Morphosemantic Number: From Kiowa Noun Classes to UG Number Features

MORPHOSEMANTIC NUMBER: FROM KIOWA NOUN CLASSES TO UG NUMBER FEATURES Studies in Natural Language and Linguistic Theory...

Author: Daniel Harbour

60 downloads 700 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

MORPHOSEMANTIC NUMBER: FROM KIOWA NOUN CLASSES TO UG NUMBER FEATURES

Studies in Natural Language and Linguistic Theory VOLUME 69 Managing Editors Marcel den Dikken, City University of New York Liliane Haegeman, University of Lille Joan Maling, Brandeis University Editorial Board Guglielmo Cinque, University of Venice Carol Georgopoulos, University of Utah Jane Grimshaw, Rutgers University Michael Kenstowicz, Massachusetts Institute of Technology Hilda Koopman, University of California, Los Angeles Howard Lasnik, University of Maryland Alec Marantz, New York University John J. McCarthy, University of Massachusetts, Amherst Ian Roberts, University of Cambridge

The titles published in this series are listed at the end of this volume.

MORPHOSEMANTIC NUMBER: FROM KIOWA NOUN CLASSES TO UG NUMBER FEATURES by

DANIEL HARBOUR

Library of Congress Control Number: 2008923537

ISBN 978-1-4020-5039-8 (PB) ISBN 978-1-4020-5037-4 (HB) ISBN 978-1-4020-5038-1 (e-book)

Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands. www.springer.com

Printed on acid-free paper

All Rights Reserved c 2008 Springer Science+Business Media B.V. No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

Abbreviations and notational conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Kiowa phonemes and orthography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv 1. FRAMEWORK

1

1.1 Aim: A morphosemantic theory of number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

1.2 Main claim and overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5

1.3 Conceptual clarification: ‘Morphosemantic’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9

1.4 Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.5 Overview of Kiowa structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.5.1 Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.5.2 The people . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.5.3 The language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.5.4 Orthography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2. KIOWA’S NOUN CLASSES

21

2.1 Overview: Meaning and features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.3 The noun classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.3.1 The SDP class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3.2 The SDI class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3.3 The IDP class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.3.4 The IDS class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.3.5 The IDI class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.3.6 The SDS class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.3.7 The PPP class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.3.8 The SSS class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 2.3.9 The SII class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2.4 Semantic coherence of the classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 2.4.1 The animate classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 2.4.2 The main vegetal classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 2.4.3 Symmetric non-constant classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.4.4 Symmetric constant classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 2.4.5 The default class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 2.4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 2.5 Against a tenth class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

v

vi

TABLE OF CONTENTS 2.6 Phonological incoherence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 2.6.1 Inverse allophones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 2.6.2 Thematic nouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

2.6.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 3. NUMBER FEATURES 61 3.1 Referential cardinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.1.1 Natural classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.1.2 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 3.2 Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 3.2.1 Classification by cardinality features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 3.2.2 ClassP and number on D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 3.2.3 Derivations I: Mnemonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 3.2.4 Derivations II: More inverse marking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 3.2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 3.3 Mnemonic naturalness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 3.4 Spurious s/p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 3.4.1 Types of grouphood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 3.4.2 Derivations III: Grouphood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 3.4.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 3.5 Mass nouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 3.5.1 Conjunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 3.7 Appendix: Missing mnemonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 3.7.1 Beyond Merrifield’s method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 3.8 Appendix: Formal demonstrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 3.8.1 Cardinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 3.8.2 Grouphood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4. AGREEMENT AND SUPPLETION 117 4.1 Suppletion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 4.1.1 Number-sensitive predicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 4.1.2 Clarification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 4.2 Analysis of the Basic Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 4.2.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 4.2.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 4.2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 4.3 Inversive mismatches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 4.3.1 Minus-valued classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 4.3.2 Plus-valued class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

TABLE OF CONTENTS

vii

4.4 Group-induced mismatches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 4.4.1 Collectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 4.4.2 Mass nouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 4.5 Harder cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 4.5.1 Pluralia-tantum-induced mismatches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 4.5.2 Animate- and reflexive-induced mismatches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 4.7 Appendix: Adverbs built on suppletive roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 5. THE AGREEMENT PREFIX

157

5.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 5.1.1 Theoretical assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 5.2 Reduction of explicanda . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 5.2.1 ‘any’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 5.2.2 More x deletion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 5.2.3 Animate/reflexive deletion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 5.2.4 Impoverishment: Person deletion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 5.2.5 Impoverishment: More person deletion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 5.3 Segmentation I: Subregularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 5.4 Segmentation II: Segments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 5.4.1 Ditransitives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 5.4.2 Transitives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 5.4.3 Intransitives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 5.6 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 6. CONCLUSIONS AND CONSEQUENCES

193

6.1 Noun classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 6.1.1 Gender and declension class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194 6.1.2 Gender-number systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 6.2 Parting comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 A A Hunting Story

201

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

Abstract The current work argues for the necessity of a morphosemantic theory of number, that is, a theory of number serviceable both to semantics and morphology. The basis for this position, and the empirical core of the book, is the relationship between semantically based noun classification and agreement in Kiowa, an indigenous, endangered language of Oklahoma. The central claim is that Universal Grammar provides three number features, concerned with unithood, existence of homogeneous subsets, and properties of those subsets. The features are used to analyze a wide variety of data. Semantic topics include the difference between granular and non-granular mass nouns, collective, non-collective and distributive plurals, and cardinality. Syntactic topics include the structure of DP, noun marking, agreement and suppletion. Morphological topics include the inventory of morphological operations, the featural basis of complex syncretisms, the difference between agreement and suppletion, and the nature of the Kiowa/Tanoan inverse.

ix

Acknowledgements Tempting though it is to dwell on the numerous ways I have become indebted to those whose names follow, to do so would only trivialize with details how much indeed I owe them. This applies to none more than to Ken Hale, Vincent (Sun Boy) Bointy and Dorothy (White Feature) Kodaseet, whose absence I note with fondness and regret. The ideas presented below have reached their current form only through the generous criticism of many teachers and colleagues. Foremost amongst these are Morris Halle, Irene Heim, Alec Marantz, and Norvin Richards, who supervised the work during its dissertation phase. Thanks are also due to David Adger, Susana Béjar, Noam Chomsky, Ken Hale, David Pesetsky, Winfried Lechner, Barry Schein, Laurel Watkins, and two anonymous SNLT reviewers for questions, suggestions, and, in Watkins’ case, data, that have permitted substantial improvements on earlier formulations. Of course, none of this would have been possible without the willing participation of many Kiowa speakers—Vincent Bointy, Dorothy Delaune, Georgia Dupoint, Dorothy Gray, Ellafay Horse, Dorothy Kodaseet, and Florene Taylor—and the kind support of many members of their community—Dennis Belindo, Grace Bointy, David Geimausaddle, Bobby and Ann Guoladdle, Carrie Guoladdle, Carl and Vanessa Jennings, Gus Palmer Jr., Glenda Redbird, and George and Margie Tahbone. Funding for fieldtrips was generously provided by the Ken Hale Field Fund and MITWPL. Finally, I wish to thank my family—parents, brothers, grandparents, aunts, uncles, cousins, and others—for their support across two oceans, their interest in the face of incomprehensibilities, and their constant love, the details of which, like my debt to my teachers, I cannot expand upon without diminishing. Ich widme dieses Buch meinem Mann, Dirk Hannemann.

xi

Abbreviations and Notational Conventions * 1 2 3 1–11 1 2 3 a adv agr bas conj d detr distr emph ex freq fut hab hsy i imp impf in inv loc masc name neg nom nom p pf

of agreement prefixes, causing subsequent verb to have low tone first person second person third person noun classes (Kiswahili) referential cardinality 1 referential cardinality 2 referential cardinality 3 animate plural agreement type adverbial agreement (Kiswahili) basic (not inverse marked) conjunction (intersententially: switch reference) agreement/suppletion type, typical of 2 detransitive distributive emphatic exclusive frequentative future habitual hearsay agreement type, typical of inverse marked nouns imperative imperfective inclusive inverse (marking on nouns) locative masculine proper name form of kin terms negative nominal nominative (Sanskrit) agreement/suppletion type, typical of 3 perfective

xiii

xiv pres priv q real recip refl rel s stat th top voc

ABBREVIATIONS AND NOTATIONAL CONVENTIONS present privative question particle nominal suffix denoting prototypicality reciprocal reflexive relative agreement/suppletion type, typical of 1 stative theme topic marker (Japanese) vocative

Kiowa Phonemes and Orthography Consonants Labial Dental Alveolar Palatal Velar Laryngeal (P) k p t Voiceless stop kh ph th Aspirated stop k! p! t! Ejective stop g b d Voiced stop x [c] Voiceless affricate x! Ejective affricate h s sy [C] Voiceless fricative z Voiced fricative m n Nasals l Liquid y Glide w

Vowels Short

High Mid Low

Front i e a [5,a]

Back u o O [6]

Long

Front ii ei aa

xv

Back uu ou OO

Diphthong

Front

ai [5e]

Back ui oi Oi [6e]

KIOWA PHONEMES AND ORTHOGRAPHY

xvi

Vowel diacritics High tone Oral Nasal

Short

Long

´ V

´V ´ V

´M V

´M V ´M V

Falling tone Short ˆ V ˆM V

Long ˆV V ˆM VM V

Low tone Short

Long

V

VV

V M

V M VM

Chapter 1 FRAMEWORK 1.1.

Aim: A morphosemantic theory of number

The current investigation aims to lay the conceptual and basic theoretical foundations of a morphosemantic theory of number. Until now, morphological and semantic theories of number have addressed very different problems and have produced answers of little mutual relevance (see below). So, the notion that the study of Universal Grammar requires a theory that covers both the semantics and the morphology of number constitutes, I believe, a substantial departure from previous work. To be specific, morphological theories of number have been primarily concerned with inventories of pronouns and agreement of the world’s languages (see Corbett 2000, Cysouw 2003 for excellent overview and synthesis). For instance, the morphologist might wonder whether there exist cardinally exact trial, quadral or quintal number, or whether the forms that permit such readings are really paucals, and hence cardinally inexact. More rarely, morphologists concern themselves with the relationship between members of such inventories. For instance, in a system with singular, dual and plural, one can wonder whether each of these is sui generis, or whether, for instance, dual is a type of ‘expanded’ singular or ‘restricted’ plural. Even more rarely, observations that numbers in some languages are not sui generis have led to attempts to specify the features that underlie them. An important example of this work is Noyer (1992), which the current investigation follows both in spirit and in content. So, morphological theories of number examine languages’ pronoun inventories and agreement categories, aiming to explain why only certain ones are attested or how members of the inventory are related. 1

2

CHAPTER 1

Such results are of little interest to semanticists, whose theories of number are concerned with how to represent the meaning of plurality so as to capture similarities between mass nouns and plural count nouns, or with the relationship between plurality of nouns and plurality of events, or with collectivity and distributivity of plurals. This lack of interest is reciprocated: it is notable that Link’s (1983) use of lattice theory to provide an ontological basis for mass∼plurality similarities, though seminal in its area, has had little, if any impact, on morphological theories of number. To parody the situation, given the sentence ‘We [you and I] carried a piano upstairs’, the morphologist would worry about how much like ‘they (two)’ ‘we [you and I]’ is, whereas the semanticist would worry about how to determine the quantity of pianos transported, and this while both claiming to work on number. As measures of the morphologists’ and semanticists’ lack of interest in each other’s work, consider Corbett (2000) and Winter (2001). In a 350-page survey of number and its morphology, Corbett devotes barely a page to semantic work on the topic; and Winter’s highly articulated theory of the semantics of number shows no concern with the difference between agreement with a head and the head itself, a difference that is basic to morphology (see his comments on Hungarian).1 The untenability of this mutual uninterest is thrown into relief by Kiowa, the language that forms the empirical core of this study (see Section 1.5 for general information). Kiowa displays two typologically noteworthy characteristics. On the one hand, it possesses a complicated rich agreement system, that is, an object of interest to morphologists. On the other hand, it possesses a semantically-based noun class system. As some of these noun classes pick out sets of nouns of traditional interest to semanticists—such as, granular versus non-granular mass nouns, nouns that form collective versus non-collective plurals—Kiowa’s noun classes are of interest to semanticists. Crucially, however, noun class and agreement interact in Kiowa. Indeed, as we will see in Chapter 2, it is precisely on the basis of agreement that noun classes are primarily recognizable. This means that a semantic theory of mass and plurality or of collectivity is answerable to the morphology: whatever primitives it posits must be usable in an account of Kiowa’s complex agreement. Conversely, it means that a morphological theory of Kiowa agreement 1 These observations are not intended as real criticisms of Corbett or Winter. Indeed, I choose their work as examples because, being so good in their respective domains, they can easily take some mild ribbing. And, of course, we should also note that the division between morphological and semantic work is not absolute. See, for instance, Ojeda (1998).

FRAMEWORK

3

and syncretisms must be answerable to the semantics: whatever primitives it posits must be capable of rigorous definition and implementation in a semantic theory of the classification criteria of Kiowa nouns. Consequently, Kiowa necessitates a unified morphosemantic theory of number capable of addressing both core morphological and core semantic issues. Let me spell this argument out a little less abstractly. To do this, I will show how the classic morphological problem that Kiowa agreement poses quickly turns into a semantic issue, thereby illustrating the interrelatedness in Kiowa of the semantic and morphological aspects of number. One of Kiowa’s most striking characteristics is its rich agreement system. Unlikely as it may seem, three distinct arguments in (1)—the third person plural giver, the third person plural recipient, and the third person singular gift—are in some way encoded by the agreement prefix ´ a∗ (‘∗’ marks that the prefix lowers the tone of the following verb). This encoding can be appreciated by changing each of the arguments in turn, to third person dual, say: each change in argument results in a different agreement prefix (2)−(4). (1)

(2)

´ A∗OM OM 3a:3a:3s-give.pf ‘They gave it to them’ ´M ∗E OM OM 3D:3a:3s-give.pf ‘They two gave him to them’

(3)

MéÓM ÓM 3a:3D:3s-give.pf ‘They gave him to them two’

(4)

EtÓM ÓM 3a:3a:3D-give.pf ‘They gave them two to them’

This system is immediately fascinating. At the most general level, there is the issue of how so much meaning can get into so little sound. One can show that prefixes are not unanalyzable wholes; rather each of the arguments makes a distinct and regular phonological contribution to the prefix (Chapter 5). So, one can ask, more specifically, why each change to an argument has the precise phonetic effect that it has: why et in (4) and m´ e in (3), and not vice versa? Furthermore, one can wonder what (3) and (4) share in virtue

CHAPTER 1

4

of which the verb has high tone, in contrast to the low tone of the earlier examples. This is, therefore, practically the paradigm case of a morphological problem: which phonetic features of the prefix realize which syntactic features of which argument? Now, the syntactic features that prefixes realize do not start their syntactic existence huddled together beside the verb. Rather, taking a standard Minimalist view, the features that comprise the agreement cluster are copies of the feature content of the D0 of each argument DP. In particular, the number features of each argument will be copied from D0 . Number features do not begin their syntactic existence there either, though. We will see, in Chapter 3, that the number content of D0 is determined by two lower, DP-internal heads. We can sketch the flow of number features through the syntax, as follows (assuming IP to be the locus of the fully inflected, agreeing verb). (5)

Syntactic transmission of number information IP .. . DP D semantically significant number information

Inflected Agreeing Verb morphologically significant number information D 6

6

(The account of inverse marking in Chapter 3 shows that the information does indeed flow through D.) The content of the DP-internal heads is determined—and this is the crucial point, where things turn semantic—by the noun itself. Kiowa nouns belong to one of nine classes. I will argue that the agreement behavior of these classes is readily explicable if we suppose that Kiowa nouns bring with them into the syntax one or more number features (rather like gender features of many Indo-European nouns). The semantic connection is this: Kiowa’s nouns classes are strongly semantically based, in two distinct senses.

FRAMEWORK (6)

a. b.

5

Each class is internally coherent, that is, all classmate nouns share particular semantic characteristics. The number feature(s) that nouns of a given class bring with themselves into the syntax is, or are, appropriate to the semantic property of that class.

So, for instance, if the defining property of a class is the type of plurality its nouns form, collective versus non-collective, then nouns of that class bring into the syntax a feature that distinguishes (non-)collectivity of plurals. (The collective versus non-collective distinction is at stake in class membership for much vegetation in Kiowa.) Or, if the defining property of a class is that its nouns generally do not occur in pluralities, then such nouns bring non-plural number features into the syntax. (This is the case for self-propelling entities, including animates, in Kiowa.) Given the semantic properties that several of the classes pick out, they are of immediate and obvious relevance to core research into the semantics of number: those consisting non-granular mass nouns, or of granular mass nouns, or of collective plurals, or of nouns that permit a ‘collection of collections’ reading. So, we have traced a path from a classical morphological problem, the constituents of the Kiowa agreement prefix, via agreement relations with D0 , through the DP-internal heads that determine the content of D0 , to the relation between number features and such properties as granularity, masshood, collectivity, and so on, a classical semantic problem. Consequently, it will not do to offer a morphological analysis of the prefix that posits features which cannot serve as the foundation of an analysis of the semantic properties just listed. Nor can one countenance a semantic analysis with primitives that can in no way be correlated with the phonetic units that comprise the agreement prefix. Rather, a unified morphosemantic theory of number must be offered that does justice to the concerns of morphologists and semanticists alike. The current investigation aims to present core elements of such a theory. 1.2.

Main claim and overview

The core number-theoretic claim of this investigation is that Universal Grammar provides three binary number features: [±singular] concerns unithood; [±augmented], the existence of (homogeneous) subsets; [±group], properties of those subsets, should they exist. The investigation concentrates on motivating these features, on rigorously defining them, and on showing how

6

CHAPTER 1

their manipulation, syntactically and postsyntactically, leads to a tightly constrained theory with wide and diverse empirical coverage. The topics and phenomena that motivate and deploy these features are as follows. Chapter 2 presents the empirical core of the investigation, Kiowa’s noun class system. First, it shows how Kiowa’s four agreement types can be used to distinguish nine noun classes. On the basis of this agreement behavior, a new system of noun class mnemonics is proposed, to facilitate the discussion. The nine classes are shown to be internally semantically coherent, that is, that the nouns in each class share semantic characteristics: (7)

Characteristics First person Animates and animate-like entities (physically similar to animates or capable of self-propulsion or determining direction of motion) Naturally regarded both as individual and as occurring in collections; permits ‘different types of’ reading with s-agreement Individuable, non-shape-inductive Non-granular mass nouns Pluralia tantum, composite nouns (and granular mass nouns for some speakers); abstract nouns Default for vegetation and implements; granular mass nouns for some speakers Vegetation forming natural collections in which members are not readily individuable; implements that act as a group to produce a single effect Default; no unifying properties

The chapter also introduces one of Kiowa’s most fascinating phenomena, inverse marking. Kiowa nouns, like English nouns, if drawn straight from the lexicon and uttered unaltered can be used to refer only to a restricted number of tokens. For instance, English table refers to a single table, chair to a single chair, and so on. To talk of not-one-table or not-one-chair, one adds something to the noun, the plural -s. In Kiowa, the range of meanings of nouns drawn straight from the lexicon is wider: some nouns refer to exactly two (k!ˆ On ‘two tomatoes’), others to two or more (´ a´ a ‘two or more sticks’),

FRAMEWORK

7

others to one or two (t´ og´ ul ‘one young man or two’), and so on. Interestingly, Kiowa has a morpheme, just as English has the plural, that permits one to talk of a quantity of nouns other than what noun by itself means. This is traditionally called the inverse. So, k!ˆ OˇOˇdO, ‘tomato’+inv, means ‘one tomato’ or ‘three or more tomatoes’; ´ a´ adO, ‘stick’+inv, means ‘one stick’; t´ og´ uu ´ d´ O, ‘young man’+inv, means ‘three or more young men’. Note that the morpheme is the same dO in all cases.2 Chapter 2 shows that inverse marking is a major indicator of class membership. For the most part, the chapter is expository and empirical. Some readers may prefer to skim through it just to gain a feel for the facts and system as a whole, rereading it more thoroughly at a later point. Chapter 3 presents the theoretical core of the investigation. It motivates and defines the unithood feature [±singular], the subset-existence feature [±augmented], and the subset-property feature [±group]. Positing a non-novel DP-structure, it claims that these number features occur in two DP-internal positions both of which are semantically contentful: Number is determined by cardinality (singular, dual, plural), and Class, by semantic properties of the noun. A computation over Number and Class determines the number features on D. (8)

DP

NumberP ⎡Class: Noun ⎤ (±singular) ⎣(±augmented)⎦ (±group)

D depends on content of Class and Number

Number ±singular ±augmented

The number-feature content of D is not itself semantically interpreted. However, it is syntactically active and, sometimes, phonetically realized. Syntactically, it is the head that triggers agreement and so contributes features to the agreement prefix. Phonetically, it is realized as inverse marking, but is otherwise generally null. Several more complex DPs, such as adjectivally modified ones, are examined, to show that inverse marking is located at D. 2

The variation in tone is purely phonological. See Watkins (1984) or Harbour (2002). The morpheme itself is subject to phonological conditioning (Section 2.6.1).

CHAPTER 1

8

The chapter shows further that the relationship between a class’ mnemonic and the semantic characteristic of the nouns it subsumes is principled, in contrast to the semi-arbitrary gender classification of Indo-Europe. The chapter concludes with discussion of what constitutes a possible noun class of a Kiowa-like language. The problem addressed is one of generative capacity. Given the description of Chapter 2, one would think that 64 different noun classes are possible. This prompts one to wonder why Kiowa should instantiate just nine. It is shown that the theory just sketched is highly restrictive and that Kiowa all but optimally exploits the space of possibilities. Chapter 4 addresses mismatches between agreement and suppletion, which readers, depending on their inclinations, are likely to regard either as delightful or as horrifying. Kiowa has two sets of number-sensitive suppletive predicates: those distinguishing singular from dual/plural, and those distinguishing plural from singular/dual. In the vast majority of cases, agreement and suppletion match. That is, if one is talking of a single young man, agreement will reflect that singularity and any suppletive predicates will be in their singular or singular/dual form: (9)

Tóg´ ul ∅- ét young man 3S-big.S ‘The young man is big’

Similarly, talking of two tomatoes, agreement will reflect that duality, and suppletive predicates will be in their singular/dual or dual/plural form: (10)

K!Ôn nen- ót tomato 1s:3D-drop.s/D ‘I dropped two tomatoes’

Likewise, mutatis mutandis, for talk of several sticks. However, in a variety of cases, agreement and suppletion seem to indicate different numbers. Two examples are (11), where agreement indicates plurality and suppletion, singularity, and (12), where the reverse holds: (11)

´M´MZde t!´ E oM u ´M gya yáM dôi- et this shirt :1s:3P-too-big.S ‘This shirt is too big for me’

(12)

Tóu ´ ∅- sÓl house 3S-be set.P ‘There are houses standing’

FRAMEWORK

9

Theoretical arguments are presented that the features that condition suppletion are located on a different head from the features that trigger agreement. D triggers agreement whereas the interpretable features directly under D, generally Number, condition suppletion. The theory of Chapter 3 claims, independently of suppletive facts, that divergences can arise between the content of Class, Number and D. Given such divergences, we predict not only when agreement∼suppletion mismatches will occur, but precisely which otherwise unexpected combination of agreement and suppletion will result. Thus, agreement∼suppletion mismatches support the theory of Chapter 3. Chapter 5 sketches in detail an analysis of the agreement prefix. This is crucial to the project of morphosemantic research on number, given the argument that Kiowa not only presents classical morphological and semantic number problems, but demands that solutions to one be exportable to treatments of the other. Having motivated three number features primarily on semantic grounds (though with reference to DP-level morphology and, more cursorily, to the agreement prefix), it becomes crucial to show that phonological subparts of the prefix are precisely realizations of those features. It is shown, moreover, that rather complex allomorphic relations can be naturally stated in terms of these features and the natural classes they define. Chapter 6 concludes by placing Kiowa’s noun class system in crosslinguistic context, arguing that it is different from Indo-European gender but strongly similar to Bantu gender-number systems. As the solutions proposed for the morphological and semantic number problems that Kiowa poses employ the same elements, we are on the way to a unified morphosemantic theory of number. In presenting a unified analysis of diverse grammatical phenomena in Kiowa, the investigation attempts to provide some definite answers. In its theoretical aims, however, it is merely programmatic, suggesting a line of inquiry that strikes me as important and interesting and offering some elements of what a fuller morphosemantic theory of number might comprise.

1.3.

Conceptual clarification: ‘Morphosemantic’

Now, in light of the foregoing exposition, the term ‘morphosemantic’ itself requires clarification, particularly in comparison to terms such as ‘morphosyntactic’ or ‘syntacticosemantic’. ‘Morphosyntactic’ refers to the interaction

CHAPTER 1

10

between morphology and syntax, ‘syntacticosemantic’ to that between syntax and semantics, and the terms can be used to name interfaces between the relevant modules of the grammar. In arguing for a morphosemantic theory of number, I am not suggesting that there is a morphology-semantics interface involving direct communication between the conceptual and pronunciational systems. Everything I argue for is compatible with the familiar Y-model. (13) Syntax Morphology

Semantics

Phonology

Conceptual Interface

Articulatory Interface

Syntax builds hierarchical objects and transmits information (via agreement and movement operations) from one part of the structure to another, as depicted in (5). Some parts of the resulting structures and chains are relevant to the semantics (i.e., are interpreted), others are relevant to the morphology (i.e., are pronounced). It is syntax that plays the role of an interface between morphology and semantics, to the extent that there is one. And to the extent that the phenomena addressed by semantic number theories and morphological number theories are fed by the same syntactic features, a single theory of both sets of phenomena is required. What is interesting about Kiowa is that it provides a relatively straightforward argument that the overlap, and hence the need for a single ‘morphosemantic’ theory of number, is great. 1.4.

Framework

The syntactic and morphological frameworks I adopt are Minimalism and Distributed Morphology, respectively. My reason for adopting these theories, rather than any others, is that they offer a constellation of assumptions that, in my opinion, make for a particularly interesting research program. Specifically, Minimalism places heavy emphasis on the importance of interfaces and it constrains the variety of features one can posit: every feature

FRAMEWORK

11

must be interpretable, that is, for every feature, [F], there must be a head, X0 , such that, when [F] is located on X0 , [F] contributes to the interpretation of the sentence (what Pesetsky and Torrego 2001 term ‘weak functionalism’).3 And Distributed Morphology claims that ‘words’ have an internal structure and that the structure is composed, primarily, by the syntax. It claims furthermore that the phonological structures associated with syntactic feature bundles are inserted postsyntactically, at the terminal nodes. Let me illustrate how each of these assumptions is important to the investigation. My aim is, as stated, to present elements of a unified theory of number serviceable to morphology and semantics alike. The initial motivation for this theory—a path from a core morphological problem, via syntax, to a core semantic one—itself relies on several tacit assumptions about the relationship between syntax, semantics, and morphology, especially those just outlined. It is immediately obvious that this an interface project. It is concerned with what nouns bring with themselves into the syntax in virtue of their meaning—this is the lexicon-syntax interface. It is concerned with the exact relationship between phonological pieces of agreement prefixes and the feature content of syntactic heads—this is the morphology-syntax interface. And most obviously, it is concerned simultaneously with two different modules of the grammar, morphology and semantics. Consequently, both the methodology and the aims of the project itself only make sense in a world where interfaces are central to research. Minimalism offers such an approach. Second, the motivation of the project began with morphological pieces and ended up at semantic ones. Conversely, the more detailed chapter summary began with semantic features in Chapter 3 and ended with morphological ones in Chapter 5. Now, everyone expects that we will be able to trace paths from one module of the grammar to another: no module is an island. However, there is no reason to expect a priori that semantic and morphological atoms will be one and the same. Sameness of atoms is expected, indeed forced, if we assume, with Distributed Morphology, that the pieces of inflection are the phonetic realization of the pieces of syntax, and if we assume, with Minimalism, that all features have an interpretation. Third, I argue below, and sketched above, that the content of D is determined by the content of two lower heads, Number and Class, evidence for the content of which is semantic. The content of D determines agreement and D’s phonetic realization (as inverse marking). Thus, both agreement and 3

This is not the same as assuming that all instances of all features are interpreted.

CHAPTER 1

12

inverse marking depend on a syntactic computation. This postsyntactic dependence is completely natural if we assume late insertion, with Distributed Morphology. However, it is quite unnatural if we adopt a morpheme-based theory, where the presence of each morpheme is justified by its contribution the meaning of the word. This is because the semantically interpreted heads are Class and Number. D, where inverse marking is located, is uninterpreted. In other words, inverse marking does not have a meaning, but occurs when a constellation of features is distributed over Number and Class. Consequently, the inverse is a sound without a meaning; it is not a Saussurean sign. 1.5.

Overview of Kiowa structure

In the remainder of this chapter, I present the basics of Kiowa grammar and apply some of the framework assumptions above. 1.5.1.

Sources

All Kiowa sentences and word lists cited in the current work are from four sources: Harrington (1928), Watkins (1984), Watkins (p.c.), and my own fieldwork (August 2001, July 2002, January and December 2003, November 2004, August and December 2005). Sentences from the first three sources are always cited as such; sentences without cited sources are from my fieldnotes. The fieldwork was conducted with speakers, aged 70−85, from Anadarko, Carnegie, and Mount Scott: Vincent Bointy, Georgia Dupoint, Dorothy Kodaseet (the primary consultants for this study); Dorothy Delaune, Dorothy Gray, Ellafay Horse, and Florene Taylor (who provided supplementary data— our collaboration has chiefly focused on other linguistic domains). All were monolingual in Kiowa until commencing school, aged 7−9, and retained fluency into adulthood. Though English is their dominant idiom, all revert to Kiowa with ease, save for occasional lexical blockage (I have not been able to record the word for ‘tickle’). Dialectal variation between these speakers, and those with whom Harrington and Watkins worked, though existent, is confined, I believe, to minor phonological, morphological, and lexical details, noted, when relevant, below. This work is not an ethnographic or historical study of the Kiowas. However, a number of such sources have, of necessity, been consulted and their contents, referred to, in passing, below. These are Mooney (1898/1979), Boyd (1983), Ellis (1996), and Merrill, Hanson, Green, and Reuss (1997).

FRAMEWORK 1.5.2.

13

The people

Kiowa is a Tanoan-related language of Oklahoma.4 It is spoken fluently now only by some 40 people, aged 70−95. Members of the next oldest generation often have good comprehension of the language, but they rarely have so intimate a grasp of its grammatical, lexical and stylistic subtleties. Knowledge of the language declines sharply as one moves into the younger generations. The prognosis for the language is therefore not good. However, comprehensive documentation is underway and it is likely that future generations of Kiowas will inherit a substantial record of their language in its cultural context, even if direct inheritance of the language becomes impossible. The Kiowa Tribal Complex is located in Carnegie, Oklahoma, and members of the tribe live mostly in Caddo, Kiowa and Comanche counties. The community’s distribution over three counties is the result of deliberate US Government policy. By 1876, White exploitation of resources had precipitated the collapse of the buffalo population, on which the Kiowas depended not only for food, but also for clothing, implements, and shelter (in the shape of hides for teepees). This made the continuation of their traditional lifestyle impossible. Instead, in return for government rations, they (finally) agreed to be settled on a joint Kiowa-Comanche-Apache reservation. The US government nonetheless quickly realized that the reservation provided the tribes with autonomy and coherence enough for the maintenance of their identities, cultures and languages. To force assimilation (and to gain access to the valuable grasslands that comprised the reservation), Congress passed the General Allotment Act of 1887 and, more controversially, the Jerome Act of 1901, which, by 1907, had had the effect of transferring ownership of 80% of the former reservation into White hands and interspersing the allotments assigned to Indians, not only amongst each other’s tribes, but also amongst the Whites. The consequent lack of a geographically coherent 4 The family is called Kiowa-Tanoan. The double barrelled appellation primarily reflects geographic and cultural differences, rather than linguistic ones: the Kiowa are, geographically and culturally, of the Northern Plains, in contrast to their linguistic relatives, all of whom are Pueblo peoples, resident in the South West. Given these differences, the relationship of Kiowa to Tanoan could not be assumed, but required proof. So it was that the Kiowa-Tanoan family was posited (e.g., Harrington 1910). Since Hale’s definitive demonstration of the relationship (Hale 1962) and his subsequent partial reconstruction of the proto-phonology (Hale 1967), the double barrelled name has become otiose and is linguistically misleading. However, the nomenclature is entrenched and the best we can do now is, as above, to avoid the cacophonic phrase ‘Kiowa is a Kiowa-Tanoan language’.

CHAPTER 1

14

Kiowa community appears to have contributed to the decline in the tribe’s language and certainly militates against efforts to revitalize it. Kiowas enter Whites’ historical record several centuries ago, in Montana. Their migration southwards occurred after introduction of the horse, which became central to their culture; it took over from the dog the designation xˆ eˇ ˇi, with dogs becoming known as xˆ eˇ ˇihiˇˇi ‘dog’+real and later as x´ egun.5 Kiowas’ earlier history is unclear, and, indeed, quite intriguing, given the geographic and cultural distance between them and their linguistic relatives in the South West. Kiowa folk memory speaks of a white bear and suggests in other ways an existence even further north. And earlier still, Kiowas had lived underground, escaping only with the assistance of Séndé, the creatorprankster, who pulled them out through an owl hole in a cottonwood tree, having heard their knocking when he chanced to rest against it. 1.5.3.

The language

Basic word order Kiowa is a rich agreement language with relatively free word order. A basic (informationally unmarked) order is nonetheless discernible: (14) (15)

Particles Subject Indirect–Object Direct–Object Verb HÓn Paithal´Z´Z P!ÓÓthÓpdek!ii áádO Óthêm- OM OM mOO neg Vincent Daniel stick.inv 3s:3s:3i-break-make.neg ‘Vincent didn’t make Daniel break the stick’

Sentences like (15) are rare for two reasons. First, Kiowa permits pro-drop of any argument DP, as in (1)−(4) (Watkins 1990), and sentences with three overt arguments are infrequent. Second, DPs, as well as other constituents, are frequently dislocated to the left or right edge of the sentence. (16)

HÓn máthOn ∅- xáM áM nÔO neg girl 3s-arrive.neg ‘The girl didn’t arrive’

(17)

MáthOn hÓn ∅- xáM áM nÔO girl neg 3s-arrive.neg approximately ‘The girl, she didn’t arrive’

5

Watkins (1984) gives the origin of gun as associated with poles or travois.

FRAMEWORK (18)

15

HÓn ∅- xáM áM nÔO máthOn neg 3s-arrive.neg girl approximately ‘She didn’t arrive, the girl.’

These dislocations correlate with information structure, leftward dislocation for topic and focus, rightward for old information, as the approximate translations attempt to capture. The rightmost of the particles in (14) can be regarded as marking the right edge of part of the clause that hosts leftwards displaced constituents. These dislocations, and their nature, effect and limits, are the topic of ongoing joint research. Like DPs, the particles in (14) occur in a relatively fixed order. Semantically, they express a variety of aspectual, modal and evidential meanings, as well as negation. Many obligatorily cooccur with inflection suffixes on the verb. (See Watkins 1984 for a thorough overview and Adger and Harbour 2007 for discussion of phrase structure implications.) (19)

BéthOO hÓn ám em-dÓM ÓM -mÔ O- hel unbeknownst neg 2 2s- be- neg-hsy ‘I didn’t realize it wasn’t you’

(20)

Háyátto hÓn ∅- deM Mi- héM´MZ-mÔ O- t!OO maybe neg 3s-sleep-die- neg-fut ‘Maybe he won’t fall asleep’

(21)

an Óbôu- hOnx!ou- yii- t!OO- dei BéthOO unbeknownst hab :3s:3i-always-come late-impf-fut-hsy ‘I didn’t realize he was going to keep on coming late’

As illustrated by the underlining, particles in selectional relations with inflectional suffixes of the verb occur in the opposite order from those suffixes. A variety of word orders and particles occur in the example sentences of the chapters that follow. However, word order is never a factor in the grammatical properties relevant below. Nouns Nominal morphology is sparse in Kiowa. There is no case marking either for DPs or pronouns, and the only marking for number is inverse marking: Taken straight from the lexicon and uttered unaltered, nouns are limited in the number of tokens they can refer to. For instance, t´ og´ ul means ‘one or two

CHAPTER 1

16

young men’, ´ a´ a means ‘two or more trees’ and k!ˆ On means ‘two tomatoes’. When this inherent number is not the same as the number of tokens talked of, the noun is inverse marked. This results in the curious situation that one and the same suffix attaches to the nouns just given for the plural, the singular and the non-dual: t´ og´ uu ´ d´ O ‘young men’, ´ a´ adO ‘a tree’ and k!ˆ OˇOˇdO ‘a tomato’ or ‘more than two tomatoes’. The form of inverse marking and its effect on agreement is discussed at length in Chapter 2. Given Kiowa’s freedom of word order, it is, perhaps, not overly surprising to find split constituents. Compare (22), which was spontaneously uttered, with its split-free paraphrase (23). (22)

Dˆ oiette an p´ enhaa gya- ôugu xó´Zkya too much hab sugar 1s:3s-pour.impf coffee.loc ‘I’m always putting too much sugar in my coffee’

(23)

An xó´Zkyá dˆ oiette p´ enhaa gya-ôugu hab coffee.loc too much sugar 1s:3s- pour.impf ‘I’m always putting too much sugar in my coffee’

Three more examples are: (24)

(25)

(26)

P´ a´ agO gya- b´ oM u ´M k!y´ a ˇ´ a ˇ hˆˇZ ˇi one 1s:3s-see.pf man ‘I saw the one man’ ´ˇ´ˇZde hâtêl k!´ E Od´ al ádÓÓ? that who.q car :3s:3s-be ‘Whose car is that?’ ´ Zg´ E´ utk!o hâtêl y´Z´Z eM hân? biscuit who.q two 3s:3d-devour.pf ‘Who ate two biscuits?’

