1 [DP^ [Nùmí ]] [TP1 [T1 ø] [TP2 [T2 nɔ́ʔ] [vP [<DP>] [vP [v ∅ ] [VP [V jʉ́n ] [Wh-XP^ á wʉ́]   III. Merge <CQ; TP1> [CP [C á] [TP1 [DP^ [Nùmí ]] [TP1 [T1 ø] [TP2 [T2 nɔ́ʔ] [vP [<DP>] [vP [v ∅ ] [VP [V jʉ́n ] [Wh-XP^ á wʉ́]  70    IV. Move TP1 to Spec-C [CP [TP1 [DP^ [Nùmí ]] [TP1 [T1 ø] [TP2 [T2 nɔ́ʔ] [vP [<DP>] [vP [v ∅ ] [VP [V jʉ́n ] [Wh-XP^ á wʉ́]]]]]]] [CP [C á] [<TP1>]  2.3.2.2 In-situ wh/focus allows entailment relations  In-situ wh-/focus allows the entailment relations in Medumba. In the question below in (99a) in which the wh-phrase is in-situ, the corresponding in-situ answer in (99b) entails (99c-d). This confirms that in-situ construals are non-exhaustive in Medumba.    71  (99) a. Q: Nùᵑgɛ̀ nɔ́ʔ  fá á kʉ́ Wàtɛ̀t à?    Nuga AUX.T2 give FOC WH Watat C.Q.L     T.H  V.H     ‘What did Nuga give to Watat?’  b. A1: à nɔ́ʔ  fá á bɔ̀ búù ʧʉ̀ʔ jí     3SG.L AUX.T2 give FOC bag 3PL hat 3SG.H      T.H  V.H     ‘He gave [the bag and the hat]FOC to him’  c. A2: à nɔ́ʔ  fá á bɔ̀ jí     3SG.L AUX.T2 give FOC bag 3SG.H      T.H  V.H       ‘He gave [the bag]FOC to him’  d. A3: à nɔ́ʔ  fá á ʧʉ̀ʔ jí     3SG.L AUX.T2 give FOC hat 3SG.H     T.H  V.H        ‘He gave [the hat]FOC to him’ 2.3.2.3 In-situ wh/focus allows universal quantification  Universal quantifiers are allowed with in-situ focus in Medumba. (100) shows that the focused quantified DP [á ⁿʤɔ̀ɔ́ŋ ʧʉ̀ʔ fɛ́] ‘every hat/FOC’ can be construed in-situ in the complement of V position. This confirms that in-situ focus is non-exhaustive in Medumba. (100) Nùᵑgɛ̀ nɔ́ʔ  fá á ⁿʤɔ̀ɔ́ŋ ʧʉ̀ʔ fɛ́ Wàtɛ̀t   Nuga AUX.T2 give FOC every hat all Watat   T.H  V.H    ‘Nuga gave [every hat]FOC to Watat?’ 2.3.2.4 In-situ wh/focus allows additive particles  Additive particles and additive follow-ups are compatible with in-situ focus in Medumba. (101) shows that the additive particle ᵐbà ‘even, also, too’ is allowed in-situ with the focused DP and (102) shows that an additive follow-up is felicitous with in-situ focus. This confirms the non-exhaustiveness of Medumba in-situ focus.   72  (101) Nùᵑgɛ̀ nɔ́ʔ  fá á ᵐbà ʧʉ̀ʔ Wàtɛ̀t   Nuga AUX.T2 give FOC even hat Watat   T.H  V.H    ‘Nuga gave [even the hat]FOC to Watat?’ (102) a. Nùᵑgɛ̀ nɔ́ʔ  fá á ʧʉ̀ʔ Wàtɛ̀t       Nuga AUX.T2 give FOC hat Watat    T.H  V.H        ‘Nuga gave the hatFOC to Watat?’  b. ŋ́ŋ̀!  ᵐbà bɔ̀      yes  even bag      ‘yes! He also gave him the bag’ 2.3.2.5 In-situ wh/focus allows functional readings  Unlike ex-situ wh-questions, the natural function denotation and the pair list reading are felicitous as an answer to an in-situ wh-question in Medumba.  This is shown in (103) and (104) respectively. (103) a. ᵐbà     jíìt     ʃùm  ló  nɔ́ʔ  tʃɔ̀ʔdə̀ə́  á       jíìt      ᵑgùn à                  even  which  boy    ?  AUX.T2 greet.H   FOC   which  girl C.Q.L         T.H  V.L        Lit: Each boy greeted which girl? b. ᵑgù-máɰáp       girl-mother.3POSS        ‘His sister’  (104) a. Q:  ᵐbà     jíìt     ʃùm  ló  nɔ́ʔ tʃɔ̀ʔdə̀ə́     á       jíìt      ᵑgùn           à                        even  which  boy    ?   AUX.T2 greet.H      FOC   which  girl        C.Q.L      T.H V.L       Lit: Each boy greeted which girl?  b. A: Pítà     nɔ́ʔ         tʃɔ̀ʔdə̀ á Màrjà, Nùᵑgɛ̀ nɔ́ʔ        tʃɔ̀ʔdə̀ á Nùmí           Peter   AUX.T2   greet FOC Mary Nuga AUX.T2   greet               FOC Numi             T.H V.L    T.H    V.L            ‘Peter greeted MaryFOC, Nuga greeted NumiFOC’   It appears from the above that in-situ wh-/focus is non-exhaustive in Medumba. Thus, in-situ wh-/foci are correctly predicted to be compatible with an overt exhaustification marker as illustrated below in (105a) for wh-question and in (105b) for focus. 73  (105) a. Nùᵑgɛ̀ nɔ́ʔ  jʉ́n á ⁿdɔ̀ɔ́ʔ wʉ́ á       Nuga AUX.T2 see FOC only WH C.Q.H    T.H  V.H       ‘Nuga saw only who?’  b. Nùᵑgɛ̀ nɔ́ʔ  jʉ́n á ⁿdɔ̀ɔ́ʔ Wàtɛ̀t        Nuga AUX.T2 see FOC only Watat     T.H  V.H       ‘Nuga saw only WatatFOC’ Moreover, unlike ex-situ wh-/focus, event modifying adverbs are also predicted to be allowed in-situ in Medumba. This prediction is confirmed. See (106-109). The (a) examples are wh-questions and the (b) examples are their focus counterpart. (106) Locative  a. Nùᵑgɛ̀ nɔ́ʔ  jʉ́n Wàtɛ̀t á já á       Nuga AUX.T2 see Watat FOC WH C.Q.H    T.H  V.H       ‘Nuga saw Watat where?’  b. Nùᵑgɛ̀ nɔ́ʔ  jʉ́n Wàtɛ̀t á ↓ⁿtɔ́n        Nuga AUX.T2 see Watat FOC market    T.H  V.H      ‘Nuga saw Watat at the marketFOC’ (107) Temporal  a. Nùᵑgɛ̀ nɔ́ʔ  jʉ́n Wàtɛ̀t á sʉ́ á       Nuga AUX.T2 see Watat FOC WH C.Q.H    T.H  V.H      ‘Nuga saw Watat when?’  b. Nùᵑgɛ̀ nɔ́ʔ  jʉ́n Wàtɛ̀t á ᵑgɣúʔ-mùʔ       Nuga AUX.T2 see Watat FOC year-other    T.H  V.H       ‘Nuga saw Watat last year’    74  (108) Manner  a. Nùᵑgɛ̀ nɔ́ʔ  sʷɛ̀ɛ́n Wàtɛ̀t á ⁿʤʰʉ́-kʉ́ á       Nuga AUX.T2 sell.H Watat FOC manner-WH C.Q.H    T.H  V.L       ‘Nuga betrayed Watat how?’  b. Nùᵑgɛ̀ nɔ́ʔ  sʷɛ̀ɛ́n Wàtɛ̀t á búù jɛ́ɛ̀-ⁿʧʰʉ́        Nuga AUX.T2 sell.H Watat FOC PL hurt-heart    T.H  V.L        ‘Nuga betrayed Watat with anger’ (109) Rationale  a. Nùᵑgɛ̀ nɔ́ʔ  sʷɛ̀ɛ́n Wàtɛ̀t á nùúm-kʉ́ á       Nuga AUX.T2 sell.H Watat FOC for-WH  C.Q.H    T.H  V.L      ‘Nuga betrayed Watat why?’  b. Nùᵑgɛ̀ nɔ́ʔ  sʷɛ̀ɛ́n Wàtɛ̀t á ⁿdɛ̀ɛ́n-ᵐbʉ́ á bɔ̀-n-í        Nuga AUX.T2 sell.H Watat FOC know-that 3SG.H hate-N-3SG.H    T.H  V.L        ‘Nuga betrayed Watat [because s/he hate him/her]FOC’ 2.3.3 Argument 3: Q/A pairs are conditioned by in-/ex-situ contrast  In this section, I show that the information-theoretic structure of their focus answers must match the information-theoretic structures of wh-questions in Medumba 2.3.3.1 Ex-situ wh-questions are answered by ex-situ focus In Medumba, if a question is asked with the wh-XP ex-situ, the answer must likewise contain the counterpart to the wh-XP as an ex-situ focus. The examples below, which illustrate object extraction, show that a felicitous answer to an ex-situ wh-question (110a) is when the constituent in the answer corresponding to the wh-XP in the question is an ex-situ focus (110b), and not an in situ focus (110c).     75  (110) a. Q: á kʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  Wàtɛ̀t à?    FOC WH Nuga AGR.AUX.T2    N-AGR.give Watat C.Q.L        T.HL      V.HL     ‘What did Nuga give to Watat?’  b. A1: á bɔ̀ à nɔ́ɔ̀ʔ  ᵐ-fáà  jí lá    FOC bag 3SG.L AGR.AUX.T2    N-AGR.give 3SG.H C.-Q       T.HL         V.HL    Lit: ‘the bagFOC he gave to him’  c. A2: #à nɔ́ʔ  fá á bɔ̀ jí       3SG.L AUX.T2 give FOC bag 3SG.H      T.H  V.H        ‘He gave the bagFOC to him’ The same pattern is found when a PP-complement is extracted as shown in (111).  (111) a. Q: á múùm wʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  bɔ̀      à?    FOC PREP WH Nuga AGR.AUX.T2    N-AGR.give bag    C.Q.L         T.HL     V.HL     ‘To whom did Nuga give the bag?’  b. A1: á múùm Wàtɛ̀t à nɔ́ɔ̀ʔ  ᵐ-fáà  bɔ̀ lá    FOC PREP Watat 3SG.L AGR.AUX.T2    N-AGR.give bag C.-Q        T.HL         V.HL    Lit: ‘to WatatFOC he gave the bag’  c. A2: #à nɔ́ʔ  fá bɔ̀ á Wàtɛ̀t       3SG.L AUX.T2 give bag FOC Watat     T.H  V.H         ‘He gave the bag to WatatFOC’ 2.3.3.2 In-situ wh-questions are answered by in-situ focus In-situ wh-questions in Medumba require their answers to be an in-situ focus. This is illustrated in (112) for in-situ objects and in (113) for in-situ PP-complements. In both cases, a felicitous answer to the wh-question is the one in which the constituent in the answer that corresponds to the wh-XP in the question is an in-situ focus, but not an ex-situ focus (see 112c and 113c).   76  (112) a. Q: Nùᵑgɛ̀ nɔ́ʔ  fá á kʉ́ Wàtɛ̀t à?    Nuga AUX.T2 give FOC WH Watat C.Q.L     T.H  V.H     ‘What did Nuga give to Watat?’  b. A1: à nɔ́ʔ  fá á bɔ̀ jí     3SG.L AUX.T2 give FOC bag 3SG.H      T.H  V.H       ‘He gave the bagFOC to him’  c. A2: #á bɔ̀ à nɔ́ɔ̀ʔ  ᵐ-fáà  jí lá      FOC bag 3SG.L AGR.AUX.T2    N-AGR.give 3SG.H C.-Q       T.HL         V.HL    Lit: ‘the bagFOC he gave to him’ (113) a. Q: Nùᵑgɛ̀ nɔ́ʔ  fá bɔ̀ á wʉ́  à?    Nuga AUX.T2 give bag FOC WH C.Q.L     T.H  V.H       ‘To whom did Nuga give the bag?’  b. A2: à nɔ́ʔ  fá bɔ̀ á Wàtɛ̀t     3SG.L AUX.T2 give bag FOC Watat     T.H  V.H       ‘He gave the bag to WatatFOC’  c. A1: #á múùm Wàtɛ̀t à nɔ́ɔ̀ʔ  ᵐ-fáà  bɔ̀ lá      FOC PREP Watat 3SG.L AGR.AUX.T2    N-AGR.give bag C.-Q        T.HL        V.HL      ‘He gave the bag to WatatFOC’ 2.3.4 Argument 4: fragment answers are conditioned by in-/ex-situ contrast  2.3.4.1 Fragment answers to ex-situ wh require focus-marking Medumba wh-phrases are always preceded by the invariable high-tone focus particle á, but, fragment answers pattern differently with regard to the presence or absence of the focus particle in Medumba. With ex-situ wh-questions, the felicitous fragment answer must be preceded by the focus particle as shown in (114) for object extraction and in (115) for a PP-complement. A fragment answer without the focus particle is infelicitous as an answer to an ex-situ wh-question. See (114c) and (115c). 77  (114) a. Q: á kʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  Wàtɛ̀t à?    FOC WH Nuga AGR.AUX.T2    N-AGR.give Watat C.Q.L        T.HL      V.HL     ‘What did Nuga give to Watat?’  b. A1: á bɔ̀     FOC bag       ‘The bagFOC’  c. A2: #bɔ̀       bag        ‘The bag’ (115) a. Q: á múùm wʉ́ Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fáà  bɔ̀      à?    FOC PREP WH Nuga AGR.AUX.T2    N-AGR.give bag    C.Q.L         T.HL      V.HL     ‘To whom did Nuga give the bag?’  b. A1: á múùm Wàtɛ̀t     FOC PREP Watat         ‘To WatatFOC’  c. A2: #Wàtɛ̀t       Watat    2.3.4.2 Fragment answers to in-situ wh- do not require focus-marking Medumba in-situ wh-questions require a bare (non-focus marked) fragment answer. This is illustrated in (116) for in-situ object and in (117) for in-situ complement PP. A fragment answer with the focus particle is infelicitous as an answer to an in-situ wh-question as confirmed by the infelicity of (116c) and (117c). (116) a. Q: Nùᵑgɛ̀ nɔ́ʔ  fá á kʉ́ Wàtɛ̀t à?    Nuga AUX.T2 give FOC WH Watat C.Q.L     T.H  V.H     ‘What did Nuga give to Watat?’  b. A1: bɔ̀     bag      ‘The bag’  78   c. A2: #á bɔ̀       FOC bag         ‘The bagFOC’ (117) a. Q: Nùᵑgɛ̀ nɔ́ʔ  fá bɔ̀ á wʉ́  à?    Nuga AUX.T2 give bag FOC WH C.Q.L     T.H  V.H       ‘To whom did Nuga give the bag?’  b. A2: Wàtɛ̀t      Watat   c. A1: #á múùm Wàtɛ̀t       FOC PREP Watat           ‘To WatatFOC’ 2.3.4.3 Implications for the syntax of fragment answers  Fragment answers confirm that there is an ex-situ/in-situ partition in Medumba both in terms of the form of the question and the form of the answer; this has implications for syntactic theories of fragments. In the standard approach for fragment answers, the remnant phrase undergoes focus movement to a peripheral position before deletion takes place (Merchant 2004). But there are also approaches to fragment answers where the remnant is in-situ (Lobeck 1995, Abe 2016). The Medumba facts suggest that there might be two ways of deriving fragment answers. Teasing these two apart is beyond the scope of the present dissertation.  (118) a. Possible derivation of focus-marked fragment answers   [CP [FocP [Foc ] [DP]] [CP [C ] [TP [T ] [vP [v ] [VP [V ] <[FocP [Foc ] [DP]>  b. Possible derivation of bare fragment answers    [CP [C ] [TP [T ] [vP [v ] [VP [V ] [FocP [Foc ] [DP]  In sum, interpretation-driven movement captures the fact that ex-situ and in-situ wh-/focus in Medumba are semantically distinct. In addition to the interpretive and the structural difference, there are two other arguments in favor of semantic and syntactic non-parallelism between in-situ 79  and ex-situ wh-/focus in Medumba. The first argument relates to A′-agreement, a crucial diagnostic property of A′-movement in Medumba. Notably, in-situ wh-/focus construals in Medumba do not show A′-agreement as shown in (119). The (a) and the (b) examples are wh- and focus respectively. (119) a. Wàtɛ̀t nɔ́ʔ  jʉ́n á wʉ́ á?       