Features, Positions and Affixes in Autonomous Morphological Structure

Robert Rolf Noyer

This thesis develops a theory of the mapping from the output of syntactic rules to the input to phonology within a component of Morphology. Three types of information are shown to be independently manipulable in the algorithm deriving fully inflected words: morphosyntactic features, phonological material (affixes), and positions-of-exponence.

To capture categorial neutralizations at the morphological level, a filter-based approach to the Impoverishment of morphosyntactic representations is proposed. Filters are argued to be automatically assumed unless positive evidence suppresses them in the learner. Neutralizations are also shown to occur at three other levels: (1) underspecification of realization rules/affixes, (2) disjunction of affixes competing for a position-of-exponence, and (3) transparency of redundant morphosyntactic values. A hierarchy of features is proposed, which, in the unmarked instance, governs the direction of categorial neutralization and the ordering of affixes and morphological rules, where not determined by syntax or by the principle of descending complexity due to Panini.

The mapping to phonological form is proposed to take place in two phases. In the first phase, morphosyntactic well-formedness is established through the Impoverishment of morphosyntactic features and the Linearization, Rebracketing, Merger (Marantz 1988) and Fusion of syntactic heads, giving a level of constituents termed M0s. In the second phase, the M0 string is mapped to the well-formed morphological word through rules and affixes. Morphological words are shown to display idiosyncratic non-phonological conditions of well-formedness (obligatory or limited positions-of-exponence) which comprise autonomous morphological structure. Particular attention is paid to cases of M0 splitting, where features from a single morphosyntactic source are realized at multiple positions.

Features in input morphosyntactic representations are argued to be discharged through the instantiation of affixes, sometimes at obligatory positions-of-exponence, which are also then discharged. This discharge constitutes spell-out or realization of phonological material. Because both features and positions may be discharged independently, it becomes possible to express disjunctions across position classes, termed here discontinuous bleeding. Further, the Rule-Affix Hypothesis states that rules change phonological material and do not discharge morphosyntactic features; affixes add phonological material and do discharge such features.

The Introduction situates the current proposal within prevailing theories of inflection, Lieber (1980, 1992), Anderson (1981, 1992), and Halle (1989a, 1992). Chapter 1 exemplifies the theory through a detailed analysis of the Afroasiatic prefix-conjugation. Chapter 2 gives an explicit theory of person-number features and their interpretation, in particular the inherent number system of Kiowa-Tanoan. Chapter 3 treats complex multiple argument agreement systems with particular attention to the transitive clitic sequences of Nunggubuyu and Ket.

Thesis Supervisor: Dr. Morris Halle
Title: Institute Professor


Table of Contents

Abstract						  2
Acknowledgments						  3
Table of Contents					  5

INTRODUCTION	
0.  Prefatory Remarks					  8
0.1	Theoretical Assumptions: 
	Towards a Theory of Morphology after Syntax	 14
	0.1.1  	Well-formedness Conditions on Words	 15
	0.1.1.1 	Halle's 1992 Model		 19
	0.1.1.2 	Strict Lexicalism: Affixes and 
			Morpholexical Rules		 21
	0.1.1.3 	Paradigm Theories: Rules 
			Without Affixes			 24
	0.1.1.4 	Disjunctivity and Bleeding: 
			The Discharge Effect		 26
	0.1.1.5	Defining the Rule-Affix Distinction	 30
0.2    Morphosyntactic Well-formedness			 39
	0.2.1  	Syntactic X0s and the Positions of 
		Exponence of their Features		 40
	0.2.1.1  	Splitting of M0s 		 43
	0.2.2	Capturing Systematic Neutralizations	 44
	0.2.2.1  	The Feature Alphabet and the 
			Category Alphabet		 45
	0.2.2.2  	Feature Hierarchies and Feature 
			Trees				 49
	0.2.2.2.1	Filters are not Constraints 
			on the Content of Affixes	 55
	0.2.2.3 	Breaking and Negation		 58
	0.2.2.4  	Enhancement and Feature 
			Transparency at Morphology	 60
0.3	Summary 					 64