(Harrington 1928, p. 45)

Quantifiers may be pre- or postnominal: ´ eˇ´ˇZde t´ ou ´ ∼ t´ ou ´´ eˇ´ˇZde ‘this house’, y´Z´Z t´ ou ´ ∼ t´ ou ´ y´Z´Z ‘two houses’, ´ ett´ e t´ ou ´ ∼ t´ ou ´´ ett´ e ‘many houses’. Attributive adjectival modification is not an overly common strategy, as the language generally prefers verbs, particular in headless relatives, to play this role. However, when adjectives do occur, they are postnominal: (27)

x!óu ´-êl rock-big.s ‘big rock’

xégun-syan dog- small ‘little dog’

tou- g´ ul house-red ‘red house’

k!´Z´Z- t!áM´MZ- de male-white-bas ‘White man’

FRAMEWORK

17

DP modifiers Consistent with its head finality (default verb finality, postnominal modification), Kiowa has postpositions, rather than prepositions: (28)

k!ÓÓ- pa bank-against ‘against the bank’

Indian Fair-kya Indian Fair-at ‘at the Indian Fair’

a´lOO- aadO- yOO apple-tree.inv-from ‘from the apple tree’

Similarly, focus modifiers are suffixal. (They differ from quantifiers in being inseparable from the noun.) (29)

nÓÓ-(x!)al 1- too ‘me or us too’

Indian Fair-deki Indian Fair-only ‘only the Indian Fair’

háote- kÓÓ several-just ‘just a few’

Pronouns Kiowa has only two pronouns: n´ O´ O for all first persons, whether singular, dual or plural, inclusive or exclusive; and ´ am for second person, whether singular, dual or plural. There are no third person pronouns, but deictics can be used instead, ´ eˇ´ˇZde∼´ eˇ´ˇZgO ‘this’∼‘this.inv’ and ´ o´Zde∼´ o´ZgO ‘that’∼‘that.inv’. The complex verb Verbs consist of two parts, an agreement prefix (next subsection) and a complex verb. The latter, in which only the root is obligatory, consists of: (30)

Incorporates Root Distributive Aspect/Negation Future Evidential

Incorporates can be adverbs: (31)

Gyat- hón-pOl- /t!Om-pÓl- tOO 1s:3p-last- eat-/first- eat-fut ‘I’ll eat last/first’

(32)

Gyá- hágyá- ton :2s:3s-already-fat ‘You’ve already got it fat’

—or verbs, with or without complement nouns:

(Watkins 1984, p. 241)

CHAPTER 1

18 (33)

TÓÓdé de- t!ÓÓthâl-dou long time 1s:3a-listen- hold ‘I listened for a long time’

(34)

A- hot- kôm- hap- do yáM olsÓ- mOOgO 1s-travel-distr-freq-because :1s:3p-belongings-lay.p-expert ‘Because I travel so much, I’m expert at packing’

—or nouns alone: (35)

A- pénhaa-ou 2s:3s-sugar- pour.imp ‘Pour the sugar’

(36)

P!él- heM Mi hétÓ gya-olsÓl drop.p-priv still 3p- belongings-lie.p ‘The groceries are still lying [in the car] unloaded’

(37)

Bé- k!ÔO- saa 2s:3i-knife-cut.imp ‘Cut it with a knife’

Object incorporation is not valence reducing: (35) is overtly transitive. With regard to the inflectional particles, the main points were mentioned above in tandem with sentential particles. Note, though, that aspect and negation are in complementary distribution (though particles that govern them are not; in fact, h´ On neg, which cooccurs verbal -neg, and ´ an hab, which cooccurs verbal -impf, frequently cooccur in that order with the meaning ‘never’; the verb is inflected for negation in these cases, not aspect). The suffixes show allomorphy for a variety of different properties, including agentivity, stativity and transitivity. For instance, the future suffix, fut, has an agentive form tOO (31) and a non-agentive form t!OO (20). The hearsay suffix, hsy, too shows variation, appearing as dei after impf-fut in (21), as ˆ ei fused with impf as in t´ o ˇu ´ ˇ nˆ ei ‘say.impf.hsy’, and elsewhere, as in (19), as hel. For full exposition, see Watkins 1984, Harbour 2004. The agreement prefix Along with the root, the prefix is only other obligatory part of the verbal complex. It registers agreement for up to three DPs: external argument, indirect object and direct object. Kiowa has some 100-to-160 prefixes depending

FRAMEWORK

19

how one counts certain homophones. Harbour (2003a) shows that this prefix is phonologically independent from the rest of the verb. The only exception is that some prefixes, notated with a ‘∗’, lower the tone of the subsequent verb; compare (1)−(4). Prefixes, discussed at length in Chapter 5, are glossed as follows. In z-, z is the subject of an unaccusative predicate, as in (20). In x :z -, x is the agent of a transitive verb and z, the direct object as in (31). In x :y :z -, x is the agent of a (di)transitive verb and y, the indirect object, and z, the direct object, as in (1)−(4). Finally, in :y :z -, z is the subject of the unaccusative (it triggers however agreement identical to that triggered by the z direct object in (36)), and y is the indirect object, such as the possessor of z or a benefactor of the event, as in (21). Thus, in prefix glosses, something of the form :n: is an indirect object; something of the form :n is a direct object; and something of the form n is a subject/agent. Phonology Kiowa phonology is not overly complicated, though it does throw up some interesting surprises, such as Dental-velar switching, summarized, together with a variety of alternations in Section 5.1.1. Appendix A contains brief comments on tonal interactions. Note that glosses always show surface tones, not underlying tones. (This accounts, amongst other things, for neg’s having falling tone in (15), mˆ OO, but low tone in (19), mOO.) The language’s phoneme inventory is described immediately below together with the orthography. 1.5.4.

Orthography

The orthography used here is that of Harbour and Guoladdle (in prep.). Brief descriptions of the sounds they represent follow. See Watkins (1984) for greater detail. Phoneme charts are provided on pp. xv–xvi. The consonants b d g h m n s w y have their IPA (International Phonetic Alphabet) values (Lagefoged 1993). The palatal affricate, IPA [ts], is written x, and the alveolo-palatal fricative, IPA [C], sy. The latter representation acknowledges that sy is very occasionally realized as s+y, that is, as IPA [sj]. Kiowa l is generally preceded by a laterally released [d], sometimes somewhat devoiced; hence it varies between IPA [dl l], [d◦ l l] and [d◦ l ◦l ] . The degree of devoicing of [d] varies to the extent that it sometimes approaches

20

CHAPTER 1

IPA [l] (particularly when intervocalic and morpheme internal, as in ´ alOO ‘apple’ and b´ OlO ‘butter’); the [l] pronunciation appears to be more pervasive in Anadarko than in Carnegie. Finally, as concerns consonants, stops show a four-way contrast: voiced, ejective, voiceless unaspirated, voiceless aspirated. The voiced stops g d b and the voiceless unaspirated k t p have their IPA values. Aspiration, IPA [h ], is marked by h as in kh th ph. Ejectivity, IPA [P], is marked by ! as in k! t! p!. X has an ejective counterpart, x!, but no aspirated one. Kiowa has six vowels, high, mid and low, front and back: i e a u o O. These have their IPA values, except for u, IPA [U]; O, IPA [6]; and a, IPA [5], fronting to [a] after palatalized velars (unless part of the diphthong ai). All the back vowels and the front low vowel form diphthongs with i; however, this is pronounced as IPA [e] if the first vowel is low: hence ai [5e] and Oi [6e]. Vowels contrast for length, tone, and nasality (but diphthongs only for tone and nasality). For high and low vowels, length is indicated by repetition: ii aa uu OO. For mid vowels, which diphthongize when long, it is indicated by the appropriate high vowel: ei ou. Nasality is indicated by a Polish hook, e.g.: a ˇ a ˇa ˇ . However, nasality is predictable, and so unmarked, on vowels cosyllabic with a nasal stop (e.g.: m´ a not m´ a ˇ ). Tonally, vowels are either high (acute), falling (circumflex), or low (unmarked). Illustrating with the low front vowel, these are: ´ aˆ a a for short vowels, ´ a´ aˆ a a aa for long. Observe that, in long falling vowels, only the first of the pair bears a circumflex (as the second segment is primarily low). Tonology is discussed at length by Watkins (1984) and Harbour (2002). Unless pertinent to the discussion, I abstract away from the causes and effects of tone interactions.

Chapter 2 KIOWA’S NOUN CLASSES 2.1.

Overview: Meaning and features

This chapter presents Kiowa’s noun class system, the empirical core of the current study. The key insight into noun classification in Kiowa was provided by Wonderly, Gibson, and Kirk (1954): (1)

Kiowa nouns are classified according to number properties

That is, number plays in Kiowa the role that gender plays in Indo-European languages; it is the primary classifier of nouns. The substantive difference between the Kiowa and Indo-European systems lies in their degree of arbitrariness. Gender-based classification is necessarily largely arbitrary, for most of a language’s nouns, such as implements, most vegetation and places, lack gender. We will see below and in the next chapter that, in Kiowa, members of each noun class share semantic characteristics with a natural nexus to number concepts. A second insight was provided by Merrifield (1959a), who refined Wonderly, Gibson and Kirk’s four-class system. Whereas the latter focused on number marking on the noun, Merrifield focused on number agreement in the verb prefix (cf., Harrington 1928). (2)

Kiowa noun classes are revealed through their effect on the verb prefix

In this way, he was able to argue for a seven-class system. Watkins (1984), the most thorough description of Kiowa grammar to date, upheld Merrifield’s 21

CHAPTER 2

22

classification, and Takahashi (1984) provided insight in the semantic coherence of the seven classes. The aims of this chapter are: (3)

a. b. c.

To explain how Kiowa noun classes are distinguished on the basis of number agreement in verbal prefix. To argue that this methodology reveals nine classes and to present a new nomenclature for these. To show that membership of the nine classes is semantically coherent.

The later sections of the chapter treat two residual issues: whether the verbal agreement prefixes support a tenth class, as Harrington (1928) believed, and whether phonology plays a role in Kiowa’s noun class system. The features that underlie the system, their semantics and their manipulation in the syntax are explored in subsequent chapters. The important issue of how Kiowa compares to other languages with rich noun class systems, such as those of Bantu, is left until Chapter 6. 2.2.

Preliminaries

The basis of noun classification in Kiowa, and of the mnemonic nomenclature adopted below, is the correlation between agreement and singularity / duality / plurality. As preliminaries, therefore, the relevant properties of agreement and number are highlighted. By number, I intend referential cardinality, the cardinality of the set of things that a noun is used to refer to on a particular occasion. Formally: (4)

Definition: Referential Cardinality Let U be an utterance containing a noun N and let S be the subset of the universe of discourse that N denotes. The referential cardinality of N is defined as | S |, i.e., the cardinality of the set that N denotes.1

So, in ‘I saw two men’, the referential cardinality of the subject is 1 and the referential cardinality of the object is 2. 1 This definition glosses over certain technicalities, such as whether proper nouns, like John, should be semantically represented as singleton sets or as individuals (on this issue, see the Schwarzchild’s 1996 appendix on Quinean set theory) and how to accommodate the contention that DPs are referential and NPs predicative (Winter 2001).

KIOWA’S NOUN CLASSES

23

Kiowa distinguishes singular, dual, and plural, that is, the language distinguishes referential cardinalities 1, 2, and 3 or more. In order to keep morphological and semantic terminology separate, the following notational convention is adopted: (5)

Notation: Referential cardinality 1 denotes referential cardinality 1 (singular). 2 denotes referential cardinality 2 (dual). 3 denotes referential cardinality ≥ 3 (plural).

Now consider the morphological form of verbal agreement in Kiowa, abstracting away from person. Agreement on the verb covaries with referential cardinality as well as with morphological marking on the noun. Four types of agreement are observed: (6)

Notation: Agreement types s, d, p, i

As a rule of thumb, s, mnemonic for ‘singular’, occurs when referential cardinality is 1; d, mnemonic for ‘dual’, occurs when referential cardinality is 2; and p, mnemonic for ‘plural’, occurs when referential cardinality is 3. i, mnemonic for ‘inverse’, occurs when the verb agrees with an inverse-marked noun (see Section 1.5.3 and below); i-agreement can occur when referential cardinality is 1, 2, or 3. Kiowa nouns do not all trigger the same agreement type for a given referential cardinality. For instance, for referential cardinality 1, some nouns trigger s-agreement, but others, i. A learner of Kiowa must learn the agreement behavior of each noun, just as a learner of French, German or Russian must learn the gender properties of each noun they encounter. Information about the correlation between referential cardinality and agreement type is represented below in the form of mnemonics. In each triliteral mnemonic, the first letter indicates the type of agreement triggered when referential cardinality is 1; the second letter, the type of agreement triggered when referential cardinality is 2; and the third letter, the type of agreement triggered when referential cardinality is 3. So, a noun, N, with the fictitious mnemonic dip would trigger d-agreement for referential cardinality 1 (‘Here’s an N ’ would show d-agreement); i-agreement for referential cardinality 2 (‘Here are two N ’s’ would show i-agreement); and p-agreement for referential cardinality 3 (‘Here are three N ’s’ would show p-agreement).

CHAPTER 2

24 2.3.

The noun classes

Merrifield (1959a) observed that Kiowa’s noun classes are revealed through covariation between referential cardinality and agreement on the verb. I will now show that this methodology reveals the existence nine noun classes. (There are, of course, 43 = 64 mnemonically possible classes. Section 3.7 addresses why 55 are unattested.) The nine classes, in mnemonic form, are: (7)

a. b. c. d. e. f. g. h. i.

sdp sdi sii idp ids idi sds sss ppp

A noun from, say, the sdp class will be referred to as an sdp noun. To illustrate the mnemonics, consider an sdp noun. For referential cardinality 1, we find s-agreement on the verb; for referential cardinality 2, d-agreement; and for referential cardinality 3, p-agreement. Consider also an ids noun. For referential cardinality 1, we find i-agreement; for 2, dagreement; and for 3, s-agreement. To justify the existence of nine classes and the appropriateness of their mnemonics, it is necessary to show that each mnemonic corresponds to one or more nouns and that each noun is described by one of the mnemonics in (7), proceeding through each mnemonic in turn. Immediately below, I will concentrate only on showing that each mnemonic corresponds to one noun. Fuller lists of members of each noun class are given during the discussion of the classes’ semantic coherence (Section 2.4), supporting the claim that all Kiowa nouns are described by one of the listed mnemonics. As the following sections are rather data rich, it should be pointed out that the reader is not required to remember any individual fact or form given below. Empirical specifics will be repeated in the theoretical chapters as and when they become relevant. The main purpose of the discussion that follows is to acclimatize the reader to Kiowa, as it were, familiarizing them with morphological and semantic characteristics of the noun class system


25

that will be treated in subsequent chapters. A summary of classes and their characteristics is presented in (77). 2.3.1.

The SDP class

Consider t´ ou ´d´ e ‘shoe’. When referential cardinality is 1, the verb shows 2, 3 s-agreement. (8)

Tóu ´dé ∅- dÓÓ shoe 3S-be ‘It’s a shoe’

When referential cardinality is 2, the verb shows d-agreement. (9)

Tóu ´dé eM - dÓÓ shoe 3D-be ‘It’s two shoes or a pair of shoes’

And when referential cardinality is 3, the verb shows p-agreement. (10)

Tóu ´dé gya-dÓÓ shoe 3P- be ‘They’re shoes’

By triggering s-agreement when referential cardinality is 1, d-agreement when referential cardinality is 2 and p-agreement when referential cardinality is 3, t´ ou ´d´ e ‘shoe’ shows itself to be an sdp noun and so justifies sdp as a noun class mnemonic. Subsequent mnemonics are illustrated by running them through the three sentence frames above: ‘It’s a ’, ‘It’s two s’, ‘They’re s’. 2.3.2. The SDI class T´ og´ ul ‘young man’ is an sdi noun. 2

It is to be assumed in all of the following examples that the agreement shown is the only option permissible. 3 It will quickly be noticed that the phonological form of the agreement prefixes bears no obvious relation to its gloss. The rather opaque relationship between prefixes’ sound and meaning is a classic problem in Kiowa-Tanoan linguistics. See Chapter 5 for a discussion of the Kiowa case.

CHAPTER 2

26 (11)

Tóg´ ul ∅- dÓÓ young man 3S-be ‘It’s a young man’

(12)

Tóg´ ul eM - dÓÓ young man 3D-be ‘It’s two young men’

(13)

Tóg´ uu ´dÓ e- dÓÓ young man.inv 3I-be ‘They’re young men’

2.3.3.

The IDP class

K´ utaa ‘pencil’ is an idp noun. (14)

K´ utaadO e- dÓÓ pencil.inv 3I-be ‘It’s a pencil’

(15)

K´ utaa eM - dÓÓ pencil 3D-be ‘It’s two pencils’

(16)

K´ utaa gya-dÓÓ pencil 3P- be ‘They’re pencils’

2.3.4.

The IDS class

´ a ‘tree’ is an ids noun. A´ (17)

(18)

(19)

´ adO A´ e- dÓÓ tree.inv 3I-be ‘It’s a tree’ ´ a eM - dÓÓ A´ tree 3D-be ‘It’s two trees’ ´ a ∅- dÓÓ A´ tree 3S-be ‘They’re trees’

KIOWA’S NOUN CLASSES 2.3.5.

The IDI class

K!ˆ On ‘tomato’ is an idi noun. (20)

K!ÔM OM dO e- dÓÓ tomato.inv 3I-be ‘It’s a tomato’

(21)

K!Ôn eM - dÓÓ tomato 3D-be ‘It’s two tomatoes’

(22)

K!ÔM OM dO e- dÓÓ tomato.inv 3I-be ‘They’re tomatoes’

2.3.6.

The SDS class

P!´ O´ O ‘river’ is an sds noun. (23)

P!ÓÓ ∅- dÓÓ river 3S-be ‘It’s a river’

(24)

P!ÓÓ eM - dÓÓ river 3D-be ‘It’s two rivers’

(25)

P!ÓÓ ∅- dÓÓ river 3S-be ‘They’re rivers’

2.3.7.

The PPP class

H´ olda ‘dress’ is a ppp noun. (26)

Hólda gya-dÓÓ dress 3P- be ‘It’s a dress’ or ‘It’s two dresses’ or ‘They’re dresses’

27

CHAPTER 2

28 2.3.8.

The SSS class

The appropriateness of the mnemonic sss cannot be illustrated via the sentence frames used for the previous classes. Nouns in this class are all mass nouns and, so, are, strictly speaking, uncountable. Nonetheless, the mnemonic can be justified, by mass-to-count conversion and conjunction.4 The significance of employing means different from those used above in order to justify the current mnemonic is discussed at the end of this section. As a preliminary to conjunction of sss nouns, observe that an increase in referential cardinality by conjunction affects agreement the same way implicit numeral modification above does. Thus compare (27), seen above, with (28). (27)

(28)

Tóg´ ul eM - dÓÓ young man 3D-be ‘It’s two young men’ ´M´MZde tóg´ E ul gO ó´Zde eM - dÓÓ this young man and that 3D-be ‘It’s this young man and that (one)’

In both cases, referential cardinality is 2, in (27), owing to implicit numeral modification, in (28), owing to the conjunction of ´ eˇ´ˇZde t´ og´ ul ‘this young man’ and ´ o´Zde ‘that’ (1 + 1 = 2). Similarly, compare (29) and (30). (29)

(30)

Tóg´ uu ´dÓ e- dÓÓ young man.inv 3I-be ‘They’re young men’ ´M´MZde tóg´ E ul gO ó´Zde y´Z´Z e- dÓÓ this young man and that two 3I-be ‘It’s this young man and those two’

In both cases, referential cardinality is 3, in (29), owing to implicit numeral modification, in (30), owing to the conjunction of ´ eˇ´ˇZde t´ og´ ul ‘this young man’ and ´ o´Zde y´Z´Z ‘those two’ (1 + 2 = 3). Having illustrated that conjunction and numeral modification are equivalent for the purposes of noun class identification, we can now use conjunction to justify the sss mnemonic.5 4 5

Conjunction justifies also the assignment of other mass nouns to the ppp class. Th´ oˇ´ u ˇ ‘water’, th´ oˇ´ u ˇÓlkhÓ´Z ‘whisky’ [lit.: crazy water], x´ o´Z ‘coffee’ and x´ o´Zg´ ul ‘tea’ are

KIOWA’S NOUN CLASSES (31)

Th´ oM u ´M ∅- dÓÓ water 3S-be ‘It’s water’

(32)

Th´ oM u ´M gO xó´Z ∅- /*eM - dÓÓ water and coffee 3S-/ 3D-be ‘It’s water and coffee’

(33)

Th´ oM u ´M gO xó´Z gO xó´Zg´ ul ∅- /*gya-dÓÓ water and coffee and tea 3S-/ 3P- be ‘It’s water and coffee and tea’

29

Further motivation for the sss mnemonic comes from mass-to-count conversion, that is, use of one water to mean ‘one portion / helping of water’. Though the mechanisms of mass-to-count conversion themselves require clarification, it is sufficient to defer this clarification for the moment and to observe that mass-to-count converted sss nouns are still sss. First, observe the typical sss agreement of a noun in this class. (34)

HÓ th´ oM u ´M ÓlkhÓ´Z an athónmO q whisky hab 2s:3S-drink.impf ‘Do you drink whisky?’

Second, observe that, in (35), ‘whisky’ has been converted to a count noun. Notwithstanding, we find s-agreement; d-agreement is unacceptable. (35)

Y´Z´Z th´ oM u ´M ÓlkhÓ´Z gya /*nen- thóm two whisky 1s:3S-/ 1s:3D-drink.pf ‘I drank two whiskies’

Even when the explicit measure phrase k!Oˆ al ‘dish, cup’ is included, the verb shows s-agreement. (36)

Y´Z´Z k!Oâl th´ oM u ´M ÓlkhÓ´Z gya /*nen- thóm two cup whisky 1s:3S-/ 1s:3D-drink.pf ‘I drank two glasses of whisky’

(37)

Y´Z´Z k!Oâl gya- thóm th´ oM u ´M ÓlkhÓ´Z two cup 1s:3S-drink.pf whisky ‘I drank two glasses of whisky’

sss nouns in virtue of their semantics: all liquids are in this class, as discussed below.

CHAPTER 2

30 (38)

PháM áM o k!Oâl an gya /*gyat- hânmO three cup hab 1s:3S-/ 1s:3P-devour.impf ‘I used to get through three glasses [of whisky]’

The irrelevance of the measure phrase k!Oˆ al ‘dish, cup’ in determining the s-agreement above is further underlined by the following sentence. (39)

Pá a´gO k!Oâl xó´Z gya- thóm one cup coffee 1s:3S-drink.pf ‘I drank one cup of coffee’

What is surprising here is the absence of inverse marking on k!Oˆ al ‘dish, cup’. It is an idp noun:6 (40)

K!OáttO bé- ól x!ep dish.inv 2s:3I-drop.s/d-lay.s/d ‘You dropped the plate’

(41)

K!Oâl men- ót dish 2s:3D-drop.s/d ‘You dropped two plates’

(42)

K!Oâl bat- p!ét dish 2s:3P-drop.p ‘You dropped some plates’

As an idp noun, it is expected to show inverse marking for referential cardinality 1. Yet, in (39), it is not inverse marked. This suggests that nouns used as measure phrases do not function as normal nouns do for the purposes of 6 The reader may notice that (40) contains two verb roots, ‘drop’ and ‘lay’, whereas (41) and (42) contain just one, ‘drop’. I have not been able to discover the difference that ‘lay’ makes. However, it should be noted that it is independent of referential cardinality.

(i)

HÔndé aót? what 2s:3s-drop.s/d.pf ‘What did you drop?’

(ii)

´ e gyat- ólEtt´ k!op many 1s:3p-drop.s/d-lay.p.pf ‘I dropped a lot’

See Section 4.7 for discussion of the surprising combination of incorporated s/d ‘drop’ with main verb p ‘lay’ in (ii).


31

number marking and agreement. Consequently, it is appropriate to attribute the s-agreement in the preceding sentences to th´ o ˇu ´ ˇ´ Olkh´ O´Z ‘whisky’ and x´ o´Z ‘coffee’ and to conclude that these nouns remain sss whether they are pure mass nouns or have undergone mass-to-count conversion. We therefore have two means of justifying the sss mnemonic. However, one might regard these as means of justifying only an s mnemonic. The problem is that mnemonics are meant to track agreement under changes of referential cardinality. As all putative sss nouns are mass nouns, and as mass nouns lack referential cardinality (one cannot say *‘It’s a water’ or *‘It’s two waters’, et cetera), the mnemonic seems to track the untrackable. Though the strength of this objection is real, it reveals at worst a slight equivocation in the use and interpretation of triliteral mnemonics. Even if mass noun mnemonics measure something other than referential cardinality, what they measure is worth measuring, as it reveals that the agreement behavior of mass nouns is not universal.7 Furthermore, qualms about the meaning of mass mnemonics could have been sidestepped entirely by using, for all classes, conjunction-based sentence frames, such as ‘It’s a and a ’, without mention of referential cardinality per se. The conclusion to be drawn is that the mnemonics are mere expository, or organizational, conveniences. It is in this way that they are used here, preparatory to the theoretical investigation of the next chapter. 2.3.9. The SII class The sii class is special. Its sole member is the first person and it is the only class triggering i-agreement for referential cardinality 2. Its existence is justified on the basis of morphological syncretism. The full data set is considered in Section 3.2.3. For now, I present the subpart that most immediately motivates an sii class: for agents of (di)transitives and the subjects 7 Compare the non-singular agreement of English (i) with Kiowa (ii). The English shows the plural agreement typical of any conjunction, whether of mass nouns or not, the Kiowa shows s-agreement, the agreement triggered by each conjunct alone.

(i)

Tea and coffee taste(*s) good

(ii)

X´ o´Zg´ ul gO x´ o´Z ∅ /*eM - t!´ olOM OM tea and coffee 3s/ 3d-tasty ‘Tea and coffee taste good’

CHAPTER 2

32

of unaccusatives, the first person exclusive dual and first person exclusive plural trigger i-agreement. This is shown below for unaccusatives.8, 9 (43)

NÓÓ a- dÓÓ 1 1S-be ‘It’s me’

(44)

NÓÓ e- dÓÓ 1 1I-be ‘It’s me and him or me and her’

(45)

NÓÓ e- dÓÓ 1 1I-be ‘It’s me and them’

Compare (44) and (45) with (46) which has i-agreement in virtue of the inverse-marked t´ og´ uu ´ d´ O ‘young men’. (46)


The crucial point is the emergence of i-agreement, inverse agreement, for referential cardinalities 2 and 3, which strongly suggests an sii class.

2.4.

Semantic coherence of the classes

We have seen that each mnemonic corresponds to some noun. It must now be shown that each noun falls under some mnemonic. Ideally, to show this, we would assign every noun in the language to one class or another. Such thoroughness belongs, however, to a project of dictionary writing and is excessive here. Instead, I establish the following claim: (47)

8

Kiowa nouns classes are internally coherent in that there are semantic properties common to members of a given noun class.

Unergatives exhibit object agreement and are classed with transitives. Kiowa has only two pronouns. NÓÓ, glossed as ‘1’, is used for first person singular, dual ´ and plural, inclusive and exclusive. Am, glossed as ‘2’, is used for second person singular, dual and plural. For third person, deictics are used (Watkins 1984). 9


33

The reader will observe as we progress through the classes that there is unlikely to be a noun that cannot be assigned to any class, especially given that, e.g., sdi is the default for animates, idp for vegetation and implements, and sdp is default in general. Consequently, establishing (47) implies that every noun in Kiowa does in fact fall into some class or other. Let me now clarify what constitutes success in showing that the noun classes are semantically coherent, or rather, let me forestall three likely misunderstandings of what constitutes failure. Non-uniqueness. The claim that classmates cohere semantically does not entail the converse, that semantically coherent nouns are classmate. Falsehood of the converse is not surprising, as one noun can possess characteristics of two classes. For instance, rivers are moving bodies of water. Consequently, p!´ O´ O ‘river’ might reasonably be assigned to one of two classes: to sdi, which contains many things that move autonomously, or to sds, which contains many bodies of water. As nouns must be assigned to one class or another—class membership is not determined ‘on-line’ according to the property most salient at the moment—‘river’ cannot be classmate with every noun with which it shares semantic characteristics. Interestingly, Dr McKenzie, with whom Watkins worked, assigned ‘river’ to the sdi class, whereas the speakers with whom I have worked assign it to the sds class. Arbitrariness. Semantic generalizations over members of a class need not be exceptionless. Exceptions are typical of noun class systems. For instance, in Russian, there is a strong implication from real-world gender to grammatical gender. Thus, nouns denoting females generally end in -a in the nominative singular, as do ˇ zenˇ sˇ cina ‘woman’, devuˇ ska ‘girl’, tsarina ‘czarina’, Karenina ‘female of the Karenin family’; such nouns end in -u in the accusative singular. By contrast, nouns denoting males generally end in -a in the accusative singular, but in a consonant in the nominative, as do starik ‘old man’, maljˇ cik ‘boy’, tsar j ‘czar’, Karenin ‘male of the Karenin family’. Despite these robust generalities, muˇ zˇ cina ‘man’, ending in -a in the nominative singular, patterns with female-denoting nouns. At the other end of the classificatory spectrum are languages like Arapesh in which class membership is primarily a matter of phonology (Foley 1986, Aronoff 1994). Here too nouns are found in classes without exhibiting the class’ phonological characteristic. Such classificatory residue does not undermine the claim that noun classification in Russian or Arapesh is systematic. The same will be true of Kiowa, in a very few cases.

CHAPTER 2

34

Subclasses. A single semantic characteristic need not be common to all the nouns in a given class. Instead, they can divide into subclasses where the nouns in each subclass share a single characteristic. We will see examples of this, for instance, in the ppp class, which subsumes both abstract and pluralia tantum nouns, and in the idi class, which comprises a hair and a fruit subclass. Subclasses do not undermine the claimed internal coherence of the classes for the following reason. Classes are defined by a classifying number feature. Let P and P be two properties and let [F] be a number feature. If there are conceptually natural relations both between P and [F], and between P and [F], then [F] can serve as the classificatory feature for nouns exhibiting either property. Consequently, the noun class defined by [F] will comprise two subclasses, one for each property, P and P . Examination of the internal semantic coherence of noun classes is a means to discover the semantics of the number features. Having two semantic characteristics that correlate with a single number feature is beneficial, not detrimental. To reiterate, then, the aim is to show that nouns in each (sub)class share a semantic characteristic, subject to the provisos immediately above. In the next chapter, where we turn to the number features that underlie the noun classes, I argue that the noun classes are ‘mnemonically’ coherent; that is, idi nouns, say, cohere not merely in virtue of Property X, but, furthermore, their mnemonic is semantically appropriate to Property X. In other words, it is not accidental that the semantic property shared by idi nouns correlates with the idi agreement profile, rather than with, say, the ppp profile. 2.4.1.

The animate classes

The SII class This class is trivially semantically coherent, having a single member. Moreover, its member, the first person, is clearly semantically distinct from all other class members. The SDI class The sdi class subsumes all animate nouns and several inanimate nouns that share certain interesting properties with animates, centering on motion (cf., ‘motility’, Noyer 1992). Let us begin with animacy.

KIOWA’S NOUN CLASSES Human animates. (48)

Gloss baby child child boy girl young man young woman man woman old, old man maternal uncle father’s mother wife husband women’s sister man’s sister

35

Words for humans are all sdi. For example: 1/2 iip!ÓÓgya ´Z´Z sân thal´Z´Z máthOn tóg´ ul yÓkÓ´Z k!yáM áM hˆMZMi maay´MZ k!yápthOO têMMi thál thá a´ k!´Z´Z p!´Z´Z tháM áM

3 iip!ÓÓgOt yyó´Z sâM aM dO thalyóp máthOM OM dO tóg´ uu ´dÓ yÓkÓ´Zg´ u k!yáM áM hyóp (/hyôm/hyôM Mi) maayóp k!yápthOOgO têMMidO thá a´lyóp thêi k!yó´Z p!yó´Z tháM áM dO

Similarly, the sdi class includes designations for humans formed by attaching a gender suffix, -k!ii ‘male’ or -maa ‘female’, to a group name, such as ‘Apache’, ‘Comanche’, ‘Mexican’, ‘White’, or adjective or noun, such as ‘crazy’, ‘chief’, ‘big, old’. (49)

10

Gloss Apache man/woman Black man/woman Black man/woman Comanche man/woman Mexican man/woman White man/woman chief elder10 madman/madwoman

1/2 ThOgˆ uik!ii/maa Paidômdek!ii/maa Kh´ oM u ´M g´MZo´M u ´M k!´Z´Z/má a´ Kyâik!ii/maa K!ópt!ÓkhÓ´Zk!´Z´Z/má a´ T!ÓkhÓ´Zk!´Z´Z/má a´ k!yátá´Zk!ii êlk!ii/maa ÓlkhÓ´Zk!´Z´Z/má a´

3 ThOgˆ ui PaidôM uM bO Kh´ oM u ´M g´MZo´M u ´M Kyâigu K!ópt!ÓkhÔi T!ÓkhÔi k!yátâi bˆMZMidO ÓlkhÔi

‘Old’, from which comes ‘elder’, is suppletive for number in Kiowa. See Chapter 4.

CHAPTER 2

36

When referential cardinality is 3, the gender suffix is omitted and is replaced by the inverse (Watkins 1984, p. 196). Alternatively, there are gender specific inverse forms, -k!yoi for males, -maimO for females. Non-human animates. Also in the sdi class are animal, bird, reptile and insect names. McKenzie (n.d.a) lists 135 birds and reptiles and McKenzie (n.d.b) lists 154 mammals, all but one sdi nouns.11 A sample, and some insects and similar creatures, is given below (not all from McKenzie’s lists). (50)

Gloss bird butterfly cedar waxwing dog frog grasshopper horse kid quail rattle snake snake spider

1/2 t!eM Miné khOibatôulé zont!OOiyátmaa xégun khOOlék!yâalé k!OOlÓtkÔM OM y´Z xêMMi ká a´boliiii pêMMisyan saM aM néhiMMi saM aM né k!OM OM nÓÓthOM OM

3 t!eM Minóp khOibatôulóp zont!OOiyátmaimO xéguMuM dO khOOlék!yâ alóp k!OOlÓtkÔM OM yóp xêMMigO ká a´boliiyyoi pêMMisyaM aM dO saM aM néhyoM Mi saM aM nóp k!OM OM nÓÓthOM OM gO

Inanimates. Although all animates are sdi, not all sdi nouns are animate. One such is hˆ eˇ ˇiii ‘doll’. Its subsumption suggests that animatelikeness is sufficient for sdi membership. In the case of ‘doll’, the nature of this likeness is clear, but it leads one to ask what other core properties of animacy inanimates may share. Self-propulsion or the ability to determine the course of motion is one such property. Several sdi inanimates are explained in this way, including heavenly bodies, machines and implements.12 11

The unique exception is the sdp noun kÓl, for which McKenzie writes: A cow or cows; a buffalo or buffaloes; usu., in the sense of “beef” or “beeves.” Sometimes, “herd” or “herds.” (The term has only one form-like English “sheep.”) (McKenzie n.d.b, p. 3)

The availability of the collective ‘herd(s)’ makes this word different from sdi animals. 12 Watkins gives p!ÓÓ ‘river’ as an sdi noun, which might be explicable on these lines. For my consultants, it is sds, however.


Gloss moon star sun car wheel, wagon, car awl, (bone) needle axe comb eyebrow tweezers hoe knife scissors

37 1/2 p!ÓÓ táM áM pá´Z khÔO k!Ódál x´ oM u ´M hÓM ÓM thoM uM Óls´ oM u ´M tá aÓlxon dOmkˆ uu k!ÔO t!áM´MZkhÓthá a´

3 p!ÓÓgO táM áM gO pá´Zg´ u khÔM OM gO k!ÓdáttO x´ oM u ´M gO hÓM ÓM thoM uM gO Óls´ oM u ´M gO tá aÓlxoM uM dO dOmkˆ uugO k!ÔOgO t!áM´MZkhÓthá a´gO

However, three terms given by Watkins do not lend themselves to this way of thinking: k!ˆ OOsoMu ˇ ‘whetstone’ [lit.: knife-grind], hˆ oux!o ‘decorative silver button (worn on head)’, and t!ÓˇÓˇ ‘spoon’. She suggests for the latter that being made of animal material is relevant—Kiowas frequently made spoons from horn, an sdi body part. Several body parts belong to the sdi class. Watkins lists the following:13 (52)

13

Gloss buttock ear egg eye gall bladder heart hide hip horn kidney knee leg liver

1/2 t!él t!ÓÓdé t!áM´MZ tá a´de t!ÓM ÓM dé thén khÓ´Z p´Z´Zthel g´ uM u ´M dé t!ÓlthOn t!elbôM uM th´ oM u ´M dé t!Ólel

3 t!éttO t!ÓÓgÓ t!áM´MZmÓ tá a´gO t!ÓM ÓM gÓ théM´MZdO khÓ´ZgO p´Z´ZthettO g´ uM u ´M gÓ t!ÓlthOM OM dO t!elbôM uM gO th´ oM u ´M gÓ t!ÓlettO

She also includes mOnx!´ oˇ ‘fingernail’. In my fieldnotes, it is sds, as are Onx!´ oˇ and Onk!Ón, both of which mean ‘toenail’.