Watat AUX.T2 see FOC WH C.Q.H    T.H  V.H       ‘Watat saw who?’   b. Wàtɛ̀t nɔ́ʔ  jʉ́n á Nùᵑgɛ̀        Watat AUX.T2 see FOC Nuga    T.H  V.H        ‘Watat saw NugaFOC’   Crucially, in-situ construals in Medumba are ungrammatical if they surface with A′-agreement. These facts are compelling evidence against the movement theories of wh-in-situ. Whether it is the head-deletion theory of wh-in-situ (see a.o Chomsky 1995, Pesetsky 2000, Bobaljik 2002, Bošković and Nunes 2007) or disguised movement (Munaro, Poletto and Pollock 2001) as under these theories, wh-in-situ would actually be wh-movement and so incorrectly predict wh-in-situ to show A’-agreement in Medumba.  (120) a. *Wàtɛ̀t nɔ́ɔ̀ʔ  ⁿ-dʒʉ́ʉ̀n á wʉ́ á?         Watat AGR.AUX.T2    N-AGR.see FOC WH C.Q.H    T.HL       V.HL         [Watat saw who?]     b. *Wàtɛ̀t nɔ́ɔ̀ʔ  ⁿ-dʒʉ́ʉ̀n á Nùᵑgɛ̀         Watat AGR.AUX.T2    N-AGR.see FOC Nuga       T.HL       V.HL          [Watat saw NugaFOC]   The second argument in favor of a non-parallelism between ex-situ and in-situ construals in Medumba relates to subjacency and intervention effects. In fact, in-situ wh-questions are not sensitive to islands in Medumba. This is illustrated in the following examples where wh-/focus 80  constituents are construed in-situ within a wh-island (121), an adjunct island (122), and a complex DP (123). (121) Wh-island  Nùᵑgɛ̀ nɔ́ʔ-bɛ́ttə́ [ᵐbʉ́ʉ̀  Wàtɛ̀t nɔ́ʔ  tʃɔ̀ʔdə̀ə́  á wʉ́] á  Nuga Aux-ask  C.HL Watat AUX.T2 greet.H FOC WH C.Q.H        T.H  V.L  Lit: Nuga ask if Watat greeted who? (122) Adjunct Island  Nùᵑgɛ̀   nɔ́ʔ  nɛ̀ɛ́n      ⁿ tɔ́n      [káà     Wàtɛ̀ɛ́t       tʃɔ̀ʔdə̀ə́    á    wʉ́] á?   Nuga    AUX.T2 go.H    market    before  Watat.H     greet.H    FOC WH C.Q.H   T.H  V.L     V.L        Lit: Nuga go to the market before Watat greeted who?’ (123) Complex DP (Relative clause)  Nùᵑgɛ̀   nɔ́ʔ         jʉ́n    [mɛ́n   zə̀       à          nɔ́ɔ̀ʔ               ⁿ-tʃɔ́ɔ̀ʔdə́    á      wʉ́] á?   Nuga   AUX.T2  see   child   C.CL1 3SG.L    AGR.AUX.T2   N-greet.H   FOC  WH C.Q.H                 T.H           V.H                                             T.HL                      V.L              ‘Nuga saw the child that greeted who? With regard to intervention effects (Beck 1996, 2006), elements such as negation behave like interveners and such block LF-movement of wh-in-situ. However, in Medumba, negation does not seem to block wh-in-situ as illustrated in (124). This is evidence against the LF-movement theory of wh-in-situ (Huang 1982, Aoun and Li 1993, Watanabe 2001, Cheng 2009) as it incorrectly predicts negation to block wh-in-situ in the language. The (a) and the (b) examples are wh- and focus respectively. (124)  a. Wàtɛ̀t nɔ́ʔ  kʉ̀ jʉ́n á wʉ́ á?        Watat AUX.T2 NEG see FOC WH C.Q.H    T.H   V.H        Lit.: Watat didn’t see who?  b. Wàtɛ̀t nɔ́ʔ  kʉ̀ jʉ́n á Nùᵑgɛ̀        Watat AUX.T2 NEG see FOC Nuga    T.H   V.H        ‘Watat didn’t see NugaFOC’ 81  To sum up, the Medumba facts suggest the wh-in-situ does not undergo any movement of some kind in Medumba and are derived in place as predicted by the interpretation-driven movement approach. However, it is worth mentioning that this approach is different from other no movement approaches to wh-in-situ which require extra mechanism such as (unselective) binding or absorption (see a.o Baker 1970, Higginbotham and May 1981, Heim 1982, Pesetsky 1987, Reinhart 1998).  2.3.5 Argument 5: subjects are (predictably) different  Subject wh-questions and focus in Medumba pose a problem in terms of their derivation and their interpretation. Extracted subjects permit an “apparent” gap or resumption. As for their interpretation, subject wh-questions and focus are always exhaustive in Medumba; this raises the question of how non-exhaustive subject wh-questions and foci are expressed in Medumba. I show that the apparent gap/resumption split with subject extractions correlates to positional differences. With regard to their interpretation, the different exhaustivity tests indicate that subjects wh-/focus are indeed exhaustive in Medumba and that non-exhaustive subjects are clefted. 2.3.5.1 Temporal auxes and resumption diagnose in-situ versus ex-situ subject Temporal auxiliaries and resumption diagnose the in-situ/ex-situ partition with subject wh-/focus in Medumba. In-situ subject wh-/focus needs T-support; that is require the presence of a temporal auxiliary at T. (125) and (126) shows that in-situ subject wh-/focus are ungrammatical if there is no auxiliary in the structure.   (125) a. *á wʉ́ kɛ́ɛ̀  ⁿʤʷɛ́n  á       FOC WH AGR.fry chips  C.Q.H    V.HL       [Who fried the chips?]  82   b. á wʉ́ nɔ́ɔ̀ʔ  ᵑ-kɛ́ ⁿʤʷɛ́n  á      FOC WH AGR.AUX.T2    N-fry chips  C.Q.H    T.HL  V.H       ‘Who fried the chips?’ (126) a. *á Nùmí kɛ́ɛ̀  ⁿʤʷɛ́n        FOC Numi AGR.fry chips      V.HL       [NumiFOC fried the chips]  b. á Nùmí nɔ́ɔ̀ʔ  ᵑ-kɛ́ ⁿʤʷɛ́n        FOC Numi AGR.AUX.T2    N-fry chips      T.HL  V.H       ‘NumiFOC fried the chips’ Ex-situ wh-/focus subject require resumption instead as show in (127) for ex-situ subject wh-question and in (128) for ex-situ subject focus. (127) a. *á wʉ́ kɛ́ɛ̀  ⁿʤʷɛ́n  á       FOC WH AGR.fry chips  C.Q.H    V.HL       [Who fried the chips?]  b. á wʉ́ à kɛ́ɛ̀  ⁿʤʷɛ́n  á      FOC WH 3SG.L AGR.fry chips  C.Q.H     V.HL       Lit.: ‘Who did [s/he] fry the chips?’ (128) a. *á Nùmí kɛ́ɛ̀  ⁿʤʷɛ́n        FOC Numi AGR.fry chips      V.HL       [NumiFOC fried the chips]  b. á Nùmí à kɛ́ɛ̀  ⁿʤʷɛ́n        FOC Numi 3SG.L AGR.fry chips       V.HL       ‘NumiFOC [s/he] fried the chips’ 2.3.5.2 Deriving subject wh/focus  As shown above, A′-extracted subjects in Medumba alternate with what I call an ‘apparent gap’ and a resumptive pronoun. I propose that these two strategies correspond to different structures in 83  Medumba. The apparent gap strategy is derived by movement of the subject from the vP-internal position to Spec-T. In contrast, the resumptive strategy is derived by movement of the subject to Spec-C. This is illustrated in (129) and (130) for subject wh-question and subject focus respectively.  (129) Subject wh-XP at Spec-T á wʉ́ nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀  Nùᵑgɛ̀  à?  FOC WH AGR.AUX.T2    N-give  bag  Nuga  C.Q.L T.HL  V.H            ‘Who gave the bag to Nuga?’          CP2   ei                     CQ                   CP1                                 á       ei                                    CEXH          TP                                          ei                          DP           TP                   4          ei          á  wʉ́ T  vP       nɔ́ɔ̀ʔ	 			ei        spec  vP           <á  wʉ́>     ei                  v  VP               6           ᵐ-fá bɔ̀ Nùᵑgɛ̀ à        84  (130) Subject focus XP at spec-T á Wàtɛ̀ɛ́t nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀   FOC Watat.H AGR.AUX.T2    N-give  bag Nuga            T.HL  V.H               ‘WatatFOC gave the bag to Nuga’                       CP2   ei                     C                   CP1                                        ei                                    CEXH          TP                                          ei                          DP           TP                   4          ei         á Wàtɛ̀ɛ́t T  vP       nɔ́ɔ̀ʔ	 			ei        spec     vP         <á Wàtɛ̀t>     ei                    v  VP               6       ᵐ-fá bɔ̀ Nùᵑgɛ̀ When the subject moves to Spec-C, Spec-T is spelled out as a resumptive pronoun. This is illustrated in (131) and (132) for subject wh-question and subject focus respectively.  (131) Subject wh-XP at Spec-C á wʉ́ à nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀  à?   FOC WH 3SG.L AGR.AUX.T2    N-give  bag Nuga  C.Q.L  T.HL  V.H         ‘who [he] gave the bag to Nuga?’              85     CP2      ei            CQ        CP1    a  ei                     spec                   CP                               á  wʉ́        ei                                    CEXH           TP                                           ei                          spec           TP                      à            ei                T  vP       nɔ́ɔ̀ʔ	 			ei        spec  vP           <á  wʉ́>     ei                  v  VP               6     ᵐ-fá bɔ̀ Nùᵑgɛ̀  (132) Subject focus XP at spec-C   á Wàtɛ̀t à nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀  lá FOC Watat 3SG.L AGR.AUX.T2    N-give  bag Nuga  C.-Q T.HL  V.H         Lit: WatatFOC [he] gave the bag to Nuga’   CP2      ei            C        CP1    lá  ei                     spec                CP                             á  Wàtɛ̀t      ei                                      CEXH         TP                                          ei                          spec           TP                      à            ei                T  vP       nɔ́ɔ̀ʔ	 			ei        spec  vP        <á Wàtɛ̀t>     ei                  v  VP               6    ᵐ-fá bɔ̀ Nùᵑgɛ̀  86  2.3.5.3 Subject wh/focus is always exhaustive  Exhaustivity tests introduced in section 2.3.1 are used in this section to demonstrate that subject wh-questions and foci are indeed exhaustive in Medumba. (133) illustrates the entailment test; (133b) can be used as a corresponding focused answer to the subject wh-question in (133a).    (133) a. Q: á wʉ́ nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Wàtɛ̀t à?    FOC WH AGR.AUX.T2    N-give  bag Watat C.Q.L       T.HL  V.H           ‘Who gave the bag to Watat?’  b. A1: á Nùmí búù Nùᵑgɛ̀ɛ́ nɔ́ɔ̀ʔ  ᵐ-fá bɔ̀ jí     FOC Numi 3PL Nuga.H AGR.AUX.T2    N-give bag 3SG.H          T.HL              ‘[Numi and Nuga]FOC gave the bag to him The statement in (133b) does not entail (133c) or (133d), confirming that subject wh-questions and foci are exhaustive in Medumba.  c. A2: #á Nùmí nɔ́ɔ̀ʔ  ᵐ-fá bɔ̀ jí     FOC Numi AGR.AUX.T2    N-give bag 3SG.H       T.HL  V.H              ‘[Numi]FOC gave the bag to him  d. A3: #á Nùᵑgɛ̀ɛ́ nɔ́ɔ̀ʔ  ᵐ-fá bɔ̀ jí     FOC Nuga.H AGR.AUX.T2    N-give bag 3SG.H        T.HL  V.H              ‘[Nuga]FOC gave the bag to him  Similar to ex-situ objects, universal quantifiers are illicit with subject focus as illustrated in (134). But if the subject is not focused, universal quantifiers are allowed (135). (134) *á ⁿʤɔ́ɔ̀ŋ bʉ̀n  fɛ́ nɔ́ɔ̀ʔ  nɛ́ɛ̀n ⁿtɔ̀n      FOC every PL.person all AGR.AUX.T2    N-go market        T.HL               ‘[[Everyone]FOC went to the market]’ (135) ⁿʤɔ́ɔ̀ŋ bʉ̀n  fɛ́ nɔ́ʔ  nɛ̀ɛ́n ⁿtɔ́n     every PL.person all AUX.T2 go.H market                 T.H  V.L    ‘Everyone went to the market’ 87   Additive particles are prohibited with focus-marked subjects. (136) shows that the additive particle ᵐbà cannot be associated with focused subjects. But if the subject is not focus-marked, additive ᵐbà is licit (137). (136) *á ᵐbà Nùᵑgɛ̀ɛ́ nɔ́ɔ̀ʔ  nɛ́ɛ̀n ⁿtɔ́n      FOC even Nuga.H AGR.AUX.T2    N-go market       T.HL              ‘[[Nuga also]FOC went to the market]’ (137) ᵐbà Nùᵑgɛ̀ nɔ́ʔ  nɛ̀ɛ́n ⁿtɔ́n    even Nuga AUX.T2 go.H market    T.H              ‘Nuga also went to the market]’ Likewise, additive follow-ups are infelicitous if they follow a subject focus (138) but felicitous if they follow a non-focused subject (139). (138) a. á Nùᵑgɛ̀ɛ́ nɔ́ɔ̀ʔ  nɛ́ɛ̀n ⁿtɔ́n        FOC Nuga.H AGR.AUX.T2    N-go market      T.HL               ‘NugaFOC went to the market’  b. #ŋ́ŋ̀!  ᵐbà Sɛ̀ɛ́mí        yes  even Sami       ‘yes! Sami also went to the market’ (139) a. Nùᵑgɛ̀ nɔ́ʔ  nɛ̀ɛ́n ⁿtɔ́n        Nuga AUX.T2 go.H market                  ‘NugaFOC went to the market’  b. ŋ́ŋ̀! ᵐbà Sɛ̀ɛ́mí     yes even Sami       ‘yes! Sami also went to the market’ A natural function is infelicitous as an answer to a subject wh-question (140).   (140) a.  á       jíìt   ⁿgùn  ᵐbà   jíìt  ʃùm  ló ʧúùp      ᵐbʉ̀  à         nɔ́ɔ̀ʔ              nɛ́ɛ̀n   ⁿtɔ́n      á                 FOC  wh   girl   even   wh  boy   ? AGR.say   C.L 3SG.L  AGR.AUX.T2   N-go  market C.Q.H                V.HL           T.HL  ‘Which girl did each boy say that [she] went to the market?’  