On the Nature of Ergativity
B. C. Levin, 1983
This item is not currently available.
Languages showing some form of ergativity have posed a problem for theories of grammatical relations. This thesis examines a particular hypothesis concerning the nature of at least some ergative phenomena: Marantz"s Ergativity Hypothesis (1981). The importance of this hypothesis stems from the specific claim it makes about the nature of ergativity: the ergativity of a language depends on the particular association between semantic roles and grammatical relations that the language utilizes. In particular, the following assignments of semantic roles to grammatical relations characterize ergative and accusative languages, respectively.
(A): Accusative (E): Ergative
agent-subject agent-object
patient-object patient-subject
A number of properties predicted by this hypothesis for ergative languages are investigated in order to provide a picture of ergative languages. These predictions involve case assignment, semantic composition, reflexive constructions, passive and anti-passive constructions, and control constructions. Evidence in support of these properties is presented. In addition, an explanation of many of the properties traditionally associated with so-called ergative languages will be given.
The status of three representatitive "ergative" languages, Basque, Dyirbal, and Walpiri, is examined in detail in light of the Ergativity Hypothesis. Evidence will be presented that Walpiri, described as a "morphologically" ergative language with an ergative system of case marking and an ergative system of person marking, is an accusative language in the sense of the Ergativity Hyptohesis with an ergative system of case marking. Dyirbal, cited as the canonical example of a "syntactically" ergative language, will be shown to be an ergative language in the sense of the Ergativity Hypothesis. Basque, a second "morphologically" ergative language with ergative systems of case and person marking, will also be an accusative language. But, the Basque system of case marking turns out to be neither ergative or accusative. An alternative account of Basque case marking will be proposed.
Thesis Supervisor: Kenneth Locke Hale
Title: Ferrari P. Ward Professor of Linguistics
1 Introduction 9
1.1 The Ergativity Hypothesis 9
1.2 Theoretical framework 14
1.2.1 D-structure and semantic role assignment 15
1.2.2 Semantic role assignment properties of verbs 18
1.2.3 The mapping between D-structure and S-structure 24
1.2.4 S-structure and the theory of case 26
1.3 Semantic roles and verb classes 31
1.4 Overview 34
2 The Ergativity Hypothesis 36
2.1 Agent-patient verbs: the motivation for the Ergativity Hypothesis 36
2.2 Delining the scope of the Ergativity Hypothesis 43
2.3 Exceptions to the Ergativity Hypothesis 50
3 Properties of ergative languages 58
3.1 Predictions concerning case marking 58
3.1.1 Case systems in ergative and accusative languages 60
3.1.2 Case marking with expressions of duration 65
3.1.3 Case marking with expressions of goal 68
3.1.4 The adjacency condition on case assignment 70
3.1.5 The genitive of negation 72
3.2 Predictions involving properties of D-structure 74
3.2.1 Range of meanings 76
3.2.2 The structure of verbal idioms 77
3.2.3 Quirky case 83
3.3 The passive and anti-passive 87
3.3.1 The passive construction in an ergative language 88
3.3.2 The anti-passive construction 89
3.3.3 Impersonal passives 96
3.4 Predictions involving reflexive verb forms 100
3.4.1 The reflexive-passive ambiguity 103
3.4.2 Inherent reflexives 108
3.4.3 The anti-causative alternation 111
3.4.4 Middle constructions 120
3.5 Predictions involving the binding theory 121
3.5.1 The controlled argument 123
3.5.2 Control verbs in ergative languages 130
3.5.3 Reflexive and reciprocal anaphors 131
4 Walpiri 137
4.1 Simple verbal sentences 138
4.2 The system of case marking and person marking in Walpiri 140
4.3 Single argument verbs 146
4.4 Other verb classes 151
4.4.1 The ERG-ABS-DAT verbs 152
4.4.2 The ABS-DAT verbs 154
4.4.3 The ERG-DAT verbs 158
4.5 Summary of case array and person marking facts 160
4.6 Evidence that Walpiri is accusative 163
4.6.1 Evidence from the binding theory 165
4.6.1.1 Evidence from control structures 168
4.6.1.2 Evidence from the reflexive construction 171
4.6.2 The structure of idioms and range of meanings tests 174
4.6.2.1 The range of meanings test 175
4.6.2.2 The idiom test 178
4.6.2.3 Expressions of physiological states 181
4.7 Ergativity tests that are not applicable to Walpiri 187
4.7.1 Tests involving the reflexive verbal form 187
4.7.2 Quantifier tests 190
4.7.3 Implications for LS structure 192
4.8 The subject and object in Walpiri 194
4.8.1 The subject in Walpiri 194
4.8.2 The object in Walpiri 200
4.9 The Walpiri system of case marking reconsidered 209
4.10 Conclusion 213
4.11 Appendix: Sample lexical entries 214
5 Dyirbal 223
5.1 Simple verbal sentences 225
5.2 The system of case marking 229
5.3 Dyirbal as a "syntactically" ergative language 233
5.4 Evidence from the reflexive-passive ambiguity 240
5.5 Simple sentence syntax reconsidered 244
5.6 Further evidence for Dyirbal as an ergative language 251
5.6.1 Further evidence involving reflexive morphology 251
5.6.2 Lexical properties 254
5.6.3 The binding theory 258
5.6.3.1 Reflexive and reciprocal anaphor evidence 258
5.6.3.2 Control in purposive constructions 259
5.6.3.3 -ngurra clauses 267
5.6.4 The pro drop parameter 270
5.6.5 Word order and configurationality 272
5.7 Other so-called "ergative" phenomena 276
5.7.1 The topic chain construction 276
5.7.2 Relative clauses 280
5.7.3 The imperative construction 283
5.8 Conclusion 287
5.9 Appendix: Verbs of giving 287
6 Basque 290
6.1 Simple verbal sentences 290
6.2 The system of case marking and agreement 294
6.3 The composition of the NOR class 296
6.3.1 Verbs taking two case arrays 300
6.3.2 The N EGIN construction 302
6.3.3 Verbs of communication 306
6.3.4 Indefinite object deletion 307
6.3.5 Summary 312
6.4 Evidence that Basque is accusative 312
6.4.1 ZERIK distribution 313
6.4.2 Evidence from idioms and range of meanings 319
6.4.3 Reflexive tests 323
6.5 A reassessment of case marking and auxiliary distribution 328
6.5.1 Auxiliary distribution 331
6.5.2 The system of case marking 332
6.6 Does Basque have a notion of S-subject 334
6.6.1 Control structures with the verb agindu 335
6.6.2 Control structures with nahi ukan 340
6.6.3 The significance of the control phenomena 345
6.7 S-structure and case assignment in Basque 346
6.7.1 A proposal for case assignment 346
6.7.2 Evidence from antecedent-anaphor pairs 348
6.8 Conclusion 352
6.9 Appendix: Other phenomena involving a notino of "s-subject" 352
6.9.1 Ari izan clauses 353
6.9.2 Causatives 356
6.9.3 Imperatives 359
7 Conclusion 361