CHAPTER 2

38 Gloss neck toe tongue spinal cord spine tooth tripe

1/2 k!ól Onthál dén pâisen gómthoM z´ oM u ´M ÓÓb´MZ´MZ

3 k!óttO OntháttO déM´MZdO pâiseM MidO gómthoM gO zêm ÓÓb´MZ´MZgÓ

Some of these body parts again suggest salient properties of animate beings, such as motion and movement (‘legs’, ‘spine’, etc.), perception and speech (‘ear’, ‘eye’, etc.), being a major organ (‘heart’, ‘liver’, etc.). Others (‘horn’, ‘tooth’) are implement-like (in particular, cutter-like), as are some items in (51). For the most part then, we see that sdi nouns are animates or inanimates that share certain salient properties with animates, such as ability to move or determine direction of motion. There is a residue of cases not readily explained by such principles. However, their relative rarity suggests that the principles are broadly correct and that the exceptions may be no more than the slight arbitrariness expected of any class system. Alternatively, the correct analogy with animate entities may simply have eluded me; these elements may be subsumed under a separate subclass in virtue of a different semantic property. In either case, they do not undermine the claim that members of the sdi class semantically cohere. 2.4.2.

The main vegetal classes

The idp and ids classes are, with sdi, the largest in the language. They subsume ‘plants and plant material, natural and man-made objects and a small number of body parts’ (Watkins 1984, p. 85). Examples of each are:14

14

Many thanks to Laurel Watkins for supplying inverse forms missing from my notes.


Gloss apple, peach, plum tree corn cottonwood cottonwood elm elm grape vine grass hackberry Osage orange pecan trees persimmon tree peyote, cactus pine tree redbud skunkberry bush tree walnut tree watercress weed willow

(54)

Gloss bean cantaloupe firewood flower leaf nut onion pecan nut seed walnut

15

39 1 álOOaadO é´ZtháttO/é´Zthâ atO a´ a´hyoM Mi a´ a´thap kÓlaadO thOOá a´dO t!éttéeibO/gO s´ oM u ´M dO a´ aéphep zépgutk!oaadO OM OM kuá a´dO á a´hyalaadO séM´MZgO(t) zónaadO khˆMZMigulaadO táM´MZpeeibO/gO á a´dO phohónaadO th´ oM u ´M é´ZbÓ/gÓ thépsOyaadO séa´ a´dO

1 t!ÓlthOM OM dO ph´ZdátkÓt k!´Z´ZbÓ/gÓ á a´khiMMigOt á´ZdeM MigO zémk!ÓÓdop sôutO/sôl15 OM OM kué´ZbÓ/gÓ é´ZbÓ/gÓ phohóneibO/gO

2/3 álOOaa é´Zthâl á a´hiMMi a´ a´thap kÓlaa thOOá a´ t!éttéei són á aéphep zépgutk!oaa OM OM kuá a´ á a´hyalaa séM´MZ zónaa khˆMZMigulaa táM´MZpeei á a´ phohónaa th´ oM u ´M é´Z thépsOyaa séá a´

2/3 t!ÓlthOn ph´Zdátkyá k!´Z´Z a´ a´khiMMigya á´ZdeM Mi zémk!ÓÓde sôl OM OM kué´Z é´Z phohónei

Dr McKenzie gives the first form, which is regular given (89), Mrs Kodaseet the second, which unexpectedly shows zero marking after l.

CHAPTER 2

40 (55)

Gloss arrow belt blanket breech cloth bucket cradle door ember fan gourd rattle gun lamp match paper peg, stake pipe saw sheet for bed shield

1 zêibOt tOnphâM aM gO khÓÓdÓ tÔigop dOâ atO/dOáttO pháM áM top xát ph´Z´ZdÓ sálpákomgOt thÔM OM gO hÓM ÓM gOt ph´Zb´ oM á a´dO thá a´dóp mÓx!áM´MZmO tóp sÓÓtóp á a´thagO t!áM´MZmÓ khyôM Mi

(56)

Gloss bone feather (downy) feather (hard-stem) foot head large intestine lymph gland nose, beak penis rib scalplock sinew udder wing

1 th´ oM u ´M segO x!´ oM u ´M gOt á a´gÓt Onsôi Ólthêm sékhôM Mi a´ aˆ atO mOOk!Ôn sóp g´ uu ´dÓ kyâip!OM OM dO téM´MZgÓt OOzâi x!ôutO

2/3 zêiba tOnphâM aM khÓÓ tÔide dOál pháM áM tol xát ph´Z´Z sálpákomgya thÔM OM hÓM ÓM gya ph´Zb´ oM á a´ thá a´dé mÓx!áM´MZ tól sÓÓtól á a´tha t!áM´MZ khˆMZMi

2/3 th´ oM u ´M se x!´ oM u ´M gya a´ a´ Onsóu ´ Ólth´ oM u ´M sékh´ oM u ´M á a´l mOOk!Ón sóu ´ g´ uu ´ kyâip!On téM´MZ OOzá a´ x!ól


41

Given the classes’ size, membership is best defined negatively, as excluding all plants, natural objects, man-made objects and body parts belonging to other classes. Definition by exclusion suggests that these classes are to some extent default. However, this is not to say that they are incoherent. Watkins observes that compounds built on ´ a´ a ‘stick, pole, tree, brush, timber’ provide insight into the difference between the basis of idp∼ids classification. Trees that grow in spinneys and groves are naturally grouped and hard to individuate, so that the collection of trees itself can by thought of as a single unit. Such trees tend to be ids nouns, as, indeed, is ´ a´ a ‘tree’ itself. (57)

IDS

Gloss elm hackberry pecan pine tree

2/3 thOOá a´ a´ aéiphep OM OM kuá a´ zónaa á a´

This is corroborated by a spontaneous comment of Mrs Dupoint’s: the following sentence would be said when one sees a grove. (58)

Té´Z phohónaa ∅- dÓÓ all walnut tree 3s-be ‘They’re all black walnuts’

By contrast, idp trees tend to be easily individuable even when there are several of them. This can be due either to their relative smallness or distance from other trees of the same kind. (59)

IDP

Gloss redbud weed willow

2/3 khˆMZMigulaa thépsOyaa séá a´

As an idp noun, ´ a´ a means ‘pole, stick’, which is smaller than a tree and, as an implement, bound up with human intentions, tends towards individuability. As there are a few discrepancies between Watkins’ idp/ids classification of plants and mine, I shall not assign all the words in (53) to one or other category. The grounds for the discrepancy are unclear. Possible sources are artifacts of judgment elicitation methodology, regional variation, or difference

CHAPTER 2

42

in speakers’ knowledge of the relevant properties of the trees in question. Uncertainty with respect to some items does not, however, undermine the validity of the observation that a concept of grouphood distinguishes ids from idp nouns, especially as the same concept pertains to the classification of idi, sds and sss nouns. Indeed, differences between idp and ids implements plausibly reflect grouphood, too. Two types of implements, primarily, are ids: those that give off light or heat (Takahashi 1984), such as embers, lamps, guns, and pipes, and those that act as covers, such as blankets, door flaps, and shields. All of these are capable of collective action, that is, a single outcome emerges from the group. So, the idp and ids classes are semantically coherent in two ways: they are the main classes for trees and other natural entities, and they are distinguished by a property of grouphood. 2.4.3.

Symmetric non-constant classes

There are four classes with symmetric mnemonics, i.e., mnemonics of the form xyx. Two of these, ppp and sss, are constant, whereas as the other two, idi and sds, are non-constant. We turn to the latter pair now. Section 2.4.4 addresses ppp and sss. The IDI class The idi class is very small, with only eight members (four, according to previous reports), mostly hair and fruit.16 (60)

16

Gloss eyebrow eyelash hair (of head) apple / plum blackberry brain17 orange tomato

2 tá aÓl tá a´phO Ól a´lOO saM aM néei k!yagóp th´ oM t!ólOM OM k!Ôn

1/3 tá aÓOdO tá a´phOgO ÓÓdÓ álOOgO/bO saM aM néeigO k!yagóp th´ oM t!ólOM OM gO k!ÔM OM dO

P´ ou ´ ei ‘strawberry’ may also be idi. On one occasion, Mr Bointy assigned ph´Zd´ atky´ a ‘cantaloupe’ to this class, though at all other times, he and others treat this as idp.


43

Fundamental to this class is, I suggest, the observation that the Italian capelli is a count noun whereas its nearest English counterpart, hair, is a mass noun (Chierchia 1998). Hence the following contrast, where the singular is preferred to the plural in English, but vice versa in Italian. (61)

a. My hair is long b. ??My hairs are long

(62)

a. ??Il mio capello the.masc.s my.masc.s hair b. I miei capelli sono the.masc.p my.p hair are ‘My hair is long’

è lungo is long.masc.s lunghi long.masc.p

Chierchia observes that ‘hair’ vacillates between a mass noun and count noun crosslinguistically, in contrast to strongly count nouns, such as ‘man’ or ‘woman’, or strongly mass nouns, such as ‘sand’ or ‘water’. idi nouns are those that can be easily regarded either as individual or as part of collections. With regard to hair, this arises because strands of hair are clearly individual, the minimal parts of bodies of hair. Yet, in a head of hair, or a fringe of eyelashes, or an eyebrow, the individual parts are not readily recognizable, and so the body too is a salient individual, albeit a group-like one. idi fruits—apples, plums, oranges, blackberries, tomatoes—are like hair in that they are clearly individuable and yet grow clustered together in collections that are themselves generally more salient than the individuals that comprise them. However, when gathered or plucked, they, unlike, say, grain, are salient individually. Thus, like hair, idi fruits can naturally be conceived of both as individuals and as part of collections. This line of thought does not readily extend to brains. They are certainly composed of parts poorly distinguishable in the whole; however, I do not know whether their typical mode of preparation in Kiowa cookery would break them into distinguishable subparts.18 Leaving brains aside (until p. 95), 17

When a noun takes a zero allophone of the inverse marker, as is common for pfinal nouns like k!yag´ op ‘brain’, class membership is still detectable through Merrifield’s method of examining the agreement the noun triggers on the verb for referential cardinalities 1, 2, and 3. 18 Mrs Dupoint said: An k!yag´ op d´ e-pÓttO. T!eˇ ˇin´ exeiyot!a ˇ ˇimO gO k!yag´ op ´ etkhOleitOnmO. E-ÓÓt!olOˇOˇ ‘I eat brains. They cook cook them together with chicken eggs.

CHAPTER 2

44

idi nouns share two properties: salient individuality, but occurrence in large collections where the collection itself is salient. Watkins has observed a third, morphosemantic, property of this class. The non-inverse noun, with s-agreement, yields a ‘different types of’ reading. Note, again, the relevance of grouphood in connection to s-agreement. (63)

´ AlOO bâ a∗- bOO apple 2p:1s:3S-bring.imp ‘Bring me apples [of three or more sorts]’ (Watkins 1984, pp. 88–9)

(64)

sÓM ÓM m´Z´Z HÓndé Ól bÓwhat hair :2p:3S-interesting ‘What interesting [kinds of] hair you all have’

idi nouns, therefore, are semantically coherent in virtue of three properties. The SDS class Like the idi class, the sds class is comparatively small. It subsumes:19 (65)

Gloss cloud house (not teepee) land(holding) path, road pond puddle river song fingernail ring toenail toenail

1/2/3 phán tóu ´ dÓm h´ oM On séxó xenth´ oM u ´M p!ÓÓ dÓÓgya mÓnx!´ oM sôude Onk!Ón Onx!´ oM

Small as it is, the list comprises two subclasses. The last four nouns share the surprising property of ‘being digit appendages’. Interestingly, several They’re really delicious.’ The implication was that brains and eggs were scrambled. 19 Watkins gives x´ o´Z ‘liquid’ as sds. I have this word only as sss with the meaning ‘coffee’. Given that all liquids I have recorded are sss, to find ‘liquid’ in this class would be surprising, unless it meant ‘body of liquid’, a sense similar to several items in (65).


45

implements also trigger s-agreement for 3, those capable of collective action, of working as a group to produce a single outcome or effect (ids, p. 42). Regarded as implements for grasping and grappling, and for protecting fingerand toe-tips, nails, interestingly, share this property; though they differ from the ids nouns in being body parts, rather than inanimate implements. The other sds nouns are ‘non-shape-inductive’ (cf., Takahashi 1984). That is, although all are clearly individuable (one can tell one river from another, one song from another, and so on), the shape of one is a poor indicator of the shape of the next. One cannot extrapolate from the shape of one to the shape of the next. This is an unexpected criterion of classhood. However, strong support for its actuality comes from my having discovered the sds nouns ‘cloud’, ‘puddle’ and ‘song’ by thinking of things that are individuable but non-shape-inductive. 2.4.4.

Symmetric constant classes

Kiowa has two more classes with symmetric mnemonics, ppp and sss. These are strongly semantically coherent. The SSS class The sss class consists exclusively of non-granular mass nouns, such as ‘milk’ and ‘honey’, or mass nouns of dubious granularity, such as ‘snow’ and ‘sleet’, which, if granular, are so primarily when descending. (66)

Gloss beer coffee juice, soda milk soda tea water whisky honey honey, syrup sugar

1/2/3 ph´Ztth´ oM u ´M xó´Z th´ oM u ´M t!ólOM OM kÓlzep th´ oM u ´M gul xó´Zg´ ul th´ oM u ´M th´ oM u ´M ÓlkhÓ´Z t!elséppenhaa a´ a´penhaa pénhaa

CHAPTER 2

46 Gloss ash hail, sleet ice rain snow eyebutter, sleepy dust snot, mucous

1/2/3 sOphán t!én téM´MZgya sép t!ól tá a´sek!On sén

All the names for liquids in my fieldnotes are in this class. Some items closely akin to particular liquids (for instance, semi-liquid or frozen forms of precipitation), and sugar (which, when Kiowas first encountered it, came in blocks and not granularly) are included by association. The class’ semantic coherence is obvious. The PPP class The ppp class is more heterogeneous than the sss class, subsuming two or, depending on the speaker, three subclasses: pluralia tantum and composite nouns, abstract nouns, and, for some, granular mass nouns. Pluralia Tantum and Composite Nouns. Objects composed of several parts are constant plurals in Kiowa. These include items familiar as pluralia tantum from English, such as ‘trousers’. (67)

Gloss belongings, stuff book, letter, picture book, letter choker clothing, shirt hoop game necklace pack saddle quilt roach (headdress) teepee trousers war headdress

1/2/3 o´l k´ ut t´ oM u ´M k´ ut k!ólphaM t!´ oM u ´M gya k!ÓdálOOgya k!ólphaM aM olpháM thOOgya pál k!ókÓ´ZouphOl tóu ´ khÓÓdé aat!OhÓ´Z


47

Granular Mass Nouns. Granular mass nouns are distinguished from non-granular mass nouns in having natural minimal parts. Thus, sand, which comes in grains, is granular, whereas water, which does not, is not. Some examples are: (68)

Gloss flour pepper rice salt sand

1/2/3 é´ZtOM xó´ZséM oM uM gya é´Zyóguei ÓÓtháM t!aM Mi pé´Zgya

These nouns are classified as ppp on the basis of conjunction, analogously to sss nouns: (69)

O´ gO

gya-kót gO ÓgO gya-phÓÓ dé gya-/*eM - mOsótdOO rel 3p- expensive and rel 3p- cheap nom 3p- / 3d-be mixed ‘The expensive and the cheap (salt) are mixed up’

Although two quantities of salt are talked of, we find p-agreement, not d. It is, however, to be noted that these nouns are not ppp for all speakers. Rather, they appear to be idp nouns, with the inverse-marked noun corresponding to the minimal part: (70)

Xó´ZséM oM uMgOt mÓÓ e- ét pepper.inv like 3i-big.s ‘The pepper grain is kind of big’

Such speakers permit a corresponding dual reading for the granular sense. (71)

´M´MZde y´Z´Z xó´ZséM oM uM gya eM - bˆZn E this two pepper 3d-big.d/p ‘These two grains of pepper are big’

However, speakers for whom granular mass nouns are ppp strongly reject sentences like (70) and (71).20 Instead, they offer sy´ Onde x´ o´Zs´ eˇ o ˇu ˇ gya ‘a little pepper’ as a Kiowa equivalent for ‘a grain of pepper’. 20

Watkins gives no granular mass noun as ppp but assigns them to her Class II, a superclass of my idp (her IIa) and ids (her IIb). Her exposition of the noun class system does not dwell on mass nouns and so the crucial sentences are not supplied.

CHAPTER 2

48

ppp classification of granular mass nouns is on a semantic par with the treatment of pluralia tantum: both trousers and piles of salt are single objects composed of complex parts. Note, however, that this composition of parts differs from the group-like idi and sds plurals (and ids below). There, the plurality consists of naturally salient parts, such as apples or hairs, or clouds or rivers. In the case of pluralia tantum nouns, the constituent parts are non-uniform and so there is no minimal part; and, in the case of granular mass nouns, though there are minimal parts, a grain of sand is not salient to the same extent as, say, an apple is. Abstract Nouns. The second ppp subclass common to all speakers is reserved for abstract nouns. Watkins observes that abstract nouns, such as ‘zero’, ‘biology’, ‘adverb’ are ppp (though, with the discussion focused on productivity of noun classification, Kiowa examples are not provided; Watkins 1984, p. 92). The following from my fieldnotes fit this description. (72)

Gloss problem dance word, language sleep work

1/2/3 mOÓ´Zde k´ ungya t´ oM u ´M gya déM´MZgyá sÓÓtégya

She hypothesizes that this is a natural extension of the use of p-agreement for unidentified or unspecified arguments, exemplified below. (73)

Gya-g´MZ´MZ- sal 3P- night-hot ‘It’s a hot night’


(74)

kÔm há a´gyá gya-k!´ ul YáM 2s:1s:3P-indicate.imp where 3p- be lying.p ‘Show me where they are’ (Watkins 1984, p. 235)

(75)

p´MZ´MZÓmtÓÓ nO dá al mán∗- pOO NÓÓ mán1 1s:2d:3P-cook.fut conj must 2d:3P-eat.imp ‘I will or If I cook for you, you must eat’ (ibid., p. 237)

If so, ppp classification of abstract nouns is independent of pluralia tantum, composite nouns and granular mass nouns. Consequently, the class comprises two disjoint subclasses, each semantically coherent.

KIOWA’S NOUN CLASSES 2.4.5.

49

The default class

Speakers of languages for whom Kiowa’s noun class system is somewhat foreign are likely to anticipate sdp as the most natural class, given that sdp agreement transparently reflects referential cardinality. Yet the class is small: (76)

Gloss arrow smoother boot earring key moccasin nail (for hammering) rock, stone rope shoe

1/2/3 zêibat!OM MiOM OM ba tók´MZ´MZn´Z´Z t!Oá a´ pháM áM xo touhˆMZMi hÓM ÓM t!Okuu x!óu ´ yá´ZpÓ tóu ´dé

Naturally, it excludes elements that ought to be in other classes, such as animates, vegetation, mass nouns, and non-shape-inductive items. However, there is little in the way of positive criteria for this class, beyond a collection of footwear. This suggests that sdp is the default class, to which nouns are assigned when they do not meet the membership criteria of other classes. 2.4.6.

Summary

The foregoing discussion of the noun classes has shown that they are internally semantically coherent, that is, that there are properties that members of each class generally share. These are summarized below.

CHAPTER 2

50 (77)

Class sii sdi

idi

sds sss ppp idp ids

sdp

Characteristics First person Animates and animate-like entities (physically similar to animates or capable of self-propulsion or determining direction of motion) Naturally regarded both as individual and as occurring in collections; permits ‘different types of’ reading with s-agreement Individuable, non-shape-inductive Non-granular mass nouns Pluralia tantum, composite nouns (and granular mass nouns for some speakers); abstract nouns Default for vegetation and implements; granular mass nouns for some speakers Vegetation forming natural collections in which members are not readily individuable; implements that act as a group to produce a single effect Default; no unifying properties

In the coming chapter, we turn to a second way in which the classes are semantically coherent, namely the connection between the semantic characteristics of class members and the class mnemonic. That is, I will argue that, given the internal semantic coherence of the classes, there could not be a language, Kiowa , in which the mnemonics and characteristics are randomly permuted. Before that, the coming two sections discuss and dismiss the possibility of a tenth noun class and the notion that class membership is to some extent phonologically, rather than purely semantically, determined. Readers of unbridlable enthusiasm may skip these sections. 2.5.

Against a tenth class

Noun classes were distinguished above by examining the correlation between referential cardinality and agreement, following Merrifield (1959a). Harrington (1928), employing a similar method, identified a class not represented above. On current terms, it would be sda, where a-agreement is animate plural agreement.21 Watkins explains that speakers typically use a-agreement to 21

Harrington distinguished between ‘animate major’ (sda) and ‘animate minor’ (sdi) categories; pp. 14, 237ff.


51

refer to fellow tribal members but i-agreement for members of other tribes, as illustrated below. Observe that, in both cases, the nouns are inverse marked. (78)

(79)

KÓ´Zg´ u á- k´ uu ýÓ Kiowa.inv 3A-be lying.p.distr ‘Kiowas are camped about’

(Watkins 1984, p. 84)

uu ýÓ Kyâygu e- k´ Comanche.inv 3I-be lying.p.distr ‘Comanches are camped about’

(ibid.)

Unlike Harrington, Watkins does not treat sda as a separate class however. Watkins’, I believe, is the correct response. Nearly any noun, or for some speakers, any noun, capable of triggering a- is also capable of triggering i-agreement. So, on the assumption that classes are pairwise disjoint, sda is not a class distinct from sdi. Rather, nouns triggering a-agreement are sdi nouns with a special property. On the basis of the following types of complementary examples, I suggest that this property is ‘empathy’ or ‘degree of identification’ with the sdi noun. In a hunting story related to me by Mr Bointy in August 2001 (see Appendix), k!y´ a ˇ´ a ˇ hy´ op ‘men’ triggers both i- and a-agreement. It relates an incident in which Mr Bointy was not a participant. The three opening sentences state where the men were and what they were doing—a simple reporting of facts, with little scope for empathy. The verbs show i-agreement. (80)

PháM áM o k!yáM áM hyóp e- ´MZ´MZ- aM aM - hel á a´- dom. SyÓn- de gya-h´ oM Onthree man.inv 3I-hunt-come-hsy tree-loc small-adv 3p- pathdOM OM -mei dé- em té´Z e- khOlé´Z- aM aM - hel. GO e- xándebe- impf.hsy nom-loc all 3I-together-come-hsy conj 3I-reachhel ÓgO á a´dO e- bôn- dOM OM -mei dé- eM Mi hsy rel tree.inv 3i-bend-be- impf.hsy nom-loc ‘Three men were hunting in the woods. They were all going along a narrow path when they arrived where the path lay under a bent tree.’

Then follow a few sentences of reported speech, concerning what should be done about a panther lying in ambush. These relate an experience with which Mr Bointy, an experienced hunter, could well empathize. Thereafter, the verbs show a-agreement.

CHAPTER 2

52 (81)

NO conj

thÓp beyond

utdehel. a´- khˆ 3A-pop/flee out-hsy

∅- T!Om-áM áM - dé 3s-first- come-nom

gôm back

gya- b´ oM u ´M - hêl nO héM´MZ ∅- dÓM ÓM -mêi. Tékhop a´∗3s:3p-look-hsy conj gone 3s-be- impf.hsy everywhere 3A:3sdon- hel né hÓn a´∗- thOM OM -mOO. seek-hsy but neg 3A:3s-find- neg ‘So they passed [quickly] to the other side. The one at the head looked back and it was gone. They looked for it all over but didn’t find it.’ Another pair of examples on similar lines concerns animals. In (82), which is about horses, the verb bears a-agreement, whereas in (83), which is about ants, it bears i-agreement. (I assume (82) to have a pro argument; this is irrelevant to agreement type; Watkins 1990, Adger and Harbour 2007.) (82)

´ zéM´MZma A3P-move.p.impf ‘They (horses in a pasture) are milling around, moving around’ (Watkins, p.c.)

(83)

EmhâM aM mop e- x´ oM u ´M - zéM´MZma ant.inv 3I-crawl-move.p.impf ‘Ants are crawling around’

(Watkins, p.c.)

a-agreement in (82) does not motivate an sda class, as, generally, xˆ eˇ ˇigO ‘horses’ triggers i-agreement, so that xˆ eˇ ˇigO is best regarded as the inversemarked form of the sdi noun xˆ eˇ ˇi ‘horse’. A typical example is: (84)

ˆ OihyOO xêMMigO ÓphÓÓhel then horse.inv :3s:3I-get.pf.hsy ‘That’s when he got the horses’

Once again, I propose that a-agreement in (82) should be explained by appeal to empathy, given the centrality of horses to traditional Kiowa culture—a conclusion drawn on the basis of numerous stories and facts told me by Mr Bointy in the course of our work together—a centrality, or empathy, in which ants by no stretch of the imagination participate. Grammaticalized empathy is in part a cultural phenomenon. Indeed, Watkins observes that i-agreement has a slightly disrespectful connotation


53

when used for adults. a-agreement is all but obligatory for K´ O´Zg´ u ‘Kiowas’: (85)

KÓ´Zg´ u né- /??nÓ- xán Kiowa.inv :1s:3A/ :1s:3I-arrive.pf ‘The Kiowas came to me’

The constant occurrence of K´ O´Zg´ u ‘Kiowas’ with a-agreement may be seen as a default, cultural empathy between any Kiowa speaker and their kin.22 Given that putative sda nouns are just empathized with sdi nouns, it is incorrect to posit a tenth class (with fluid membership). 2.6.

Phonological incoherence

Crosslinguistically, membership of noun classes is frequently phonologically determined. Phonological classes are of two types: there are Bantu-style classes (see Chapter 6), where noun and adjective begin with m- if the noun is in Class 1, but with wa- if it is in Class 2; and there are Indo-Europeanstyle classes, where nouns with a particular ending are all classmate, as are Latin nouns in -i¯ o. Bantu-style classes are apparent from phonological elements external to the noun itself; Indo-European-style classes, from elements internal to the noun. Dramatic instances of the latter are the Arapesh and Yimas languages of Papua New Guinea (Foley 1986, Aronoff 1994). A natural question is whether Kiowa’s noun classes are (to some extent) phonologically defined and whether, then, they are not organized purely semantically. However, Kiowa noun classes are not phonologically defined. First, the ten or so allophones of inverse marking distribute freely across all inversive classes (Section 2.6.1), so classes are not defined Bantu-style, by selecting particular affixes. Second, thematic nouns (defined in Section 2.6.2), which could imaginably yield Indo-European-style classes, cross-cut the number classes and show semantics to be the key determinant of class membership. 22

The notion of ‘default empathy’ may help to explain why a-agreement is systematically ambiguous between animate plural agreement and reflexive agreement: (i)

De- h´ ol 1s:3a-kill.pf ‘I killed myself or them (people)’

The reflexive ‘oneself’ might be regarded as involving obligatory empathy, rather than as being in some mysterious sense obligatorily animate plural.

CHAPTER 2

54 2.6.1.

Inverse allophones

The only morpheme capable of yielding Bantu-style noun classes in Kiowa is the inverse, as it is the only nominal affix that triggers agreement. For instance, in (86), the idp direct object in the relative clause is inverse marked and triggers inverse marking on the subordinate verb, on the relative clause itself and on the matrix verb. (86)

[P´MZá a´dO e- ét ]-gO dé- hÓÓgya table.INV 3I-big.s -INV 1s:3I-get.pf ‘I bought a table that is big’

(Watkins 1984, p. 231)

Similarly, demonstratives agree for inverse with their noun. This is shown below for the sdi noun sˆ an ‘child’. In (87), where referential cardinality is 1, there is no inverse marking; however, in (88), where referential cardinality is 3, there is inverse marking on the demonstrative, the noun and the verb. (87)

(88)

´M´MZde /ó´Zde sân ∅- khópdOO E This/that child 3s-sick ‘This/that child is sick’ ´M´MZgO E /ó´ZgO sâM aM dO e- khópdOO This.INV/that.INV child.INV 3I-sick ‘These/those children are sick’

(Watkins 1984, p. 97)

(ibid.)

The inverse marker has some ten distinct forms. So, it is reasonable to ask whether any are proprietary to particular classes. To show that they are not, it suffices to show that all are purely phonologically conditioned. The inverse marker has the forms in (89).23 Thematic nouns (-th) such as ‘tobacco’ are discussed in Section 2.6.2.

23

This follows Watkins, except that -∅ is my addition. Some phonological comments are needed. First, note three informally stated processes: (i) VN→V M VM in the context of inv dO; (ii) l→t in the context of tV; (iii) V→∅ in the context of V V. Second, note that the tone of the inverse marker need not be specified, contra Watkins’ practice, as it is predictable (Harbour 2002; -gu resists local tone lowering; hence the final high in KÓ´Zg´ u, in contrast to final lows in KÓ´Zmaa ‘Kiowa woman’, KÓ´Ztoˇ gya ‘Kiowa language’).


Suffix Form Context V M´MZ -mO -bO m n -dO l -tO l -Ot th -gu Vi ˆ-i

Vi

-oi -op

ii e

-∅

p/t

-gO

elsewhere

55 Example Basic∼Inverse Gloss t!áM´MZ∼t!áM´MZmO egg kóm∼k´ oM u ´M bO friend k!Ôn∼k!ÔM OM dO tomato tóg´ ul∼tóg´ uu ´dÓ young man tâl∼tâttO skunk thá a´bá∼thá a´bÓt tobacco KÓ´Z-∼KÓ´Zg´ u Kiowa K!ópséná´Z-∼K!ópsénâi Mexican ÓlkhÓ´Z∼ÓlkhÔi madman p!´Z´Z∼p!yó´Z (female’s) sister saM aM né∼saM aM nóp snake k!óp∼k!óp mountain xát∼xát door xêMMi∼xêMMigO horse th´ oM u ´M se∼th´ oM u ´M segO bone

(Note that -gu∼ˆ-i are in competition and must be learned on a case-bycase basis.) Tone change too may occur, sometimes with vowel change or suffixation. (90)

Processes tone change only tone + vowel change tone change + suffixation

Basic∼Inverse mOOk!Ón∼mOOk!Ôn thâa∼thêi zêiba∼zé´ZbOt

Gloss nose wife arrow

And there is a unique l∼p alternation in words built on t´ ol ‘peg’ (55). The point to observe, relating to (89) and (90), is that no form of the inverse marker is proprietary to any one noun class. Each is found in any class with a phonologically appropriate noun in it. Hence, we find -tO on sdi tˆ al ‘skunk’ and idp k!Oˆ al ‘dish’; -∅ on sdi t!´ ap ‘deer’, ids k!´ op ‘mountain’ and idi k!yag´ op ‘brain’; -dO on idi k!ˆ On ‘tomato’, sdi sˆ an ‘child’, and idp s´ on ‘grass’; -mO on sdi t!´ ou ´t!´ a ˇ ´ˇZ ‘white faced cattle’, idp m´ Ox!´ a ˇ ´ˇZ ‘paper’, ids t!´ a ˇ ´ˇZ ‘sheet’; -op is sdi sa ˇa ˇ n´ e ‘snake’ and idp m´ Ons´ ou ´de ‘bracelet’; and so on. Even the competing -gu∼ˆ-i do not define classes, as, by lexical coincidence,

CHAPTER 2

56

they are almost entirely confined to animates, which are classmate.24 Particularly striking in this regard are words that serve as the base for compounds or metaphorical extensions of meaning. Though the phonological shape that conditions the inverse allophone is identical, the class membership varies according to semantic principles. So, t!´ OlthOn is sdi as ‘kidney’ but idp as ‘bean’; th´ o ˇ t!´ olOˇOˇ is sss as ‘soda, fruit juice’ but idi as ‘orange’; ´ a´ a is idp as ‘stick, pole’ but ids as ‘tree’, and it forms the idp compound khˆˇZˇigulaa ‘redbud’ and ids z´ epgutk!oaa ‘Osage orange’; t!´ a ˇ ´ˇZ ‘white’ forms sdi t!´ a ˇ ´ˇZ ‘egg’ and t!´ ou ´t!´ a ˇ ´ˇZ ‘white faced cattle’ and ids t!´ a ˇ ´ˇZ ‘sheet’; ppp hÓˇÓˇgya ‘metal’ forms ids hÓˇÓˇgya ‘gun’ and sss ´ OlhÓˇÓˇgya ‘money’. Similarly, the inverse form of noun+adjective compounds is determined by the phonology of the adjective, to which the inverse suffix attaches, and not by the noun: hence, in k!´ op ‘mountain’, the inverse suffix is -∅, but in K!´ op´ ettO ‘Mount Scott’ [lit.: big mountain], it is -tO, attaching to ˆ el ‘big.s’; compare also the inverse markers of ´ a´ akyo ˇˇimO ‘long pole’ and ´ a´ adO ‘pole’. Consequently, no form of the inverse marker can be regarded as constituting the phonological signature of any particular class. This means that affixal phonology does not define classhood, as it does in other languages. 2.6.2.

Thematic nouns

In Indo-European, nouns with the same ending frequently belong to the same gender class. Many Kiowa nouns end in the ‘themes’ -gya, -de and -ba, which might be reminiscent of gender-specific noun endings in Romance. However, a noun’s number properties correlate with the theme it takes, suggesting that phonology is again irrelevant. Moreover, these number properties are not always the same as the semantic principles that determine membership of the classes discussed above. Where they conflict, the semantic principles are decisive, not the theme. Themes I begin with morphological and semantic properties of the themes.25 24

Note though khÓ´Zg´ u ‘cloth.inv’ from khÓ´Z; and Onsˆ oi ‘foot.inv’ from Ons´ ou ´ , OOzˆ ai ‘udder.inv’ from OOz´ a´ a, and s´ ekhˆ oˇ ˇi ‘large intestine’ from s´ ekh´ oˇ´ u ˇ , by apparent overapplication of ˆ-i. 25 The discussion follows Watkins’ Section 3.131, though aspects of the morphological decomposition are my own.


57

Kiowa has a variety of nominal roots that cannot occur independently. They must be suffixed, for instance, by -gya ‘in’ (91), or incorporated (92). (91)

(92)

Tá a´-gya hÓndé éMxél eye- in something :1s:3s-be set.s/d ‘There is something in my eye’ tá- hot- guM uM eye-travel-distr ‘watch everywhere’

(Watkins 1984, p. 93)

(ibid.)

When without a postposition or when unincorporated, thematic nouns end in a special -CV(C) suffix, the ‘theme’; t´ a´ a- ‘eye’ takes -de for referential cardinalities 1 and 2, and -gO for 3. (Agreement correlates -gO with inverse.) The themes -de∼-gO occur with: (93)

Gloss ear eye half horn (of cow) leg shoe trousers

de-Form t!ÓÓdé tá a´de zâide g´ uM u ´M dé th´ oM u ´M dé tó u ´dé khÓÓdé

These nouns naturally occur in pairs. Interestingly, e is a frequent exponent of d-agreement in the verbal prefixes (Chapter 5), and O, of i-agreement: (94)

n´ e = :1s:3d; m´ enˆ ei∗ = 2d:1s:3d; . . . n´ O = :1s:3i; m´ enˆ OO∗ = 2d:1s:3i; . . .

This suggests that these thematic nouns have e when non-inverse because they denote inherent pairs. Consequently, the vowel appears to be independent from the consonant in the thematic CV suffix and is connected to number. The consonant may be supposed to be g: g+O trivially yields gO and g+e yields de by regular Kiowa phonology.26 The other themes support this decomposition. Consider -gya: 26

Dentals become velar before i/y and velars dental before e. See Halle (2005) for a recent treatment.

CHAPTER 2

58 (95)

Gloss beads cantaloupes downy feathers fabric, dry goods sand seeds shirt

gya-Form p´ oM u ´M gya ph´Zdátkyá x!´ oM u ´M gya phátkyá pé´Zgya d´ oM u ´M gya t!´ oM u ´M gya

These form their inverse in -gOt. Decomposing onset and rhyme yields -gand -ya / -Ot. Again, the rhymes are recognizable from the verbal prefixes: (96)

gy´ at = :1d/p:3p; gy´ at = 3s:2s:3p; . . . d´ Ot = :1d/p:3i; g´ Ot = 3s:2s:3i; . . .

The theme’s non-inverse form correlates with p-agreement morphology and the nouns in (95) naturally occur in groups or are pluralia tantum. Similar observations can be made of the handful of nouns that take the theme -ba. Decomposing, we have the theme -b- and the plural-associated rhymes -ya / -Ot.27 Again, the relevant nouns plausibly occur in groups. (97)

Gloss arrows reed pipes tobacco

ba-Form zêiba (inv = zé´ZbOt) tóM u ´M bá thá a´bá

The resulting picture of Kiowa thematic nouns is that they consist of a root followed by a thematic consonant, b or g, followed by a number-related coda, ya/Ot or e/O. Themes and classes The relationship between themes and the meaning of the nouns to which they attach shows that themes are not indicative of phonological noun clas(i) 27

⎤⎡ ⎤ ⎡ −consonantal +consonantal [αback] → [−αback] / ⎣−sonorant ⎦ ⎣+sonorant ⎦ αback

The combination b+y standardly simplifies to b; for instance compare (96)’s gy´ at, :1p:3p, from d+y. . . , with b´ at, :2p:3p, from b+y. . . .


59

sification. The ‘pairhood’ that conditions e/O is not a conditioning factor for noun class membership. Such cases might open the door to phonologically determined class membership, but instead underline the semantic basis of the system. In such cases, the semantic principles of each noun class determines class membership: of the nouns in -de, ‘eyes’, ‘ears’, ‘legs’ and ‘horns’ are sdi, ‘trousers’ ppp, and ‘shoes’ sdp; of the nouns in -gya, ‘shirts’ and ‘sand’ are ppp, ‘beads’, ‘cantaloupes’ idp, and ‘guns’ ids. Indeed, two words in this last group for etymological reasons, iip!´ O´ Ogya ‘baby’ and b´ elk´Ztkya ‘screech owl’, are animates and, so, sdi. Thus, nouns that share thematic endings fall into different noun classes and do so on purely semantic grounds. 2.6.3.

Conclusion

Kiowa noun classes are not phonologically defined. Rather, the semantic classification in (77) stands.

This page intentionally blank

Chapter 3 Number Features The previous chapter showed that nouns in Kiowa are organized into nine number-related classes, according to semantic properties of their referents. Such a system poses several questions: (1)

a. b. c.

d.

e.