88   b. #á ᵑgù-máɰáp         FOC girl-mother.3POSS  ‘His sister’  Similarly, a pair list reading is also incompatible as an answer to a subject wh-question (141). (141) a. á       jíìt ⁿgùn  ᵐbà   jíìt  ʃùm  ló ʧúùp      ᵐbʉ̀  à         nɔ́ɔ̀ʔ              nɛ́ɛ̀n   ⁿtɔ́n      á                 FOC  wh  girl   even  wh  boy   ? AGR.say   C.L 3SG.L  AGR.AUX.T2   N-go market  C.Q.H              V.HL       T.HL       ‘Which girl did each boy say she went to the market?’  b. #á Màrjà Pítà ʧúùp      ᵐbʉ̀  á         nɔ́ɔ̀ʔ         nɛ́ɛ̀n   ⁿtɔ́n         lá      FOC Mary Peter AGR.say  C.L  3SG.H  AGR.AUX.T2 N-go    market  C.-Q     V.HL                  T.HL    á Nùmí Nùᵑgɛ̀ ʧúùp      ᵐbʉ̀  á         nɔ́ɔ̀ʔ              nɛ́ɛ̀n   ⁿtɔ́n      lá    FOC Numi Nuga AGR.say  C.L 3SG.H  AGR.AUX.T2   N-go market  C.-Q       V.HL            T.HL  ‘Peter said that MaryFOC went to the market, Nuga said that NumiFOC went to the market’  In sum, subject wh-questions and subject foci are exhaustive in Medumba as confirmed by the various exhaustivity tests.  2.3.5.4 Clefted subject wh/focus is non-exhaustive  A question that arises is that if subject wh-questions and subject foci are always exhaustive in Medumba, how are non-exhaustive wh-questions and foci expressed in the language?. I show that non-exhaustive subject wh-questions and foci are expressed as clefts.   I start with the entailment test and show that a clefted subject wh-question (142a) can be answer by the corresponding clefted focused answer (141b).  (142) a. Q: à         bʉ́    á wʉ́   zə̀         à         nɔ́ɔ̀ʔ             ᵐ-fá bɔ̀ Wàtɛ̀t à?           3SG.L   BE   FOC WH  C.CL1   3SG.L   AGR.AUX.T2    N-give bag Watat C.Q.L                T.HL  V.H          ‘Who is it that gave the bag to Watat?’   89   b. A1: à bʉ́ á Nùmí búù Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fá bɔ̀ jí    3SG.L   BE FOC Numi 3PL Nuga AGR.AUX.T2    N-give bag 3SG.H          T.HL  V.H            ‘It was Numi and NugaFOC who gave the bag to him’ Unlike subject wh-questions and their corresponding focused answers, the clefted subject focus statement used as an answer to the question above entails (142c) and (142d). This is only possible if clefts are assumed to be non-exhaustive in Medumba.  c. A2: à bʉ́ á Nùmí nɔ́ɔ̀ʔ  ᵐ-fá bɔ̀ jí    3SG.L   BE FOC Numi AGR.AUX.T2    N-give bag 3SG.H        T.HL  V.H            ‘It was NumiFOC who gave the bag to him’  d. A3: à bʉ́ á Nùᵑgɛ̀ nɔ́ɔ̀ʔ  ᵐ-fá bɔ̀ jí    3SG.L   BE FOC Nuga AGR.AUX.T2    N-give bag 3SG.H        T.HL   V.H            ‘It was NugaFOC who gave the bag to him’  Likewise, universal quantifiers (143), additive particles (143) and additive follow-ups (145) are also compatible with clefted subject focus. (143) Universal quantifier  à bʉ́ á ⁿʤɔ́ɔ̀ŋ bʉ̀n  fɛ́ nɔ́ɔ̀ʔ  nɛ́ɛ̀n ⁿtɔ̀n    3SG.L BE FOC every persons all AGR.AUX.T2    N-go market          T.HL             ‘*It is everyone that went to the market]’ (144) Additive particle  à bʉ́  á ᵐbà Nùᵑgɛ̀ɛ́ nɔ́ɔ̀ʔ  nɛ́ɛ̀n ⁿtɔ́n    3SG.L BE FOC even Nuga.H AGR.AUX.T2    N-go market         T.HL             ‘*It was even NugaFOC who went to the market’ (145) Additive follow-up   à  bʉ́  á Nùᵑgɛ̀ɛ́ nɔ́ɔ̀ʔ  nɛ́ɛ̀n ⁿtɔ́n      3SG.L  BE FOC Nuga.H AGR.AUX.T2    N-go market         T.HL             ‘It was NugaFOC who went to the market’ 90   b. ŋ́ŋ̀!  ᵐbà Sɛ̀ɛ́mí        yes  even Sami       ‘yes! Sami too’  With regard to function denotations, both the natural function (146b) and the pair-list reading (146c) are used as felicitous answers to a clefted subject wh-question. (146) a. à    bʉ́ á jíìt ⁿgùn ᵐbà jíìt ʃùm ló ʧúùp  ᵐbʉ̀                   3SG.L  BE FOC which girl even which boy  ? AGR.say C.L               V.HL     á nɔ́ɔ̀ʔ  nɛ́ɛ̀n ⁿtɔ́n  á     3SG.H AGR.AUX.T2    N-go market  C.Q.H      T.HL   ‘It was which girl that each boy said she went to the market?’  b. à bʉ́ á ᵑgù-máɰáp  3SG.L BE FOC girl-mother.3POSS  ‘It was his sister’   c. à    bʉ́ á Màrjà Pítà ʧúùp      ᵐbʉ̀  á         nɔ́ɔ̀ʔ               nɛ́ɛ̀n   ⁿtɔ́n     lá      3SG.L    BE FOC Mary Peter AGR.say  C.L  3SG.H  AGR.AUX.T2   N-go market  C.-Q       V.HL             T.HL      à    bʉ́ á Nùmí Nùᵑgɛ̀ ʧúùp      ᵐbʉ̀  á         nɔ́ɔ̀ʔ              nɛ́ɛ̀n   ⁿtɔ́n      lá      3SG.L    BE FOC Numi Nuga AGR.say  C.L 3SG.H  AGR.AUX.T2   N-go market  C.-Q       V.HL            T.HL      ‘It was MaryFOC Peter said went to the market,        it was NumiFOC Nuga said went to the market …’  Subject wh-questions and subject foci are always exhaustive in Medumba whereas their non-exhaustive counterparts are always clefted. The behaviour of subject clefts in Medumba has theoretical significance. Clefts are usually described as exhaustive (Hole and Zimmermann 2009) –– even though it is not clear what the nature of their exhaustiveness is –– but the fact that subject clefts are not exhaustive in Medumba raises the question whether clefts have the same structure, semantics, and pragmatics cross-linguistically. The answer to this question is beyond the scope of this dissertation and will be subject of future investigation. 91  2.3.6 A loose end: the semantics of exhaustivity  The in-situ/ex-situ contrast found with Medumba wh-questions and focus construals has semantic correlates: while in-situ configurations are non-exhaustive, ex-situ ones are exhaustive and exclusively individual-denoting as seen by the unavailability of all event-modifying adjuncts (e.g. how, where, when, how wh-questions) and function denotations in ex-situ contexts. Building on this, a further investigation of other non-individual denoting elements is needed for a unifying semantic analysis of exhaustivity marking. My initial proposal in this chapter is that the exhaustivity operator or “Max” operator applies only to individuals, but not to functions, whether natural functions or random functions (i.e. pair-list readings). This is of theoretical significance as it appears that the denotational semantic properties of some elements might determine their syntactic configuration (see also Szabolcsi 1997). 2.4 The broader landscape of A′-movement Interpretation-driven movement has implications both for the syntax of A′-dependencies within the Minimalist framework and for the in-situ/ex-situ partition cross-linguistically. One of the outstanding questions in modeling Move in this framework includes the mechanisms that force it and why it applies at all. As pointed out by Rouveret (2011), more work needs to be done to develop a better understanding of the properties of Agree that distinguish it from Move. Interpretation-driven movement predicts that wh-ex-situ and wh-in-situ would not always mean the same thing cross-linguistically. Moreover, it predicts that if a language has an in-situ/ex-situ partition, there should be an interpretive difference as well.  92  2.4.1  Wh-ex-situ doesn’t always “mean” the same thing  2.4.1.1 Wh-ex-situ can be just inquisitive: English English ex-situ wh-questions are inquisitive. That is, the basic interrogative form used for information seeking as given in (147).  (147) Context: John is doing a survey about linguistic diversity in his neighbourhood. He meets a resident on the street and asks:    a. What is your native language ___ ?  b. #Your native language is what?     2.4.1.2 Wh-ex-situ can be inquisitive and exhaustive: Medumba As demonstrated in this chapter, Medumba ex-situ wh-questions are inquisitive and exhaustive. That is, the speaker when asking an ex-situ wh-question, is not only seeking information, but also is requesting from the addressee an exhaustive list as answer. This is illustrated in (148).   (148) Context: Watat and Nuga went to the market while their mother was out of town, to buy groceries for the feast happening in the neighbourhood. When she returns in town, she wants to know the full list of things they have bought. So, she asks:   a. á kʉ́ bìn ʒʷíìn  ___ ⁿtɔ́n  á?        FOC WH 2PL.L AGR.buy market C.Q.H     V.HL       ‘What did you buy at the market?’  b. #bìn ʒʷín á kʉ́ ⁿtɔ́n  á?          2PL.L buy  FOC WH market C.Q.H    V.H            ‘Lit.: You bought what at the market?’ 93  2.4.2 Wh-in-situ doesn’t always “mean” the same thing  2.4.2.1 Wh-in-situ can be just inquisitive: Medumba In-situ wh-questions in Medumba can only be inquisitive. That is, the basic form of interrogative used to seek information in the language.  (149) Context: Watat and Nuga went to the market to buy groceries. When they return, their mother wants to know what they have bought. So, she asks:   a. bìn  ʒʷín á kʉ́ ⁿtɔ́n  á?        2PL.L buy  FOC WH market C.Q.H    V.H        ‘Lit.: You bought what at the market?’   b. #á  kʉ́ bìn ʒʷíìn  ___ ⁿtɔ́n  á?          FOC WH 2PL.L AGR.buy market C.Q.H      V.HL       ‘What did you buy at the market?’ 2.4.2.2 Wh-in-situ can be supplemented with reprise “a.k.a echo” particle: Medumba  A question that arises is whether in-situ wh-questions are used for reprise questions (a.k.a echo-questions). More precisely, how are reprise questions expressed in Medumba? This is of interest because some languages that use the in-situ strategy for information seeking also use the same strategy for reprise questions (see Engdhal 2006 for French). Interpretation-driven movement predicts that the inquisitive semantics (i.e. wh-in-situ) is the “base form” of information seeking questions in Medumba. All other forms are additive. We have already seen that exhaustive inquisitive questions requires wh-XP + Move.  As for reprise questions, I show that they require the reprise question particle dí, that is, they involve wh-XP + dí. (150) illustrates the different wh-questions strategies and semantics in Medumba.   94  (150) Wh-question strategies and their semantics in Medumba  a. wh-XP (i.e. wh-in-situ)  = inquisitive Q  b. wh-XP + dí    = reprise inquisitive Q  c. wh-XP +  Move   = inquisitive exhaustive Q  d. wh-XP +  Move + dí  = reprise inquisitive exhaustive Q A reprise question is a replay of (part of) a previous utterance (Engdhal 2006: 102); (see also Bolinger 1978; Ginzburg and Sag 2000). The type of reprise question addressed in this section is the one in which part of the utterance is inaudible. The form of a reprise question in Medumba depends on the part of the utterance that is inaudible. That is, whether it is a subject, an object, or a full CP. The reprise particle is always in final position as illustrated in the following examples for reprise inquisitive question. (151) for object, (152) for adjuncts, and (153) for a full CP. (151) Object   A: Nùᵑgɛ̀ nɔ́ʔ  jʉ́n  [inaudible]       Nuga AUX.T2 see    T.H  V.H      ‘Nuga saw [inaudible]’  B: Nùᵑgɛ̀ nɔ́ʔ  jʉ́n á wʉ́ ↓dí       Nuga AUX.T2 see FOC WH C.ECHO    T.H  V.H       ‘Nuga saw who?’ (152) Adjuncts a. Locative  A: Nùᵑgɛ̀ nɔ́ʔ  jʉ́n Wàtɛ̀t [inaudible]       Nuga AUX.T2 see Watat    T.H  V.H        ‘Nuga saw Watat [inaudible]’  95   B: Nùᵑgɛ̀ nɔ́ʔ  jʉ́n Wàtɛ̀t á já ↓dí       Nuga AUX.T2 see Watat FOC WH C.ECHO    T.H  V.H      ‘Nuga saw Watat where?’ b. Temporal  A: Nùᵑgɛ̀ nɔ́ʔ  jʉ́n Wàtɛ̀t [inaudible]       Nuga AUX.T2 see Watat    T.H  V.H         ‘Nuga saw Watat [inaudible]’  B: Nùᵑgɛ̀ nɔ́ʔ  jʉ́n Wàtɛ̀t á sʉ́ ↓dí       Nuga AUX.T2 see Watat FOC WH C.ECHO    T.H  V.H       ‘Nuga saw Watat when?’ c. Manner  A: Nùᵑgɛ̀ nɔ́ʔ  sʷɛ̀ɛ́n Wàtɛ̀t [inaudible]       Nuga AUX.T2 sell.H Watat    T.H  V.L        ‘Nuga betrayed Watat [inaudible]’   B: Nùᵑgɛ̀ nɔ́ʔ  sʷɛ̀ɛ́n Wàtɛ̀t á ⁿʤʰʉ́-kʉ́ ↓dí       Nuga AUX.T2 sell.H Watat FOC manner-WH C.ECHO    T.H  V.L       ‘Nuga betrayed Watat why?’ d. Rationale  A: Nùᵑgɛ̀ nɔ́ʔ  sʷɛ̀ɛ́n Wàtɛ̀t [inaudible]       Nuga AUX.T2 sell.H Watat    T.H  V.L        ‘Nuga betrayed Watat [inaudible]’   B: Nùᵑgɛ̀ nɔ́ʔ  sʷɛ̀ɛ́n Wàtɛ̀t á nùúm-kʉ́ ↓dí       Nuga AUX.T2 sell.H Watat FOC for-WH  C.ECHO    T.H  V.L       ‘Nuga betrayed Watat why?’   96  (153) CP  A: [inaudible]  B: ᵐbʉ̀ kʉ́ ↓dí       C WH C.ECHO       Lit: that what?         ‘What?’ The reprise particle can also appear with exhaustive questions in Medumba as shown in (154).  (154) Reprise exhaustive inquisitive Q a. Subject  A: [inaudible] nɔ́ɔ̀ʔ  ⁿ-ʤʉ́n  Wàtɛ̀t      AGR.AUX.T2    N-see  Watat    T.HL  V.H         ‘[inaudible] saw Watat?’  