CHAPTER 1:  	The Afroasiatic Prefix Conjugation and 
		Discontinuous Bleeding	
1.0   Introduction					 66
1.1   The Prefix Conjugation Paradigm			 67
1.2   Affixes and their Content				 68
1.3   The Extended Word-Paradigm Model			 70
1.4   Allomorphy vs. Discontinuous Bleeding		 76
1.5   Affix-based Models				 80
1.6   Abstract Morpheme Models and Word Templates	 91
1.7   Qs and their Labels				 94
1.8   Qs with Variable Labels: Proposal			100
1.9   Explaining Suffix Allomorphy			105
1.10   Hierarchy among the AGR and T suffixes		113
1.11   Zero Affixes					115
1.12   Historical Reanalysis and Discontinuous Bleeding	116
1.13   The Fate of -uuna  and Gender in the Plural	116
	1.13.1  Gender in the Plural			117
	1.13.2  The Plural Suffix as Specifically 
		non-1st Person				121
1.14   Loss of a Filter:  1st Person Duals		123
1.15   The y-  and t-  Prefixes as Defaults 		128
1.16   The Loss and Reappearance of t-  'f' in Akkadian	138
1.17    Feature-changing Syncretism			141
1.18	Discontinuous Bleeding in the Berber 
	prefix conjugation				144
	1.18.1   The Prefix Conjugation in Tamazight	145
	1.18.2   An Analysis of Tamazight 		148
	1.18.3   A Word-Paradigm Analysis of Tamazight	149
	1.18.4   Dialect Differences within Berber	152
	1.18.5   The Extension of /-y/ in 
		 Three Tuareg Dialects			154
1.19	Summary 					156

CHAPTER 2:  Toward a Theory of Person and Number Features	
2.0	Introduction					157
2.1	Person Features					158
	2.1.1	Defining the Semantic Question		158
	2.1.2	Natural Language Person Syncretisms	160
	2.1.3	The Value [+participant]		163
	2.1.4	A Universal Hierarchy of Persons?	165
	2.1.5	The Putative 'syou' of Algonquian	168
	2.1.6	The Person Affixes of Mam		170
		2.1.6.1  Emphatic Possessor		174
		2.1.6.2  Enclitics in Patient-Agent 
			 Agreement			175
		2.1.6.3  Processing and Ambiguity	178
	2.1.7	A Hierarchy of Person Feature Filters	181
	2.1.8	Two Cases of Person Syncretism		184
	2.1.9	On the Markedness of [-participant]	186
	2.1.10  Summary					190
2.2	Number Features					191
	2.2.0	The Simple Cardinality Features		192
	2.2.1	Number and the 1st Person Inclusive	193
	2.2.2	Augmentation vs. Nonsingularity		196
	2.2.3	Composite Pronouns			198
	2.2.4	Functional Inference and Ambiguity	204
	2.2.5	Trials and Paucals			205
	2.2.6	Inherent Dual and Trial in Nunggubuyu	207
	2.2.7   The Person-Number Affixes of 
		Sierra Popoluca				211
		2.2.7.1  The 1st Person Categories	212	
		2.2.7.2  The Limited Inclusive as Dual	213
		2.2.7.3   Predictions of Usage		215
2.3	Inherent Number and Agreement Class in 
	Kiowa-Tanoan					216
	2.3.1	Jemez					216
	2.3.2	Kiowa					228
	2.3.3	Tiwa and Tewa				237
2.4	Summary						245

CHAPTER 3: Problems in Multiple Argument Agreement
3.0	Introduction					246	
	3.0.1	The Neutralization Problem		246
	3.0.2	Placing					247
	3.0.3	Licensing				253
	3.0.3.1  Impoverishment and Competition for 
		Non-obligatory Positions		253
3.1.1	Strict and Free Licensing			256
3.1.2	Strict and Free Placing				261
3.2	Transitive agreement in Nunggubuyu		270
	3.2.1	Overview of Categories and 
		Intransitive Agreement			271
	3.2.2	Neutralizations in Transitive Agreement	275
		TABLES I-IV and notes			276
		3.2.2.1  Object Impoverishment		281
		3.2.2.2  Subject Impoverishment		284
		3.2.2.3  Gender Impoverishment		286
		3.2.2.4  Summary of Neutralizations 
			 in A forms			293
	3.2.3	Placing in Nunggubuyu			294
	3.2.4	Neutralization and Placing in the 
		B forms					302
	3.2.5	Discussion				308
3.3   Participant Neutralizations in Kiowa-Tanoan 
      and Iraqw						310
3.4  The Position and Types of Verbal Inflection in Ket	319
	3.4.1	The Constituent Structure of the 
		Verb Word				321
	3.4.2	A Preliminary Morphosyntax of the 
		Ket INFL				329

3.5  Summary						334

Concluding Remarks					335
Bibliography						338
Abbreviations						354