What are the features that do the classifying? How do classifying features interact with referential cardinality to yield inverse marking on the noun? How do the classifying features and referential cardinality interact with the syntactic mechanisms responsible for agreement to create patterns like sdp, idi, ppp, et cetera? What is the relationship between a class mnemonic and the properties of the nouns it subsumes? (Is it coincidence that animates are sdi and hair types idi and not vice versa?) Why are there nine classes and why no did, sip, or psi?

These questions are the focus of this chapter. Answering them consists in providing an inventory of number features, properly semantically defined, together with a theory of their distribution and interaction in the syntax and their treatment in the morphology. To do this, I make two claims. Claim #1. There are three number features, motivated on semantic grounds: [±singular] and [±augmented] by referential cardinality (Noyer 1992; also Conklin 1962, Matthews 1972, McKay 1978, 1979; cf., Krifka 1992) and, later, by properties of noun classes, [±group] purely by properties of noun classes (cf., Lasersohn 1995, Winter 2001, Krifka 1992). Thus, all three features determine noun class and the first two additionally 61

CHAPTER 3

62

determine referential cardinality; the first two, further, are phonologically realized as inverse marking on the noun and agreement on the verb. Claim #2. DPs have the structure below (cf., Carstens 1991, Ritter 1991, 1993), where Class is the locus of classificatory features, Number the locus of referential cardinality, and D the locus of the features that trigger agreement and inverse marking. (2)

DP NumberP

D

Class: Noun Number Given Claim #1, this means that features are distributed as in (3). (3)

DP

NumberP ⎡Class: Noun ⎤ (±singular) ⎣(±augmented)⎦ (±group)

Number

D ±singular ±augmented

±singular ±augmented

The features in (3) are not all of the same status. Those on Class and Number are intrinsic to the meaning of the head, that is, they are interpretable in the sense of Chomsky (2000, 2001). The number features on D are not intrinsic to its meaning. When merged, D bears uninterpretable number, notated [usingular uaugmented], which must be valued. Valuation (Section 3.2.2 and following) involves matching the feature content of D with that of Class and Number. In a departure from other work, such as Chomsky (op. cit.), uninterpretability is argued to involve overspecification, rather than underspecification, of the feature. That is, [uF] is argued to abbreviate specification of both [−F] and [+F], which must be matched with interpretable counterparts or else deleted, rather than specification of a valueless [ F]. This view permits valuation of D to produce the feature conflicts, [−F +F], on which the analysis of the inverse hinges.

NUMBER FEATURES

63

(4)

DP

NumberP

D

u singular u augmented

⎡Class: Noun ⎤ (±singular) ⎣(±augmented)⎦ (±group)


Verbal agreement is with D, once valued; the agreeing heads do not enter into a direct relationship with Number or Class. The evidence for the content of Class and of Number is semantic. Evidence for the feature content of D is morphosyntactic and is discussed below, primarily in relation to inverse marking on nouns, and in Chapter 5. I will show that the features’ semantics explain the relationship of class characteristics to class mnemonics and that Kiowa’s class system makes near optimal use of the inventory of classificatory features at its disposal. Claim #1 is particularly important in relation to the current work’s aim of establishing the core elements of a morphosemantic theory of number. Specifically, as the references cited on p. 61 show, the features argued for below have precedents both in morphological and semantic work. However, what is particularly striking is the precise formulation of the features (the significance of the formal definitions is discussed below): (5)

Feature definitions x, a free variable a. [+singular] = atom(x) b. [+augmented] = λP∃y[P(x) ∧ P(y) ∧ y x] c. [+group] = λP[+augmented] ∃Q[P(x) ∧ P = Q ∧ Q-atom(x)]

(5a−b), although based on previous morphological work (especially Noyer 1992), are in fact variants of definitions proposed by Krifka (1992) in his treatment of aspect and Aktionsart. Their formal equivalence is demonstrated in Harbour (2006b). Furthermore, the notion of Q-atomicity, crucial to (5c), is also in common with Krifka. This convergence complements, and concretizes, the purely architectural argument of Chapter 1 by showing that, not merely is a morphosemantic theory necessary, but that its core elements have begun to emerge in the seemingly disjoint work of both subfields.

64

CHAPTER 3

Structure of the chapter This chapter contains the core ideas of the current study: proposals concerning the feature inventory of Universal Grammar, these features’ distribution and operations affecting them. In consequence, the chapter is rather full. Readers are, therefore, advised to return to this section if, at any time, they find the progress of the argument to be obscured by its detail. The chapter falls into three parts. The first, Sections 3.1−3.3, concerns the features [±singular] and [±augmented] and their associated classes. The second, Sections 3.4−3.5, concerns the feature [±group] and its associated classes. The third, Section 3.7, concerns the totality of possible classes. The first part of the chapter is structured as follows. Section 3.1 argues for two features, [±singular] and [±augmented], on the basis of referential cardinalities and the natural classes that they form. Proofs relating referential cardinalities to particular feature-value combinations are presented in Appendix 3.8. Section 3.2 comprises several subsections. Section 3.2.1 shows that Kiowa nouns bring with them into the syntax different specifications for the features [±singular] and [±augmented] and that noun classification consists in such feature-value specifications. The content of Class is specified, in terms of [±singular] and [±augmented], for the classes sii, sdi, idp, idi, and sdp (and, partially, for ids). Section 3.2.2 proposes a simple principle by which referential cardinality and nouns’ inherent number specifications interact to yield inverse marking on the noun. The basis of this claim is that D is valued by a computation over the features located at Class and Number and that this computation leads to inverse marking / i-agreement when Class and Number have opposite specifications for a single feature. Finally, Sections 3.2.3−3.2.4 present further derivations. Examples of each of the classes discussed in Section 3.2 are given, and it is shown how positing an sii class for first person derives syncretic properties of first person agreement. DPs modified by adjectives, demonstratives, relative clauses and ‘only’ are also derived. (Readers may prefer to skip these technical excursions on early readings of the chapter.) Section 3.3 argues that the relationship between a noun class’ mnemonic and the semantic properties of the nouns it subsumes is non-arbitrary. The basis of this explanation is the conclusion that mnemonics are not primitive, but result from nouns’ Class specification. Thus, the question is not

NUMBER FEATURES

65

how class characteristics relate to mnemonics, but how they relate to Class. Since the features at Class have by this stage been rigorously defined, the relationship between Class and nouns’ semantic properties are readily explicable. Consequently, we are able to say why animates are sdi and hair types idi, not vice versa. This concludes the first part of the chapter. The second part of the chapter concerns s-agreement that occurs when referential cardinality is not 1 and p-agreement that occurs when referential cardinality is not 3. Section 3.4.1 argues that these are complementary phenomena, to be implemented as different values of a single feature. The feature [±group] is then rigorously defined and argued to predicate of [+augmented]. The Class specifications for sds, ids, and ppp (and for idi on its ‘different types of’ reading) are then given. Section 3.4.2 proposes how [±group] affects the valuation of D. Derivations of various referential cardinalities of sds, ids, ppp, and idi nouns are given. The section also explains the mnemonic naturalness of sds, ids, ppp. Finally, Section 3.5 analyzes the specification of Class for mass nouns, arguing that these crucially involve the feature [±group]. This section also returns to the topic of agreement triggered by conjunctions, discussed in Section 2.3.8, where mass nouns and their mnemonics were introduced. This concludes the second part of the chapter. The third part of the chapter asks why Kiowa has only nine of 64 mnemonically possible classes. It argues that Kiowa almost optimally exploits the space of possible mnemonics generated by the features at its disposal. 3.1.

Referential cardinality

The main claim of this section is that 1, 2 and 3, the values of referential cardinality, are not atomic. Rather, they are composed of two features, [±singular] and [±augmented], to be defined below. The evidence for this claim comes from the distribution of i-agreement and inverse marking. (This section recapitulates, in part, arguments from Noyer 1992 and Harbour 2003a. The analysis of cardinality is essentially Noyer’s.) 3.1.1.

Natural classes

The values 1, 2 and 3 tempt one to posit three monovalent number features, [singular], [dual], [plural]. However, such privative features predict that there

CHAPTER 3

66

are no natural classes of number features, that no process of the grammar refers both to 1 and 2 or to 2 and 3. However, such processes do exist, both in the morphology (Chapter 5) and in the syntax, as inverse marking shows: Recall that there are four agreement types in Kiowa: s, d, p, i. iagreement always cooccurs with inverse marking on the noun. This is exemplified below for the sdi noun t´ og´ ul ‘young man’, for the idp noun k´ utaa ‘pencil’ and the idi noun k!ˆ On ‘tomato’. (6)

Tóg´ uu ´dÓ e- dÓÓ young man.INV 3I-be ‘They’re young men’

(7)

K´ utaadO e- dÓÓ pencil.INV 3I-be ‘It’s a pencil’

(8)

K!ÔM OM dO e- dÓÓ tomato.INV 3I-be ‘It’s a tomato’ or ‘They’re tomatoes’

As the examples of the previous chapter show, there is no noun marking concomitant with and proprietary to any of the other agreement types. A natural question, then, is what inverse marking means. However, class mnemonics cast doubt on the coherence of this question, as i-agreement occurs with referential cardinalities 1 (idp), 2 (sii), and 3 (sdi).1 Moreover, if we examine the sets of referential cardinalities for which inverse marking occurs, the result is not enlightening: {1} (idp/ids), {3} (sdi), {1, 3} (idi), and {2, 3} (sii). Nothing seems to unite the classes beyond idi’s being the union of the first two.2 By contrast, a quite neat picture emerges if we ask what non-inverse marked members from each class mean. This corresponds to the natural assumption that the unsuffixed form of the noun is basic. The sets of ref1

Other agreement types do not pose this problem. d-agreement occurs exclusively with 2. s-agreement occurs primarily with 1, the exception being objects that are group-like and hence 1-like. And p-agreement occurs primarily with 3, the exception being objects that are always composed of parts, such as pluralia tantum nouns, which are 3-like. 2 Strictly, one can posit [±F], where [+F] := {1}, and [±G], where [+G] := {2}. The required classes follow: [−F] = {2, 3}, [−G] = {1, 2}, [−F −G] = {3}. However, when we come to consider the relationship between the nouns’ semantic properties and their class mnemonic, it will become apparent that these are not the desired features.

NUMBER FEATURES

67

erential cardinalities that emerge are {1} (sii), {2} (idi), {1, 2} (sdi), and {2, 3} (idp/ids). These classes follow if referential cardinalities are taken to be composed as below:3 (9) 1 2 3

[±singular] + − −

[±augmented] − − +

{1, 2} is the natural class defined by [−augmented]; {2, 3}, the natural class defined by [−singular]; {1}, [+singular]; and {2}, [−singular −augmented]. For this feature composition to be more than a formal nicety, [±singular] and [±augmented] must be properly defined and their connection to inverse marking explained. Answering these questions lays the foundations for explaining the relationship between nouns’ semantic properties and their class mnemonic, e.g., why hair types are idi and animates, sdi and not vice versa. 3.1.2.

Definitions

I assume a general lattice-theoretic semantics, following Link (1983), according to which the denotation of nominal predicate is a set of atoms and the set of subsets of the atoms. Technically, this constitutes a join-complete semi-lattice. It is graphically represented as:

. . .

s

t

u

v

w

x

. . .

The bottom stratum represents the atoms. If the lattice corresponds to the nominal predicate chair(x), then s, t, and so on are individual chairs; if the predicate is person(x), then they are individual persons; and so on. The points of intersection of lines emanating from the atoms represent sets of atoms. So, the point between s and t represents {s, t}, the point directly 3

The combination [+singular +augmented] corresponds to no referential cardinality. See Appendix 3.8.

CHAPTER 3

68

above t, {s, t, u}.4 More generally, the lowest row of points of intersection represents dyads, above that are triads, and so on. The cardinality features [±singular] and [±augmented] apply to the lattice, partitioning it into [−singular] and [+singular], and [−augmented] and [+augmented], regions. As a convention, I will define only [+F].5 Then: (10)

Feature negation [−F] = ¬[+F]

The definitions build on previous morphological work (especially Conklin 1962, Matthews 1972, McKay 1978, 1979, Noyer 1992). However, in their current form, they most closely resemble definitions from Krifka (1992). The surprising finding that morphologists concerned with pronoun and agreement categories and semanticists concerned with aspect and Aktionsart have converged only logical equivalent discoveries is reported in Harbour (2006b). It strikes me as highly significant for an incipient joint theory of morphosemantic number. The feature definitions are: (11)

Definition: [±singular] [+singular] = atom(x)

x, a free variable

Correspondingly, [−singular] = ¬atom(x). So, the feature partitions the lattice into its atomic and non-atomic regions.

4

The lattice is simplified in that not all such sets (e.g., {s, u}) are represented. The simplification aids readability. 5 The convention should not be mistaken for a claim that there is a universal correlation between ‘+’ and markedness. What the marked value of the feature is, is an empirical matter. For instance, based on p-agreement’s being the default agreement in Kiowa, it appears that minus is the unmarked value of [±singular] but that plus is the default value of [±augmented] (Harbour 2003a). Now, one can transform a feature inventory into one with a uniform representation of markedness as plus, by negating the definition and switching the sign of any feature with the marked value minus (Harbour 2003b, p. 135). However, this is possible only if marked values are invariant language internally and crosslinguistically; otherwise ‘the marked value of [±F]’ and ‘[+F]’ are non-equivalent, no matter how we redefine or switch signs, as the former is variable, the latter not.

NUMBER FEATURES

69

[−singular]

. . .

. . .

[+singular]

The definition of [±augmented] is slightly more complex: (12)

Definition: [±augmented] [+augmented] = λP∃y[P(x) ∧ P(y) ∧ y x]

x, a free variable

That is, [+augmented] guarantees that whatever properties hold of x hold also of some proper subset of x. By contrast, [−augmented] guarantees that no such subset exists, that x is a minimal element satisfying P. Note that join-complete semi-lattices lack bottom elements (⊥), the lattice-theoretic equivalent of the emptyset. So, the emptyset is never a possible value for x in [−augmented]. I note this now and omit reference to the condition below. (13)

x = ∅

Therefore, for common nouns, such as chair(x), [±augmented], if acting in isolation, induces the same partition as [±singular].

[+augmented]

. . .

. . .

[−augmented]

However, the features are non-equivalent in at least three contexts. The first, not relevant here, concerns personal pronouns and minimal∼augmented versus singular∼plural systems (such as Ilocano versus English; see Noyer 1992 or Harbour 2006b). The second, also irrelevant here, concerns their use in Krifka’s treatment of Aktionsart. And the third, of major importance here, concerns the interpretation of feature bundles in which both features occur, as already intimated in (9). When the features cooccur, they are interpreted by function application:6 6

Composition in the other order is possible, but the learner would never have semantic evidence for it, as the resulting system is identical to one in which [±augmented] is the only number feature (Harbour 2006a).

CHAPTER 3

70 (14)

Semantic composition [±singular ±augmented] = [±augmented]([±singular])

We can now justify (9) intuitively, as follows. An atom is clearly singular and, as a singleton, lacks proper subsets. So, the singular is [+singular −augmented]. On the other hand, pluralities are clearly non-atomic, and, so, [−singular]; and any plurality like {s, t, u, v, w} has a proper subset, {s, t, u, v} for instance, that is also non-atomic, and, so, is [+augmented]. So, the plural is [−singular +augmented]. The limiting case, though, are dyads. Clearly, {s, t} is non-atomic and, so, [−singular]. However, its proper subsets, {s} and {t}, are not non-atomic. Thus, dyads are [−augmented], and, so, the dual is [−singular −augmented]. For more formal demonstrations, see Appendix 3.8. The relevant lattice is provided below.

. . .

. . .

[−singular +augmented] [−singular −augmented] [+singular −augmented]

Summary The referential cardinalities 1, 2 and 3 have been shown not to be primitive. Rather they are composed, as shown in (9), by the features [±singular] and [±augmented], defined in (11) and (12). Adopting the proposal that Number constitutes a separate projection within the DP (Carstens 1991, Ritter 1991; see also the suggestive discussion in Dryer 2005), we have arrived at: (15)

DP NumberP Noun Number ±singular ±augmented

D

NUMBER FEATURES 3.2.

71

Class

We are now well placed to answer, in part, several of the questions in (1). With respect to the classes sii, sdi, idp, idi, and, to some extent, ids, we can say what the classifying features are. A simple hypothesis about the nature of inverse marking further enables us to say how the inverse arises. Furthermore, we can explain the relationship between the mnemonics of the classes just listed and the semantic properties of the nouns they subsume. 3.2.1.

Classification by cardinality features

Let us begin by taking stock of the noun class mnemonics as a whole. There is a broad correlation between referential cardinality and agreement type: 1 with s-agreement, 2 with d-agreement, 3 with p-agreement. Therefore, we can extend the correlation between features and referential cardinalities in (9) to include also agreement types: (16)

Referential Cardinality 1 2 3

Features [±singular] [±augmented] + − − − − +

Agreement Type s d p

So, sdp nouns, which exhibit no robust, unifying semantic commonalities, trigger agreement as expected on the basis of referential cardinality. Other mnemonics diverge from this correlation in two ways. First, some have i, where s, d, or p is expected. For instance, replacing p by i in sdp gives sdi; replacing s gives idp; replacing both gives idi. Second, some, like ppp, exhibit p where s or d are expected, or, like sds, s where d or p are. We return to unexpected s/p in Section 3.4 and focus now on i. An ideal explanation of the mnemonics sii, sdi, idp and idi would make the occurrence of s, d and p as straightforward as possible and would subsume all occurrences of i under a single principle. This is what we shall now do, by means of the features [±singular] and [±augmented]. As emphasized in the preceding section, the referential cardinalities where agreement is as expected form natural classes with respect to [±singular] and [±augmented]. The table below expresses when s, d and p occur as expected, given (16); the middle column gives the features that must occur under Num-

CHAPTER 3

72

ber and the right column gives the corresponding agreement type(s). (17)

Class sii sdi idp idi

Feature [+singular] [−augmented] [−singular] [−singular −augmented]

Agreement Type s s/d d/p d

For instance, for sii, the feature [+singular] must occur under Number. This happens only with [+singular −augmented], which, according to (16), yields s-agreement for referential cardinality 1. Consequently, s appears in the first mnemonic position. For idp, the feature [−singular] must occur under Number. This happens with [−singular −augmented] / [−singular +augmented], which, according to (16), yields d-agreement for 2 / p-agreement for 3. Consequently, d/p appear in the second/third mnemonic positions. I suggest that we take (17) as representing the classificatory features of each of the classes shown. By so doing, we can state a simple generalization about the expected occurrence of s, d and p and another concerning occurrence of i, for these classes. (18)

Generalization: S, D, P s, d, p occur when Class ⊆ Number.

(19)

Generalization: I i occurs when Class ⊆ / Number.

The first states that s/d/p occur when the features that make up referential cardinality subsume the class features. The second states that, if a class feature is not also a referential cardinality feature, then i occurs. Note that these generalizations apply also to sdp and, partially, to ids. Consider the classificatory features below: (20)

Class sdp ids

Feature ∅ [−singular . . . ]

If Class is empty for sdp, then it is trivially a subset of Number. Consequently, i never occurs. If ids is [−singular], then i will arise for referential cardinality in the first mnemonic position. (Neither principle accounts for s in third position and so Class is at present incomplete; hence, the ‘. . . ’.)

NUMBER FEATURES

73

So far, all generalizations have been stated at the level of the mnemonic: “A mnemonic has i in such-and-such a position if so-and-so holds”. Such generalizations are superficial, as the mnemonics are mere expository devices. We must now ask what mechanisms and structures underlie them.

3.2.2.

ClassP and number on D

I now present a syntactic structure and a computational mechanism that produces s, d, p and i in accord with the generalizations above. The broad view of syntax adopted is that of Chomsky (2000, 2001), which has been applied in detail to Kiowa by Adger and Harbour (2007). Other theoretical assumptions concern the location of class features, to which I now turn. Following ideas of Marantz (1997), Kihm (2002) and Ouhalla (2005) (see also Ritter 1993, Kihm 2005), I propose that classificatory features occupy their own projection, which I label Class. Syntactically, Class serves to make the featureless root nominal and visible to the computational system. Like Number, it is a locus of the features [±singular] and [±augmented]. (21)

DP NumberP Class: Noun (±singular) (±augmented)

D


To address questions of agreement, I adopt the framework of Chomsky (2000, 2001; see Adger and Harbour 2007 for application to Kiowa). Verbal agreement is the result of an Agree (feature matching) relation between heads in the extended verbal projection and the D head of the verbal arguments. As agreement is number sensitive, D must bear number features. However, semantically, D is the locus of definiteness, to which number is irrelevant. So, number on D is misplaced, or, technically, uninterpretable.

CHAPTER 3

74 (22)

DP

NumberP

D


Class: Noun (±singular) (±augmented)


The eventual specification of uninterpretable features depends on those heads for which the features are interpretable. In the current case, the number content of D depends on Class and/or Number. Now, if Number alone valued D, all nouns would be sdp, and if Class alone did, agreement would be invariant across referential cardinality. So, D crucially depends on both. This is possible given that Class and Number are in a mutual dominance relation and, so, equidistant from D. The mechanism of valuation involves replication of the content of Class and of Number on D. (23)

Valuation of D Uninterpretable number on D is valued by a computation over Number and Class. The features of both are replicated on D.

To make this concrete, consider an sdi noun of referential cardinality 1: [−augmented] is located at Class and [+singular −augmented] at Number. (24)

DP

NumberP Class: Noun Number −augmented

D u singular u augmented

+singular −augmented

Valuation is straightforward, as Class is a subset of Number:

NUMBER FEATURES

75

(25)

DP


D



Bearing [+singular −augmented], D will trigger s-agreement (16). Exactly the same mechanisms account for an sdi noun’s d-agreement when referential cardinality is 2. Class is [−augmented], as before, and Number is [−singular −augmented]. Prior to valuation, we have: (26)

DP



−singular −augmented

The content of Class and Number is replicated on D, yielding: (27)

DP

NumberP

Class: Noun Number −augmented

D −singular −augmented


Again, by (16), this is the correct feature content to trigger d-agreement. Matters are more interesting for sdi nouns of referential cardinality 3, where, in contrast to the preceding cases, Class is not a subset of Number:

CHAPTER 3

76 (28)

DP


D


−singular +augmented

Copying yields opposing specifications on D: it is both [+augmented] and [−augmented]. (29)

DP



D ⎤ −singular ⎣−augmented⎦ +augmented ⎡

Although surprising at first glance, I suggest that such feature conflicts are not only possible, but are the precise configuration behind the inverse: (30)

Kiowa inverse [−F +F] ⇔ i/inv

That is, inverse marking on nouns and i-agreement on verbs is the vocabularic reflex of feature conflict (abstracting away from actual phonological content). Now, this requires slight revision of the standard notion of uninterpretability. In general, [uF] is taken to be [±F] without a value, that is, [ F], an underspecified version of [±F]. However, if valuation involves simply ‘filling in the blank’, then it is impossible to value [uF] as [−F +F]. Introduction of an extra [αF] violates Inclusivity. Instead, I propose that uninterpretability is overspecification: [uF] means [−F +F]. Valuation involves pairing these features with interpretable counterparts. Only if paired in this way do the features survive at the interface, to be interpreted and pronounced. Depending on one’s view of the interfaces, uninterpretable features that have not been paired with an interpretable counterpart must either be deleted, or else

NUMBER FEATURES

77

they are not visible and, so, are ignored.7 So, D in (22) bears [−singular +singular −augmented +augmented]. In (26), the interpretable number features are [−singular] and [−augmented]. So, the plus values are invisible/deleted and only the minus ones survive to be pronounced as agreement. In (29), all but [+singular] survive, creating a feature conflict, which, by (30), is pronounced as inverse marking and goes on to trigger i-agreement. Taking sdi ´ O´ Op´ˇZ´ˇZ ‘fish’ as the root noun, the plural DP is pronounced as: (31)

DP NumberP

D

dÓ

Class: fish Number ÓÓp´ ˇZ´ ˇZ

∅

The nature of uninterpretability and valuation is not intended to be a major focus of the current investigation. However, two observations are worthwhile. First, there is independent, semantic, evidence for the reality of conflicting feature specifications. Harbour (2006a), a thoroughgoing crosslinguistic investigation of number categories and number features, proves that the most complex categories, such as unit augmented, minor paucal and trial, result from multiple specification of the feature [±augmented]. For instance, the trial is [−singular −augmented +augmented], subject to (14) and (32). (32)

[±singular −augmented +augmented] = [−augmented]([±singular +augmented])

It is only possible for categories so defined to trigger agreement if feature conflicts are generally possible on agreeing heads. Second, default agreement, which is problematic for the underspecification approach, is trivial for the overspecification approach. Default agreement is whatever agreement type arises when there is nothing to agree with. Given underspecification, failure to establish an Agree relation means that [uF] receives no value. Rather than triggering default agreement, this should cause the derivation to crash. 7

The general view is the former, that everything must be legible at the interface; illegibility causes crashes. However, this seems to me to be an unnecessary assumption: the interfaces ‘ignores’ what it cannot read, in the way that the human eye ‘ignores’ infrared.

CHAPTER 3

78

But, with overspecification, if no Agree relation is established, both [−F] and [+F] delete (or are ignored) and only the host head is pronounceable, yielding featureless, hence default, agreement. So, the DP structure (22), the overspecification approach to uninterpretability and the feature conflict view of the inverse (30) constitute an account of the mechanisms of inverse marking and i-agreement. Furthermore, (25), (26) and (28), show that, if Class is [−singular], then s-agreement results for referential cardinality 1, d-agreement for 2, and i-agreement for 3. That is, they derive the sdi mnemonic from the Class specification [−augmented]. Given (17) and (20), we can derive several other mnemonics, as the coming section shows.

3.2.3.

Derivations I: Mnemonics

This section applies the mechanisms of Section 3.2.2 to the Class features of Section 3.2.1 to derive other class mnemonics: (33)

Class sii sdi idp idi sdp

Feature [+singular] [−augmented] [−singular] [−singular −augmented] ∅

(Details of the derivations for sdp, idp and idi, though not sii, are much the same as for sdi. Readers satisfied with the latter may wish to skip the former. The reader may refer back to the example sentences in Section 2.3.)

The SDP class The sdp class is the simplest case. Only Number bears interpretable number features. Class bears only categorial features (Kihm 2002). Consequently, the content of Number is always replicated on D and agreement reflects referential cardinality. The general case of a fully valued DP is shown below.

NUMBER FEATURES

79

(34)

DP

NumberP

D

Class: Noun Number

αsingular βaugmented


∅

For s-agreement (1), α is plus and β is minus; for d-agreement (2), α and β are both minus; for p-agreement (3), α is minus and β is plus. The IDP class Here, Class is [−singular]. For referential cardinality 1, Number is [+singular −augmented]: (35)

DP

NumberP Class: Noun Number −singular



Class and Number oppose for [±singular], creating feature conflict on D: (36)

DP

NumberP

Class: Noun Number −singular


D ⎤ −singular ⎣+singular ⎦ −augmented ⎡

The result is an inverse-marked noun and i-agreement.

CHAPTER 3

80

For 2, no such incompatibility arises. Number, [−singular −augmented], subsumes Class and no conflict arises on D. The valued DP is: (37)

DP

NumberP


Class: Noun Number −singular

D


The noun will not be inverse marked and will trigger d-agreement. (Note that (36) and (37) also derive agreement and inverse marking for ids nouns; (38), however, does not represent ids nouns of referential cardinality 3.) For 3, Number, [−singular +augmented], again, subsumes Class: (38)

DP

NumberP Class: Noun Number −singular

D −singular +augmented


The noun will not be inverse marked and will trigger p-agreement. This derives the idp class.

The IDI class Here, Class is [−singular −augmented]. D, therefore, avoids conflicting specifications only when Number equals class:

NUMBER FEATURES

81

(39)

DP

NumberP Class: Noun −singular −augmented

D


Number −singular −augmented

Such DPs trigger d-agreement, and so only for 2 do nouns of this class trigger agreement that reflects referential cardinality. For 1 (40), Class and Number oppose for [±singular], and, for 3 (41), they oppose for [±augmented]. (40)

DP


D ⎤ −singular ⎣+singular ⎦ −augmented ⎡

Number


(41)

DP


Number −singular +augmented


Given the conflicts, both nouns will be inverse marked and trigger i-agreement. This derives the idi class. The SII class: Justification by syncretism As mentioned in Section 2.3.9, the sii class is less readily observable than the others, being bound up with person exponence. However, its sole member,

CHAPTER 3

82

the first person, exhibits four distinct sets of syncretisms and it is in the explanation of these syncretisms that positing an sii class yields dividends. Syncretism #1. For agents of (di)transitives and the subjects of unaccusatives, the first person exclusive dual and first person exclusive plural trigger i-agreement. (42)

NÓÓ a- dÓÓ 1 1S-be ‘It’s me’

(43)

NÓÓ e- dÓÓ 1 1I-be ‘It’s me and him or me and her’

(44)

NÓÓ e- dÓÓ 1 1I-be ‘It’s me and them’

Compare (43) and (44) with (45) which has i-agreement in virtue of the inverse-marked t´ og´ uu ´ d´ O ‘young men’.8 (45)


i-agreement for referential cardinalities 2, 3 immediately suggests an sii class. Syncretism #2. For agents of (di)transitives and the subjects of unaccusatives, the first person inclusive dual and first person inclusive plural trigger the same agreement as second person plural. That is, the same pattern as #1, but syncretic with second, not third, person. This agreement is notated 2i, reflecting the assignment of second person to the animate, sdi class. (46)

NÓÓ ba-dÓÓ 1 2I- be ‘It’s me and you.s’

(47)

NÓÓ ba-dÓÓ 1 2I- be

8 d-agreement is always nasal in Kiowa, unless a structurally higher position is nonsingular (see Harbour 2003a, or Chapter 5, for details). Consequently, the lack of nasality in (43) and (46) is genuinely noteworthy and stands in need of explanation.

NUMBER FEATURES

83

‘It’s me and you.d/p’ Compare (46) and (47) with second person plural agreement in (48). Note that (48) does not have a dual reading; the dual would have the prefix ma-. (48)

´ Am ba-dÓÓ 2 2I- be ‘It’s you.p’

Syncretism #3. For indirect and direct objects, the first person inclusive dual, the first person inclusive plural, the first person exclusive dual, and the first person inclusive plural all syncretize. Moreover, these forms are distinct from the inverse and second plural as well as from other persons.9 (49)

DÓb´ oM u ´M 3p:1d/p-see.pf ‘They saw me and you or me and him or me and them’

(50)

BáougO dÓtÓM ÓM cat.inv 3p:1d/p:3i-give.pf ‘They gave cats to me and you or me and him or me and them’

Syncretism #4. First person agreement never distinguishes dual from plural. This can be verified by examining Syncretisms #1−3. Nonetheless, the difference is syntactically represented and is evident in the suppletion they condition in number-sensitive predicates (Chapter 4). (51)

A- x!ó´Zgyá 1s-fall.s/d.pf ‘I fell’

(52)

E- x!ó´Zgyá 1i-fall.s/d.pf ‘She and I fell’

9

More specifically, for given values of object agreement (s, d, p, i), some prefixes in this group are homophonous with others. E.g.: d´ et means both 3s:1d/p:3d (as in ‘He gave us two cats’) and 3s:2s:3d (as in ‘He gave you.s two cats’). However, this is phonological coincidence: second singular and first non-singular indirect object agreement diverge for other values of object agreement: for instance, 3s:1d/p:3i (as in ‘He gave us many cats’) is dÓt, whereas 3s:2s:3i (as in ‘He gave you.s two cats’) is gÓt.

CHAPTER 3

84 (53)

E- k!´ u´Zgyá 1i-fall.p.pf ‘They and I fell’

Given the mechanisms argued for above, we can account for these syncretisms by positing two rules of postsyntactic deletion (impoverishment). Now, it is important to note that positing such rules does not in any sense vitiate the explanatory role that the sii class plays. Assigning Class the value [+singular] in the current theory causes the syntax to ‘set things up’ in such a way that the morphological component need do only the basic types of things that it generally does (e.g., deletion; Bonet 1991, Halle and Marantz 1993) in order to produce the desired syncretisms. That is, the morphological component need not do formal somersaults to produce second plural agreement from first inclusive dual arguments; the syntax all but gives it this, and other, syncretisms. To the extent that the morphological component must do something, similar deletions are required elsewhere in the language (Chapter 5); they are not mere stipulations about the sii class. I assume person features [±author], [±hearer] (Hale 1973, Noyer 1992). These yield first person inclusive and exclusive, and second person thus:10 (54)

Person first inclusive first exclusive second

[±author] + + −

[±hearer] + − +

We begin with the unproblematic first person singular. The structure before valuation of number on D is:

10 [−author −hearer] corresponds to third person. However, as here, one may also represent third person as absence of person (Adger and Harbour 2007, references therein). My own investigations of person systems lead me to posit three features, [±author], [±participant] and privative [hearer]. Two features are adopted here for simplicity.

NUMBER FEATURES

85

(55)

DP

NumberP ⎡Class: I ⎤ +author ⎣−hearer ⎦ +singular

D ⎤ u author ⎢u hearer ⎥ ⎢ ⎥ ⎣u singular ⎦ u augmented ⎡

Number +singular −augmented

For parity with earlier examples, first person is treated as the root noun, with [+author −hearer] located underneath, for readability. Also on Class is the first person’s class feature [+singular]. Number for first person singular is naturally [+singular −augmented]. Observe that D bears uninterpretable person as well as number. Without this, person agreement would not arise; it was irrelevant to, and so omitted from, earlier derivations. D is valued by matching the uninterpretable features with interpretable counterparts on Class and Number: (56)

DP


Number


D ⎤ +author ⎢−hearer ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented ⎡

This straightforwardly triggers first person singular agreement. Next, consider the first person exclusive. Dual and plural are conflated as [−singular ±augmented]. Class is as before and, so, Class and Number oppose for [±singular], resulting in feature conflict on D:

CHAPTER 3

86 (57)

DP


D ⎤ +author ⎢−hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ ±augmented ⎡

Number −singular ±augmented

Finally, consider first person inclusive. Here, [±hearer] is specified as plus, and, because second person, like nearly all animates, is sdi, [−augmented] is also present on Class. 2 and 3 are shown separately, as (58) and (59). (58)

DP

NumberP Class: I, you⎤ ⎡ +author ⎢+hearer ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented

Number


(59)

D ⎤ +author ⎢+hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented ⎡

DP

NumberP

Class: I, you⎤ ⎡ +author ⎢+hearer ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented


D ⎤ +author ⎢+hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented⎦ +augmented ⎡

Given (57)−(59), we can now move on to the syncretisms. Agents of (di)transitives and subjects of unaccusatives (without indirect objects) fall under Syncretisms #1−2; indirect objects and direct objects fall under Syncretism #3. For simplicity, let us label these two contexts

NUMBER FEATURES

87

Nominative and Non-Nominative (see Adger and Harbour 2007 for an analysis of Case and Case-related syncretisms in Kiowa). To derive Syncretisms #1−3, I propose that [±author] and [±hearer] can be deleted postsyntactically dependent on the Case context; that is, the bundle containing these features is impoverished. Nom. (60) [αauthor] → ∅ / ⎡

(61)

⎤ Non-nom. [αhearer] → ∅ / ⎣ +author ⎦

To see how these work, consider the agreement that DPs (57)−(59) trigger. Agreement involves matching the features on D with the agreeing verbal head. Suppose that head is in a Nominative context. Then we have: (62)

Nominative(57) ⎡ ⎤ +author ⎢−hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ ±augmented

Nominative(58) ⎡ ⎤ +author ⎢+hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented

Nominative(59) ⎡ ⎤ +author ⎢+hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented⎦ +augmented

Author deletion (60) applies in this context and yields: (63)

Nominative (57) ⎡ ⎤

−hearer ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ ±augmented


+hearer ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented


+hearer ⎢−singular ⎥ ⎢ ⎥ ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented⎦ +augmented

There are no specific vocabulary items for [−hearer], so Nominative (57) is equivalent to [−singular +singular ±augmented], which is simply a personless inverse [−F +F]. On the other hand, Nominative (58) and Nominative (59) are equivalent to an inverse with the second person feature [+hearer]. Since second person, like most animates, is sdi, this is the representation of second person plural agreement. So, we have derived that, in Nominative contexts, 1ex.d/p syncretize with 3i and 1in.d/p with 2i (Syncretisms #1−2). Finally, consider first person (in)direct objects, both dual and plural.

CHAPTER 3

88 (64)

Non-nominative(57) ⎡ ⎤ +author ⎢−hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ ±augmented

Non-nominative(58) ⎡ ⎤ +author ⎢+hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented

Non-nominative(59) ⎡ ⎤ +author ⎢+hearer ⎥ ⎢ ⎥ ⎢−singular ⎥ ⎢ ⎥ ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented⎦ +augmented

Hearer deletion (61) applies in this context and yields: (65)

Non-nominative ⎡ ⎤(57) Non-nominative ⎡ ⎤(58) Non-nominative ⎡ ⎤(59)

+author ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ ±augmented

+author ⎢−singular ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented

+author ⎢−singular ⎥ ⎢ ⎥ ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented⎦ +augmented

With respect to person, all are identical, but are distinct from second and third. Moreover, by (30), all are identical with respect to number, constituting a [+author] inverse, but are distinct from the singular. So, 1in.d, 1in.p, 1ex.d and 1ex.p all syncretize, but remain distinct from other persons of the same number and other numbers of the same person (Syncretism #3). Syncretism #4 follows as the consequence of the previous three. It should be noted that, in contrast to other DPs of the language, the pronouns n´ O´ O 1 and ´ am 2 are never inverse marked. I assume that this is a specially conditioned form of zero marking. 3.2.4.