B: á wʉ́ nɔ́ɔ̀ʔ  ⁿ-ʤʉ́n  Wàtɛ̀t  dí       FOC WH AGR.AUX.T2    N-see  Watat  C.ECHO    T.HL  V.H      ‘Who saw Watat?’ b. Object  A: [inaudible] Nùmí  nɔ́ɔ̀ʔ  ᵑ-kɛ́ɛ̀  ___ lá          Numi  AGR.AUX.T2    N-AGR.fry  C.-Q      T.HL  V.HL          ‘[inaudible] Numi fry’  B: á kʉ́  Nùmí  nɔ́ɔ̀ʔ  ᵑ-kɛ́ɛ̀  ___ dí?       FOC WH Numi  AGR.AUX.T2    N-AGR.fry  C.ECHO      T.HL  V.HL        ‘What did Numi fry? 2.4.2.3 Wh-in-situ can be a reprise question: English In English a wh-in-situ form is typically used for reprise questions as shown in (155). (155) A: Lucy saw [inaudible]  B: Lucy saw who?  97  2.4.3 Wh-questions always and only contrast in-situ versus ex-situ 2.4.3.1 Prolific inquisitive forms in French: seven ways of asking a question? French is also a language where the wh-XP can stay in-situ or move to the clause left-peripheral position. The hypothesis adopted in this chapter predicts that there must be a difference between in-situ and ex-situ wh-questions in French. It is reported that there are several strategies (4 to 7 strategies) to form wh-questions in French (Chang 1997; Adli 2006; Deprez et al. 2012; Shlonsky 2012; Tailleur 2013). Those strategies are listed in (156), and are adapted from Chang 1997, Tailleur 2013 and Shlonsky 2012.   (156) a. Tu  as vu qui                         2SG  have see wh       ‘You saw who?’  b. Qui  tu as vu __ ?             WH  2SG have see      ‘Who did you see?’  c. Qui  as-tu  vu __ ?              WH  have-2SG see        ‘Who did you see?’  d. Qui  que tu as vu __?              WH  C 2SG have see       ‘Who that you saw?’  e. C’est qui que tu as vu __ ?           It be WH C 2SG have see       ‘It is who that you saw?’  f. Qui c’est que tu as vu __ ?           WH it be C 2SG have see      ‘Who it is that you saw?’  g. Qui est-ce que tu as vu __ ?            WH be-it C 2SG have see      ‘Who is it that you saw?’ 98  2.4.3.2 The French paradigm reduces to an in-situ/ex-situ partition The various wh-question patterns in French can be reduced to two major strategies namely, an in-situ/ex-situ partition.  The in-situ strategy is illustrated in (157). The ex-situ strategy can be divided into three different categories: The simple ex-situ –– which includes ex-situ with no T-to-C movement (158a), ex-situ with T-to-C movement (158b), and ex-situ with doubly-filled Comp (158c) ––; the clefted wh-questions (which include what I call cleft type 1 (159a) and cleft type 2 (159b)); and the base-generated form or reinforced interrogative, which is argued to be different from wh-clefts (160); (see Chan 1997 and Tailleur 2013). (157) In-situ   Tu as vu qui                      2SG have see wh  ‘You saw who?’ (158) Simple ex-situ wh-questions a. Qui tu as vu __ ?   Ex-situ with No T-to-C movement      WH 2SG have see  ‘Who did you see?’ b. Qui as-tu  vu __ ?   Ex-situ with T-to-C movement      WH have-2SG see  ‘Who did you see?’ c. Qui que tu as vu __?   Ex-situ with Doubly-filled Comp     WH C 2SG have see  ‘Who that you saw?’ (159) Wh-clefts a. C’est qui que tu as vu __ ? Cleft type 1      It be WH C 2SG have see  ‘It is who that you saw?’ b. Qui c’est que tu as vu __ ? Cleft type 2      WH it be C 2SG have see  ‘who it is that you saw?’ 99  (160) Reinforced interrogative  Qui est-ce que tu as vu __ ?       WH be-it C 2SG have see  ‘who is it that you saw?’   It is still unclear whether there is an interpretative difference between in-situ and ex-situ wh-questions in French; the description in the literature is controversial. Bošcović 1998 argues for optionality in French wh-questions; that is, in-situ and ex-situ wh-questions in French are optional variants of the same structure (see also Adli 2006). Matthieu 1999 claims that French wh-in-situ forms are not restricted only to echo-questions but are also used as standard questions to ask new information and suggested no difference in interpretation between French in-situ and ex-situ construals. Cheng and Rooryck 2000 argue that the optionality is only apparent, but their argument is only in terms of derivation. Shlonsky 2012 argues that French wh-clefts are associated with an existential presupposition whereas their in-situ counterparts are not necessarily associated with such a presupposition. One important thing worth mentioning is that not all strategies may be available in a single dialect of French. But it is clear that each dialect has the in-situ strategy plus some version of the ex-situ strategy. Other factors that could also condition the variation in French are register and social class (See also Tailleur 2013, Adli 2017). A careful study including control of dialect differences, possible answers, and information structure is needed in order to examine the complexity of French wh-questions. Chang 1997 argues that there is a three way distinction among the types of questions. They are questions neutres (information seeking questions), question de précision (detail seeking questions), and questions de reprise (echo questions); (Chang 1997: 45). Hamlaoui 2011 in this respect argues that there are information-structural differences related to givenness that tease apart in-situ and ex-situ wh-questions in Francilian French, the variety of French spoken in Paris and its suburbs. 100  2.4.4 A prediction about wh-interrogative and wh-relative in Medumba  In English, a wh-question is only inquisitive, and thus not exhaustive. Relativization too is not exhaustive. As such wh-movement in English is compatible with questions and operator-movement (i.e. relativization) as shown in (161). (161) a. Who did Jane talk to?  b. The boy who Jane talked to In Medumba, wh-movement is inquisitive and exhaustive; relativization is not exhaustive and so, relativization is predicted to have a different form in Medumba. Crucially there is no wh-relative in the language as shown in (162).  (162) a. á wʉ́ Nùmí jʉ́ʉ̀n  á     FOC WH Numi AGR.see C.Q.H     V.HL          ‘Who did Numi see?’  b. *má-ⁿʤùm wʉ́ Nùmí jʉ́ʉ̀n lá        SG-male WH Numi see C.-Q        [The boy who Numi saw]  Relativization in Medumba is introduced by two complementizers: An optional clause-internal C that agrees in noun class with the head noun and an obligatory external C (lá) that surfaces in clause-final position. When the optional C surfaces in the structure, an additional high tone is added to the head noun. (163a) shows the C (zə̀) which agrees in noun class 1 with the head noun má-ⁿdʒùm ‘boy’, whose final syllable bears an additional H-tone. (163b) in contrast, shows that when the internal C is absent, there is no additional H-tone on the head noun.    (163) a. má-ⁿdʒùúm  zə̀  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n lá …        SG-male.H  C.CL1  Watat  AGR.AUX.T2    N-AGR.sell C.-Q  T.HL       V.HL                  ‘The boy that Watat betrayed…’  101   b. má-ⁿdʒùm  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n lá …        SG-male  Watat  AGR.AUX.T2    N-AGR.sell C.-Q       HL       HL                   ‘The boy Watat betrayed…’  (164a) illustrates the plural exponent (tˢə̀) of the C which agrees in noun class 6 with the head noun bá-ⁿdʒùm ‘boys’. (164b) lacks the internal C. (164) a. bá-ⁿdʒùúm  tˢə̀ Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n lá …        PL-male.H  C.CL6 Watat  AGR.AUX.T2    N-AGR.sell C.-Q T.HL      V.HL                  ‘The boys that Watat betrayed…’  b. bá-ⁿdʒùm  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n lá …        PL-male  Watat  AGR.AUX.T2    N-AGR.sell C.-Q       T.HL       V.HL                 ‘The boys that Watat betrayed…’  (165a) shows that the complementizer (zə̀) which is used for class 1 is also used for a class 3 noun such as kʰù ‘foot’.  If the agreeing C surfaces in the structure, the head noun carries an additional H-tone. (166) shows its class 4 plural counterpart (mì).  (165) a. kʰùú  zə̀ Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sə́ə̀m   lá …        foot.H C.CL3 Watat  AGR.AUX.T2    N-AGR.massage C.-Q       T.HL       V.HL                  ‘The foot that Watat massaged…’  b. kʰù  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sə́ə̀m   lá …        foot  Watat  AGR.AUX.T2    N-AGR.massage C.-Q      T.HL       V.HL                  ‘The foot that Watat massaged…’ (166) a. ᵑ-kʰùú mì Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sə́ə̀m   lá …        PL-foot.H C.CL4 Watat  AGR.AUX.T2    N-AGR.massage C.-Q       T.HL      V.HL                   ‘The feet that Watat massaged…’  b. ᵑ-kʰù Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sə́ə̀m   lá …        PL-feet Watat  AGR.AUX.T2    N-AGR.massage C.-Q      T.HL       V.HL                  ‘The feet that Watat massaged…’ 102   (167) shows the complementizer (sə̀) agreement in noun class 5 with the head noun sò ‘tooth’ and (168) is its class 4 plural form. (167) a. sòó  sə̀ Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-ʧúùʔ   lá …        tooth.H C.CL5 Watat  AGR.AUX.T2    N-AGR.remove C.-Q       T.HL       V.HL                  ‘The tooth that Watat removed…’  b. sò  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-ʧúùʔ   lá …        tooth Watat  AGR.AUX.T2    N-AGR.remove C.-Q      T.HL       V.HL                  ‘The tooth that Watat removed…’ (168) a. ⁿ-sòó  mì Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-ʧúùʔ   lá …        PL-tooth.H C.CL4 Watat  AGR.AUX.T2     N-AGR.remove C.-Q       T.HL       V.HL                  ‘The teeth that Watat removed…’  b. ⁿ-sò  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-ʧúùʔ   lá …        PL-tooth Watat  AGR.AUX.T2    N-AGR.remove C.-Q      T.HL       V.HL                  ‘The teeth that Watat removed…’  The agreeing C is not always optional in Medumba. It is obligatory in relative clauses where the head noun is not spelled out. This is illustrated in the following examples with C inflected for the noun class of the head noun. (169) a. zə̀  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n lá …       C.CL1 Watat  AGR.AUX.T2    N-AGR.sell C.-Q      T.HL       V.HL                  ‘The one/CL1 that Watat betrayed…’  b. tˢə̀  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n lá …        C.CL6 Watat  AGR.AUX.T2    N-AGR.sell C.-Q      T.HL       V.HL                 ‘The ones/CL6 that Watat betrayed…’ (170) a. zə̀  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sə́ə̀m   lá …        C.CL3 Watat  AGR.AUX.T2    N-AGR.massage C.-Q      T.HL       V.HL                  ‘The one/CL3 that Watat massaged…’  103   b. mì  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-sə́ə̀m   lá …        C.CL4 Watat  AGR.AUX.T2    N-AGR.massage C.-Q      T.HL       V.HL                  ‘The ones/CL4 that Watat massaged…’ (171) a. sə̀  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-ʧúùʔ   lá …        C.CL5 Watat  AGR.AUX.T2    N-AGR.remove C.-Q      T.HL       V.HL                  ‘The one/CL5 that Watat removed…’  b. mì  Wàtɛ̀t  nɔ́ɔ̀ʔ  ⁿ-ʧúùʔ   lá …        C.CL4 Watat  AGR.AUX.T2    NAGR.remove  C-Q      T.HL       V.HL                  ‘The ones/CL4 that Watat removed…’  In sum, the internal C can be optional if the head noun is overt, but obligatory if the head noun is covert. One way of understanding this variability is in terms of recoverability: if the head noun is not spelled out, the only way to recover its features for the purposes of interpretation is through the agreeing features of C, making C obligatory in such contents14. This makes the complementizer obligatory in this context.   The fact that the internal C found in relative clauses agrees in noun class with the head noun is reminiscent of the connection between the relative marker and demonstrative in other Bantu languages, where the same form (C/D) is used for relative clauses and demonstratives. In Lingala for instance, the form of C involves in relative clauses is similar to that of the demonstrative D (Henderson 2006). This is illustrated in (172a) and (172b) where the same class 5 form (muye) is used for the relative C and for the demonstrative D.    14 Another way of understanding this is in terms of Koopman’s 1997 Principle of Projection Activation which states that a projection is interpretable if and only if it is activated by lexical material (Koopman 1997:32). That is for a projection to be interpretable, lexical material must associate either as head or specifier of that projection in the course of the derivation. Thus, the condition on recoverability can be reduced to the Principle of Projection Activation.  