Derivations II: More inverse marking

We have now seen how the mnemonics sdp, sdi, sii, idp, and idi derive from the interaction between feature conflict exponence (30) and the Class specifications in (33). How this derives an inverse-marked noun from an underlying DP was illustrated for sdi ‘fish’ (31); the same processes apply, mutatis mutandis and modulo sii, to DPs of other classes. We now turn to derivations of more complex inverse-marked objects: relative clauses, third person possessed kin terms, and DPs modified by deictics, adjectives, and ‘only’. Essentially, this section concerns the location of uninterpretable number throughout the clause. To some extent, it serves to justify the assumption that D is one such locus. Readers more concerned with the class system and its underlying feature structure may skip this section.

NUMBER FEATURES

89

Let us begin with adjectival modification. Matters differ little from the DPs above. Consider: (66)

k!yáM áM h´MZ-k!oM uM -bO man- old- inv ‘old men’

(Watkins 1984, p. 99)

Assuming uncontroversially that adjectives are below D, and adjoined, we have the DP below, once number on D has been valued: (67)

DP

NumberP NumberP Class: man

−augmented


Adjective

Number


After vocabulary insertion we have: (68)

DP NumberP

NumberP

D

bO

Adjective

k!´ om

Class: man Number

k!y´ a ˇ´ a ˇ hˆˇZˇi

∅

Standard phonology yields k!y´ a ˇ´ a ˇ h´ˇZk!o ˇu ˇ bO. Note that adjectival modification serves to emphasize that the inverse marker and, hence uninterpretable number, is D-level, rather than lower in the structure. Conversely, we can place an upper bound on the height of inverse marking by considering other forms of modification. Let us begin with demonstratives. Kiowa has two deictic roots, proximal,

CHAPTER 3

90

´ eˇ´ˇZ, and distal, ´ o´Z. They are found, for instance, in the locatives, ´ eˇ´ˇZhOO ‘here’ and ´ o´ZhyOO ‘there’. Demonstratives comprise a deictic root and a numbersensitive suffix, -gO if modifying an inverse-marked noun, -de otherwise. (69)

éM´MZde/ó´Zde thal´Z´Z this /that boy ‘this/that boy’ or ‘these/those (two) boys’

(70)

éM´MZgO /ó´ZgO thalyóp this.INV/that.INV boy.INV ‘these/those (several) boys’

For thal´Z´Z ‘boy’, an sdi noun, the inverse-marked demonstratives occur when referential cardinality is 3. For an idp noun, say, they occur for 1. (71)

éM´MZgO /ó´ZgO á a´dO this.INV/that.INV stick.INV ‘this/that stick’

(Demonstratives occur before the noun, as above, or after it, discontiguous from it, or without an overt noun.) To derive (70), we can posit the following structure. It is simply (31) embedded under a demonstrative, where the demonstrative, like D, has uninterpretable number that must be valued. (72)

DemonstrativeP

DP

Demonstrative u singular u augmented

NumberP

Class: boy

−augmented

Number



As second cycle of probing and matching values the higher number:

NUMBER FEATURES

91

(73)

DemonstrativeP

DP

NumberP

Class: boy

−augmented

Demonstrative ⎡ ⎤ −singular ⎣−augmented⎦ +augmented


Number


By (89), p. 55, we have (74): (74)

DemonstrativeP

Demonstrative

DP

gO

NumberP

D

op

Class: boy Number

thalii

∅

Inserting ´ o´Z, the distal deictic, we have thaly´ op ´ o´ZgO. Variant word orders are derived by syntactic means irrelevant here. The non-inverse-marked form, DP+´ o´Zde, follows if -de is a default realization of number on Demonstrative. The same suffixes, basic -de and inverse -gO, are found with relative clauses, depending on the class and number properties of the head noun: (75)

(76)

[ám gyát´ oM u ´M - khÔM OM mO] -de anph 1s:2s:3s-talk-name.impf-BAS ‘who I was speaking to you about’

(Watkins 1984, p. 231)

[p´MZá a´dO e- ét] -gO table.inv 3I-big.s-INV ‘a big table’

(Watkins 1984, p. 231)

The simplest way to derive common number marking between relative clauses

CHAPTER 3

92

and deictically modified DPs is to suppose that the two are structurally identical at the relevant level or that -de/gO are default exponents. Both are plausible. However, it was suggested, in Section 2.6.2, that -de/gO decompose into a D-related element, g, with allophone d, and pure number e/O, for bas/inv. If they are morphologically complex, with number conditioned by D, then -de/gO are not defaults. So, syntactic similarity is a more likely explanation of the common exponence. Specific possibilities are that relatives are D-embedded CPs (Vergnaud 1974, Kayne 1994) or that demonstratives are highly reduced relatives; however, I do not pursue the matter here. Basic -de and inverse -gO feature in yet other constructions. They form part of Noun+‘only’, though, here, the inverse alternant is optional:11 (77)

(78)

a.

KÓ´Zg´ ude /gO- ki Kiowa.inv-bas/inv-only ‘only Kiowas’

b.

a´ádO- de /gO- ki tree.inv-bas/inv-only ‘only a tree’

a.

KÓ´Z- mathOn-de /*gO- ki Kiowa-girlbas/ inv-only ‘only a Kiowa girl’

b.

á a´- de /*gO- ki tree-bas/ inv-only ‘only trees’

They also form part of third person kin possession (sometimes with -de devoiced; see Watkins 1984 for possession in general): (79)

aíi- te 3poss-son-BAS ‘his/her son or two sons’

(80)

aýyoi- gO 3poss-son.inv-INV ‘his/her sons’

11

After numerals, -kOO is used, as in y´Z´Z-kOO ‘only two’ (*y´Z´Z-de-ki). After personal pronouns, irrespective of number, only -de-ki is permitted, consistent with the impossibility of inverse marking on these elements.

NUMBER FEATURES

93

In (80), both the possessed noun and the possessor suffix are inverse marked. Further modification with ‘only’ leads, optionally, to triple inverse marking: (81)

[áyyoigO] -gO- ki 3poss-son.INV-INV-INV-only ‘only his/her sons’

Such structures warrant two observations. First, the feature clash generated by the base noun spreads through the syntax in a manner reminiscent of argument-verb agreement and can clearly be handled by positing further instances of uninterpretable number. Second, if Kiowa is the head-final version of English phrases such as only the X, then ‘only’ places an upper bound on the locus on nominal inverse marking, for which D (English ‘the’) is a clear possibility. (Possessive structures, only partially described above, are too complicated for treatment here.) This complements the lower bound for inverse marking, again D, motivated by adjectival constructions. 3.2.5.

Summary

The preceding subsections have motivated the following distribution of DP: (82)

DP

NumberP

D


Class: Noun (±singular) (±augmented)


The features at Class are determined by the root noun: (83)

Class sii sdi idp ids idi sdp

Feature [+singular] [−augmented] [−singular] [−singular ...] [−singular −augmented] ∅

CHAPTER 3

94

D in (82) bears uninterpretable number features, which must be valued and then go on to trigger verbal agreement (as well as number marking on demonstratives, relative clauses, possessives, and ‘only’). Uninterpretable features involve specification of both values; however, only those matched with same value on Class or Number are legible at the interface. The result is replication of Class and Number on D. In some cases, this leads to feature conflict, [−F +F], on D. Exponence of such conflicts leads to inverse marking and i-agreement. This system generates several of the mnemonics of the previous chapter: sdp, sii, sdi, idp and idi. The economy of this account is worth emphasizing. It relies on a variety of theoretical assumptions concerning DP structure, feature organization and agreement. However, beyond slight modification of uninterpretability, these assumptions are neither novel nor specific to this analysis. The most arbitrary element so far is the claim that noun N occurs with Class [αF], whereas N occurs with [α F ]. However, the relationship between a noun and its classificatory features is, in fact, far from arbitrary, as we will now see. 3.3.

Mnemonic naturalness

With the underlying classificatory features of five (and a half) classes uncovered, we can consider the naturalness of their mnemonics given the semantic properties of the nouns they subsume. The sdp class represents the minimal case. Absence of inherent number features explains why no positive properties unite members of this class, an assortment of implements, footwear, and natural items (2.4.5): lack of positive properties is the semantic correlate of the lack of classificatory number features. The sii mnemonic is entirely natural: it claims that first person, the sole sii member, is inherently [+singular]. This is not to claim first person dual or plural ineffable. However, these are not multiple instances of the author, in the way that chairs denotes multiple instances of chairs. Rather, they denote groupings of people of which the author is one element, and always a unique, hence conceptually [+singular], one. The sdi mnemonic is also appropriate for animates and other nouns that determine or influence the course of their own motion. Elements that independently determine their course of motion do not readily form homogeneous groups. The feature [+augmented], however, ensures that properties of the whole, P(x), are properties, P(y), of the subpart, y x. By contrast, then,

NUMBER FEATURES

95

[−augmented] ensures non-homogeneity. So, there is a natural nexus between sdi nouns and the Class specification [−augmented]. In this light, k´ Ol, the other exception to the implication from animacy to sdi membership also appears natural. In addition to ‘cattle’ and ‘buffalo’, it means ‘herd(s)’. Herds are crucially collective and so describable as [+augmented], for properties of herds are properties of subparts of herds. For nouns that are ambiguously herd-denoting or individual-denoting, the classification [−augmented], and with it sdi class membership, is inappropriate. The idp mnemonic results from [−singular] at Class. Now, [+singular] defines atomicity, which is conceptually close to individuality. So, inherently [−singular] nouns are those that are not salient, or recognizable, as individuals. This is obvious for vegetation, and for other members of the class, such as implements. The same reasoning applies to i and d of the ids mnemonic. The idi mnemonic is at first sight mysterious. Given that [−singular −augmented] is the featural correlate of referential cardinality 2, it seems that these nouns are conceptualized as inherently dual. However, this makes scant sense. First, of all idi nouns, only eyebrows come in pairs—apples, hair, blackberries, brains, eyelashes, plums and tomatoes do not.12 Second, there are good many things that are inherently paired—the thematic nouns in -de (Section 2.6.2) for instance—but these are not idi nouns. Third, it is doubtful that ‘two hairs’, ‘two apples’, et cetera are frequently enough uttered and heard for idi membership to be acquirable in this way. The naturalness of the mnemonic becomes apparent, however, if each class feature is considered separately. Recall, from 2.4.3, Chierchia’s observation that hair is classified as count in some languages and as mass in others (this applies also to ‘brain(s)’, which, in English, varies between count and mass). This crosslinguistic classificatory equivocation can be expressed in terms of the number features motivated above. The individuability of count 12

One might think that the inherent pairhood of eyebrows or, possibly even, of brains, owing to their hemispheres, reflects true inherent duality. However, eyebrows are idi nouns by conceptual parity with other hair types. And the irrelevance of hemispheres to the classification of brains can be seen in (i), where ‘brain’ triggers i-agreement. The point is that only one brain is being talked of; if k!yag´ op really meant ‘pair of brain hemispheres’ rather than simply ‘brain’, we would see d-agreement. (i)

K!yag´ op há áxo Ó há´ı-dOO! brain.inv how :3s:3i-q- be ‘What a brain he has!’ (Idiomatic expression of intelligence)

96

CHAPTER 3

nouns implies [−augmented]; however, salience of the collection, and hence deemphasis the parts, implies [−singular]. Alternatively, thinking of apples, et cetera, like all vegetation, they are [−singular] because inherently nonindividuable. However, given the clusters in which they grow, the constituent fruit are clear and evident, like the strands in a head of hair, and, so, the collection is non-homogeneous [−augmented]. Consequently, [−singular], on its own, indicates a property of idi nouns, and [−augmented], on its own, does so too. Of course, copied onto D, they trigger d-agreement; but this apparent inherent duality is accidental. So, ‘two hairs’, ‘two apples’, et cetera, of implausible frequency, are not requisite data for acquisition of this class. This implies that the interpretation of [−singular −augmented] is subtly different for Class and for Number. For Class, each feature applies to the root noun directly and, so, separately. For Number, [−augmented] applies to [−singular] and [−augmented]([−singular]) then applies to the root noun (14) (see also Lemmas 1−3, p. 113 f.). This is not surprising, however. First, it is clear that the interpretation of number features in the two cases is somewhat different: Number has a direct, truth-conditional effect on interpretation, but Class affects interpretation more marginally (as when one root has two interpretations class-dependently, e.g., ´ a´ a is ids ‘tree’ but idp ‘stick, pole’, k´ Ol is i ‘heads of cattle’ but p ‘herds of cattle’). Now, if the effect of Class is to determine the precise meaning of the root, then, naturally, all Class features apply directly to the root, rather than to each other. On Number, however, locality forces interpretation of the bundle [−singular −augmented] before the features can apply to the neighboring head, {Class, Noun}. The preceding chapter argued that the classes are internally semantically coherent, in that the nouns each class subsumes share properties. We now see that these shared properties are naturally expressible in terms of number features. Moreover, these number features are responsible, in combination with the mechanisms for valuing and pronouncing number on D, for the agreement types s, d, p and i. Consequently, the system of mnemonics accords with the semantic properties of the classes. As we discuss the remaining classes and their underlying features and values, the semantic naturalness of those other mnemonics will also become clear. This concludes the first part of the chapter. The discussion now moves away from the features [±singular] and [±augmented] and the classes SDP, SDI, SII, IDP, and IDI to the feature [±group] and the classes SDS, IDS, SSS, and PPP.

NUMBER FEATURES 3.4.

Spurious

97

S/P

The first part of the chapter (3.1−3.3) established and explained the broad correlation between s-agreement and referential cardinality 1, d-agreement and 2, and p-agreement and 3. It further explained how i-agreement interferes with this correlation. The classes under consideration in this section, and in the second part of the chapter generally, also interfere with the correlation. However, they do so by virtue of what might be termed spurious s and spurious p, occurrences of s-agreement and p-agreement that do not correlate with their typical referential cardinality values: sds, ids and idi in its ‘different types of’ reading show s-agreement when referential cardinality is 3, and ppp nouns show p-agreement for referential cardinalities 1 and 2. Below, I argue that spurious s and p should be attributed to a feature [±group]. As before, the feature is given a clear semantics, associated with the relationship between wholes and parts. The feature has an effect on the valuation of D that reflects its interpretation.

3.4.1.

Types of grouphood

Nouns that trigger spurious s-agreement form pluralities in which the whole is more salient than the part. For sds nouns, this results because they are non-shape-inductive. Faced with a single cloud, or a small number of them, it is generally clear where the boundaries are. However, in a large group, boundaries of the single clouds are unclear and one is likely to view the whole as a single mass itself. For ids, similar reasoning holds. A single mountain or a single tree may be well defined. However, in a mountain range or in a grove or spinney, the boundaries of individuals are indistinct and the range or grove is viewed as a whole itself. For idi, Watkins describes the reading they attain under s-agreement as one of ‘plural sets’: ‘three or more separate collections of a single type, e.g., varieties of apples in separate piles or bags’, or ‘more than two sets of hair, i.e., heads of hair belonging to different individuals’ (Watkins 1984, 88−89). The key notion of ‘collection’ or ‘set’ again exhibits the property of not exhibiting clear or inductive boundaries. So: (84)

Generalization: Spurious S nouns form pluralities in which the whole is more salient than the part.

CHAPTER 3

98

Pluralia tantum nouns exhibit a property almost the opposite of (84). Even when singular, trousers, war headdresses, necklaces, and so on are clearly composed of parts. That is: (85)

Generalization: Spurious non-plural.

P

nouns have salient parts even when

Given the complementarity of (84) and (85), we should define a single binary feature for which the whole∼part relations correspond to the plus and minus values. The solution that I will shortly offer is that the feature predicates of [±augmented]. This solution arises naturally from the failure of some more obvious possibilities, from which we begin. At first, one might think to define simply a predicate Parts(x): (86)

Parts(x) is true if, and only if, x has salient subparts.

However, this feature is too coarse. We are concerned not with properties of x per se but with properties of the pluralities that x forms. Dependence on plurality might be captured by including an implication in the definition: (87)

[+group] if [+augmented] → Parts(x) [−group] if [−augmented] → ¬Parts(x)

However, (87) violates the definition of feature negation, [−F] = ¬[+F], and cannot be fixed by choosing some truth functor other than ‘→’ (Appendix 3.8). The technical problem can be avoided by use of α-notation: (88)

[αgroup] = [αaugmented] → αParts(x)

However, this notation merely disguises that one cannot define one value and deduce the other, and, so, is of a richer type than we have needed until now. Instead, I propose that [±group] be a ‘partial’ feature. That is, it has a definition in the standard style but predicates of [+augmented]. (89)

[+group] = λP[+augmented] [P(x) ∧ ¬Parts(x)]

Now, it is possible to dispose of the novel notion Parts(x), in favor of more conventional predicates, as follows. Observe first, that Parts(x) is almost the converse of atomhood, as atoms lack (salient) parts and things with salient parts are non-atomic. So, we can think [+group], as defined in (89), as introducing a new predicate, Q, and asserting that x is atomic with respect

NUMBER FEATURES

99

to Q. Incorporating Q-atomicity into the definition yields: (90)

Definition: [±group] [+group] = λP[+augmented] ∃Q[P(x) ∧ P = Q ∧ Q-atom(x)] x, a free variable

Importantly, the notion of Q-atomicity is adopted from Krifka (1992) and so is not a novel assumption of the current theory. Rather, it is a legitimate resource to call on in a morphosemantic theory of number. (91)

Definition: Q-atom Q-atom(x) = Q(x) ∧ ¬∃y[Q(y) ∧ y x]

According to (90), [+augmented] x is [+group] if it lacks salient (distinguishable) subparts and so is atomic with respect to some other predicate (namely, the collective); and it is [−group] if it has salient subparts, and so is nonatomic with respect to other predicates. Note, crucially, the generality of the phrase ‘other predicates’, which arises by negating (−) the existential of [+group]. This captures the fact that the composite parts of pluralia tantum nouns are non-uniform, that is, there is no single predicate with respect to which they are atoms. If [+augmented] is absent, then [αgroup] does not predicate of anything. Now, this might be thought to cause a presupposition failure. However, as discussed on p. 96, class features do not affect the truth conditions directly, but constrain the interpretation of the root noun. So, if [+augmented] is absent, [±group] does not constrain the noun at all, and so is inert, both for interpretation and syntax, that is, both conceptually and computationally. Note an immediate consequence of restricting [±group] according to (90). Watkins observed that idi nouns, when they trigger s-agreement, refer to three or more groupings (p. 97 above). The noun has this [+group] reading when Number is [+augmented]. The reading does not arise for other values of Number: ´ alOO cannot mean ‘two groups of apples’ when Number is [−singular −augmented]. This follows from the restriction of [+group] to [+augmented]: without it, [+group] is inert and the ‘collections’ reading unavailable. We can now state the classificatory features of several more classes.

CHAPTER 3

100 (92)

Class sds ids idi ppp

Feature [+group] [−singular +group] [−singular −augmented (+group)] [+augmented −group]

For idi, [+group] is parenthetic because the group reading is optional. Observe that ppp’s classificatory features must include [+augmented]. Otherwise, for referential cardinalities 1 and 2, there would be no [+augmented] feature and [−group] would be inert, resulting in sdp. Its inclusion is conceptually natural, given, on the one hand, the composite nature (hence, plurality) of pluralia tantum and granular mass nouns, and, on the other, that [+augmented] is the crucial defining feature of plurality. 3.4.2.

Derivations III: Grouphood

Having argued that there is a third feature on Class, we must ask whether there is a third uninterpretable feature on D. If so, then the valued DP of an sds noun of referential cardinality 3 would be: (93)

DP

NumberP Class: Noun Number +group


D ⎡ ⎤ −singular ⎣+augmented⎦ +group

If uninterpretable number is uniform throughout the language, then [+group] would propagate from D onto verbs, demonstratives, relative clause, and the other elements of Section 3.2.4. However, this is problematic, as [±group] is never itself pronounced. This does not undermine the claim that it exists; its semantic motivation still stands. Instead, it suggests that [±group] is never present in the feature bundles comprise the verbal prefix and so on. Rather, the effect of [±group] is on the valuation of D, triggering s-agreement for referential cardinality 3 and p for 1/2. Naively, then:

NUMBER FEATURES (94)

101

Computational effect of [±group] Suppose [±group] occurs (with [+augmented]) on Class/Number. Then D is valued as [±singular ∓augmented].

Essentially, then, [±group] acts like a decoy for valuing [usingular] and [uaugmented]: [usingular] acquires the value of [±group], [uaugmented] the opposite. A summary of the divergence [±group] forces between Number and D (in those cases where its effect is non-trivial) makes this apparent: (95)

[±group] [−group] [−group] [+group]

Number

+singular −augmented −singular −augmented −singular +augmented

D −singular +augmented −singular +augmented +singular −augmented

Now, permitting non-matching features to match values is, of course, a highly questionable complication of the feature matching mechanism. However, so radical a revision is not necessary if we consider again the semantics [±group]. Conceptually, groups are atomic pluralities: that is, they are plural with respect to one predicate, such as tree(x), but atomic with respect to the corresponding collective, such as grove(x). Conversely, collections of diverse objects are atomic with respect to a pluralia tantum predicate, such as trousers(x), but there is no natural corresponding non-collective, as the parts that comprise trousers are non-uniform. The semantic effect of [±group], then, is to shift interpretation of the root noun between collective and non-collective. In so doing, it recalibrates the relevant notion of atomhood, with pluralities becoming atomic and non-pluralities becoming non-atomic. This is notion of Q-atomicity in the definition (90). So seen, the computational effect of [±group] is more natural. It shifts the root noun to or from a collective, inducing a new set of atoms, and it is with respect to the new atoms that the feature concerned with atomicity, [usingular], on D, is valued. It is valued as [+singular] if [+group] induces a set of atomic plurals, and as [−singular] if [−group] induces a set of nonatomic plurals. (The value of [±augmented] could be incorporated into the syntactic computation. However, I believe that it is a default in these cases,

CHAPTER 3

102

arising via the mechanisms discussed by Noyer 1998, Harbour 2003a, a digression profitably avoided here. See Section 4.5.1 for crucial evidence that the representation of [±augmented] in these cases is indeed different from others.)

A note on mnemonic naturalness The semantics of the feature [±group] has been motivated directly on the basis of the semantic properties of the nouns that it classifies. Consequently, nothing more need be said about the naturalness of the role that [+group] and [−group] play in classification. Similarly, its effect on valuation of D has already been discussed. Consequently, nothing more need be said about the mnemonic naturalness of the classes in (92).

Spurious S-agreement We now consider derivations for sds and ids nouns. We begin with referential cardinalities 1 and 2 for the sds class. Class is [+group] and Number is [±singular −augmented]. (96)

DP


D


±singular −augmented

As (96) contains no (interpretable) instance of [+augmented], [±group] is conceptually, and, so, computationally inert: it does not affect interpretation of the root noun or the valuation of D. So, D simply replicates Number:

NUMBER FEATURES

103

(97)

DP

NumberP


Class: Noun Number +group

D


This is all but identical to an sdp DP for referential cardinalities 1/2. s/dagreement result, as desired. For ids nouns of referential cardinality 1/2, Class is [−singular +group]: (98)

DP

NumberP Class: Noun Number −singular +group



Again, without interpretable [+augmented], [+group] is conceptually and computationally inert. So, (98) is in practical terms identical to an idp DP of referential cardinality 1/2. So, i/d-agreement result, as desired. Now consider an sds noun of referential cardinality 3: (99)

DP

NumberP


D



Here, [+group] is rendered active by [+augmented] on Number. So, by (94), D is valued as [+singular −augmented], triggering s-agreement, as desired.

CHAPTER 3

104 (100)

DP

NumberP



D


The derivation is identical for an ids noun of referential cardinality 3. The extra feature, [−singular], on Class, has no effect on the computation: (101)

DP

NumberP

Class: Noun Number −singular +group

D +singular −augmented


Again, the DP triggers s-agreement.

Spurious P-agreement Now consider a pluralia tantum ppp noun, for any referential cardinality. Class is [+augmented −group]; Number is expressed in the general form [±singular ±augmented]. (102)

DP

NumberP Class: Noun +augmented −group


D


NUMBER FEATURES

105

Irrespective of the value of [±augmented] on Number, the [+augmented] of Class renders [−group] active. Therefore, D will be valued as [−singular +augmented], triggering p-agreement, for all referential cardinalities. (103)

DP

NumberP Class: Noun +augmented −group


D


3.4.3. Summary We can now expand the list of noun classes and classificatory features. (104)

Class sii sdi idp idi ids sds ppp sdp

Feature [+singular] [−augmented] [−singular] [−singular −augmented (+group)] [−singular +group] [+group] [+augmented −group] ∅

The feature [±group] pertains to salience of objects’ subparts. It predicates of [+augmented] and is conceptually and computationally active only in the presence of that feature. That is, if [+augmented] is found either on Class or Number, [±group] affects the computation. Otherwise, it is inert. The computational effect of [±group] is to value D [±singular ∓augmented]. It has this effect in virtue of introducing a new predicate, the (non-)collective counterpart of the root noun, and recalibrating the notion of atomhood in accord with it. Uninterpretable number on D is valued in accord with the recalibrated notion of atomhood; hence [±group] and [±singular] share signs.

CHAPTER 3

106 3.5.

Mass nouns

We now have a firm enough grasp of [±singular], [±augmented] and [±group] to move from the semantic properties of mass nouns to their class features. For both granular and non-granular mass nouns, divisibility (any subpart of water is water) entails both non-atomicity and augmentation. So, they are either negatively, or zero, specified for [±singular] and are [+augmented]. Further, granularity means salience of subparts, and so corresponds to [−group]; conversely, non-granularity corresponds to [+group]. We can, therefore, ascribe to mass nouns the following classification: (105)

Class ppp sss

Feature [−singular +augmented] [−singular +augmented +group]

The classification makes two non-crucial assumptions. The first is that mass nouns do not project NumberP. This is justifiable in terms of the well noted incompatibility between mass nouns and count expressions. It ensures constant agreement for both granular and non-granular nouns. If we abandoned the assumption, the combination [+augmented +group] would ensure constant s-agreement for non-granular mass nouns; addition of [−group] to Class would ensure constant p-agreement for granular mass nouns. The second assumption is that mass nouns have negative, rather than zero, specification for [−singular]. If zero specified, then [−singular], necessary for p-agreement, can be assumed to arise postsyntactically, by insertion of unmarked feature values (Noyer 1998, Harbour 2003a). It is included in (105) for purely expository reasons, so that valued D resembles that of previous trees. Fully valued sss/ppp DPs are shown, in accord with both assumptions, in (106)−(107): (106)

DP ⎡Class: Noun⎤ −singular ⎣+augmented⎦ +group

D +singular −augmented

NUMBER FEATURES (107)

107 DP Class: Noun −singular +augmented

3.5.1.


Conjunction

The foregoing is enough to justify constant s/p-agreement for mass nouns. However, the sss/ppp mnemonics were justified mainly by their agreement behavior under conjunction. A brief examination of conjunctions of non-mass nouns establishes several generalizations from which the agreement behavior of mass noun conjunctions follows directly.13 Conjunctions can have greater referential cardinality than any of their individual conjuncts. For instance, 1 + 1 = 2 (but 1 + 3 = 3). Recall from Chapter 2, examples (27)−(30), that, for sdi nouns, increasing in referential cardinality through conjunction has the same effect on agreement as increase in referential cardinality by numeral modification. That is, ‘this young man and that’ triggers d-agreement, ‘this young man and those two’, i-agreement. By contrast, conjunctions of sdp nouns trigger agreement purely according to referential cardinality of the conjunction. Therefore, the contents of Class and Number of each conjunct contribute to the agreement triggered by the 13 Agreement triggered by conjunction is a topic of my ongoing research. It is relevant to the number theory for several reasons. First, observe that languages have mechanisms for determining the referential cardinality of a conjunction from the referential cardinalities of its individual conjuncts. Clearly, some form of computation is involved and it is the task of a theory of number to account for such mechanisms. Second, intra- and inter-class conjunction provide an opportunity to test the content proposed for Class and Number here; the agreement type triggered by a conjunction

C=

n

∧ Ci

i=1

for some integer n > 1, should follow from simple generalizations concerning the featural content of Class and Number for each conjunct Ci . This is no small task, as the total number of intra- and inter-class conjunctions, even excluding the sii class, is (8 × 3)2 +(8 × 3)3 = (8 × 3)2 × 25 = 122 × 100 = 14, 400. The reason that this is a topic of ongoing research is not that I intend to sift through each one of these cases. Rather, it is that each modification to one’s view of the feature content of Class, Number and D in turn changes what are the crucial test-case conjunctions, requiring further trips to the field.

CHAPTER 3

108

conjunction as a whole. I implement this by proposing that the functional projection that hosts gO ‘and’ cooccurs with Number and D projections. (108)

DP

NumberP

D


⎡Class: andP ⎤ (±singular) ⎣(±augmented)⎦ (±group)


Number’s features are determined by a computation over the c-commanded Number heads, i.e., those of the conjuncts under andP. Likewise, those of Class are determined by a computation over the c-commanded Class heads. The computation of Number is simply ‘addition’ (recall that 3 is the equivalence class of all cardinalities greater than two). Based on the behavior of conjunctions of sdi nouns, let us tentatively assume that, when all conjuncts are classmate, Class in (108) is simply copied from the lower Class nodes.14 Therefore, in a conjunction of sss or ppp nouns, Class in (108) will be as for the conjuncts. Number in (108) is slightly trickier. One possibility, is that, in the absence of lower Number heads, Number is valued by a computation over Class heads. Alternatively, absence of Number on the conjuncts might entail absence of Number on the conjunction. Assuming that some such mechanism guarantees either the absence of Number or its identity with Class, we can disregard Number for conjunctions of mass nouns. In 14 In fact, matters are slightly more complex, as some intra-class conjunctions permit different agreement options. For instance, both s-agreement and p-agreement are acceptable for conjunctions of ids nouns of referential cardinality 3.

(i)

P!ÓÓ gO á ´ a gya /gyat- g´ uttO river and tree 1s:3s/1s:3p-draw.impf ‘I’m drawing rivers and trees’

A second type of agreement variation is that, in Mr Bointy’s dialect, ‘this young man and those two’ triggers either d- or i-agreement. This is not a right-conjunction effect, as, more generally, d-agreement and i-agreement are both possible for any conjunction of sdi nouns, where each conjunct has referential cardinality 1 or 2. Other speakers do not have this variation. On the d∼i alternation in Kiowa generally, see Section 4.5.2.

NUMBER FEATURES

109

consequence, such conjunctions agree just as the unconjoined nouns do. This concludes the second part of the chapter.

3.6.

Conclusion

Noun class systems are often seen as bastions of the arbitrary. However, Kiowa’s noun class system is highly principled. We saw in the last chapter that the nouns subsumed under each class mnemonic are semantically coherent. In this chapter, the feature system that underlies the classification has been examined. The results are that [±singular] and [±augmented], the features that compose referential cardinality, are also used for noun classification along with a third feature, [±group], which qualifies type of augmentation. (109)

Class sii sdi idp idi ids sds ppp sdp sss ppp

Feature [+singular] [−augmented] [−singular] [−singular −augmented (+group)] [−singular +group] [+group] [+augmented −group] ∅ [−singular +augmented +group] [−singular +augmented]

There is a principled relationship between a noun’s semantic properties and its classificatory feature. Furthermore, class features interact with referential cardinality in a consistent and straightforward fashion, valuing D and so generating the class mnemonics. Consequently, the relationship between class semantics and class mnemonics is non-arbitrary. The theory above employs well justified and/or standard assumptions concerning the structure of DPs and agreement. The main points are reiterated below. Number features occur on Class, Number and D:

CHAPTER 3

110 (110)

DP

NumberP

D


⎡Class: Noun ⎤ (±singular) ⎣(±augmented)⎦ (±group)


Uninterpretability is taken, novelly, to involve overspecification: [uF] means [−F +F]. Valuation involves matching each uninterpretable [αF] with an interpretable correspondent. Only matched [αF] are visible at the interface, for pronunciation and interpretation. The loci of interpretable number features are Class and Number, which are equidistant from D and so equally accessible to it. The effect of valuation, then, is to replicate the feature content of both heads on D. This view of valuation permits conflicting feature specifications to arise on D. Such conflicts are pronounced on D and related heads as inverse marking, and on verbal heads as i-agreement. Valuation of D is also affected by [±group]. The feature predicates of [+augmented]. When no [+augmented] is present, it modifies an aspect of the root noun that is irrelevant to the DP and so is conceptually and computationally inert. When active, it introduces a new predicate and recalibrates the notion of atomhood. It is with respect to this new set of atoms that uninterpretable number on D is valued. So, in essence, [±group] acts like a surrogate for the valuing of [±singular] and, from that, of [±augmented]. 3.7.

Appendix: Missing mnemonics

Chapters 2−3 have concentrated on the classes that Kiowa has and not on the classes it hasn’t. Yet, there are 64 mnemonically possible classes (111) and it is natural to ask why only nine are attested (sdp, sdi, idp, ids, idi, sds, ppp, sss, sii). Where are the missing 55, such as ddd, pip, pds? ⎧ ⎫⎧ ⎫⎧ ⎫ s ⎪ s ⎪⎪ s ⎪⎪ ⎪ ⎪ ⎬ ⎬⎪ ⎨ ⎪ ⎬⎪ ⎨ ⎪ ⎨ ⎪ d d d (111) p ⎪ p ⎪⎪ p ⎪⎪ ⎪ ⎪ ⎭ ⎭⎪ ⎩ ⎪ ⎭⎪ ⎩ ⎪ ⎩ ⎪ i i i

NUMBER FEATURES

111

In fact, most of these classes are excluded on principled grounds and we find that Kiowa makes near optimal use of its classificatory resources. (See Harbour 2006c for further discussion and inter-Tanoan comparison.) An attempt to explain the scarcity of attested mnemonics might concentrate on the inventory of mnemonics itself: Consider d-agreement. This arises only when D is valued as [−singular −augmented]. This in turn requires that Number be specified as [−singular −augmented] (and Class as some subset thereof, [(−singular) (−augmented)]). The requirement that Number be specified in this way restricts d-agreement to the referential cardinality corresponding to the feature combination [−singular −augmented], i.e., 2. Therefore, d is confined to the middle position of the mnemonic. The number of possible mnemonics reduces from 64 to 3 × 4 × 3 = 36. If the third position of the mnemonic is s, then the class is [+group]. If the first position is p, it is [−group]. Taking [−group +group] to be semantically impossible, we exclude p. . . s and reduce the number of possible mnemonics to 36 − 4 = 32. If the second position of the mnemonic is i, then the class is either [+singular] or [+augmented]. If [+augmented], then the first and second positions are identical, unless the class is specified as [−group]. But [−group] would block i-agreement when referential cardinality is 2. So, pis and pip are excluded, and pii. Likewise, iis is impossible as the last position requires [+group], the first two [+augmented]; but [+augmented +group] is sss. The number of possible mnemonics reduces to 28. One can imagine how such reasoning would continue, but it is rather superficial, emphasizing the mnemonics first and the features that generate them second. This study focuses on features first and, so, a better test of the theory developed above is to see how it constrains the space of mnemonics. The three class features are [±singular], [±augmented], [±group]. For each class, they are specified as plus, minus, or zero. This yields 27 classes: ⎧ ⎫ ⎫⎧ ⎫⎧ ⎨ [+singular] ⎬⎨ [+augmented] ⎬⎨ [+group] ⎬ [−singular] [−group] [−augmented] (112) ⎩ ⎭ ⎭⎩ ⎭⎩ ∅ ∅ ∅

CHAPTER 3

112

Therefore, 37 mnemonically possible classes are featurally impossible. The remaining classes are tabulated in (113). Of the 27 possibilities, [(±singular) −augmented ±group] and [+singular +augmented (±group)] are excluded below: as [±group] predicates of [+augmented], the specification [−augmented ±group] is semantically questionable; and [+singular +augmented] is simply contradictory (Corollary, p. 114).15 ‘0’ represents absence of a feature; boldfacing marks classes not attested above. (113)

[±singular] + + + + − − − − − − − 0 0 0 0 0 0 0

[±augmented] − 0 0 0 + + + − 0 0 0 + + + − 0 0 0

[±group] 0 + − 0 + − 0 0 + − 0 + − 0 0 + − 0

Class sii SIS SIP

sii sss ppp IIP

idi ids idp idp sss ppp IIP

sdi sds sdp sdp

Only three of the resulting mnemonics are new: sip, sis and iip, corresponding to [+singular ±group] and [(−singular) +augmented]. Two observations are appropriate. First, these classes are attested in related languages, such as Jemez (Sprott 1992, Yumitani 1998; see Noyer 1992, Harbour 2006c for crosslinguistic analysis). So, the system does not overgenerate. Second, observe that all are based on plus specification of [±singular] and [±augmented]. As a general rule, Kiowa avoids such specification: it occurs only for the first 15

Non-granular mass nouns are the nearest approximation to such nouns. However, [(−singular) +augmented +group] is more satisfying, especially when faced with two liquids at once.

NUMBER FEATURES

113

person singular (sii [+singular]), or in conjunction with [±group]. So, sip, sis and iip, are naturally absent from Kiowa, avoiding ‘local’ overgeneration. 3.7.1.

Beyond Merrifield’s method

Throughout this work, classes have been identified on the basis of Merrifield’s method (cf., Harrington 1928), examining correlations between referential cardinality and agreement type. However, in the discussion of subclasses (p. 34), attention was drawn to the possibility that featurally distinct classes can be mnemonically indistinguishable, as with ppp, which comprises pluralia tantum and granular mass subclasses. Given (113), we can move beyond Merrifield’s method, associating semantic properties of nouns, not with mnemonics, which underdetermine class features, but with the class features themselves directly. For instance, footwear might be sdp not because it is default, but because it is [−group]; and subclasses of idp nouns might be distinguished by zero versus minus specification for [±group] (vegetation versus implements and body parts, say). This still permits many-to-one correspondences between semantic characteristics and feature sets. However, the method of detecting such subsets is more direct than Merrifield’s method permits. I leave this for future research. 3.8. 3.8.1.

Appendix: Formal demonstrations Cardinality

I now present four simple proofs justifying (9) and the exclusion of [+singular +augmented]. Lemma 1.

[+singular −augmented] corresponds to 1.

Proof. By definition, atom(x) entails ¬∃y[y x]. So, in particular, ¬∃y[atom(y) ∧ y x]. Now, if P |= Q, then P |= Q ∨ ¬P. So: atom(x) |= ≡ ≡ ≡

¬∃y[atom(y) ∧ y x] ∨ ¬atom(x) ¬∃y[atom(x) ∧ atom(y) ∧ y x] [¬λP∃y[P(x) ∧ P(y) ∧ y x]](atom(x)) [−augmented]([+singular])

CHAPTER 3

114

By definition, 1 corresponds to atomicity. So, [+singular −augmented] = 1. Corollary. Lemma 2.

[+singular +augmented] is contradictory. [−singular −augmented] corresponds to 2.

Proof. It suffices to show that only dyadic x satisfies: [−singular −augmented] ≡ ≡ ≡ ≡

[−augmented]([−singular]) [¬λP∃y[P(x) ∧ P(y) ∧ y x]](¬atom(x)) ¬∃y[¬atom(x) ∧ ¬atom(y) ∧ y x] ∀y[atom(x) ∨ atom(y) ∨ y x]

This simplifies to ∀y[atom(y) ∨ y x], given the monotonicity of standard logics (specifically, x such that atom(x) is disregarded as a means of satisfying [−singular −augmented], as it contradicts the prior [−singular]). Consider the logically equivalent formula ∀y[y x → atom(y)]. Now, if x = {a, b}, then {a} and {b} are the only y such that y x; and atom(y). However, for any larger x, such as x = {a, b, c}, there is non-atomic y, such as {a, b}, such that y x. So, only dyadic x satisfies the formula. Lemma 3.

[−singular +augmented] corresponds to 3.

Proof. It suffices to show that triadic x, or larger, satisfies: [−singular +augmented] ≡ [+augmented]([−singular]) ≡ [λP∃y[P(x) ∧ P(y) ∧ y x]](¬atom(x)) ≡ ∃y[¬atom(x) ∧ ¬atom(y) ∧ y x] The formula requires that x be non-atomic with a non-atomic subpart. The smallest value for y, then, is a dyad. So, the smallest value for x is a triad. Theorem. (9) accurately represents, and exhausts, the combinatorial possibilities of [±singular] and [±augmented]. Proof. Note.

Lemma 1, Lemma 2, Lemma 3, Corollary. See p. 70 for a graphic interpretation in the shape of a lattice.

NUMBER FEATURES 3.8.2.

115

Grouphood

On p. 98, we entertained the [±group] definition: (114)

[+group] if [+augmented] → Parts(x) [−group] if [−augmented] → ¬Parts(x)

However, this violates the feature negation condition, [−F] = ¬[+F]. I now prove that this is not fixed by choosing a more complex truth functor than ‘→’. Rather, it is simply impossible to define the feature in this way. Theorem. There is no feature / truth functor [+group] = ∗(A, B), such that [−group] = ¬[+group] and: ∗ (A, B) ≡ ¬A → B ¬ ∗ (A, B) ≡ A → ¬B

(3.1) (3.2)

Proof. Suppose, for a contradiction, that ∗ exists. By (3.2): | ¬ ∗ (A, B) | = 0 iff | A → ¬B | = 0 iff | A | = | B | = 1

(3.3)

However, it follows from the definition of negation that: | ¬ ∗ (A, B) | = 0 iff | ∗(A, B) | = 1 So, by (3.1): | ¬ ∗ (A, B) | = 0 iff | ¬A → B | = 1

(3.4)

Combining (3.3) and (3.4) produces the desired contradiction: | A | = | B | = 1 iff | ¬A → B | = 1 Comment. Taking A as [+augmented] and B as Parts(x), the theorem shows that there is no logical manipulation of [+augmented] and Parts(x) that permits [±group] to be defined as in (114).


Chapter 4 Agreement and Suppletion Several predicates in Kiowa are sensitive to the number of their inner argument; that is, Kiowa has predicates that supplete for number. Consequently, sensitivity to number is a property of two parts of Kiowa grammar: agreement and suppletion. Mostly, these two operate in tandem: if referential cardinality is 1, the verb will occur in its s-suppletive form bearing s-agreement, or if referential cardinality is 3, the verb will occur in its p-suppletive form bearing p-agreement. However, at times, agreement and suppletion mismatch, with agreement implying one referential cardinality and suppletion, another. This chapter is concerned with the mechanisms of suppletion in general and with agreement∼suppletion mismatches in particular. I argue that these facts receive a natural analysis given the theory developed in Chapter 3. The aspect of that theory relevant to mismatches is the divergence permitted between the feature content of Number and D (and between Class and D). By providing divergent feature specifications in a single syntactic structure, the theory permits divergence between number-sensitive phenomena, provided they depend on different heads. For instance, if agreement depends on D, as argued in Chapter 3, and if suppletion depends on Number or Class, then agreement and suppletion will be conditioned by different features whenever grouphood or inverse marking affect D. The discussion begins with the introduction of Kiowa’s number-sensitive predicates and clarification of the notion of suppletion, as opposed to allomorphy and phonological readjustment. I then suggest some syntactic and morphological principles following Sportiche (1997) and Adger, Béjar, and Harbour (2001), according to which it is natural for agreement and suppletion to be sensitive to different heads. With these principles in hand, basic cases 117

CHAPTER 4

118

of suppletion without mismatches are derived. These are followed by treatment of inversive mismatches, mass nouns, [±group]-induced mismatches, and reflexive-induced mismatches. An appendix provides information on further uses of suppletive pairs and their interesting differences in meaning. 4.1.

Suppletion

This section introduces Kiowa’s suppletive predicates and defines suppletion. 4.1.1.

Number-sensitive predicates

Kiowa’s number-sensitive predicates fall into two classes. Those in (1) display an S∼D/P opposition, those in (2), S/D∼P. (1)

(2)

Gloss big, old, important small, young tall, long short Gloss be sitting be lying be sitting.inan set, put in lay land, fall against, fall drop, fall sever wander

S

D/P

êl syÓn ky´ oM´MZ xé´Z

bˆZn syán k´MZ´MZn´Z´Z xá a´dóu ´

S/D

P

a´M áM gya k!ÓÓ xél xé´Z x!óu ´ x!ó´Zgyá o´l t!ál thóu ´

k!´ ul k!´ ul sÓl sÓÓ k!´ uu ´ k!´ u´Zgyá p!él thá a´ zéM´MZ

In featural terms, the predicates in (1) are sensitive to the value of [±singular], those in (2), to [±augmented].1 Terminologically, these features are said to condition suppletion of the respective predicates. The semantic division is between individual-level (1) and stage-level (2). It is natural to conjecture as to semantic principles behind this correlation: perhaps because stages 1

Some predicates in (2) share roots: ‘lay’ ∼ ‘land, fall against, fall’; ‘be sitting.inan’ ∼ ‘set, put in’. See Watkins (1984) on the semantics of -l, -i and -gy´ a.

AGREEMENT AND SUPPLETION

119

can be in a subpart relation, they are associated with the subpart feature [±augmented]; and because permanent states are not in a subpart relation, they are associated with the atomic feature [±singular]. However, I leave this aside, as supporting evidence is scant (but see Appendix 4.7.) Number-conditioned suppletion can be illustrated by combining ‘big’ from (1) and ‘be lying’ from (2) with the sdp noun t´ ou ´d´ e ‘shoe’. In (3), referential cardinality is 1 and we find s-agreement and the s-form of the predicate, ˆ el. (3)

Tóu ´dé ∅- ´ et shoe 3s-big.s ‘The shoe is big’

In (4)−(5), for referential cardinalities 2−3, we find the d/p-form of the predicate, bˆZn. The two are distinguished by d∼p-agreement. (4)

Tóu ´dé eM - bˆZn shoe 3d-big.d/p ‘The two shoes are big’

(5)

Tóu ´dé gya-bˆZn shoe 3p- big.d/p ‘The shoes are big’

Now consider ‘be lying’. In (6)−(7), for referential cardinalities 1−2, we find s/d-form of the predicate, k!´ O´ O, distinguished by s∼d-agreement. (6)

Tóu ´dé ∅- k!´ O´ O shoe 3s-be lying.s/d ‘The shoe is lying’

(7)

Tóu ´dé eM - k!´ O´ O shoe 3d-be lying.s/d ‘The two shoes are lying’

In (8), for 3, we find, the p-form of the predicate, k!´ ul. (8)

Tóu ´dé gya-k!´ ul shoe 3p- be lying.p ‘The shoes are lying’

Both patterns can be shown in tandem (cf., Hale 1997 for Hopi). Here, ‘big’ is used attributively; ‘drop’ is sensitive to the number of its internal

CHAPTER 4

120

argument. (T´ ou ´d´ e ‘shoe’ loses its -de when modified; Section 2.6.2.) (9)

Tóu ´- ˆ el gya- ´ ot shoe-big.s 1s:3s-drop.s/d.pf ‘I dropped the big shoe’

(10)

Tóu ´- bˆZn nen- ´ ot shoe-big.d/p 1s:3d-drop.s/d.pf ‘I dropped the two big shoes’

(11)

Tóu ´- bˆZn gyat- p!´ et shoe-big.d/p 1s:3p-drop.p.pf ‘I dropped the big shoes’

Lastly, before clarifying some concepts relevant to the analysis of such sentences, consider some examples of agreement∼suppletion mismatches: iagreement with the s/d- and p-form of the predicate; and p-agreement with the s-form, and s-agreement with the p-form, of the predicate. (12)

a.

b.

(13)

a.

b.

E- k!ÓÓ 1/3I-be lying.S/D ‘It is lying’ or ‘We two are lying’ NÓ- k!´ ul :1s:3I-be lying.P ‘Mine are lying’ YáM - dôi- et :1s:3P-too-big.S ‘It is too big for me’ ∅- SÓl 3S-be sitting.P ‘It is sitting’

These examples may seem to undermine the classification of the predicates as s/d/p-forms. However, we will see that the mismatches are principled and follow naturally from the theory of Chapter 3. 4.1.2.

Clarification

There now follows a definition of suppletion and a comparison of suppletion, readjustment and allomorphy, as well as some description of Kiowa phonology


121

relevant to distinguishing suppletion from forms that are related but by nonobvious means. Only the definition of suppletion is central to what follows and even this may be skipped by readers familiar with the concept. Suppletion, readjustment and allomorphy ‘Suppletion’ has been used to cover three phenomena that probably deserve explanation by different theoretical mechanisms. English examples are: (14)

a. b. c.

Past tense morphemes -ed∼∅ Present∼past forms sing∼sang Present∼past forms go∼went

Only the last of these is suppletion in the sense relevant here. The other two differ from it in ways detailed below. A pair ψ∼ψ is suppletive if the two forms realize the same root in different grammatical contexts, but are not related by synchronic phonology. (15)

Definition: Suppletion An item is said to supplete if and only if: a. it is a root b. it has (at least) two phonologically distinct exponents, ψ and ψ , such that: (i) ψ is conditioned (by, say, tense, number or animacy) (ii) no phonological process of the language is capable of generating ψ from ψ (or vice versa).

Consequently, a suppletive item is a vocabulary item of the form below: a single root with two or more exponents, conditioned by morphosyntactic features (with the possible exception of an unconditioned elsewhere form). (16)

√

⇔ ψ / [F] .. . ⇔ ψ

On this definition, go∼went is suppletive:2 2

I leave aside the issue of whether went is, in fact, the root wend together with the past tense morpheme of send∼sent, lend∼lent, et cetera.

CHAPTER 4

122 (17)

√

go ⇔ went / [+past] ⇔ go

Cases of suppletion should be distinguished from alternations like -ed∼∅ and sing∼sang for the following reasons. Chomsky and Halle (1968/1991) √ reject a suppletive analysis for sing∼sang ( sing ⇔ sang / [+past]; ⇔ sing). Rather, simplifying slightly, they posit a phonological rule that is triggered by [+past]. On such an account, only one form of the root is stored. √ sing ⇔ sing (18) Chomsky and Halle observe that the phonological processes that derive sang from sing can be called on elsewhere in the grammar, as in the superficially unique ¯ ai∼æ alternation of satisfy∼satisfaction (pp. 201−202). See also Yang (1999, 2002) for more recent discussion of consequences of distinguishing go∼went from sing∼sang alternations. Now consider -ed∼∅, which I regard as allomorphy: (19)

Definition: Allomorphy An item is said to exhibit allomorphy if and only if: a. it is not a root b. it has (at least) two phonologically distinct exponents, ψ and ψ , such that: (i) ψ is conditioned (by, say, tense, number or animacy) (ii) no phonological process of the language is capable of generating ψ from ψ (or vice versa).

Terminology aside, allomorphy differs from suppletion in exactly one respect: suppletion is confined to roots, allomorphy to non-roots. Adger, Béjar, and Harbour (2001) argue that this terminological difference is warranted because the apparent locality conditions on suppletion and allomorphy differ. Suppletion requires sisterhood, whereas allomorphy does not. This reflects the syntactic difference between roots and affixes. Affixes are feature bundles that can enter Agree relations; these Agree relations are non-adjacent and can provide the conditioning context that licenses ψ as opposed to ψ. Roots, by contrast, are primitive, and so do not enter into Agree relations; consequently, they have no access to non-adjacent features. (This difference between affixal allomorphy and root suppletion is exportable to other frameworks making a principled distinction between affixes and roots.)


123

It is clear that the alternations in (1) and (2) could not be allomorphy, as they concern roots. Moreover, as is clear from the thorough description of segmental alternations in Harrington (1928) and, especially, Watkins (1984), (1) and (2) cannot be the result of readjustment rules. Therefore, I conclude that these are all cases of suppletion. Suppletion versus phonology As just emphasized, what counts as suppletion in a given language depends on that language’s phonology. A skeptic may wonder whether, say, x´ e´Z ‘set.s/d’ and s´ O´ O ‘set.p’ might not be related by, for instance: (20)

xé´Z → sé´Z / Context1 sé´Z → sÓÓ / Context2

To make such a claim, one must have evidence for each step. In (20), the steps are otherwise unattested and the context unspecified. It should be noted, however, that the demand for evidence is not setting the standard impossibly high, for, first, the child must likely have such evidence, and, second, the standard can be met in some quite daunting cases, as I now illustrate. (This illustration is not crucial to the analysis of suppletion that follows and readers without a taste for phonological oddments may wish to proceed directly to subsequent sections.) Watkins (1984, p. 164) lists four root∼perfective pairs as suppletive. I suggest that they are actually phonologically regular. (21)

Gloss die disappear exit, carry out pull

Root hˆMZMi y´Z´Z khˆZi k´MZ´MZ

Perfective hêm yâi thép tém

To explain ‘disappear’, observe that -i∼-iˇ is an infrequent form of the perfective (the nasal form occurs with nasal vowels): (22)

Gloss roast sound speak, say

Root tÓM ÓM p´ oM u ´M t´ oM u ´M

Perfective tÓM´MZ p´ oM´MZ t´ oM´MZ

CHAPTER 4

124

Rightwards spreading of high tone onto -i∼-iˇ and shortening of the root vowel are phonologically regular (Watkins 1984, Harbour 2002). Second, note the alternation (y)i∼ya: (23)

Gloss fire; kindling God; Jesus uncle.voc; uncle.name

i -Form ph´Z´Z dOOk!´Z´Z segˆZi

ya-Form phyá´Zsón dOOk!ya´Z´Z segyâi

Though not entirely understood, we may hypothesize that a sufficient condition for the alternation is that i be short and followed by another i (an OCP effect, not to be confused with the orthographically identical long ii). Last, note that anything following the root y´Z´Z has low tone (e.g., y´Z´ZhOO ‘keep on vanishing’, y´Z´Zt!OO ‘will vanish’). The perfective now follows from √ root+pf, y´Z´Z-i: (24)

Concatenation yields: Low tone assignment yields: Shortening of the root vowel yields: yi → ya before i yields:

y´Z´Z-i y´Z´Z-`ı y´Z-`ı y´ a-`ı

´V ` is written VV. ˆ So, the desired yˆ By orthographic convention, V ai results. To derive the other three forms, several of the same mechanisms are called on. First, -p∼-m is another infrequent form of the perfective; it, too, conditions vowel shortening (and, again, the nasal form occurs with nasal vowels): (25)

Gloss drink hit lay.p pour recall, be aware of

Root th´ oM u ´M gˆ uu k!´ uu ´ oú ´ hÓÓ

Perfective thóm góp k!óp o´p hÓp

Observe the lowering of the high vowel uu in ‘hit’ and ‘lay.p’ in (25). Generalizing lowering to all high vowels, we derive ‘die.pf’ from hˆˇZˇi-m: (26)

Concatenation yields: Vowel shortening yields: High vowel lowering yields:

hˆˇZˇi-m hˆˇZ-m hˆ eˇ -m (= hˆ em)


125

Recall that velars become dental before e (Section 2.6.2). This switching, together with processes affecting hˆˇZˇi ‘die’, apply to k´ˇZ´ˇZ ‘pull’: (27)

Concatenation yields: Vowel shortening / lowering yields: Velar-to-dental switching yields: Nasalization yields:

k´ˇZ´ˇZ-p k´ eˇ -p t´ eˇ -p t´ eˇ -m (= t´ em)

Finally, observe that falling tones cannot be realized over the sequence Vp; hence, the tonal simplification of gˆ uu to g´ o in (25). Given this, the perfective of khˆZi ‘pull’ is derivable: (28)

Concatenation yields: Vowel shortening / lowering yields: Velar-to-dental switching yields: Tone simplification yields:

khˆZi-p khˆ e-p thˆ e-p th´ e-p (= th´ ep)

Returning to (1) and (2), the claim is that none of these pairs is explicable by synchronic phonology in the manner just illustrated and for that reason all are related by suppletion, rather than readjustments.3 4.2.

Analysis of the Basic Cases

We now analyze the basic cases of Section 4.1.1. 4.2.1.

Assumptions

The analysis is based on the theory of allomorphy and suppletion of Adger, Béjar, and Harbour (2001). The conclusions relevant here are: (29)

a. b.

Suppletion requires adjacency of root to conditioner. Uninterpretable features do not condition suppletion.

These two conclusions tightly constrain the syntactic structure in terms of which suppletion must be explained. There are three heads capable of bearing 3

The emphasis on ‘synchronic’ is important. Some pairs may have been synchronically related at an earlier stage of the language. For instance, consider x!´ ou ´ ∼k!´ uu ´ ‘lay’. X!´ ou ´ has a homophone x!´ ou ´ ‘rock’. Interestingly, ‘rock’ in Rio Grande Tewa is k’u:; compare also Kiowa x!´ ol ‘wing’ with Rio Grande Tewa k’un, x!´ e´Z ‘thick’ with k’a:’i’; but k!´ ol ‘neck’ with k’´ e: (Kroskrity 1993, Appendix 1, citing Randall and Anna Speirs, p.c.).

CHAPTER 4

126

number features: Class, Number and D. For an sdp noun, Class is empty. So, suppletion depends on Number or D. The number features on D are uninterpretable, however, and so cannot condition suppletion (29b). This leaves Number, the highest head with interpretable number features, as the possible source of suppletion conditioning. The adjacency requirement (29a) requires the following structure: (30)

VP NumberP

V

Class: Noun Number If so, D cannot immediately dominate Number, as assumed in the Chapter 3. Instead, a position nearer to that of Sportiche (1997) is required, according to which D and its complement NP do not begin as complements in the syntax. Rather, NP begins as the complement of the verb and both are lower than D; the NP moves to D later, creating a DP. For current purposes, it must be NumberP that is the complement of V and that moves to D: (31)

DP .. . VP NumberP

D


V

Class: Noun Number This structure has the requisite adjacency relations for Number to condition V-suppletion, as we shall now see.4 If Noun is a mass noun, then Number is absent and, so, the verb is adjacent to Class. So, for mass nouns, suppletion will reflect the class features.5 4

Sportiche suggests that Number, like D, is higher than V. It seems to me that his arguments really address the position of D and are neutral on number. 5 Readers disinclined to accept the foregoing assumptions need not despair. Early predictions below may be viewed as demonstrating what suppletion and agreement depend on. Several predictions concerning the specifics of mismatches will still be left.

AGREEMENT AND SUPPLETION 4.2.2.

127

Analysis

Consider first the general case of an sdp noun. Number, represented in the abstract form [αsingular βaugmented], merges with the verb: (32)

VP NumberP

V

Class: Noun Number αsingular βaugmented

∅

The value α will condition suppletion of the verb if it is [±singular]-sensitive, or β will if it is [±augmented]-sensitive. D is Merged higher in the structure and attracts NumberP. (Movement can be implemented via an EPP feature D, in the framework of Chomsky 2000.) Movement and valuation yield: (33)

DP

NumberP Class: Noun Number ∅

D αsingular βaugmented


D will trigger agreement that reflects referential cardinality, hence sdp. As a specific example, consider (4), repeated below. (34)

Tóu ´dé eM - bˆZn shoe 3d-big.d/p ‘The two shoes are big’

For referential cardinality 2, Number is [−singular −augmented]. The predicate ‘big’ is [±singular]-sensitive:6 6

In (35), as in suppletive vocabulary entries below, both forms have conditioning contexts, [+singular] for ˆ el, [−singular] for bˆZn. This may be excessive, as one could be the elsewhere form and so be uncontextualized. I avoid this for two reasons. First, there is no

CHAPTER 4

128 (35)

√

big ⇔ ˆ el / [+singular] ⇔ bˆZn / [−singular]

In (36), the value of [±singular] is minus and bˆZn is inserted. (36)

VP

V

NumberP

bˆZn

Class: shoe Number −singular −augmented

∅

Raising NumberP to D and valuing of D’s uninterpretable number yields: (37)

DP

NumberP Class: shoe Number ∅

D



D triggers d-agreement, pronounced, for a simple intransitive, as eˇ . The resulting complex verb, then, is eˇ -bˆZn, as desired. As an example of the other predicate class, consider (8), repeated below. (38)

Tóu ´dé gya-k!´ ul shoe 3p- be lying.p ‘The shoes are lying’

Number is [−singular +augmented]. The predicate ‘be lying’ is [±augmented]sensitive, so its plus form (39) is used (40). √ (39) be lying ⇔ k!´ O´ O / [−augmented] ⇔ k!´ ul / [+augmented] clear √evidence for which√is the elsewhere form in each case. Second, s might be elsewhere for big, and d/p for small. However, the predicates of each class pattern alike with respect to suppletion; so, something more systematic than elsewhere forms is required.

AGREEMENT AND SUPPLETION (40)

129 VP

NumberP

V

k!´ ul

Class: shoe Number −singular +augmented

∅

Raising NumberP to D and valuing of D’s uninterpretable number yields: (41)

DP

NumberP Class: shoe Number ∅



D triggers p-agreement, yielding the complex verb gya-k!´ ul. Last, consider (10), repeated below, as an example of adjective with verb.

(42)

Tóu ´- bˆZn nen- ót shoe-big.d/p 1s:3d-drop.s/d.pf ‘I dropped the two big shoes’

The adjective ‘big’ is adjoined to NumberP (Section 3.2.4). So, adjective and verb are adjacent to Number. By (35) and (43), we have (44) (where, for the sake of brevity, the whole structure is shown with vocabulary items): (43)

√

drop ⇔ o ´l / [−augmented] ⇔ p!´ el / [+augmented]

CHAPTER 4

130 (44)

VP NumberP

NumberP

V

ot ´

Adjective bˆZn

Class: shoe Number t´ ou ´-


NumberP moves to D, yielding the DP t´ ou ´bˆZn ‘big shoes’. D is valued as [−singular −augmented] and triggers d-agreement, yielding the complex verb nen-´ ot ‘I dropped them two’. 4.2.3. Summary The system of agreement and suppletion triggering stated and illustrated in this section, with suppletion dependent on Number but agreement dependent on D, derives the correct agreement and suppletion forms for sdp cases. However, it may seem overcomplicated, particularly in its reliance on Adger, Béjar, and Harbour (2001), which forces suppletion and agreement, two rather similar phenomena, to depend on different parts of the structure. For sdp nouns, where the feature content of D is always that of Number, it would indeed be sufficient for both to look at one. Chapter 3, however, presented a variety of cases in which Number and D differ. The full variety of agreement∼suppletion mismatches is summarized below: (45)

Cardinality, Class 1 idp, ids, idi 1 ppp 1 sss 2 sii 2 sss 3 sdi, idi 3 sds, sss, idi 3 sdi

Agr−Adj.S∼D/P i−s p−s i−d/p i−d/p s−d/p a−s

Agr−V.S/D∼P i−s/d s−p i−s/d s−p i−p s−p a−s/d

AGREEMENT AND SUPPLETION 4.3.

131

Inversive mismatches

i-agreement is opaque to referential cardinality. It can occur when it is 1 (idp, ids, idi), 2 (sii), and 3 (sii, sdi, idi). Since inverse marking is the reflex of feature conflict, it is overspecified and cannot condition a unique suppletive form. We predict, however, that verbs with i-agreement ignore D and supplete instead according to referential cardinality. The subsections below consider minus-valued, i.e., non-sii, and plus-valued classes separately. 4.3.1.

Minus-valued classes

To illustrate the prediction, consider the idi noun ‘hair’ and the s/d∼p predicate ‘be lying’ (Watkins 1984, p. 89). For referential cardinality 1, the noun is inverse marked and triggers i-agreement. However, the predicate, in its s/d-form, reflects referential cardinality. (46)

´M´MZgOO ÓÓ dÓ E e- k!ÓÓ there hair.inv 3I-be lying.S/d ‘There’s a hair lying there’

For 2, the predicate is still in its s/d-form but bears d-agreement and the noun is not inverse marked. (47)

O´ l

éM´MZgOO eM - k!ÓÓ hair there 3D-be lying.s/D ‘There are two hairs lying there’

For 3, the noun is again inverse marked and triggers i-agreement. Again, the predicate, now in its p-form, still reflects referential cardinality. (48)

O´ÓdÓ

hóldap nÓ- k!´ ul hair.inv dress.loc :1s:3I-be lying.P ‘I’ve got some hair on my dress’

As a sample derivation, consider (48). The VP with its NumberP complement is (40), and, by (39), [+augmented] conditions k!´ ul, as in (38). When NumberP raises to D, Class, [−singular −augmented], and Number, [−singular +augmented], create feature conflict on D, yielding inverse marking on the noun and i-agreement on the verb, as desired.

132

CHAPTER 4

As examples of the s∼d/p predicate type, consider sdi (49) and idp (51) nouns. (2, essentially the same as for sdp nouns, is omitted.) (49)

NÓÓ-baou éMné ám-baougO gÓ- syân ét, 1- cat :1s:3S-big.S but 2- cat.INV :2s:3I-small.d/P ‘My cat is big, but your cats are small’

For sdi nouns, such as b´ aou ‘cat’, Class is [−augmented]. In the first clause, where referential cardinality is 1, Number is [+singular −augmented] and conditions the s-form of ‘big’ (35). D is valued as [+singular −augmented], triggering s-agreement. The resulting verb, with agreement for the indirect object (possessor), is ´ eˇ -´ et. In the second clause, Number is [−singular +augmented], which conditions the d/p-form of ‘small’: √ (50) small ⇔ sy´ On / [+singular] ⇔ syˆ an / [−singular] Class and Number have opposite specifications of [±augmented]. So, the possessed noun, b´ aougO, is inverse marked and triggers i-agreement. The resulting verb, with agreement for the possessor, is g´ O-syˆ an, which shows number-conditioned suppletion despite number-neutral i-agreement. In contrast to (49), consider (51), with an idp noun. (51)

´M´MZgO E p´MZa´ a´dO e- ét, né ó´Zde gya-syân this.INV table.INV 3I-big.S but that 3P- small.d/P ‘This table is big, but those tables are small’

Here, class is [−singular]. So, in the first clause, where Number is [+singular −augmented], the s-form of ‘big’ is conditioned (35). However, Number and Class clash, so that demonstrative, ´ eˇ´ˇZgO, and noun, p´ˇZ´ a´ adO, are inverse marked and the verb bears i-agreement. The resulting e-´ et again shows number-conditioned suppletion despite number-neutral i-agreement. In the second clause, both agreement and suppletion are transparent to Number: [−singular +augmented] conditions the d/p-form of the predicate and triggers p-agreement via D. These, with (50), yield gya-syˆ an. Consider, lastly, an appositively used s∼d/p adjective with the idi noun ´ alOO ‘apple’. Since we are only concerned with the DP, I simplify the discussion by assuming a structure in which NumberP, with its adjectival adjunct, has raised to D. The forms below are taken from Wonderly, Gibson, and Kirk (1954, p. 6, with minor corrections to tone). The adjective suppletes


133

for number as expected and the DP as a whole is inverse marked exactly on a par with any idi noun.

(52)

a´lOO- et- tO apple-big.s-inv ‘a big apple’

(53)

a´lOO- bin apple-big.d/p ‘two big apples’

(54)

a´lOO- biMMidO apple-big.d/p-inv ‘several big apples’

Consider first ‘a big apple’. Class, [−singular −augmented], and Number, [+singular −augmented], cause feature conflict on D:

(55)

DP

NumberP NumberP Class: apple −singular −augmented

Adjective

D ⎡ ⎤ −singular ⎣+singular ⎦ −augmented

Number


Number conditions ˆ el, the s-form of ‘big’ which, in turn, phonologically conditions the inverse marker tO on D:

CHAPTER 4

134 (56)

DP NumberP

NumberP

D

tO

Adjective ˆ el

Class: apple Number +singular −augmented

´ alOO

Standard phonology yields the desired form, ´ alOOettO.

Things are slightly simpler for 2. The valued tree is:

(57)

DP

NumberP

NumberP Class: apple −singular −augmented

D −singular −augmented

Adjective

Number −singular −augmented

Because Class and Number are both [−singular −augmented], no conflict arises on D. There is no inverse marking. Number conditions bˆZn, the [−singular] form of ‘big’.


135

(58)

DP NumberP

NumberP

D

∅

Adjective

bˆZn


´ alOO

Again, standard phonology yields the desired form, ´ alOObin.

Finally, consider referential cardinality 3. The valued DP is:

(59)

DP

NumberP

NumberP Class: apple −singular −augmented

D ⎡ ⎤ −singular ⎣−augmented⎦ Adjective +augmented


This combines aspects of the preceding cases: as for 1, there is feature conflict on D, though this time for [±augmented]; and, as for 2, [+augmented] on Number conditions bˆZn for ‘big’.

CHAPTER 4

136 (60)

DP NumberP

NumberP

D

dO

Adjective

bˆZn


´ alOO

Again, standard phonology yields the desired form, ´ alOObiˇˇidO. So, the correct forms are easily derivable for all relevant noun classes under varying values of Number and in a variety of syntactic constructions. 4.3.2.

Plus-valued class

Consider now the [+singular] class sii, the only member of which is the first person, [+author]. Recall from Chapter 3 that a valued first person DP for referential cardinality 2 or 3 has the following structure: (61)

DP

NumberP ⎡Class: I (you)⎤ +author ⎢+singular ⎥ ⎢ ⎥ ⎣(±hearer) ⎦ (−augmented)

Number −singular ±augmented

D ⎡ ⎤ +author ⎢−singular ⎥ ⎢ ⎥ ⎢+singular ⎥ ⎢ ⎥ ⎢±augmented ⎥ ⎢ ⎥ ⎣(+hearer) ⎦ (−augmented)

This corresponds to first person inclusive if [+hearer] is included (and with it [−augmented]). The important point to observe in (61) is that Number distinguishes 2 from 3, but agreement does not. This differs from, say, sdi and idp nouns, where there are distinct agreement types for each referential cardinality. It also differs from idi, however. Both idi and sii trigger i-agreement for two


137

values of referential cardinality. However, for idi, these values are 1 and 3, a non-natural class for suppletive predicates. For sii, they are 2 and 3, the natural class defined by [−singular]. One might imagine, therefore, that sii simply lacks specification for [±augmented]. Given the preceding discussion, we can use suppletion to show that this is not so: [±augmented]-sensitive predicates distinguish first person dual from first person plural. For the first person exclusive, dual and plural are distinguished as in (62); (63) gives the relevant vocabulary items: (62)

(63)

a.

E- x!ó´Zgyá 1i-fall.s/d.pf ‘She and I fell’

b.

E- k!´ u´Zgyá 1i-fall.p.pf ‘They and I fell’ √ lay ⇔ x!´ ou ´ / [−augmented] ⇔ k!´ uu ´ / [+augmented] detr ⇔ -´Zgy´ a

a. b.

Similarly, for first person inclusive: (64)

a.

b.

(65)

Ba-thóu ýa 2p- move.s/d.impf ‘You and I are moving around’

(Watkins, p.c.)

Ba-zéM´MZma 2p- move.p.impf ‘You, I and (s)he/they are moving around’

(Watkins, p.c.)

√ wander ⇔ th´ ou ´ / [−augmented] ⇔ z´ eˇ´ˇZ / [+augmented]

Summary The DP structure of Chapter 3, by permitting divergence between the content of Number and D, provides the correct features to derive the agreement∼suppletion mismatches that arise with cardinally opaque i-agreement.

CHAPTER 4

138 4.4.

Group-induced mismatches

Further agreement∼suppletion mismatches are predicted to arise in virtue of [±group]. This follows because, if Class and Number bear [+augmented] and [αgroup], then D bears [¯ αaugmented], even if Number bears the opposite value [αaugmented]. We examine first the sds, ids and idi cases, then sss, then [+group] mass nouns. Although pluralia tantum nouns, ppp, fall naturally into this discussion, being [−group], they are deferred until Section 4.5, as their behavior is more complicated than [+group] nouns’. 4.4.1.

Collectives

For countable [+group] nouns, agreement∼suppletion mismatches are predicted to occur for referential cardinality 3, where the DP has the structure:7 (66)

DP


D



Number and D are oppositely specified for [±singular] and [±augmented]. Therefore, for s∼d/p predicates, like ‘big’, we predict s-agreement on the d/p-form; and for s/d∼p predicates, like ‘be set’, s-agreement on the pform. These combinations are exemplified below. First, consider the s/d∼p suppletive predicate ‘be set’. This is illustrated with the sds noun, t´ ou ´ ‘house’ (Watkins 1984, p. 90). (67)

Tóu ´ ∅- xél house 3S-be set.S/d ‘There’s a house standing’

(68)

Tóu ´ eM - xél house 3D-be set.s/D ‘There are two houses standing’

7

I abstract away from the possibility of postsyntactic feature insertion (Section 3.4.2).


139

Tóu ´ ∅- sÓl house 3S-be set.P ‘There are houses standing’

In the first two sentences, agreement and suppletion both transparently reflect referential cardinality. However, in (69), the p-form bears s-agreement. This is exactly as expected given (66) and the vocabulary items: √ (70) a. be set ⇔ x´ e´Z / [−augmented] ⇔ s´ O´ O / [+augmented] b.

stat ⇔ -l

(71)

VP NumberP

V

sÓl



The [+augmented] on Number licenses s´ Ol, but D, [+singular −augmented], triggers s-agreement, yielding (70). The same point is made for ids nouns by (Watkins 1984, p. 87): (72)

(73)

(74)

´ adO A´ e- xél tree.inv 3i-be set.s/d ‘A tree is standing / growing’ ´ a eM- xél A´ tree 3d-be set.s/d ‘Two trees are standing / growing’ ´ a ∅- sÓl A´ tree 3s-be set.p ‘Trees are standing / growing’

As details are as discussed above for sds nouns and for the inverse, these examples require no further comment. Consider, now, s/d∼p suppletive predicates. These are illustrated with idi nouns on their ‘different types of’ reading.

CHAPTER 4

140 (75)

HÓndé Ól bÓ/*ky´ oM´MZ k´MZ´MZn´Z´Z what hair :2p:3S-long.d/P/ long.s ‘What long [types of] hair you all have’

(76)

HÓndé Ól bÓxá a´dóu ´ /*xé´Z what hair :2p:3S-short.d/P/ short.s ‘What short [types of] hair you all have’

In each case, ‘hair’ triggers s-agreement. However, the s-form of the predicates is unacceptable and the d/p-form must be used. Again, this is exactly as expected given (66) and the vocabulary items: √ (77) long ⇔ k´ˇZ´ˇZn´Z´Z / [+singular] ⇔ ky´ o ˇ´ˇZ / [−singular] √ short ⇔ x´ a´ ad´ ou ´ / [+singular] (78) ⇔ x´ e´Z / [−singular] Thus, nouns that trigger s-agreement when referential cardinality is 3 trigger suppletion as expected. 4.4.2.

Mass nouns

Thus far, all mismatches have been cases in which suppletion ‘sees through’ agreement to true referential cardinality. In this regard, non-granular mass nouns are very interesting. If their uncountability indicates absence of Number (Section 3.5), then suppletion in such cases cannot depend on Number. Nonetheless, it can depend on number features, as these are present on Class. So, for mass nouns, Class-conditioned suppletion, rather than Number-conditioned suppletion is expected, in virtue of the structure: (79)

VP Class: Mass V

Given (79), the feature composition of Class is of key importance for mass nouns, especially sss. Generally, s-agreement arises because D dominates only [+singular] and [−augmented]. However, I argued that the classification [+singular −augmented] makes no sense for mass nouns; rather, they are [−singular +augmented +group]. The correctness of this argument can now be verified by the predictions it makes with respect to suppletion.