104  (172) a. mukanda muye Poso a-tind-aki …               Lingala         5letter 5REL Poso 3S-send-PST     ‘The letter that Poso sent …’             [Henderson 2006: 35a, p. 43]  b. mukanda muye       5letter 5DEM        ‘this letter’                [Henderson 2006: 35b, p. 43] The same pattern is also found in Southern Sotho and Tsonga (Zeller 2004, 2006). This is illustrated in (173) for Southern Sotho where the class 7 form (seo) is used for both the relative clause C and the demonstrative D, and in (174) for Tsonga where the class 9 form (leyi) is likewise used as C and D. The examples in (a) and (b) represent relative clause Cs and demonstrative Ds respectively. (173) a. setulo seo basadi  ba-se-rek-ile-ng  kajeno …     S. Sotho       7chair 7REL 2women 3PL-5OM-buy-PERF-RS today       ‘The chair which the women bought today …’                      [Zeller 2004:7, p. 77]  b. setulo seo      7chair 7DEM      ‘this chair’                         [Zeller 2004:10a, p. 77] (174) a. buku leyi munhu a yi hlaya-ka …          Tsonga      9book 9REL 1person 1SA 9OM read-RS      ‘The book that the person is reading …’                      [Zeller 2004:15, p. 79]  b. buku leyi      9book 9REL       ‘this book’                [Zeller 2004:17, p. 79]  In contrast, in Medumba, the internal C used for relative clauses is different from the demonstrative D. In fact, the relative clause agreeing C cannot substitute for the demonstrative D as shown in (175). (175) a. *má-ⁿdʒùúm zə̀         SG-male.H  C.CL1        [Intended: this man]  105   b. má-ⁿdʒùm jʉ́n-ní       SG-male AGR.CL1-1PROX       ‘this man’  It is worth mentioning that relative clauses lacking the head noun (which I refer to as pro-drop relative clauses) are not free relatives in Medumba because the latter have a different form as illustrated in the following examples. The head noun is always present in headed relative clauses (176). The features of the head noun are interpretable in pro-drop relative clauses (177) whereas a quantified expression is used in free relative (178). (176) Headed relative clause  bə̀k àʔ-fá  ⁿdɔ̀ʔ ᵑgùún zə̀ à ʧʰúùm  ᵐbʰə̀ mɛ̀n lá   1PL IRR-give gift girl.H C.CL1 3SG.L AGR.come first person  C.-Q         V.HL  ‘We should give the present to the girl who came first’ (177) pro-drop relative clause  bə̀k àʔ-fá  ⁿdɔ̀ʔ zə̀ à ʧʰúùm  ᵐbʰə̀ mɛ̀n lá  1PL IRR-give gift C.CL1 3SG.L AGR.come first person  C.-Q        V.HL  = ‘We should give the present to the one who came first’  # ‘we should give the present to whoever came first’       106  (178) Headless or free relative clause15  a. bə̀k àʔ-fá       ⁿdɔ̀ʔ    ᵐbà   wʉ́  ↓ló     à       ʧʰúùm ᵐbʰə̀ mɛ̀n lá     1PL IRR-give    gift     even   WH    ?      3SG.L  AGR.come first person  C.-Q              V.HL       ‘We should give the present to whoever came first’  b. bə̀k àʔ-fá       ⁿdɔ̀ʔ    ᵐbà   jíìt ᵑgùn  ↓ló    à  ʧʰúùm            ᵐbʰə̀ mɛ̀n lá      1PL IRR-give    gift     even   WH   girl    ?    3SG.L  AGR.come first person  C.-Q              V HL        ‘We should give the present to whichever girl came first’ 2.5 Conclusion This chapter has given an overview of the different A′-movement construals in Medumba such as wh-movement, focus movement and relativization. With regard to A′-movement diagnostics in Medumba, the major findings are that six diagnostics identify a cluster of properties characteristics of an A′-movement dependency, namely:  • Diagnostic 1: A′-movement from a root CP permits gapping or resumption, but A′-movement from a non-root CP requires resumption;  • Diagnostic 2: A′-movement participate in long-distance dependencies  • Diagnostic 3: A′-movement exhibits reconstruction effects  • Diagnostic 4: A-movement conditions Strong and Weak Crossover  • Diagnostic 5: A-movement is island-sensitive, and islands can be rescued by resumption  15 Headless relative clauses cannot surface with the agreeing C in Medumba i. bə̀k àʔ-fá       ⁿdɔ̀ʔ    ᵐbà   wʉ́  ↓ló     zə̀ à          ʧʰúùm ᵐbʰə̀ mɛ̀n lá            1PL IRR-give    gift     even   WH    ?      C.CL1 3SG.L  AGR.come first person  C.-Q                      V.HL       ‘We should give the present to whoever came first’ ii. bə̀k àʔ-fá       ⁿdɔ̀ʔ    ᵐbà   jíìt ᵑgùn  ↓ló    zə̀        à     ʧʰúùm           ᵐbʰə̀ mɛ̀n lá      1PL IRR-give    gift     even   WH   girl    ?    C.CL1  3SG.L  AGR.come      first person  C.-Q                   V HL        ‘We should give the present to whichever girl came first’  107   • Diagnostic 6: A′-movement conditions A′-agreement in the form of an HL tone melody  As for the derivation of A′-movement, I argued that it is driven by interpretation, and in particular, by a covert exhaustive operator at C. I marshalled the following arguments in support of the claim that movement is driven by interpretation:  • Argument 1: ex-situ wh-/focus are interpreted as exhaustive  • Argument 2: in-situ wh-/focus are interpreted as non-exhaustive  • Argument 3: ex-situ wh-questions require and ex-situ focus as answer; in-situ wh-questions require an in-situ focus as answer  • Argument 4: ex-situ wh-questions require an ex-situ fragment answer; in-situ wh-questions require an in-situ fragment answer  • Argument 5: subject wh-questions are always exhaustive; non-exhaustive subject wh-questions are clefted. 108  Chapter 3: A′-movement and A′-agreement in Medumba  3.1 What is A′-agreement? Usually referred to as wh-agreement, wh-copying or extraction morphology, A′-agreement is the ‘morphological’ reflex of A′-movement of an XP (Chung 1994, Fanselow and Mahajan 2000, Felser 2004, Carstens 2005, Reintges, LeSourd and Chung 2006, Wasike 2007, Hedinger 2008, Lochbihler and Mathieu 2010, Zentz 2016 a.o.). In the literature, this phenomenon is often referred to as wh-agreement. Choice of this term reflects the fact that previous research has focused on classical cases of wh-movement (including content questions, relative clauses and focus movement). To date, there exists no systematic investigation of wh-agreement across different A′-movement contexts. Since in Medumba this phenomenon occurs in every context where there is A′-movement of an XP –– namely with content questions, relative clauses, focus movement and topicalization –– I refer to this phenomenon as A′-agreement.   Although A′-agreement and related phenomena in many languages have been used as a diagnostic of successive cyclic movement, its cross-linguistic formalization is still not yet thoroughly established. In this chapter, I argue that A′-agreement associates with different cluster of properties when one moves from one language to another. However, all instances of A′-agreement reflect the activity of the same formal mechanism, namely Phasal-Agree: I call this the agree hypothesis (see also Carstens 2005, Reintges et al. 2006). (1) Agree hypothesis: A′-agreement is the reflex of Phasal-Agree  109  3.1.1 The locus and form of A′-agreement in Medumba 3.1.1.1 Detecting the basic form of the verb: yes/no question frame Medumba has been variously described as having a contrastive: (i)  two-tones (H vs L) system (Voorhoeve 1974; Franich 2014); (ii)  four-tones (H vs L vs M vs LH) system (Wandji 1993); or  (iii) five-tones (H vs L vs M vs HL vs LH) system (Tondji 1979, Kouankem 2012).  These descriptions are usually based on the surface forms with little attention paid to all the different surface allomorphs or to word classes. By attending to tonal distribution between lexical and functional categories in Medumba, I adopt a rather radical approach to the analysis16 of tone contrasts in Medumba: I posit that Medumba lexical categories contrast a low versus an unmarked tone (L vs Æ), and that the unmarked tone with lexical categories surfaces as a default high tone (Æ®H). With regard to functional categories, they show a three-way partition between L, H, and a toneless mora (which copies the tone of the element that precedes it). The Medumba tone inventory is given in (2).           (2) Medumba tone inventory: L-cats: {L,  Æ} (surface L/H contrast, Æ®H)      F-cats: {L, H, μ} (surface L/H contrast, μ ®H/L)   In table 3.1, I summarize the different surface tonal allomorphs of verbs in Medumba which will be key in diagnosing A′-agreement in Medumba.       16 See Barczak 2007; Déchaine 2001 and Déchaine 2015 as a precedent for this type of partition  110    BASE Surface tonal allomorphs conditions Phonological Morphological Syntax –––#(Hμ)  N-NMLZ(Lμ) N-(H) –––(H)  DP HLμ(overwrite)17 1 CVL CVVLH CVVLL CVVHL CVVLH CVVHL 2 CV CVH CVVHL CVH CVH CVVHL 3 CVCL CVCVLH CVVCLL CVVCHL CVVCLH CVVCHL 4 CVC CVCVHH CVVCHL CVCH CVCH CVVCHL  Contexts Final position Nominalization (i) non-initial V (ii) V2-form in SVCs (i) Inf. V DP (ii) VTNS DP A’-movement Table 3. 1: Verb tone classes and their surface tonal allomorphs in Medumba As shown in appendix B, there are four verb tone classes in Medumba: CVL vs CV and CVCL vs CVC. In the examples below, I use the yes/no question frame as the diagnostic of Medumba verbal base forms. The tone classes that are analyzed as unmarked surface in this context as high (default H-tone) as illustrated in (3).  (3) a. Nùᵑgɛ̀ɛ́  kɛ̀  kí    CVL ® CVL  Nuga.H choose  C.QY/N    V.L  ‘Did Nuga choose ?         b. Nùᵑgɛ̀ɛ́  ↓kɛ́  kí    CV ® CVH  Nuga.H fry  C.QY/N    V.H   ‘Did Nuga fry?         c. Nùᵑgɛ̀ɛ́  kɛ̀k  kí    CVCL ® CVCL  Nuga.H weed  C.QY/N    V.L   ‘Did Nuga weed?’         d. Nùᵑgɛ̀ɛ́  ↓kɛ́k  kí    CVC ® CVCH   Nuga.H harvest  C.QY/N    V.H    ‘Did Nuga harvest (it)?  17 Exception to this are the Aux. fə̀ and lù (which appear with a LH contour) and the future marker àʔ which doesn’t change in A-bar context. 111  3.1.1.2 Detecting the reflex of A′-agreement: HL tone overwrite with A′-movement In the yes/no question frame provided in the previous section, there is a split between the two verb tone classes is observed in Medumba. Classes that are lexically marked as low surface as low (CVL and CVCL) whereas unmarked tone classes surface as high (CVH and CVCH). When there is A′-movement of an XP in Medumba there is a complete neutralization of the different verb tone classes: all verb classes surface with an HL tonal melody. The marked tone classes CVL and CVCL all surface as CVVHL and CVVCHL and the unmarked tone classes CV and CVC also surface as CVVHL and CVVCHL. This is illustrated in the following examples in which are contrasted the verbal tone in yes/no interrogative and wh-interrogative. (4) CVL verb  a.  Nùᵑgɛ̀ɛ́  kɛ̀  kí  CVL ® CVL    Nuga.H choose  C.QY/N     V.L    ‘Did Nuga choose ?  b. á wʉ́ Nùŋgɛ̀  kɛ́ɛ̀  á  CVL ® CVVHL       FOC WH Nuga  AGR.choose  C.Q.H      V.HL         ‘Who did Nuga choose?’ (5) CV verb   a.  Nùᵑgɛ̀ɛ́  ↓kɛ́  kí  CV ® CVH     Nuga.H fry  C.QY/N      V.H     ‘Did Nuga fry?  b. á kʉ́ Nùŋgɛ̀  kɛ́ɛ̀  á  CV ® CVVHL      FOC WH Nuga  AGR.fry C.Q.H      V.HL      ‘What did Nuga fry?   112  (6) CVCL verb     a.  Nùᵑgɛ̀ɛ́  kɛ̀k  kí  CVCL ® CVCL    Nuga.H weed  C.QY/N      V.L     ‘Did Nuga weed?’  b. á kʉ́ Nùŋgɛ̀  kɛ́ɛ̀g  á  CVCL ® CVVCHL      FOC WH Nuga  AGR.weed C.Q.H      V.HL      ‘What did Nuga weed?’ (7)  CVC verb   a.  Nùᵑgɛ̀ɛ́  ↓kɛ́k  kí  CVC ® CVCH     Nuga.H harvest  C.QY/N      V.H      ‘Did Nuga harvest (it)?  b. á kʉ́ Nùŋgɛ̀  kɛ́ɛ̀g  á  CVC ® CVVCHL      FOC WH Nuga  AGR.harvest C.Q.H      V.HL      ‘What did Nuga harvest?’  The HL overwrite pattern found with the above examples is not restricted only to wh-movement in Medumba. It is actually found in all A′-movement contexts, including focus-movement (8), relativization (9), and topicalization (10).  (8) Focus movement  a. á ᵑgùn Nùŋgɛ̀ kɛ́ɛ̀  lá  CVL ® CVVHL   FOC girl Nuga AGR.choose  C.