141

With sss nouns, [±augmented]-sensitive predicates occur their in p-form. (80)

´ e /syÓndé th´ Ett´ oM u ´M ∅- sÓl dOálkya much/little water 3s-be sitting.P bucket.loc ‘There’s much / little water in the bucket’

(81)

SyÓndé th´ oM u ´M ∅- p!étkyá /*ótkyá little water 3s-fall.P.detr.pf/ fall.s/d.detr.pf ‘A little / a drop of water fell’

(82)

/*xél K!Oâlkya étté pénhaa ∅- sÓl dish.loc much sugar 3s-be sitting.P/ be sitting.s/d ‘There’s a lot of sugar in the dish’ (suggestive of a bowl)

(83)

K!Oâlkya étté pénhaa ∅- k!´ uu ´ /*k!ÓÓ dish.loc much sugar 3s-be lying.P/ be lying.s/d ‘There’s a lot of sugar on the dish’ (suggestive of a plate)

Evidence for how sss mass nouns behave with [±singular]-sensitive predicates is harder to come by, owing to the meanings of these predicates. Sentences such as ‘The sugar is short’ or ‘The whisky is long’ do not make much sense, and ‘The water is large’ or ‘The snow is small’, to the extent that they are interpretable, are so in a non-mass sense, such as ‘body of water’ or ‘snow flake’. The nearest I have come to a suitable example is ´ OlhOˇ Oˇsyan ‘dime’, from ‘money’+‘small’. This seems to support the idea that ‘money’, and so sss nouns, are [−singular]. However, here, the language tricks us, for, of the four s∼d/p predicates, precisely ‘small’ ceases to be [±singular]-sensitive when attributive: syˆ an is used in all cases (Watkins 1984, p. 99). For completeness, observe that ppp mass nouns also occur only with the p-form of predicates. This is expected given the inherent classification [−singular +augmented]. (Again, for semantic reasons, I have no examples of [±singular]-sensitive predicates with these nouns.) (84)

(85)

O´ÓtháM t!aM Mi

p´MZá a´thai gya-sÓl salt table.loc 3p- be sitting.P ‘There’s salt on the table’

K!ÓM ÓM etOM OM étté ÓÓtháM t!aM Mi gya-sÓl frybread much salt 3p- be sitting.P ‘There’s a lot of salt in the frybread’

CHAPTER 4

142 (86)

Pé´Zgya toudôm yáM sÓl sand floor.loc :1s:3p-be sitting.P ‘I’ve gotten sand on the floor’

The examples of mass nouns show that it is irrelevant to suppletion whether number features are on Number or Class, and that the sss class features [−singular +augmented +group], motivated on semantic grounds, correctly predict [+group]-induced suppletion∼agreement mismatches parallel to those of sds, ids and idi nouns. 4.5.

Harder cases

The agreement∼suppletion mismatches induced by inverse and grouphood fall under a single generalization: that the information relevant to suppletion is the feature content of Number, or Class, when Number is absent. This result has been established above for the vast majority of nouns in the language, sdp, sdi, sii, idp, idi, ids, sds, sss, and ppp mass. In this section, we examine two corners of the language where the generalization that governs the previous cases breaks down: pluralia tantum and reflexives. 4.5.1.

Pluralia-tantum-induced mismatches

Pluralia tantum nouns, such as h´ olda ‘dress’, kh´ O´ Od´ e ‘trousers’, t´ ou ´ ‘teepee’, t´ o ˇ k´ ut ‘book, letter’, and t!´ o ˇu ´ ˇ gya ‘shirt’, are ppp, [+augmented −group]. Their behavior with respect to suppletion only partly conforms to the generalization that suppletion depends on the feature content of Number, or Class, when Number is absent: [±singular]-sensitive predicates obey the generalization, [±augmented]-sensitive ones do not, displaying the p-form throughout. Consider, first, a [±singular]-sensitive predicate. (87)

´M´MZde t!´ E oM u ´M gya yáM dôi- et this shirt :1s:3P-too-big.S ‘This shirt is too big for me’

Here, referential cardinality is 1. Number is [+singular −augmented] and, so, conditions the s-form of ‘big’ (35). However, the class features [+augmented −group] cause D to be valued as [−singular +augmented], triggering pagreement. So, the mismatch is correctly predicted. Now consider:


143

Hólda yáM dôi- bin dress :1s:3P-too-big.D/P ‘My dresses are too big for me’

For referential cardinalities 2−3, Number is [−singular ±augmented]. Consequently, [±singular]-sensitive predicates occur in their d/p-form, matching p-agreement, as expected. Similar examples are: (89)

´M´MZde k!ókÓ´ZouphOl gya-dôi- kyoM Mi /xei E this roach 3P- too-long.S/short.S ‘This roach is too long/short’

(90)

Té´Z gya-dôi- kiMMinii /xaadou all 3P- too-long.D/P/short.D/P ‘They [roaches] are all too long/short’

More complicated are [±augmented]-sensitive predicates, like ‘lay’. The predicate is in its p-form, irrespective of referential cardinality. (91)

T!´ oM u ´M gya gyat- p!Ó´Z-k!op shirt 1s:3P-mis-lay.P.pf ‘I lost a shirt / some shirts’

Though this is predicted for referential cardinality 3, it is entirely unexpected for 1 and 2. For these, Number is [±singular −augmented] and so, given (63), x!´ ou ´ should occur. However, it is in fact ungrammatical. (92)

*T!´ oM u ´M gya gyat- p!Ó´Z-x!ep shirt 1s:3P-mis-lay.S/D.pf ‘I lost one/two shirt(s)’

Similar examples are: (93)

K!ókÓ´ZouphOl yáM p!étkyá /*ótkyá roach :1s:3P-fall.P.detr.pf/ fall.S/d.detr.pf ‘My roach fell off’

(94)

Té´Z k!ókÓ´ZouphOl gya- p!étkyá all roach :3a:3P-fall.P.detr.pf ‘Everyone’s roaches fell off’

The constant occurrence of p-forms contradicts the notion of Number-

CHAPTER 4

144

sensitivity. A near solution is possible, however, if we take Number to be partially defective, deprived of [±augmented], for these nouns. (95)

NumberP Class: Pluralia Tantum +augmented −group

Number

±singular

The features in (95) are the correct ones to condition suppletion: [±singular]sensitive predicates will covary with the specification of that feature on Number, whereas [±augmented]-sensitive predicates will have access only to [+augmented] on Class. Two issues require comment: First, there is the issue of the defectiveness of Number. Carstens (1991) argues the content of Number varies crosslinguistically (see Harbour 2006a for a detailed proposal): English Number bears just [±singular], as the language contrasts just singular∼non-singular, rather than singular∼dual∼plural, as Kiowa does. So, to posit (95) is to claim that crosslinguistic variation can be replicated language internally. Split ergativity can be considered in this way: some languages are ergative, others accusative, and others exhibit the variation internally. More immediately relevant is the total defectiveness of Number for mass nouns. As some languages have totally defective Number for all nouns (my 2006a treatment of Pirahã, Everett 1986), Kiowa already presents a case of crosslinguistic variation language internally. So, I regard the partial defectiveness in (95) as unproblematic. Second, there is the issue of sisterhood. The verb, if sister to Number, is not sister to Class, and so Class cannot condition verb suppletion. However, as Class and Number are not in a mutual dominance relation, it is possible to define the notion of sisterhood, in terms of dominance, so that Class can condition verb suppletion (possibly, just in cases where Number is unspecified for the relevant feature). Whether this redefinition is insightful or mere technical opportunism, however, lies beyond the scope of the current work. It is, I think, fair to conclude that the suppletion conditioned by pluralia tantum nouns constitutes a slightly harder case than those examined above, but that it is not beyond analysis.

AGREEMENT AND SUPPLETION 4.5.2.

145

Animate- and reflexive-induced mismatches

Animates and reflexives constitute a second source of unexpected agreement∼ suppletion mismatches, conditioning suppletion as though of referential cardinality 1, yet agreeing, and being interpreted, as though 2 or, more frequently, 3. However, their variable agreement behavior reveals a feature distribution (108) consistent with the theory and mechanisms above, even if somewhat semantically surprising. Animate plural agreement, ‘a’, is triggered by animate nouns, of referential cardinality 3, that are subject to ‘empathy’ (Section 2.5); so, it is near obligatory with K´ O´Zg´ u ‘Kiowas’, common with k!y´ a ˇ´ a ˇ hy´ op ‘men’, maay´ op ‘women’, xˆ eˇ ˇigO ‘horses’, and rare with children and lower animates. Semantically, the featural correlate of ‘a’ is expected to be [−singular +augmented +empathic] (where the last feature is the correlate of empathy). However, animates condition suppletion exactly conversely, as though [+singular −augmented]. So, [±singular]-sensitive predicates occur in their s-form (96), and [±augmented]-sensitive ones, in their s/d-form (97): (96)

(97)

a.

KÓ´Z- k!yakoM uM bO a´- ky´ oM´MZ Kiowa-people.inv 3A-tall.S ‘Kiowas are tall’

b.

KÓ´Z- k!yakoM uM bO á- khóu ´- xei Kiowa-people.inv 3A-body-short.S ‘Kiowas are short’ ´ pé´Z- x!ó´Zgyá A3A-dead-land.S/d.pf ‘They fell down dead’

a.

b.

KÓ´Zg´ u de- deM Mi- x!ép Kiowa.inv 1s:3A-sleep-lay.S/d.pf ‘I laid the Kiowas down to sleep’

Animate plurals share this suppletive pattern with reflexives. For instance, there are three expressions built on suppletive predicates and demanding reflexive agreement: h´ ou ´t!al ‘part ways’ (travel-sever.s/d), h´ ou ól ‘turn off’ (travel-drop.s/d), and h´ onx!oigya ‘come late’ (last-land.s/d):8 8

H´ onx!oigya ‘come late’ is a dative reflexive, with the reflexive triggering, generally, i-agreement. See pp. 146ff.

146

CHAPTER 4

(98)

∅- SÓOM OM dehel gO . . . téphOi em- tóu ´dO gO hegÓ 3s-angry.hsy and everyone 3s:3a-gather.pf and then ém- hóu ´- t!alhel 3a:3A-travel-sever.S/d.pf.hsy ‘He became angry and . . . he gathered everyone [his followers] together they parted ways [from the others]’ (Harrington 1928, p. 252) ´ Ethóu ´- ot gO e- p!Ôi 1i:3A-travel-drop.S/d.pf and 1i-lose.pf ‘We turned off and got lost’

(99)

(100)

HÓn dÓsÓ´Z- Om-then- dOM OM mOO-do dÓt- hón-x!oigya neg :1p:3s-fast-do- heart-be.neg- because :1p:3A-last- fall.S/d.pf ‘We arrived late because we didn’t want to drive fast’

In all of the preceding, the subjects are plural (‘everyone’, ‘we’). However, the predicates all appear in their s/d-forms. Indeed, [+augmented]-forms, like *h´ ou ´thaa and *h´ onk!uigya, are explicitly rejected. Consistently, if the subject is singular, the predicate appears in s/d-form, as in: (101)

Hágyáxo de- hóu ´- ot somewhere-instead 1s:3A-travel-drop.S/d.pf ‘I turned off somewhere’

It is to some extent unsurprising reflexives and animate plurals should act as a class for suppletion, as they form a class for other purposes, too. Most basically, reflexive agreement simply is animate plural agreement, resulting in the systematic ambiguity recorded by Harrington (1928); for example: (102)

De- hól 1s:3A-kill.pf ‘I killed myself or them (people)’

The one agreement type is ambiguous, for all person-number combinations between acting reflexively and acting on a group of empathic animates.9 9

They can be disambiguated with ÓˇÓˇgO ‘self’ (i); (ii) and (iii) show other uses of ÓˇÓˇgO, emphatic and adverbial. (i)

ÓMÓM gO de- hól self 1s:3a-kill.pf ‘I killed myself / *them’


147

Also of importance here is the behavior of a-agreement as the direct object of a ditransitive, as it gives significant insight into the category’s feature composition. In this case, there is no special exponent of a-agreement, in contrast to (in)transitives. Instead, such cases exhibit a d∼i alternation. Consider, first, animate plurals. For the combination of agent and benefactive below, two agreement prefixes are possible, n´ e (103a) and n´ O (104a). The former is also used when the direct object is dual (103b) and, so, is dagreement; and the latter is i-agreement, as it is used when the direct object is inverse marked (104b). (103)

(104)

a.

Néhól 2s:1s:3A-kill.pf ‘You killed them (people) for me’

b.

Y´Z´Z néhól two 2s:1s:3D-kill.pf ‘You killed two for me’

a.

Té´Z k!yáM áM hyóp nÓhá´ZgyádOO all man.inv :1s:3A-know ‘I know all the men’

b.

Zé´ZbOt nÓ- p!Ôi arrow.inv :1s:3I-lose.pf ‘I lost my arrow.’

Reflexives, too, show d∼i variation in the presence of indirect object agreement. Consider the class of dative reflexives (105): (106) shows the d-agreement of (103), (107) the i-agreement of (104) (though, for the most part, each predicate occurs either with d or with i).

(ii)

ÓMÓM gO a- x´ an self 1s-arrive.pf ‘I arrived’

(iii)

ÓMÓM gO em- tóM u ´M - hÔn self 3s:3a-talk-exhaust.pf ‘He became silent of his own accord’


CHAPTER 4

148 (105)

Gloss awake be fittingly dressed come late have one’s heart skip look nice need trip

Root án dólbé hónx!oide kh´ utdOO dótkyai ÓtkhyádOO Ony´Z´Z

(106)

HÓ nédólbé? q :2s:3A-be fittingly dressed ‘Am I fittingly dressed?’

(107)

NÓ- hónx!oigya :1s:3I-come late.pf ‘I came late’

So, again, animate plurals and reflexives behave as a natural class with respect to a morphosyntactic process, the d∼i-object-agreement alternation conditioned by indirect object agreement. The d∼i-alternation gives a strong clue as to the feature composition of a-agreement. Suppose that Number is [+singular −augmented], as suppletion suggests, and, additionally, that Class is [−singular +empathic], where [±empathic] is the feature that distinguishes sda from sdi. Valuation of D involves replication of all features: (108)

DP

NumberP Class: Animate Number −singular +empathic


D ⎤ −singular ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented⎦ +empathic ⎡

Immediately, this creates a feature conflict on D, which causes inverse marking on the noun, as in (96)−(97), and on related nominal heads (see Section 3.2.4), such as relative clauses (109) and demonstratives (110):


(110)

149

[háote á- dÓÓ]-gO several 3A-be -INV ‘the few of them’ O´´ZhyOO- gO háyá ém- hóu ´- OM OM zonhel that one-INV somewhere 3A:3a-travel-start off.pf.hsy ‘Those ones traveled off somewhere’ (Harrington 1928, p. 252)

I suggest that D in (108) triggers a/d/i-agreement as follows. For all subjects, and for objects of transitives, there are particular vocabulary items for feature bundles containing [+empathic]. These are what is labeled aagreement. For objects of ditransitives, there are no special vocabulary items. So, the next most highly specified, those for [−F +F], are used, resulting in i-agreement. However, as Watkins observes, these forms have slightly disrespectful for connotations when used for adults. If [+singular] is impoverished here, as a grammaticalized means to avoid disrespect, the remaining features [−singular −augmented +empathic] result in d-agreement. (One can, perhaps, rationalize impoverishment of [+singular] over [−singular] by observing that, if the latter cooccurs with [+empathic] on Class, then the former is the more marked value in the context of [+empathic]; so, impoverishment of [+singular] is impoverishment of the marked.) So, the surprising s∼s/d suppletion pattern of animate plurals and reflexives is amenable to treatment within the current framework. This requires a slightly enlarged feature inventory; however, all accounts will need some parallel posit. The only point requiring clarification, then, is the effect of [+empathic] on Class and Number. Given that the feature only ever qualifies pluralities, its occurrence with [−singular] on Class is natural. (In fact, one could go further and suppose that Class is fully plural in these cases, bearing [−singular +augmented +empathic]. The treatment above requires only one change, that impoverishment affect [+singular +augmented], rather than just [+singular].) The effect of [+empathic] on Number strikes me as genuinely mysterious, and, at this stage, I leave the matter open. 4.6.

Conclusion

Agreement∼suppletion mismatches are a challenge for any theory of number. Nonetheless, I believe that the theory of Chapter 3 meets this challenge more than satisfactorily. It distributes number features throughout three locations, Class, Number and D, and claims that each can diverge from the others in

CHAPTER 4

150

systematic ways. Independent theories of agreement and of suppletion ensure that agreement is sensitive to the content of D and suppletion, to Number or Class. This correctly predicts nearly all mismatches. Of the remainder, pluralia tantum nouns with [±augmented]-sensitive predicates are explicable if Number is partial defective, and animate plurals are accommodable given a natural specification for Class and a somewhat surprising one for Number.10 4.7.

Appendix: Adverbs built on suppletive roots

When the two classes of suppletive predicates were introduced, it was noted that the [±singular]-sensitive ones are individual level and the [±augmented]sensitive ones, stage level. It is interesting to wonder whether this is semantically principled.11 At present, I am not prepared to argue to matter either way. However, I wish to note some properties of [±singular]-sensitive predicates, which are interesting both in themselves, and in relation to the possibility of a semantically principled connection between predication level and suppletive sensitivity. The facts concern semantic differences between adverbs built on [±singular]-sensitive roots. Kiowa forms adverbs by affixation of -de∼-te (Watkins 1984, p. 185) and all roots in (111) form adverbs: (111)

Root big.s big.d/p long.s long.d/p short.s short.d/p small.s small.d/p

Gloss a lot a lot, much a long time a long time a short time a short time a little a little

Adverb étté bˆZnde ky´ oM´MZde k´MZ´MZn´Z´Zte xé´Zde xáádóu ´te syÓnde syânde

The glosses suggest that the meanings of each s∼d/p adverb pair are identical. However, this is not so. 10 Agreement∼suppletion mismatches are not confined to Kiowa. Hale (1975) discusses the phenomenon in Navajo. Straightforward application there of the theory developed here is hampered by agreement and suppletion’s interaction with ‘conjunct movement’, a process that itself requires clarification. A number of others are detailed in Corbett 2000. 11 It would be a gross abuse of terminology to think that [±singular]-sensitivity and individual level predication are connected because individuals are inherently singular.


151

To be sure, there are cases where the members of each pair are interchangeable. (Dˆ oi ‘too’ in (113) is irrelevant; compare (123)−(124).) (112)

Ky´ oM´MZdé /k´MZ´MZn´Z´Zte bé- thá a´ long.s.adv/long.d/p.adv 2s:3i-sever.p.imp ‘Cut long [pieces]’

(113)

K´MZ´MZn´Z´Zte /dôi- kyoM Mide an ánt´ oM zánma long.d/p.adv/too-long.s.adv hab :3s:3p-talk.impf ‘She speaks (too) long’

(114)

ThaM aM te étté /bˆZnde gyamaternal grandmother.name big.s.adv/big.d/p.adv 3s:3pé´Z- k!óp seed-lay.p.pf ‘Grandma planted a lot’

Such interchangeability is independent of the referential cardinality of the nouns involved: one can crosscut s∼p ‘short’ with 1∼3 ‘tree’: (115)

´ adO bé- t!âl A´ xé´Zde tree.inv short.S.adv 2s:3i-sever.S/D.imp ‘Cut the tree short’

(116)

´ Zde á a´ xé´Zde O´ athá a´ that tree short.S.adv 2s:3s-sever.P.imp ‘Cut those trees short’

(117)

´ ZgO O´ á a´dO xá a´dóu ´te bé- t!âl that.inv tree.inv short.D/P.adv 2s:3i-sever.S/D.imp ‘Cut that tree short’

(118)

´ Zde á a´ xá a´dóu O´ ´te athá a´ that tree short.D/P.adv 2s:3s-sever.P.imp ‘Cut those trees short’

However, interchangeability is not generally the case. Consider first (119), where both forms of ‘big’ are simultaneously attested. Discussion of similar examples with two speakers made it clear that the s-form ´ ett´ e has quantificational force, ‘many’, and the d/p-form bˆZnde, secondary predicational force, describing the pies. Compare also the two examples that follow.

CHAPTER 4

152 (119)

´ Zt!áppây bˆZnde E´ étté gyat- ÓM ÓM mé´Z pie big.d/p.adv big.s.adv 1s:3p-make.pf ‘I made lots of pies big’

(120)

BˆZnde dé- tháttetOO big.d/p.adv 1s:3i-break up.fut ‘I’m going to cut it into big pieces’ ´ e Ett´ dé- tháttetOO big.s.adv 1s:3i-break up.fut ‘I’m going to cut it into many pieces’

(121)

An initial conclusion here might be that bˆZnde simply is not quantificational. However, this is incorrect. It has at least two quantificational senses. First, Watkins (personal communication) observes that, in her examples, bˆZnde is quantificational with respect to food, as in (122). The same holds true for examples in my fieldnotes. Representative examples are foodoriented (123), where bˆZnde is acceptable, versus non-food-oriented (124), where it is not. (On the limits of food-orientation, see (114).) (122)

XêMMi hégÓ gyá- hágyá- ton k!Ot dôi- binde ahorse already :2s:3s-already-fat conj too-big.d/p.adv 2s:3smáágop feed.impf ‘You already have the horse fat and yet you feed it too much’ (Watkins 1984, p. 241)

(123)

Háote an gya- pÓttO? how much hab 3s:3p-eat.impf ´ e Ett´ /bˆZnde /dôiette /dôibinde big.s.adv/big.d/p.adv/too.big.s.adv/too.big.d/p.adv ‘How much does he usually eat?’ ‘Too much’

(124)

Háot an gya- sÓÓtétOO? how much hab 3s:3p-work.impf ´ e Ett´ /*bˆZnde /dôiette /*dôibinde big.s.adv/ big.d/p.adv/too.big.s.adv/ too.big.d/p.adv ‘How much does he usually eat?’ ‘Too much’

Second, bˆZnde is used for some types of event quantification. BˆZnde in


153

(125) emphasizes repeated gain of lots of money; I surmise that ´ ett´ e (126) is more stative, implying constant wealth. (125)

(126)

BˆZnde ÓlhÓM ÓM gya gya- dôu big.D/P.adv money 1s:3s-hold ‘I have a lot of money [on different occasions]’ ´ e ÓlhÓM ÓM gya gya- dôu Ett´ big.S.adv money 1s:3s-hold ‘I have a lot of money [generally]’

A similar difference arises for other pairs in (111). For instance, for ‘small’, the [−singular] adverb syˆ ande means ‘a little at a time’, but the [+singular] adverb sy´ Onde simply ‘a little’. (127)

´M´MZde syânde E éMmá a´gop gO a- sÓOM OM detOO this small.D/P.adv 3s:1s-feed.impf and 1s-anger ‘She’s serving me a little at a time and I’m getting mad’

(128)

SyÓnde éMôu small.S.adv 2s:1s:3s-pour.imp ‘Pour me a litte’

Similarly, for ‘long’, the d/p-form is associated with iterativity, the s-form with a simple event (which can only be interpreted iteratively owing to the habitual particle an). (129)

K´MZ´MZn´Z´Zte an em- sôugu long.D/P.adv hab 3s:3a-sew.impf ‘She sews with great distance between stitches’

(130)

Ky´ oM´MZde an em- sôugu long.S.adv hab 3s:3a-sew.impf ‘She sews for a long time’

The generalization to emerge from these examples is that adverbs built on [±singular]-sensitive predicates are semantically equivalent only in some cases. Where they differ, the [+singular] variant is nearer to object quantification and the [−singular] variant, to event quantification. However, the exact domains of quantification will have to be defined with care, as [+singular] adverbs sometimes do not quantify over individuals (130), nor [−singular] ones over events (114). Nonetheless, the generalizations are suggestive and

CHAPTER 4

154

may well impact on any attempt to make suppletive sensitivity a matter of predicate semantics. If such account is possible, it will provide an interesting source of insight into the meaning and uses of number features.

PRO In the more informal context of this appendix, I wish to register one final speculation: examples such as (115)−(118) may indicate that PRO can have any number specification in Kiowa (Norvin Richards, p.c.). If secondary predicates, then they have a more articulated internal structure than simple adverbs. Now, Watkins (1984, p. 203) observes that adverbial ‘-dé looks suspiciously like nominal dé’. As this is also the suffix for relative clauses (Section 3.2.4) adverbs might be highly reduced, agreementless relative clauses (cf., the same suggestion for demonstratives in Section 3.2.4). Their meaning would then be something like ‘being big’, and, being agreementless, they would license only PRO. Interchangeability of [±singular]sensitive predicates would then indicate that PRO can be specified with either value of the feature. Verb incorporation provides similar evidence for [±augmented]-sensitive predicates. Recall the examples of interchangeability and crosscutting from p. 30 (the slightly different translations for (131)−(132) and (133)−(134) are artifacts of the elicitation sessions): (131)

(132)

HÔndé aólx!ep? what 2s:3s-drop.S/D-set.S/D.pf ‘What did you drop?’ ´ e gyat- ólEtt´ k!op? much 1s:3p-drop.S/D-set.P.pf ‘I dropped a lot’

(133)

p!él- x!ep drop.P-set.S/D ‘knock off [e.g.: from a table]’

(134)

p!él- k!op drop.P-set.P ‘knock off [e.g.: from a table]’


155

The incorporated predicates show the same indifference to number as the adverbials in (115)−(118). Again, if incorporation is analogous to an infinitival or manner complement in English, then they contain PRO, which is, again, apparently arbitrarily variable in number. Such facts lie well beyond the scope of this study and so I merely note them without offering further analysis.


Chapter 5 The Agreement Prefix This chapter addresses one of the central problems of Kiowa (and Tanoan) linguistics—the structure and content of the verbal agreement prefix—in terms of the theory developed in the preceding chapters. The classic problem is simply put. In (1), the form of the agreement prefix depends on three arguments: the third person animate plural sender, the third person animate plural recipient, and the third person singular sendee. The reality of this dependence can be shown by changing any of the arguments, say, to third person dual, and observing that the prefix covaries with each change (2)−(4). (‘∗’ indicates that the prefix lowers the following verb’s tone.) With the agreement prefixes encoding information about three participants in as little as a single vowel, the question is how so much meaning gets into so little sound. (1)

(2)

´ A∗tot 3a:3a:3s-send.pf ‘They sent him to them (people)’ ´M ∗E tot 3D:3a:3s-send.pf ‘They two sent him to them (people)’

(3)

Métót 3a:3D:3s-send.pf ‘They sent him to them two’

(4)

Ettót 3a:3a:3D-send.pf 157

CHAPTER 5

158 ‘They sent them two to them (people)’

Related to this is the size of the inventory of agreement prefixes. Given, first, that Kiowa distinguishes thirteen person-number combinations (four persons, three numbers, inverse, animate plural), second, that these may be external arguments, indirect objects or direct objects, and, third, that verbs may be ditransitive, transitive, intransitive, and intransitive with dative, the total number of agreement combinations is naively expected to be 2, 548. (5)

Total number of argument combinations = 3n=1 total number of n-argument combinations n) = 3n=1 (3+(−1) × 13n 2 = 2, 548

Clearly, some of these 2, 548 are excluded by person-case restrictions (*‘I will bring you to him’, Adger and Harbour 2007) and binding conditions (*1s:1d). However, these considerations are insufficient to bring the number of prefixes down to the mere 160 or so that actually exist. (The exact number depends on how one counts certain homophones). So, the question is how so few agreement prefixes encode so many argument combinations. These questions are relevant to the current investigation because they permit a justification of the feature inventory and mechanisms that value D argued for in Chapter 3. That is, if agreement replicates D, then all properties of the agreement prefix must be explicable in terms of the feature content of D argued for above: [±singular], [±augmented], and person/empathy features. Structure of the analysis Answering to these questions consists in providing a syntax-to-phonology mapping that transforms bundles of syntactic features into the phonologically familiar agreement prefixes. That is, we must specify the mechanisms that, in (1) for instance, produce ´ a∗ from the feature specification of 3a:3a:3s. ⎫ ⎧ Dir. Obj. ⎪ Agent ⎤ ⎡ Ind. Obj. ⎤ ⎪ ⎪ ⎪ ⎡ ⎪ ⎪ ⎪ ⎪ −singular ⎨ −singular ⎬ ⎥ ⎥ ⎢ ⎢ +singular +singular −→ ´ a∗ (6) ⎥ ⎢+singular ⎥ ⎢ ⎪ ⎪ ⎪ ⎣−augmented⎦ ⎣−augmented⎦ −augmented ⎪ ⎪ ⎪ ⎪ ⎪ ⎭ ⎩ +empathic +empathic

THE AGREEMENT PREFIX

159

I regard an explanation of this sort as consisting in specifying: (7)

a. b. c. d.

A structure in which the agreement features of each argument are located. A list of morphological processes that affect this structure and the features it contains. A list of correspondences between feature bundles and phonological strings (vocabulary items). A list of phonological processes that yield the surface forms.

My analysis builds very much on past ones: Merrifield (1959b), Trager (1960), Watkins (1984) and later manuscript revisions, Takahashi (1984), Harbour (2003a). Watkins’ inventory of prefixes is assumed (differences from others’ are minor), but their distribution is slightly altered to reflect recent discoveries. A full analysis of the prefix system proceeds in several stages. First, the number of explicanda is reduced by appeal to morphological operations. On the one hand, these explain why the inventory of prefixes is substantially smaller than the number of possible argument combinations. On the other, they allow one to exclude from active analysis prefixes that are phonologically predictable from others. Second, the prefixes are decomposed into segments that are correlated with particular arguments or sets of arguments. Finally, a list of correlations between phonological strings and morphosyntactic features (vocabulary items) is given. This accounts for the segmentation and the regular relationship between sets of prefixes. Detailed though it is, this chapter is still preliminary to a full treatment: I leave for future work the synthesis of all exponents below into a single vocabulary list and explicit derivation of the entire inventory (see also Appendix 5.6). Such delimitation is reasonable within the confines of the current investigation as the Kiowa prefix system is at least as complicated as Catalan’s clitic clusters, which have been the topics of books in their own right. The crucial point below is that all generalizations are best stated in terms of the features proposed above, not in terms of the traditional categories ‘singular’, ‘dual’, and so on. So, the analysis, though preliminary, supports the theory of the preceding chapters. Structure of chapter The analysis below is divided as follows.

CHAPTER 5

160

Section 5.1 presents the full inventory of prefixes and explains the slightly modified prefix notation used throughout the chapter. In addition, it summarizes various theoretical assumptions, such as prefix structure, and empirical generalizations, such as phonological alternations. Section 5.2 reduces the number of prefixes to be derived by appeal to morphological rules. First, by deletion of whole agreement nodes, the section shows why the inventory of prefixes is significantly smaller than number of possible argument combinations. Second, by deletion of features (rather than of entire nodes), it shows how various syncretisms arise. Section 5.3 focuses on sets of prefixes that are phonologically predictable from others. These are important in two respects. Methodologically, predictable prefixes can be excluded form active consideration. Theoretically, they show that combinations of morphosyntactic features from Chapter 3 have constant phonological realizations. Together, Sections 5.2−5.3 reduce the number of prefixes to be derived from 160 to 64. Section 5.4 presents the segmentation of the prefixes, concentrating on syntactically natural classes, such as ditransitive prefixes, transitive prefixes with non-singular agents, with singular agents, and so on. The methodological and theoretical points of the previous paragraph apply again here. The broader view: Morphosemantic number Micromorphology is not to everyone’s taste. However, it must be emphasized that this chapter does more than decompose and segment. It completes the architectural argument which opened Chapter 1: that the theory of Universal Grammar requires a unified morphosemantic theory of number, rather than disjoint morphological and semantic theories. It achieves this by solving a classic morphological problem in terms of features that originate in the semantic treatment of noun classification. It shows, therefore, that the semantic primitives of such core oppositions as count∼mass, collective∼noncollective, granular∼non-granular, and the morphological primitives of exponence and complex syncretisms are one and the same. 5.1.

Preliminaries

Before proceeding to the analysis, I give a full list of Kiowa’s prefixes and outline some theoretical assumptions.

THE AGREEMENT PREFIX (8)

161

Kiowa Agreement Prefixes x :y :

1s:(3a:) 1ex:(3a:) 1in:(3a:) 2s:(3a:) 2d:(3a:) 2i:(3a:) 3s:(3a:) 3d:(3a:) 3i:(3a:) 3a:(3a:) ∅/2s/3s:1s: 2d:1s: 2i:1s: 3d:1s: 3i:1s: 3a:1s: any:1d/p: ∅/1s:2s: other :2s: any:2d: any:2i: 1s:3s: 1ex:3s: 1in:3s: ∅/2s/3s:3s any:3d: any:3i: 2d:3s: 2i:3s: 3d:3s: 3i:3s: 3a:3s:

∅ a e ba em ma ba ∅ eM e á éM mâa∗ bâa∗ êM Mi∗ ê i∗ âa∗ dÓ em gO mÓ bÓ

3s gya é∗ bá∗ a má∗ bá∗ ∅ éM∗ é∗ á∗ éM mâa∗ bâa∗ êM Mi∗ ê i∗ âa∗ dÓ gyá gÓ mÓ bÓ gyá êi∗ bâa∗ á mé bé mâa∗ bâa∗ êMMi∗ êi∗ âa∗

z 3d nen et bet men mén bet eM én et et né ménêi∗ bédêi∗ éM nêi∗ édêi∗ dêi∗ dét nén dét mén bét nén édêi∗ bédêi∗ én mén bét ménêi∗ bédêi∗ éMnêi∗ édêi∗ dêi∗

3p gyat ét∗ bát∗ bat mán∗ bát∗ gya én∗ ét∗ gyá∗ yáM mánˆZi∗ bágˆZi∗ éM nˆZi∗ é gˆZi∗ gyâa∗ gyát yán gyát mán bát yán égˆZi∗ bágˆZi∗ án mén bét mánˆZi∗ bágˆZi∗ éMnˆZi∗ égˆZi∗ gyâa∗

3i dé ét bét bé mén∗ bét é én ét et nÓ mÓnÔO∗ bÓdÔO∗ éM nÔO∗ é dÔO∗ dÔO∗ dÓt gÓ gÓt mÓn bÓt gÓ édÔO∗ bÓdÔO∗ Ó mén bét mÓnÔO∗ bÓdÔO∗ éMnÔO∗ édÔO∗ dÔO∗

3a de ét bé be mé bé em én ét ém né d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i

CHAPTER 5

162

The table requires the following comments: (a) Cells representing impossible agreement combinations are blank. (b) ‘3a’ represents animate plural / reflexive agreement discussed in Sections 2.5 and 4.5.2. (c) ‘any’ includes ∅, unless z =∅—the argument combination ∅:x:∅ is impossible in Kiowa. ‘any’ is also constrained by binding theory; for instance, in ‘any:2d:z ’, ‘any’ cannot be second person. (d) Agreement prefixes for transitive sentences with first / second person objects are represented as agent:object:∅. E.g.: 2i:1s:∅ would be used in ‘You all (2i) saw me (1s)’. The reason for this notation is that first / second person objects behave morphologically as indirect objects. See Adger and Harbour (2007) for discussion. (e) Intransitive agreement is represented as argument:∅. E.g., 1s:∅ would be used in ‘I arrived’. (Note that these cannot be termed either ‘agents’ or ‘external arguments’. Consequently, they are labelled simply ‘subjects’ in (8). Note also that if the z -agreement is ∅ then a must be absent in such prefixes as 1s:(3a:)z. I.e., a is the prefix only for 1s:∅, not for 1s:3a:∅, which is, in any event, an impossible argument combination in Kiowa.) (f) Third plural inanimate intransitive agreement, 3p:∅, absent from the table, is gya. Throughout this chapter, I will refer to prefix positions by letter: x, y, z.

(9)

x :y:z

x :z :y:z x :∅

x y z x z y z x

= = = = = = = =

subject of ditransitive indirect object of ditransitive direct object of ditransitive subject of transitive direct object of transitive possessor possessee subject of unaccusative

For example, in (2), we would write x =3d, y=3a, z =3s. Referring to prefix positions by letter has two advantages. First, it avoids cumbersome phrases like ‘subject of (di)transitive’. (Compare the readability of, for instance, ‘if x =3d’ with ‘if the subject of a (di)transitive is 3d’.) Second, it permits us to abstract away from the sometimes complicated syntactic and featural reality behind the prefix positions. (See Adger and Harbour 2007 for analysis.)

THE AGREEMENT PREFIX 5.1.1.

163

Theoretical assumptions

Prefix structure Following Harbour (2003a), I assume that the prefix contains as many heads as there are agreeing arguments (i.e., one, two, or three) and that these heads form a cluster dependent on the verb. (The prefix forms a phonological domain separate from the verb.) (10)

V Prefix x

y

V z

Consequently, the feature structure of the prefix in (1) is: (11)

Prefix x ⎤ −singular ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented⎦ +empathic ⎡

y ⎤ −singular ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented⎦ +empathic ⎡

z


The challenge posed by Kiowa agreement prefixes can now be clarified. It is to explain how (11) yields ´ a∗, or how general instantiations of (10) yield (8). The answer offered relies in part on the morphological and phonological rules that follow. Phonology The following phonological processes, informally described, are assumed (Harrington 1928, Merrifield 1959b, Watkins 1984; see Harbour 2002 on tone):1 1

I abstract away from rule ordering. Indeed, there is evidence of cyclic application, in the shape of an ordering paradox. On the one hand, for 2i:1/3s:3p, Nasalization must precede Dental-velar switching to derive m´ anˆZi from /b+ia+d+ia+[+nasal]/; the reverse ordering yields *m´ aNˆZi→*m´ aˆZi. On the other hand, for ∅/1s:2s:3d, Dental-velar switching must precede Nasalization to derive n´ en from /g+e+d+[+nasal]/; the reverse ordering yields *N´ en→*´ en. The new cycle might be associated with (re)syllabification.

CHAPTER 5

164 (12)

Dental-velar switching g / k(!/h) → d / t(!/h) / d / t(!/h) → g / k(!/h) /

i/y e

(13)

Glide insertion a → ya / g/k(!/h)

(14)

Glide formation i → y / g/k(!/h)

(15)

Nasalization [+nasal] spreads to all phonemes of the prefix

(16)

Engma-deletion N → ∅ (Note: engma is not a phoneme of the language.)