-Q      V.HL       ‘The girlFOC (is the one) Nuga chose’          b. á ⁿʤʷɛ́n Nùŋgɛ̀ kɛ́ɛ̀  lá  CV ® CVVHL   FOC chips Nuga AGR.fry C.-Q      V.HL   ‘The chipFOC Nuga fried’  113           c. á nà Nùŋgɛ̀ kɛ́ɛ̀g  lá  CVCL ® CVVCHL   FOC field Nuga AGR.weed C.-Q      V.HL   ‘The fieldFOC Nuga weeded’          d. á ŋʷɔ́ʔɔ́ Nùŋgɛ̀ kɛ́ɛ̀g  lá  CVC ® CVVCHL   FOC honey Nuga AGR.harvest C.-Q      V.HL   ‘The honeyFOC Nuga harvested’ (9) Relativization   a. ᵑgùún zə̀  Nùŋgɛ̀ kɛ́ɛ̀  lá … CVL ®  CVVHL   girl.H C.CL1.L Nuga AGR.choose  C.-Q       V.HL       ‘The girl that Nuga chose …’          b. ⁿʤʷɛ́n zə̀  Nùŋgɛ̀ kɛ́ɛ̀  lá … CV ® CVVHL   chips C.CL1.L Nuga AGR.fry C.-Q       V.HL   ‘The chip that Nuga fried …’          c. nàá zə̀  Nùŋgɛ̀ kɛ́ɛ̀g  lá … CVCL ®  CVVCHL   field.H C.CL1.L Nuga AGR.weed C.-Q       V.HL   ‘The field that Nuga weeded …’          d. ŋʷɔ́ʔɔ́ zə̀  Nùŋgɛ̀ kɛ́ɛ̀g  lá CVC ®  CVVCHL   honey C.CL1.L Nuga AGR.harvest C.-Q       V.HL   ‘The honey that Nuga harvested …’ (10) Topicalization18  a. ᵑgùn jʉ́ʉ̀n-ní  kí Nùŋgɛ̀ kɛ́ɛ̀  í CVL ®  CVVHL  girl AGR-1PROX TOP Nuga AGR.choose  3SG.ANIM       V.HL      ‘This girl, Nuga chose her’   18 Object pronouns with regard to topicalization show an animacy contrast in Medumba. Animate pronouns are overt whereas inanimate pronouns are null. 114          b. ⁿʤʷɛ́n ʧʉ́ʉ̀n-ní kí Nùŋgɛ̀ kɛ́ɛ̀  ∅ CV ® CVVHL  chips AGR-1PROX TOP Nuga AGR.fry 3SG.INAM        V.HL  ‘These chips, Nuga fried them’        c. nà jʉ́ʉ̀n-ní  kí Nùŋgɛ̀ kɛ́ɛ̀k  ∅	 CVCL ®  CVVCHL  field AGR-1PROX TOP Nuga AGR.weed 3SG.INAM        HL  ‘This field, Nuga weeded it?’         d. ŋʷɔ́ʔɔ́ jʉ́ʉ̀n-ní  kí Nùŋgɛ̀ kɛ́ɛ̀k  ∅ CVC ®  CVVCHL  honey AGR-1PROX TOP Nuga AGR.harvest 3SG.INAM        HL  ‘This honey, Nuga harvested it?’  The HL tonal melody found with Medumba A′-movement overwrites lexical tone. I consider this tonal melody to be a form of A′-agreement signaling A′-movement of an XP in Medumba. With this in place, I now introduce the formal mechanism that accounts for the appearance of a dedicated tone melody with A′-movement, namely phasal agree. 3.1.2 The proposal: Phasal-Agree derives A′-agreement 3.1.2.1 There are (at least) two phases: vP and CP A phase is an economy principle designed to solve derivational complexities. It is a domain within which all derivational processes operate at the same time and where all features are checked (Chomsky 2001). A phase consists of a phase head and a phase domain, also known as the phase spell out domain or the complement of the phase head. Whenever any derivation reaches a phase and all the features are checked, the phase spell out domain is sent to transfer and is invisible to further computations. Any movement must obey the Phase Impenetrability Condition (PIC) defined as:  (11) Phase Impenetrability Condition: “The domain of H is not accessible to operations outside HP. Only H and its edge are accessible to such operations” (Chomsky 2001:13).  115   The edge is any elements outside H (the phase head), which can be specifiers or elements adjoined to HP (Chomsky 2001: 13). I consider CPs and vPs as phase boundaries in Medumba. The choice of these phase boundaries19 is motivated by the fact that they are “propositional” in nature. That is, they are either verbal phrases with full argument structure or CPs with force indicators (See Chomsky 2000, 2001). A phase and its different constituents are schematized in (12). (12)          ...         ei        XP     → Phase boundary               qp             ...               XP  (Phase Edge)            qp                        X                 ZP   (Phase head)                (Phase spell out Domain)  3.1.2.2 A′-agreement is Phasal-Agree Agree is the basic dependency-building mechanism within the Minimalist framework. It involves an operation of feature checking between a probe and a goal. The operation Agree takes place only if the probe and the goal both bear uninterpretable/unvalued features, which make them active. After the operation Agree, the unvalued features of the probe and the goal are valued or deleted, making them inactive or invisible for further Agree operations. The operation Agree is schematized in (13) with α as a probe and β as a goal:    19 What constitutes a phase is subject to debate. Proposals vary from: every phrase (Müller 2010), CP and vP (van Urk 2015, Georgi 2017), just vP (Rackowski and Richards 2005), or more nuanced possibilities (Den Dikken 2007, Gallego 2007, Wurmbrand 2012, Harwood 2015, Ramchand and Svenonius 2014, Bošković 2014, Sailor 2014, Aelbrecht and Harwood 2015). 116  (13)   a. Feature matching phase     b. Agree phase                                              …                                           ...              3         3  αuF          ...         αuF                    ...            3       3          βuF    ...      βuF          ...     I propose that A′-agreement in Medumba is move-based and proceeds by phase. It creates what I call an ‘agreement chain’ within each phase spell out domain each time the moved XP reaches a phase edge and the relevant A′-feature is checked. In Medumba, when the moved XP reaches a phase edge, an HL overwrite tonal melody is added to the head of the complement of the phase head. This is Phasal-Agree: the result of an Agree operation that tracks the different movement and agreement steps of an A′-bound XP through the different phase-edges.  (14) Phasal-Agree: A phase-bound operation (OP) between a probe (P) and a goal (G), where P is a phase-head and G an A′-bound XP; applies in such a way that the reflex of OP is either on P or the complement of P.    I consider the A′-feature as a feature bundle relating to discourse or information structural features such as focus and topic (see also Lochbihler and Mathieu 2010). In some languages, it also involves ɸ-feature agreement with the moved constituent (see Carstens 2005, Baier 2018). The valuation of the A′-feature in Medumba is reflected by overwriting of lexical tone of phase-head complements. This overwrite tone surfaces in Medumba as an HL melody on relevant heads; namely, verbal heads, temporal auxiliaries and aspectual auxiliaries. I argue that A′-agreement is not only a crucial diagnostic for A′-movement but also for Phasal-Agree and for the locality of movement (cyclic phase-by-phase movement (see also Biberauer and D’Alessandro 2006, van Urk 2015, van Urk and Richards 2015). 117  3.2 How phasal agree derives A′-agreement in Medumba 3.2.1 Phasal-Agree predicts subject/non-subject asymmetry Analyzing A′-agreement in Medumba as a move-based Phasal-Agree operation predicts that subjects should pattern differently from non-subjects in Medumba. Assuming that subjects are first merged in the vP-internal position at Spec-v (Koopman and Sportiche 1991), when there is A′-movement from Spec-v, the next available position is spec-T before it reaches spec-C. Therefore, the reflex of agreement can only appear on T when the moved constituent reaches the CP phase edge in root-clauses as TP is the complement of the C phase-head. As for objects, they are first merged within the VP-domain as the complement of V. When an XP undergoes A′-movement from the object position, the next available escape hatch is to adjoin at vP-phase edge in order to avoid the PIC (Chomsky 2001). The reflex of agreement is realized on V as VP is the complement of the v phase-head. When the next phase is built the object can then continue to its landing position at the edge of the CP-phase, namely Spec-C where the reflex of agreement is spelled out on T. As stated above, the reflex of A′-agreement surfaces on the complement of phase-heads, and T and V are respectively the head of the complement of C for the CP-phase and v for the vP-phase. This is illustrated by the tree structures in (15).        118  (15) a. Subject      b. Object            The predicted patterns of A′-agreement are summarized in the table below for Medumba root-clauses: we expect A′-agreement to surface within the CP-domain for subjects and within the CP- and vP-domain for objects.   Locus of agreement Extraction site CP vP Subject ✓ ✗ Object ✓ ✓ Table 3. 2: The locus of A′-agreement in Medumba root-clauses The examples below illustrate the locus of A′-agreement in Medumba root-clauses. (16a) shows that when the subject wh-phrase á wʉ́ ‘who’ moves from the vP-internal position to Spec-C, T (in the case headed by Aux.T2 nɔ́ʔ) surfaces with an HL tone melody. As shown in (16b), in the presence of A′-movement, absence of this HL melody on the T-head leads to ungrammaticality.  (16) Subject Wh-movement   a. á wʉ́ nɔ́ɔ̀ʔ  ⁿ-ʤʉ́n  Nùᵑgɛ̀ à?                         FOC WH AGR.AUX.T2 N-see  Nuga C.Q.L     T.HL  V.H           ‘Who saw Nuga?’            CP → Phase boundary      ru                      CP                  ru              C              TP                     3                                    TP                           ru                         T[AGR]        vP  → Phase boundary                                                                                                                 ru                               SUBJ          vP                                            ru                                           v                VP                                                     ru                                                   V               DP          CP → Phase boundary    ru                      CP                 ru              C                 TP                        ru                      T[Agr]            vP → Phase boundary                                                                                                                 ru                                                  vP                                          ru                                      Subj.            vP                                                   ru                                                 v               VP                   3                V[Agr]        DP 119   b. *á wʉ́ nɔ́ʔ  ʤʉ́n Nùᵑgɛ̀ à?                           FOC WH AUX.T2 see Nuga C.Q.L     T.H  V.H            [Who saw Nuga?] With regard to object extraction, (17a) shows that when the object wh-phrase á kʉ́ ‘what’ moves, it first stops at the vP-edge to avoid violating PIC. In that position agreement is expressed on the V head jʉ́n ‘see’ which now surfaces with an HL tone melody (ⁿ-ʤʉ́ʉ̀n). The Wh-XP continues to Spec-C where it is interpreted and, in that position, agreement is expressed on T nɔ́ʔ as an HL tone melody (nɔ́ɔ̀ʔ). The absence of A′-agreement with V and T when the object is A′-moved leads to ungrammaticality (17b). (17) Object wh-movement  a. á wʉ́ Wàtɛ̀t nɔ́ɔ̀ʔ  ⁿ-ʤʉ́ʉ̀n á?                         FOC WH Watat AGR.AUX.T2 N-AGR.see C.Q.H      T.HL        V.HL         ‘Who did Watat see?’  b. *á wʉ́ Wàtɛ̀t nɔ́ʔ  ʤʉ́n á?                          FOC WH Watat AUX.T2 see C.Q.L      T.H  V.H          [Who did Watat see?] The structures in (18) illustrates the derivation of subject and object wh-movement in Medumba.         120  (18) a. Subject    b. Object               A question that arises is why the subject extraction does not trigger A′-agreement with the verb when the subject first merged at Spec-v. It appears that A′-agreement in Medumba requires movement. So, A′-agreement is not expected at first merge, that is between the verb and the subject at Spec-v. This correctly predicts that A′-agreement will not occur with in-situ XPs in general, including in-situ object XPs. This is confirmed in (19) for in-situ wh-questions, and in (20) for in-situ focus-marking. A′-agreement with in-situ XP is illicit in Medumba (see 19b and 20b). (19) In-situ wh-question  a. m-ɛ́n nɔ́ʔ  fá bɔ̀ á wʉ́ á                                         C1-child AUX.T2 give bag FOC WH C.Q.H     T.H  V.H                    Lit.: ‘the child gave the bag to who(m)?’       CP    2          C         CP   á    ru    DP              CP     4       ru  á wʉ́     CØEXH        TP                          ru                         T              vP                                                                                                       nɔ́ɔ̀ʔ     ru                             HL        DP            vP                                  4     ru                             <á wʉ́>    v             VP                                                     ru                                                    V              DP        ⁿʤʉ́n           4            Nùᵑgɛ̀	          CP    2        C         CP   á    ru       DP            CP         4     ru     á wʉ́    CØEXH        TP    2 DP          TP                          4    ru                       Wàtɛ̀t   T              vP                                                                                                                   nɔ́ɔ̀ʔ     ru                                   HL                       vP                                                        ru                                                      DP            vP                                                     4      ru                                                 <Wàtɛ̀t>  v             VP                                 3                                V             DP                           ⁿʤʉ́ʉ̀n 4                 HL           á wʉ́ 121   b. *m-ɛ́n nɔ́ɔ̀ʔ  ᵐ-fáà  bɔ̀ á wʉ́ á                                           C1-child AGR.AUX.T2 N-AGR.give bag FOC WH C.Q.H     T.HL        V.HL                    [Lit.: ‘the child gave the bag to who(m)?] (20) In-situ Focus  a. m-ɛ́n nɔ́ʔ  fá á bɔ̀  Nùᵑgɛ̀                                         C1-child AUX.T2 give FOC bag Nuga    T.H  V.H                     ‘The child gave [the bagFOC] to Nuga’     b. *m-ɛ́n nɔ́ɔ̀ʔ  ᵐ-fáà  á bɔ̀  Nùᵑgɛ̀                                         C1-child AGR.AUX.T2 N-AGR.give FOC bag Nuga    T.HL       V.HL                     [The child gave [the bagFOC] to Nuga]    Thus, it is illicit to have A′-agreement in the absence of A′-movement in Medumba. This is consistent with analyzing A′-agreement as the by-product of move-based Phasal-Agree in Medumba. 3.2.2 Phasal-movement and the absence of Superiority effects in Medumba A language like English allows only one wh-XP to be fronted in multiple wh-questions with bare wh-XPs like who/what. Any additional wh-XP must remain in-situ. Moreover, there is a constraint known as the Superiority effects which governs which wh-XP must be fronted. Thus, only the closest wh-XP to C can move to spec-C (see Kuno and Robinson 1972, Chomsky 1973). This explains the grammaticality of (21) and the ungrammaticality of (22). (21) a. Who ate what?  b. Who went where?  c. What happened to whom?  d. What did you give to whom?  122  (22) a. *What did who ate?  b. *Where did who go?  c. *To whom did what happen?  d. *To whom did you give what? Medumba does not work the same way. Multiple wh-fronting is possible in Medumba multiple wh-questions and thus violates the Superiority effect. I argue that Phasal-Agree could explain the lack of superiority effect found in Medumba. In chapter 2, I showed that subject wh-/focus does not need to be at Spec-C to be interpreted as exhaustive. In fact, subject wh-/focus can be interpreted as exhaustive in Spec-T as they just need to be within the vicinity (same phase) of the covert exhaustive operator. This makes it possible for the object wh-phrase to move to Spec-C while the subject wh-phrase is in Spec-T. This configuration is repeated in (23). (23)  CP            3    Spec            CP  wh-XPOBJ    3                     CExh         TP                             ei             Spec                  TP                    wh-XPSUBJ      3                                   T             vP  The configuration in (23) predicts the absence of superiority in Medumba. In fact, when the first phase (vP) is built, the object wh-XP can move and adjoins to vP to avoid violating the PIC as illustrated in the structure in (24).     123   (24)      TP            3           T              vP          3     wh-XP vP          3      DPSUBJ       VP (PIC application domain)      4    3               wh-XP   V             DPOBJ              4           <wh-P>    When the next phase (CP) is built the object can move to Spec-C while the subject moves to Spec-T, creating a multiple wh-question where the object and the subject are both exhaustive.  (25) CP      3   Spec         CP wh-XP    3             CExh             TP                            3           Spec          TP         wh-XP   3            T              vP          3              <wh-XP>        vP          3     DPSUB         VP       4    3            <wh-XP>  V            DPOBJ              4           <wh-XP>  This prediction is borne out in Medumba. I first show that in Multiple wh-questions in Medumba, it is possible to front one wh-XP and leave the other XPs in situ. (26a) shows a fronted subject and 124  an in-situ direct object; (26b) shows a fronted subject and an in-situ indirect object, and (26c) shows a fronted subject with all the other arguments in-situ.  (26) a. á wʉ́ nɔ́ɔ̀ʔ  ᵐ-fá  á kʉ́ Wàtɛ̀t à?       FOC WH AGR.AUX.T2 N-give  FOC WH Watat C.Q.L T.HL  V.H               ‘Who gave what to Watat?’  b. á wʉ́ nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ á wʉ́ á?       FOC WH AGR.AUX.T2 N-give  bag FOC WH C.Q.H T.HL  V.H               ‘Who gave the bag to whom?’  c. á wʉ́ nɔ́ɔ̀ʔ  ᵐ-fá  á kʉ́ wʉ́ á?20       FOC WH AGR.AUX.T2 N-give  FOC WH WH C.Q.H T.HL  V.H                ‘Who gave what to whom?’ (27) confirms that it is possible to have multiple wh-fronting21 in Medumba where two wh-phrases are fronted at the same time. In (27a), the inanimate object wh-XP kʉ́ ‘what’ and the subject wh-XP are fronted and in (27b) the animate object wh-XP wʉ́ ‘who’ and the subject wh-XP are fronted.   (27) a. á kʉ́ á wʉ́ nɔ́ɔ̀ʔ        ᵐ-fáà      á wʉ́ á?      FOC WH FOC WH AGR.AUX.T2 N-AGR.give  FOC WH C.Q.H     HL  HL              ‘*What did who gave to whom?’  b. á wʉ́ á wʉ́    nɔ́ɔ̀ʔ      ᵐ-fáà   á      kʉ́    jí        á?       FOC WH FOC WH   AGR.AUX.T2 N-AGR.give  FOC  WH    3SG.H   C.Q.H             T.HL         v.HL             ‘*Who did who gave what to him?’ Interestingly, only two wh-XPs can be fronted at the same time in Medumba and one of them must be a subject. (28) shows that it is illicit to have more than two fronted wh-XPs in  20 With multiple wh-in-situ, only the first wh-XP is focus marked in Medumba (focus absorption?). The question of this is beyond the scope of the present dissertation and will be the subject of future research.   21 It is to be noted here that what I call multiple wh-fronting in Medumba does not involve multiple A’-fronting as the subject stays in spec-TP. 125  Medumba, and (29) shows that even though two wh-phrases can be fronted, it is illicit if one of them is not a subject.    (28) a. *á        kʉ́      á     wʉ́    á  wʉ́      nɔ́ɔ̀ʔ  ᵐ-fáà             jí         á?         FOC    WH       FOC   WH     FOC   WH    AGR.AUX.T2 N-AGR.give  3SG.H   C.Q.H   b. *á        wʉ́      á     kʉ́      á      wʉ́     nɔ́ɔ̀ʔ  ᵐ-fáà              jí         á?         FOC    WH       FOC   WH     FOC   WH     AGR.AUX.T2 N-AGR.give  3SG.H   C.Q.H (29) a. *á        kʉ́      á     wʉ́    Nùᵑgɛ̀ nɔ́ɔ̀ʔ        ᵐ-fáà              jí         á?         FOC    WH       FOC   WH     Nuga  AGR.AUX.T2 N-AGR.give  3SG.H   C.Q.H  b. *á        wʉ́      á     kʉ́      Nùᵑgɛ̀     nɔ́ɔ̀ʔ  ᵐ-fáà              jí         á?         FOC    WH       FOC   WH     Nuga     AGR.AUX.T2 N-AGR.give  3SG.H   C.Q.H Multiple wh-fronting in Medumba does not create additional Spec-C positions. The lack of superiority in Medumba is a result of a conspiracy between Phasal-Agree and the ability of subject wh-XPs to move to Spec-T in Medumba. Thus, Medumba wh-questions are predicted to be sensitive to superiority if the subject wh-XP moves to Spec-C. As discussed in chapter 2, subject wh-movement to Spec-C is detected by the presence of a resumptive pronoun. This is confirmed in (30). (30) a. *á kʉ́ á wʉ́ à nɔ́ɔ̀ʔ        ᵐ-fáà        Nùᵑgɛ̀ à?        FOC WH FOC WH 3SG.H AGR.AUX.T2   N-AGR.give  Nuga C.Q.L      T.HL  v.HL                 [*What did who [he] gave to Nuga?]  b. *á wʉ́ á wʉ́    à  nɔ́ɔ̀ʔ          ᵐ-fáà           bɔ̀    jí         á?       FOC WH FOC WH   3SG.H AGR.AUX.T2    N-AGR.give   bag    3SG.H   C.Q.H               T.HL         v.HL                [*Who did who [he] gave the bag to [him]?]  3.2.3 The surface realization of A′-agreement 3.2.3.1 The problem: iterated agreement is incompatible with agree This section addresses the issue of multiple Agree and how to handle it within the Minimalist Framework. In fact, the reflex of Agree at every phase level poses a problem to the standard notion 126  of Agree namely, how can an item enter subsequent Agree relations after valuing its features. After the operation Agree, the unvalued features of the probe and the goal are valued or deleted, making them inactive or invisible for further Agree operations. This is shown below with α as a probe and β as a goal: (31)   a. Feature matching phase     b. Agree phase               ...          ...              3         3  αuF          ...         αuF                    ...            3       3          βuF    ...      βuF          ...   The problem is how to account for multiple Phasal-Agree with root clauses and non-root clauses. After feature checking, the constituents involved in that operation are inactive or invisible for further Agree operations as their unvalued features are checked. So, with a phase-based derivation of ex-situ wh-/focus XPs, the derivation will crash at the first phase level (vP) for two reasons: (i) all the features are checked, therefore the goal is invisible for further Agree operations, and (ii) the A′-moved XP is stuck at the edge of the first phase. Given that this is not a position where it is interpreted, the derivation crashes. Consider for example how Agree operates in Kilega.  Carstens (2005) proposes that C has an uninterpretable Q-feature which initiate the Agree relations and which must be deleted by a matching feature of the closest c-commanded wh-operator. In long distance extraction, she argues that the moved XP first raises to embedded Spec-v and then to the embedded Spec-C where agreement is obligatory. Then from there it can move to the specifier position of the matrix vP, and finally to the specifier position of the matrix CP where a second agreement is also obligatory as illustrated in (32).  (32) bi-kí bi-á-ténd-ílé  b-ána bi-á-gúl-ílé  nina-bó  8-what 8.Agr-V-say-PFV 2-child 8.Agr-V-buy-PFV mother-their  ‘what did the children say their mother had bought?’ 127  It is not clear from Carstens’s analysis how Agree could handle the second Agree relation with the matrix CP and, also why agree is not possible with the embedded v before V-to-C movement as it would yield the same result with only one agreement prefix on the verbal complex. In fact, under standard Agree, the wh-XP at the edge of the embedded CP cannot be an active goal for the matrix C because the features of the wh-XP are already valued and deleted. This therefore makes it invisible to the matrix probe C as represented in (33). (33) CP               CP  3  CuX     vP                   4              CP         3   [whouX] CP         3        CuX  vP               4                        <who> One possibility to account for multiple instances of Agree is to assume that once the moved XP reaches the matrix CP edge, Agree can happen across phase boundaries at the same time with the matrix and embedded CP (see also Zeijlstra 2012). The problem with this view is that if we take the Phase Impenetrability Condition seriously there is no way to go back and value features inside lower phases because these are already sent for transfer and so are invisible to Agree. Therefore delaying Agree until the highest phase will make the derivation crash if there are unvalued features at the lower phases.  3.2.3.2 The solution: Resume Agree In order to account for multiple reflexes of Agree at different phase domains, I propose a condition on Agree with I called Resume Agree: 128  (34) Resume Agree: the features of a goal (G) are resumed at each phase so that G can remain active until it reaches the position where it is interpreted. With this condition in mind, when the moved XP first reaches the vP phase edge, the first Phasal-Agree operation takes place and the reflex of Agree surfaces on V. Then, the A′-features of the moved XP are resumed so that it can become an active goal for the next Phasal-Agree operation with the probe C22. Finally, when the moved XP reaches the edge of the CP-phase where it is interpreted, the reflex of Agree surfaces on T. At this stage, for root clauses, the features of the moved XP do not need to be resumed as it is interpreted at Spec-C. But with non-root clauses, this derivation predicts the features of the moved XP to be resumed so that it can reach the highest CP edge where it is interpreted. This is illustrated in (35). (35)  CP           3  whuX           CP      ei   CuX                  vP                                         3              [whuX] ←Resume Agree ←[whuX]         vP                                                              r                                                            vuX             CP                                             3                                [whuX] ←Resume Agree ←[whuX]            CP                                           3                                        CuX           vP                                               3                                     [whuX] ←Resume Agree ←[whuX]            vP                                               3                                              vuX            VP                                              3                                             V           DP                                                             <whu>  22 Another alternative would be to assume that the probe-goal relation can be asymmetric, with only features being checked on the probe. As such, that would leave the features on the goal active. 129  3.3 A surprising result: Phasal-Agree does not predict root/non-root CP asymmetry A move-based Phasal-Agree analysis of A′-agreement in Medumba predicts an agreement symmetry between root CPs and non-root CPs. That is, root-clauses and non-root clauses are expected to display the same agreement pattern. Consider the case of object extraction, as illustrated in (36); we expect phasal cyclic-movement through the different phase edges with agreement at every phasal-domain, namely the embedded V and T, as well as on the matrix V and T.  (36) CP       ru                           CP                    ru                    C               TP                             ru                           T[AGR]              vP                                         ru                                              v                 VP                                                      ru                                                    V[AGR]         CP                                                                                    ru                                                                                          CP                                                                        ru                                                                                  C              TP                                                                                             ru                    T[AGR]            vP                 ru                            v              VP               ru                                                                                    V[AGR]          DP   Medumba non-root CPs do not exhibit the predicted pattern of agreement; in particular, the matrix V fails to show A′-agreement. The table below summarizes the expected versus attested A′-agreement pattern with root CPs and non-root CPs in Medumba.  130   Extraction site Locus of agreement Embedded Subject Embedded Object  Expected Attested Expected Attested Matrix T ✓ ✓ ✓ ✓ V ✓ ✗ ✓ ✗ Embedded T ✓ ✓ ✓ ✓ V ✗ ✗ ✓ ✓ Table 3. 3: The locus of A′-agreement in Medumba non-root clauses Non-root clause extraction from a subject position triggers A′-agreement on T in matrix and embedded clauses. This is illustrated in (37) for wh-movement, in (38) for focus movement, and in (39) for relativization in which the matrix verb ⁿ-tʃúp ‘say’ surfaces with H-tone (the (a) examples) rather than with the expected A′-agreement HL tone melody, which is ill-form in this context (the (b) examples). (37) Subject wh-movement  a. á wʉ́ Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúp  ᵐbʉ̀     FOC WH Sami AGR.AUX.T2 N-say  C.L      T.HL      V.H      á  nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀ à?     3SG.H  AGR.AUX N-give  bag Nuga C.Q.L        T.HL  V.H     ‘Who did Sami say that [he] gave the bag to Nuga?’   b. *á wʉ́ Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúùp  ᵐbʉ̀        FOC WH Sami AGR.AUX.T2 N-AGR.say C.L      T.HL      V.HL      á  nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀ à?     3SG.H  AGR.AUX.T2 N-give  bag Nuga C.Q.L        T.HL   V.H  [Who did Sami say that [he] gave the bag to Nuga?]    131  (38) Subject focus-movement  a. á mɛ́n Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúp  ᵐbʉ̀     FOC child Sami AGR.AUX.T2 N-say  C.L      T.HL      V.H      á  nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀ lá     3SG.H  AGR.AUX.T2 N-give  bag Nuga C.-Q        T.HL   V.H  ‘The childFOC Sami said that [he] gave the bag to Nuga’   b. *á mɛ́n Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúùp  ᵐbʉ̀        FOC child Sami AGR.AUX.T2 N-AGR.say C.L      T.HL      V.HL      á  nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀ lá     3SG.H  AGR.AUX.T2 N-give  bag Nuga C.-Q        T.HL   V.H  [The childFOC Sami said that [he] gave the bag to Nuga] (39) Subject relativization  a. mɛ́n  zə̀ Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúp  ᵐbʉ̀      child  C.CL1 Sami AGR.AUX.T2 N-say  C.L       T.HL      V.H     á nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀ lá     3SG.H AGR.AUX.T2 N-give  bag Nuga C.-Q       T.HL   V.H  ‘The child that Sami said that [he] gave the bag to Nuga’   b. *mɛ́n zə̀ Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúùp  ᵐbʉ̀        child C.CL1 Sami AGR.AUX.T2 N-AGR.say C.L       T.HL      V.HL      á  nɔ́ɔ̀ʔ  ᵐ-fá  bɔ̀ Nùᵑgɛ̀ lá     3SG.H  AGR.AUX.T2 N-give  bag Nuga C.-Q        T.HL   V.H  [The child that Sami said that [he] gave the bag to Nuga]  With non-root clause extraction from an object position, there is A′-agreement with the embedded V and T in the embedded clause, but A′-agreement only on T in the matrix clause. This is shown in (40) for wh-movement, (41) for focus movement, and (42) for relativization. The 132  matrix V ⁿ-tʃúp ‘say’ surfaces with H-tone (the (a) examples) rather than with the expected A′-agreement HL tone melody which is ungrammatical in this context (the (b) examples). (40) Object wh-movement   a. á wʉ́ Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúp  ᵐbʉ̀                 FOC WH Sami AGR.AUX.T2 N-say  C.L     T.HL       V.H     Nùᵑgɛ̀ɛ́  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n  í  á?     Nuga.H AGR.AUX.T2 N-AGR.give 3SG.H  C.Q.H       T.HL      V.HL      ‘Who did Sami say that Nuga betrayed [him/her]?’  b. *á wʉ́ Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúùp  ᵐbʉ̀                   FOC WH Sami AGR.AUX.T2 N-AGR.say C.L     T.HL      V.HL     Nùᵑgɛ̀ɛ́  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n  í  á?     Nuga.H AGR.AUX.T2 N-AGR.give 3SG.H  C.Q.H       T.HL      V.HL       [Who did Sami say that Nuga betrayed [him/her]?]    (41) Object focus-movement   a. á mɛ́n Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúp  ᵐbʉ̀                 FOC child Sami AGR.AUX.T2 N-say  C.L     T.HL       V.H     Nùᵑgɛ̀ɛ́  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n  í  lá     Nuga.H AGR.AUX.T2 N-AGR.give 3SG.H  C.-Q       T.HL      V.HL      ‘The childFOC Sami said that Nuga betrayed [him/her]’  b. *á mɛ́n Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúùp  ᵐbʉ̀                   FOC child Sami AGR.AUX.T2 N-AGR.say C.L     T.HL      V.HL     Nùᵑgɛ̀ɛ́  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n  í  lá     Nuga.H AGR.AUX.T2 N-AGR.give 3SG.H  C.-Q       T.HL      V.HL      [‘The childFOC Sami said that Nuga betrayed [him/her]]     133  (42) Object Relativization   a. mɛ́n  zə̀ Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúp  ᵐbʉ̀                 child  C.CL1 Sami AGR.AUX.T2 N-say  C.L      T.HL       V.H         Nùᵑgɛ̀ɛ́  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n  í  lá     Nuga.H AGR.AUX.T2 N-AGR.give 3SG.H  C.-Q       T.HL      V.HL      ‘The child that Sami said that Nuga betrayed [him/her]’  b. *á mɛ́n zə̀ Sɛ̀ɛ́mí nɔ́ɔ̀ʔ  ⁿ-tʃúùp  ᵐbʉ̀                   FOC child C.CL1 Sami AGR.AUX.T2 N-AGR.say C.L      T.HL       V.HL      Nùᵑgɛ̀ɛ́  nɔ́ɔ̀ʔ  ⁿ-sʷɛ́ɛ̀n  í  lá     Nuga.H AGR.AUX.T2 N-AGR.give 3SG.H  C.-Q       T.HL      V.HL      [‘The child that Sami said that Nuga betrayed [him/her]]      The matrix verb fails to show A′-agreement when there is A′-extraction from a non-root clause. Any attempt to force A′-agreement (in the form of the HL tone melody) with the matrix V results in ungrammaticality. In a phase-based approach to A′-agreement, this implies that in a non-root clause extraction, the matrix vP phase does not participate in A′-agreement as summarized in the table below.  Root clause extraction Locus of A′-agreement  CP vP Root YES YES  Non-root clause extraction  CP vP Matrix YES NO Embedded YES YES Table 3. 4: Phasal-Agree and clause typing in Medumba  To account for the lack of A′-agreement with matrix vP phase, two hypotheses are possible: H1: either there is local phasal movement (phase-by-phase movement) with root and embedded CPs and long movement Spec-C-to-Spec-C after the first CP with non-root clauses; or  134  H2: embedded clauses are extraposed as vP adjuncts clauses and thus movement does not proceed through the edge of the matrix vP phase. In the next section, I go through both hypotheses and argue that the second hypothesis better accounts for the Medumba facts.  3.3.1 Hypothesis 1: Spec-C-to-Spec-C movement with non-root CPs This hypothesis assumes that there are short local phase-by-phase cyclic-movement with root clauses and embedded clauses in Medumba, and then, long movement (Spec-C-to-Spec-C) after the first CP with non-root clauses. Thus, A′-agreement is expected on T with intermediate CP phases as illustrated in (43).               135  (43)   CP         ru                       CP                  ru                 C             TP                          ru                         T[AGR]        vP                                     ru                                 Subj.          vP                                            ru                                          v                VP                                                     ru                                                   V               CP                      ru                                            CP                                 ru                             C              TP                                     ru                                 T[AGR]       vP                                                                                                                                                                        ru                                                                         vP                                                       ru                                                    Subj.           vP                                                                             ru                                                               v