(17)

(y)i∼ya alternation (y)i ↔ ya (Note: the conditions for this alternation are not entirely understood.)

(18)

Final devoicing d/b → t/p / [σ V ] (Note: there are no syllable-final velars, except by assimilation.)

(19)

V Vowels in hiatus V → ∅ / (Note: the proper statement of the rule does not affect diphthongs.)

a

Morphology The following morphological processes are assumed (Bonet 1991, Noyer 1998, Harbour 2003a; see the last two for discussion of how a morphological theory with these operations succeeds in restricting the class of possible grammars). Feature deletion (impoverishment) removes features from a terminal node. Such rules, which leave the node intact, are of the form: [αF] → ∅. Applied to [αF βG], this yields [βG]. Node deletion removes a terminal node and all features located at it. Applying | → ∅ to yields [αF] βG γH αF βG γH

Feature insertion inserts the (contextually) unmarked value of a feature. Such rules are of the form ∅ → [αF]. Taking minus to be the unmarked value of [±singular] in the context of [+augmented], the rule ∅ → [−singular] would apply to [+augmented] to yield [−singular +augmented].

THE AGREEMENT PREFIX 5.2.

165

Reduction of explicanda

In this section, morphological operations are appealed to to reduce the argument combinations and prefixes that must be accounted for. Specifically, deletion of syntactic features or of whole nodes neutralizes differences between argument combinations, inducing many-to-one correspondences between argument combinations and prefixes. A variety of such deletions succeed in reducing the number of prefixes to be derived from 160 to 116. 5.2.1.

‘any’

What is the featural reality behind ‘any’ ? Concretely, consider any:2d:3s. There are no restrictions on the external argument beyond binding theory (e.g., *2s:2d:3s). So, it could be zero, as with an experiencer predicate (‘You two know him’, ∅:2d:3s): y ⎤ −author ⎢+hearer ⎥ ⎢ ⎥ ⎣−singular ⎦ −augmented ⎡

z +singular −augmented

Or the agent could be first person exclusive, any number, or third person, any number: x ⎤ (+author) ⎢(−hearer) ⎥ ⎢ ⎥ ⎣ ±singular ⎦ ±augmented ⎡

y ⎤ −author ⎢+hearer ⎥ ⎢ ⎥ ⎣−singular ⎦ −augmented ⎡

z


How are these seven argument combinations realized as a single form? I suggest that the two structures are rendered identical by deletion of the x node. When vocabulary insertion occurs, this identity results in syncretism. In the current case, deletion is: (20)

| x

→ ∅ /

[y 2d ] [z 3s ]

CHAPTER 5

166

However, deletion of the x node should apply not just when z =3s, but also for z =3d, z =3p, and so on. Furthermore, it should apply not just for y=2d, but for y=1i, y=2i, y=3d/i too. These correspond to the natural class defined by [−singular]. So, we can generalize (20) to: (21)

| x

→ ∅ /

[y −singular]

Note a welcome consequence of (21). A cursory glance at the top part of (8), specifically, at the simple transitive part, shows that the featural composition of the x node has a phonological effect on the agreement prefix—scanning down any column, i.e., keeping the direct object constant, we find that the cells vary as the external argument changes. In the ‘any’ cases, we require a means of preventing the external argument from having any phonological effect on the prefix. That is, we must derive their syncretism with absence of an external argument, ∅. Deleting the x node achieves just that. 5.2.2.

More x deletion

The syncretism ∅:2s:z ≡ 1s:2s:z is also captured by node deletion: | x

(22)

⎡ ⎤ +author ⎢−hearer ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented

y ⎤ −author ⎢+hearer ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented ⎡

→ ∅ /

For ease of illustration, we can paraphrase this as 1s → ∅ / of which is:

2s, the effect

(23) 1s

2s

z

→

2s

z

What then is the featural reality behind ‘other ’ in other :2s:z ? Again, we find the prefixes phonologically invariant despite variation of the external argument. So, x features must be deleted. However, deletion of the entire node incorrectly forces syncretism with ∅:2s:z . I follow Harbour (2003a) in regarding the relevant operation as deletion of features but not of the x node itself. Consequently, (24) applies to (25) to yield (26).


(24)

167 y ⎤ −author ⎢+hearer ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented ⎡

(+author) (−hearer) ± singular ± augmented

→ ∅ / [

]

(25) x ⎤ (+author) ⎢(−hearer) ⎥ ⎢ ⎥ ⎣ ±singular ⎦ ±augmented ⎡


z


(26)

x


z


This empty node is automatically targeted for insertion of the unmarked feature [−singular]. As this feature conditions allomorphy in z -exponence (59), it prevents syncretism of (26)/(27) with ∅:2s:z . (27)

x

−singular

5.2.3.


z ±singular ±augmented

Animate/reflexive deletion

Recall (Sections 2.5 and 4.5.2) that transitive agreement prefixes are systematically ambiguous; that is, x :3a:z ≡ x :z. Gya, for instance, means 1s:3s and 1s:3a:3s, so that one and the same verb form is ambiguous between ‘I killed it’, ‘I killed it for them’, and ‘I killed it for myself’. Formally:

CHAPTER 5

168

⎡

(28)

5.2.4.

| y

⎤ −singular ⎢+singular ⎥ ⎢ ⎥ ⎣−augmented ⎦ +empathetic

→ ∅

Impoverishment: Person deletion

In the treatment of first person syncretisms and the sii class, two rules of person deletion were introduced. These accounted for the syncretisms of first exclusive with inverse and first inclusive with second plural. In current notation, they are: (29)

[±author] → ∅ / [x

(30)

[±hearer] → ∅ / [y,z +author

] ]

5.2.5. Impoverishment: More person deletion The effect of (29), syncretism of first person with third person, arises elsewhere and again requires person deletion, though under a different conditioning environment: specifically, x :1s:z ≡ x :3s:z, for non-zero x , z :2 2 The discrepancy with respect to the ∅ column is as explained following (8). 2d:1s:∅, for example, is used in, say, ‘You two saw me’, where 1s is a direct object. It is represented in the prefix notation as an indirect object, ‘:1s:’, because first / second direct objects share certain syntactic, and, consequently, morphological properties with indirect objects (Adger and Harbour 2007). Third person direct objects do not display these syntactic properties and third person z -agreement is distinct from third person y-agreement. As a result, the agreement prefix for ‘You two saw her’ is 2d:3s, not *2d:3s:∅, which is not a possible argument combination in Kiowa. Thus, it is not that the correlation ‘x :1s:z = x :3s:z, for all non-singular x ’ breaks down for z =∅. Rather, when one bears in mind the syntactic reality behind this generally useful morphological notation, one realizes that the empty cells are impossible, and so trivially exempt from such correlations.

THE AGREEMENT PREFIX (31)

x :y : 2d:1s: 2d:3s: 2i:1s: 2i:3s: 3d:1s: 3d:3s: i:1s: i:3s: a:1s: a:3s:

169 z

∅ mâa∗

bâa∗

êMMi∗ êi∗

âa∗

3s mâa∗ mâa∗ bâa∗ bâa∗ êMMi∗ êMMi∗ êi∗ êi∗ âa∗ âa∗

3d ménêi∗ ménêi∗ bédêi∗ bédêi∗ éM nêi∗ éM nêi∗ édêi∗ édêi∗ dêi∗ dêi∗

3p mánˆZi∗ mánˆZi∗ bágˆZi∗ bágˆZi∗ éM nˆZi∗ éM nˆZi∗ égˆZi∗ égˆZi∗ gyâa∗ gyâa∗

3i mÓnÔO∗ mÓnÔO∗ bÓdÔO∗ bÓdÔO∗ éM nÔO∗ éM nÔO∗ édÔO∗ édÔO∗ dÔO∗ dÔO∗

3a d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i

Now, the cause of this correlation cannot be that y=1s and y=3s are realized by the same vocabulary items. If they were, then, for instance, ∅:1s:3s (´ eˇ ) and ∅:3s:3s (´ a) would be identical. Instead, I suggest that first person singular [y +author −hearer +singular −augmented] is rendered identical to third person singular [y +singular −augmented] by deletion of the person features [+author −hearer]. In order to prevent syncretism of ∅/2s/3s:1s:z and ∅/2s/3s:3s:z , it is necessary to contextualize this deletion to non-singular x .3 3

The deletion (32) competes with (20): in 2i:1p:3s, for example, either could apply. The correct results are achieved if (20) precedes (32). It is not clear to me that this ordering is intrinsically given, for instance, by P¯ an.ini’s Principle. And, though it is possible to avoid this problem by recontextualizing (32), to [x −singular] [y +singular] say, this is not overly satisfying. A more exciting alternative, arising from discussion with Elena Anagnostopoulou—I do not pursue it here as it lies well beyond the scope of this inquiry— departs from the observation that external argument and indirect object in Kiowa almost never both agree for person. Indeed, if one is [−singular], then it is never the case that both agree fully. In this chapter, this is treated as a purely morphological fact: whenever such argument combinations enter the morphology, some of their features are deleted. However, it is possible that the syntax itself never gives rise to structures of the relevant form, so that these morphological rules are superfluous, the true explanation stemming from a deeper syntactic fact. Interestingly, similar restrictions hold in two other rich agreement systems that I am familiar with, namely, Yimas (Foley 1991) and Georgian. To the extent that Yimas differs from Kiowa in this respect, it does so by use of portmanteau 1+2 affixes, the syntactic nature of which is not entirely clear. In all three languages, there is a case to be made that, when the (in)direct object is second person and the external argument is first, second person, rather than first, agrees. See Béjar (2007) and references therein for discussion, and Heath (1998) for an overview of several other cases.

CHAPTER 5

170 (32)

+author → ∅ / [x −singular] [y −hearer

]

With this in hand, the list of prefixes to be derived diminishes, as we can conflate two sets of prefixes in (31) under the label x :1/3s:z . Combining all of the above, we have the reduced table (33). Summary (33)


1s:(a:) 2s:(a:) 2d:(a:) 2i:(a:) 3s:(a:) 3d:(a:) i:(a:) a:(a:) ∅/2s/3s:1s: 2d:1/3s: 2i:1/3s: 3d:1/3s: i:1/3s: a:1/3s: any:1d/p: ∅/1s:2s: other :2s: any:2d: any:2i: 1s:3s: ∅/2s/3s:3s any:3d: any:i:

∅ a em ma ba ∅ eM e á éM mâa∗ bâa∗ êMMi∗ êi∗ âa∗ dÓ em gO mÓ bÓ

3s gya a má∗ bá∗ ∅ éM∗ é∗ á∗ éM mâa∗ bâa∗ êMMi∗ êi∗ âa∗ dÓ gyá gÓ mÓ bÓ gyá á mé bé

z 3d nen men mén bet eM én et et né ménêi∗ bédêi∗ éM nêi∗ édêi∗ dêi∗ dét nén dét mén bét nén én mén bét

Simplification 1

3p gyat bat mán∗ bát∗ gya én∗ ét∗ gyá∗ yáM mánˆZi∗ bágˆZi∗ éM nˆZi∗ égˆZi∗ gyâa∗ gyát yán gyát mán bát yán án mén bét

3i dé bé mén∗ bét é én ét et nÓ mÓnÔO∗ bÓdÔO∗ éM nÔO∗ édÔO∗ dÔO∗ dÓt gÓ gÓt mÓn bÓt gÓ Ó mén bét

3a de be mé bé em én ét ém né d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i d∼i


171

The effect of the morphological deletion and impoverishment operations simultaneously derives the desired syncretisms and helps to explain why the number of prefixes is significantly smaller than the number of possible argument combinations. Counting 1d and 1p as separate argument combinations, as they were when 2, 548 was calculated, the reader may easily verify that several hundred morphological distinctions have been neutralized. (I omit an exact count as the figure of 2, 548 was itself only approximate, neglecting person-case and binding-theoretic restrictions.) More specifically, we have reduced the number of prefixes under active consideration from 160 to 116. 5.3.

Segmentation I: Subregularities

By considering identity relations between prefixes corresponding to different argument combinations, it has proved possible to reduce the total number of different forms that must be derived. The discovery of identity relations is not, however, the only way to reduce the explanatory burden. Another is to find correspondences between classes of prefixes. For instance, suppose that the prefix x 1 :z is ϕ1 and that the prefix x 2 :z is ϕ1 +ϕ2 . That is, to say x 2 :z , you say x 1 :z (ϕ1 ) plus something extra (ϕ2 ). Then we want ϕ1 to realize features common to x 1 and x 2 (to derive the overlap), and we want ϕ2 to realize features of x 2 that x 1 lacks (to derive the additional element). Such correlations induce requirements on the vocabulary list and are generally phrased as such below. By the end of this section, the number of prefixes to be derived will have been reduced from 116 to 66, at which point, decomposition of individual prefixes can proceed quite easily. Duals Prefixes where x /y=3d are predictable from prefixes where x /y=3i, by nasalization of the entire prefix. Informally: 3d ≡ 3i+nasalization. For instance, consider the ditransitive prefixes any:3d:z and any:3i:z . Every cell in the top row is the nasalization of the cell directly below it: (34)

x :y : any:3d: any:3i:

3s mé bé

3d mén bét

z 3p mén bét

3i mén bét

3a mén bét

CHAPTER 5

172

For the prefixes 3d:1/3s:z and 3i:1/3s:z , the same holds (by orthographic convention, nasality is not marked on vowels cosyllabic with nasal stops). x :y :

(35)

z ∅ êMMi∗ êi∗

3d:1s: 3i:1s:

3s êMMi∗ êi∗

3d éMnêi∗ édêi∗

3p éMnˆZi∗ égˆZi∗

3i éMnÔO∗ édÔO∗

3a éMnêi∗ édêi∗

Only z =3p requires further comment. Straightforward nasalization yields *´ eˇ N˙ıi, not ´ eˇ nˆZi. However, if the underlying form is /é-d-ˆZi/, then ´ eˇ nˆZi arises by nasalization and ´ egˆZi by Dental-velar switching (Section 5.1.1). For transitives too, the correlation holds. The only difficulty here is with z =3d. Nasalization of 3i:3d, et, yields en, rather than 3d:3d ´ en, which has high tone. We return to this below (p. 181). x:

(36)

3d: 3i:

z ∅ eM e

3s éM∗ é∗

3d én et

3p én∗ ét∗

3i én ét

3a én ét

The correspondence extends to the second person dual. For example: x :y :

(37)

any:2d: any:2i:

z ∅ mÓ bÓ

3s mÓ bÓ

3d mén bét

3p mán bát

3i mÓn bÓt

3a mén bét

The correlation 2d ≡ 2i+nasalization faces two exceptions, again, from transitive prefixes, and again, with respect to tone. 2d:3d is m´ en, but the result of nasalizing 2i:3d, bet, is men. And 2d:3i is m´ en∗, but the result of nasalizing 2i:3i, b´ et, is m´ en. These results prompt (38), subject to the provisos in (39): If [x,y i ] ⇔ ϕ, then [x,y d ] ⇔ ϕ [x,y d ] ⇔ [+nasal].

(38)

a. b.

(39)

Tone 2d:3d has high tone although 2i:3d has not. 2d:3i is ∗ although 2i:3i is not. 3d:3d has high tone although 3i:3d has not.


173

We will see that much the same realizations hold for z =d (59), but that there is an extra allomorphy condition. First person ‘duals’ are excluded from (38), as they trigger inverse agreement, being sii. x:1/3S:z We consider z =∅ and z =∅ separately, as x :1s:∅ is predictable from intransitive x :∅, and x :1/3s:z, z =∅, from x :z. Recall that, by (32), x is non-singular in these prefixes. Prefixes of the form x :1s:∅ are identical to x :1/3s:3s.4 (40)

x :y : 2i:1/3s: 1/3i:1/3s: 3a:1/3s:

z ∅ bâa∗ êi∗ âa∗

3s bâa∗ êi∗ âa∗

There are two ways to capture this identity relation. Either z =3s is realized as zero in the context of y=1/3s; or whatever realizes z =3s in the context of y=1/3s is non-zero but phonologically redundant. Anticipating that tonal evidence from other :2s:∅ favors the latter solution, we can write: (41)

If [z 3s ] ⇔ ϕ / [y 1/3s ] , then x /y ⇔ ϕ. (If y=1/3s, then z =3s’s realization is redundant, given the realizations of x and y.)

For z =∅, x :1/3s:z is predictable from x :z. To illustrate this, consider 3i:1/3s:z and 3i:z . (42)

x :y :

3i: 3i:1/3s:

z

3s é∗ êi∗

3d ét édêi∗

3p ét∗ égˆZi∗

3i ét édÔO∗

The second row can be derived from first as follows. Lengthening of the 3i:3s vowel into a falling tone plus ∗ yields 3i:1/3s:3s.5 Addition of ˆ ei to 4

For some speakers, x :3s prefixes are used here. This follows immediately from (32). The specification ‘into a falling tone’ is redundant, given that 3i:3s lowers tones that follow it. A falling tone is simply a vowel that lowers itself in the middle. See Harbour 5

CHAPTER 5

174

3i:3d yields 3i:1/3s:3d. (Note also the difference in voicing of the alveolar. I follow Watkins in assuming that the same vocabulary item is involved in both and that it is /d/, which sometimes nasalizes to n and at other times devoices to t.) Addition of ı˙i to 3i:3p yields 3i:1/3s:3p, once again voicing the alveolar, and switching it to velar before i/y. Lastly, addition of ˆ OO to 3i:3i and voicing of the alveolar yields 3i:1/3s:3i. Now, the identity of the added vowels, ˆ ei, ˆZi and ˆ OO, is not at all mysterious. E is a standard component of prefixes with z =3d, as can be easily verified by scanning down the 3d column of (8). Similarly, i is a standard component prefixes with z =3p, generally in its ya∼a allophones. And O is a component of prefixes with z =3i, when there is y-agreement. Consequently: (43)

ˆ If z ⇔ -V-, then [y 1/3s ] ⇔ -VV (If y=1/3s, then the vowel that realizes z occurs twice.)

(Nasalization yields 3d:1/3s:z from 3i:1/3s:z .) The correspondences just discussed apply, mutatis mutandis, to 2i:1/3s:z and 2i:z , with one minor difference. (44)

x :y : 2i: 2i:1/3s:

3s bá∗ bâa∗

z 3d 3p 3i bet bát∗ bét bédêi∗ bágˆZi∗ bÓdÔO∗

The z =3s form follows from lengthening of the vowel, and the 3d, 3p and 3i forms from respective addition of ˆ ei, ˆZi and ˆ OO. The minor difference concerns the first vowel in 2i:1/3s:3i, which is ´ O, not ´ e, as in 2i:3i. This difference is discussed below.6 (Again, the 2d forms are predictable by nasalization.) Finally, consider: (2002) for detailed discussion of such cases. 6 In Mr Bointy’s Mount Scott dialect (possibly shared by some Carnegie speakers), there is no difference in the vowel, both forms beginning b´ e. . . . An additional difference between his speech and that of Dr McKenzie, Watkins’ primary consultant, is that the alveolar does neither voices intervocalically nor velarizes, as in Dr McKenzie’s b´ agˆZi. For Mr Bointy, this was b´ at.ˆZi, where the ‘.’ indicates the syllable boundary. Mrs White Horse Taylor, who appears in several respects to speak as Dr McKenzie did, says b´ a.gˆZi. For convenience, I give the relevant variants below:


175

x :y :

(45)

z 3s a´∗ âa∗

3a: 3a:1/3s:

3d et dêi∗

3p gyá∗ gyâa∗

3i et dÔO∗

Again, for the z =3s form, the vowel lengthens into a falling tone. For the other three, there is once again addition of the expected long falling vowel, though for z =3p, it is in its ya allophone. However, the 3d and 3i forms present a minor wrinkle. Straightforward addition of the expected vowels yields the wrong forms, respectively, *´ edˆ ei and *´ edˆ OO. In these cases, the initial vowel is not inserted, or, if inserted, does not survive the phonology. Summarizing these patterns, we have: (46)

a. b.

If x :z ⇔ ϕ, then x :1/3s:z ⇔ ϕ ˆ If x :z ⇔ V, then x :1/3s:z ⇔ VV

z =∅ z =∅

Also, the following minor deviations should be explained: (47)

a. b.

Absence of initial vowels from 3a:1/3s:z prefixes. Change from e to O of the medial vowel in 2i:1/3s:3i.

Third singular agent There is an intriguing correlation between transitive 3s:x prefixes and intransitive x :∅ prefixes.7 (48)

:∅ 3s:

3s ∅ ∅

3d eM eM

3p gya gya

3i e é

3a a´ em

For z =3s/3d/3p, x =3s makes no difference at all to the prefix. Given this, it is tempting to imagine that [x 3s ] is simply deleted. Certainly, this would x :y :

(i)

2i:1/3s: 1/3i:1/3s: 7

3d bét.êi∗ ét.êi∗

z 3p bát.ˆZi∗ ét.ˆZi∗

3i bét.ÔO∗ ét.ÔO∗

Just in this subsection, the same letter, x, is used as a variable over two different prefix positions. This is to emphasize that we are dealing with essentially the same feature bundle in the two positions.

CHAPTER 5

176

work well for 3s:1s:z (since 3s:1s:z ≡ ∅:1s:z ) and for 3s:3s:z (3s:3s:z ≡ ∅:3s:z ). However, it would not work for 3s:2s:z , which is non-syncretic with ∅:2s:z . Nor would it work for z =3i/3a in (48): intransitive i-agreement, 3i:∅, is e, but 3s:3i, ´ e, has high tone (however, observe that, on p. 185, this high tone attributed, for independent reasons, to allomorphy of y=3i in the context of x ); and intransitive 3a-agreement, 3a:∅, is ´ a, but 3s:3a is em. I am unsure how far to pursue this resemblance. Summary By applying a series of morphological operations (Section 5.2) and then by noting systematic phonological correspondences between sets of prefixes (Section 5.3), we have reduced both the number of argument combinations to be accounted for and the number of prefixes to be explained, replacing these with a series of vocabulary items and phonological and morphological processes. This explains why the inventory of attested prefixes, at 160, is substantially smaller than the inventory of logically possible argument combinations, 2, 548, and has reduced the inventory of prefixes to be accounted for from 160 in (8) to 66: (49)


1s:(a:) 2s:(a:) 2i:(a:) 3s:(a:) i:(a:) a:(a:) ∅/2s/3s:1s: any:1d/p: ∅/1s:2s: other :2s: any:2i: 1s:3s: ∅/2s/3s:3s any:i:

Simplification 2 z

∅ a em ba ∅ e á éM dÓ em gO bÓ

3s gya a bá∗ ∅ é∗ á∗ éM dÓ gyá gÓ bÓ gyá á bé

3d nen men bet eM et et né dét nén dét bét nén én bét

3p gyat bat bát∗ gya ét∗ gyá∗ yáM gyát yán gyát bát yán án bét

3i dé bé bét é ét et nÓ dÓt gÓ gÓt bÓt gÓ Ó bét

3a de be bé em ét ém né d∼i d∼i d∼i d∼i d∼i d∼i d∼i

THE AGREEMENT PREFIX 5.4.

177

Segmentation II: Segments

We now turn to the segmentation proper of prefixes. We begin with the ditransitive prefixes, where, with tone almost exceptionlessly high, segmentation is somewhat simpler. (‘Ditransitive’ means prefixes where y =3a, as morphological rules induce x :3a:z ≡ x :z , which are transitive prefixes.) 5.4.1.

Ditransitives

We are now concerned with the following prefixes. (50)

Kiowa Ditransitive Agreement Prefixes x :y :

z

∅/2s/3s:1s: any:1d/p: ∅/1s:2s: other :2s: any:2i: 1s:3s: ∅/2s/3s:3s any:i:

∅ éM dÓ em gO bÓ

3d né dét nén dét bét nén én bét

3s éM dÓ gyá gÓ bÓ gyá á bé

3p yáM gyát yán gyát bát yán án bét

3i nÓ dÓt gÓ gÓt bÓt gÓ Ó bét

3a né d∼i d∼i d∼i d∼i d∼i d∼i d∼i

Let us first consider y=singular prefixes (from which other :2s:z is excluded on principled grounds, as explained below). Singular y We now concentrate on the following prefixes. (51)

x :y :

∅/2s/3s:1s: ∅/1s:2s: 1s:3s: ∅/2s/3s:3s:

∅ éM em

3s éM gyá gyá á

z 3d né nén nén én

3p yáM yán yán án

3i nÓ gÓ gÓ Ó

CHAPTER 5

178

Let us begin with the bottom row, ∅/2s/3s:3s:z . Observe that x =2s/3s makes no phonological contribution beyond that of x =∅. Consequently, I suggest that the phonological components of these prefixes are vocabulary items realizing (some subset of) y=3s and the relevant object, z . (52)

a. b. c. d.

´ a ⇔ [y 3s ] [z 3s ] ´ en ⇔ [y 3s ] [z 3d ] ´ an ⇔ [y 3s ] [z 3p ] ´ O ⇔ [y 3s ] [z 3i ]

Now, 1s:3s:z and ∅/1s:2s:z rows are identical (modulo em) and both are derivable from the last row of (51) by standard phonology following the addition of g- to (52): g+´ a → gy´ a by Glide insertion; g+´ en → n´ en by Dental-velar switching and Nasalization; g+´ an → y´ an by Glide insertion, Nasalization, Engma-deletion; and g+´ O → g´ O transparently. The question is whether the g- is the same in both cases. If it is, then the features that g- realizes are those common to 1s:3s:z and ∅/1s:2s:z . So: ⎧ ⎫ ⎫ ⎧ y x y ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪⎡ ⎤ ⎡ ⎤⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ −author ⎨ +author ⎬ ⎬ ⎨ ⎥ +singular ⎥ ⎢+hearer −hearer g- ⇔ ⎢ ∩ ⎥ ⎢ ⎥ ⎢ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎣+singular ⎦ −augmented ⎪ ⎣+singular ⎦⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ ⎭ ⎭ ⎩ −augmented −augmented Equivalently, g- ⇔ [y +singular −augmented ] = [y 3s ]. Given (52), this incorrectly forces the occurrence of g- in the last row of (51). Unless we wish to invoke further morphological operations, we should claim that 1s:3s:z and ∅/1s:2s:z are realized by different, accidentally homophonous, g-. Since 1s:3s:z ≡ x =1s + ∅:3s:z , I attribute one g- to x =1s. (53)

g- ⇔ [x 1s ]

And since a zero agent, being featureless, is not realized by any vocabulary items, I attribute the other g-, of ∅:2s:z and 1s:2s:z , to y=2s. (54)

g- ⇔ [y 2s ]

When we turn to other :2s:z and to 1s:(3a:)z , we will see that this accidental homophony is not theoretically problematic. On the contrary, it correctly predicts two other sets of prefixes.


179

For the top row, matters are more complicated. Unlike the other rows, ∅/2s/3s:1s:z cannot be derived by adding phoneme(s) to ∅/2s/3s:1s:z . However, several properties of these prefixes are predictable if we claim that two processes, metathesis8 and unmarked feature insertion, required elsewhere in the agreement system, also apply here. De-metathesis yields:9 (55)

Onset∼coda De-Metathesis x :y : ∅ 3s éM éM ∅/2s/3s:1s:

z 3d én

3p ań

3i Ón

The distribution of codas here is different from that in the other rows of (51) in that ´ On (z =3i) has one. However, in (57), the distribution of codas is identical to (55): codas if and only if z =3d/3p/3i. This distribution of codas is sensitive to non-singular x /y (Harbour 2003a), but, here, y is singular. The mismatch can be fixed if we invoke feature insertion. That is, we delete [+singular] in 1s and then insert what Harbour (2003a) argues to be the default value, [−singular]. This yields: (56)

[y +author −hearer −singular −augmented]

Being non-singular, the distribution of codas will now be as in (57). Moreover, a second property of ∅/2s/3s:1s:z prefixes is derived. Number in (56) is dual and so, by (38), causes the prefixes to be nasal. (Such spurious duals are not confined to the current case: recall the dual-like realization of 3a in the presence of indirect objects; Section 4.5.2.) 8 Watkins (1984) relies on metathesis to derive 3p:∅ gya from /ia+d/, for instance: Metathesis yields /dia/, Dental-velar switching yields /gia/, Glide formation yields /gya/. Like Watkins, I am unsure of the conditions that trigger metathesis. Note, interestingly, that Halle and Vaux (1998) face the same problem with respect to Latin nominal morphology. They decompose the genitive plural and dative plural endings (fifth declension), -rum and -bus, into r+um and bu+s. Undoing the effect of the Latin rhotacism rule, we have s+um and bu+s. Halle and Vaux observe: ‘The affix order [in bu+s] is the reverse of that in the Genitive Plural [s+um]; there is at present no explanation for this fact.’ One might appeal to well-formedness of syllables in this case (in preference to *sbu and *ums), though this is not clearly exportable to the Kiowa cases. 9 Abstracting away from tone, y´ a ˇ derives from /d+ia+[+nasal]/. To derive the form in (55), observe that we have /ia+d+[+nasal]/, prior to Metathesis; Nasalization yields /ian/; Vowel-in-hiatus deletion yields /an/, the desired form, modulo tone.

CHAPTER 5

180 Non-singular y We now concentrate on the following prefixes. (57)

x :y : any:1d/p: other :2s: any:2i: any:i:

∅ dÓ gO bÓ

z 3d dét dét bét bét

3s dÓ gÓ bÓ bé

3p gyát gyát bát bét

3i dÓt dÓt bÓt bét

Again, as in the cases considered above, the onsets are constant across each row: d- for any:1d/p:z (with Dental-velar switching for 3p); g- for other :2s:z (with Dental-velar switching for 3d); b- for the other two rows. Similarly, the rhymes are also nearly constant throughout the columns, with two exceptions. Working right to left, in the 3i column, the rhyme is -´ Ot (except that the vowel for any:3i:3i is e); in the 3p column, the rhyme is -´ at (except that the vowel for any:3i:3p is e again); in the 3d column, the rhyme is -´ et; in the 3s column, the rhyme is -´ O (except that the vowel for any:3i:3p is e again); in the ∅ column, the rhyme is -´ O (except that the tone for other :2s:∅ is low; any:3i:∅ is, recall, an impossible argument combination). What is the relationship between these regularities and those discussed for singular y? First, let us compare the rhymes of two groups (ignoring the more involved cases any:3i:z and ∅/2s/3s:1s:z ).

z

(58)

singular y non-singular y

3s á Ó

3d én ét

3p ań át

3i Ó Ót

There are three differences between the two rows. The 3d rhymes and the 3p rhymes are nasal∼non-nasal counterparts of one another. The 3i rhyme has a -t coda in the non-singular case, though none in the singular. Third, the 3s rhymes are just different. We will see, when we turn to the transitive prefixes, that the first two differences hold there too, except that singularity of x , rather than y, is the crucial factor. The third difference does not hold of transitive prefixes for non-interesting reasons: the vowel O is confined to y prefixes. I suggest, therefore, that all three differences depend on the value of [±singular] on y or x —that is, [±singular] the higher argument conditions


181

allomorphy of z -agreement. (59)

z -allomorphy [±singular]-conditioned a. ´ a ⇔ [z 3s ] / [+singular ] ´ O ⇔ [z 3s ] / [−singular ] b. ∅ ⇔ [z 3i ] / [y +singular ] -t ⇔ [z 3i ] / [y −singular ] c. [+nasal] ⇔ [z 3d ] / [+singular ] [−nasal] ⇔ [z 3d ] / [−singular ]

A potential problem for these allomorphy relations is that other :2s:z conditions all the [−singular] allomorphs though 2s itself is [+singular]. Nor are all values of x compatible with other themselves [−singular], as other covers 3s. However, recall that the featural reality behind other is an x node from which all features have been deleted (24). Following Noyer’s (1998) argument that Universal Grammar permits the insertion of unmarked feature values, Harbour (2003a) argues that Kiowa too permits such insertion and that, in the case of number, the unmarked feature and value inserted is [−singular]. Having deleted all but the x head itself, which I take to be a root of a ϕstructure (Chapter 3), the unmarked [−singular] is inserted, as shown above in (27). Consequently, other :2s:z patterns with others that have non-singular x /y with respect to the allomorphs of z -agreement that they condition. Consider now the two rhyme exceptions. First, e in any:3i:z . Observe, in (49), that occurrence of e irrespective of the value of z is also a characteristic of 3i:(3a:)z . Consequently, I assume that this is a property of x /y=3i. (60)

Irrespective of z , x /y=3i is realized as e.

A number of technical means are available to satisfy (60): there could be a specific vocabulary item e that realizes i and d and that deletes the vowels that realize z (via the Vowel-in-hiatus rule); e could be a default; or e could result from phonological readjustment (cf., Chomsky and Halle’s 1968 treatment of sing∼sang, tell∼told, and similar). At this point, it is not necessary to decide between these options. The second rhyme exception is the low tone of gO other :2s:∅. Interestingly, the only other low tone prefix in this part system is em 1s:2s:∅. Given (32), the only features realized by em are [y 2s ]. So, one might think low tone a systematic property of [y 2s ]. However, the intransitive 2s:∅ of ‘You

CHAPTER 5

182

arrived’ is also em. This suggests that em realizes the features common to intransitive 2s-agreement and indirect object 2s-agreement. Consequently, the low tone of em is a fact about that vocabulary item per se, not about the pair other :2s:∅ / 1s:2s:∅ generally. Thus, the low tone of gO other :2s:∅ does not appear to correlate with anything else and may represent an idiosyncracy of this argument combination. Notice, however, that (41) (in dative prefixes, no vocabulary items realize z =3s beyond those realizing x and y) extends to nearly all prefixes of the form x :1/2s/d/i:∅, i.e., those used when the direct object is first or second person, as in ‘x saw me/us/you’. (61)

x :y : 2i:1/3s: i:1/3s: a:1/3s: any:1d/p: any:2i: ∅/2s/3s:1s: 1s:2s: other :2s:

z ∅ bâa∗ êi∗ âa∗ dÓ bÓ éM em gO

3s bâa∗ êi∗ âa∗ dÓ bÓ éM gyá gÓ

The two ‘exceptions’, ruled off from the others, are 1s:2s:z and other :2s:z . The first may be disregarded for the reason given in the previous paragraph, namely, that 1s:2s:∅ is some form of default agreement and so irrelevant to generalizations concerning realizations of y. However, (41) suggests an explanation for the second exception, other :2s:∅. If high tone realizes z =3s in the context of y, then every prefix of the form x :y:3s will have high tone. And if every y except 2s has high tone, then every prefix of the form x :y:∅ will have high tone, except x :2s:∅. Now, of course, there is considerable redundancy in the determining of tone here. However, as this affects only a single form, I am content to leave the exact solution open, noting only that a solution is possible. So, despite initial appearances to the contrary, all exceptions to rhyme regularities can be captured through generalizations that emerge from more careful examination of the agreement system. Now let us consider the onsets. The onset g- for y=2s is expected given the analysis of the singular y-prefixes; the same correlation was found there.


183

The onset d- for y=1d/p is straightforwardly accounted for:10 (62)

d- ⇔ [y 1d/p ]

Finally, there is the onset b-. Given that this occurs for both 2i and 3i, it is either a default or an accidental homophone. Amongst prefixes with second person x , b- is frequent; however, it does not occur in prefixes with x =3i (or other third persons). Thus, the approximate distribution is: (63) x y

2 bb-

3i ∅b-

Given (63), some accidental homophony may be unavoidable. Either b- generally realizes second person, in which case it is too highly specified to realize y=3i; instead, a second, homophonous b- must be provided. Or it is the elsewhere form for y-agreement, in which case it has the wrong specification to realize x =2; again, a second, homophonous b- must be provided. The only homophony-free possibility is to make b- a default for x and y, with the zero in (63) representing a more highly specified vocabulary item. Surprising though this situation would be, it is not impossible. However, it should be noted that it does not obviate the need for accidental homophones, as we will need more than one zero, to deal with 3s:∅ and with 2s:3s. So, little is gained by regarding any:2i:z and any:3i:z as being realized by a single b-. (64)

b- ⇔ [y 3i ] b- ⇔ [y 2i ]

10 The identification of the underlying form as dental necessitates a slight revision to earlier statements. Specifically, the rhyme for z =3p cannot be simply ´ at. Rather, it must begin underlyingly in a segment that causes dentals to become velar. The obvious options are ia. . . and ya. . . . We can rule out ya. . . owing to words like yal ‘hopefully’, y´ a´ZpÓ ‘rope’, y´ atky´ a ‘eight’: if the underlying form were ya. . . , as it is in y´ al ‘hopefully’, et cetera, then ∅:3s:3s would be y´ an. However, if it is ia. . . , then there is no need for a specific rule to delete i word-initially: the Vowel-in-hiatus rule will do this automatically. (Clearly, then, the Vowel-in-hiatus rule must be ordered after the Glide formation. Note, though, that i also deletes before b, as in 2s:3p bat, though not because by is an impossible onset in Kiowa, witness paaby´ oi ‘brother.inv’.) (1s:3s:3p and ∅/2s/3s:2s:3p y´ an have initial y-, in contrast to ∅/2s/3s:3s:3p ´ an, owing to opaque phonological processes. Y´ an derives from g+i´ an by Glide formation, gy´ an, Nasalization, Ny´ an, and Engma-deletion, y´ an. By contrast, in ∅/2s/3s:3s:3p ´ an, there is never a consonant before i´ an, so that Vowel-in-hiatus deletion applies, yielding ´ an.)

CHAPTER 5

184 Summary

This concludes the discussion of ditransitive prefixes. The important point to observe, in relation to the goals of this chapter and its context in the investigation of morphosemantic number as a whole, is that Kiowa’s verb agreement prefixes decompose, not just into traditional categories such as ‘singular’, ‘dual’, ‘first’, ‘second’, but into more finegrained ones. For instance, observe that allomorphy relations and deletion operations are stated with respect to, e.g., [−singular], the specific features argued for in Chapter 3. We will now see that the transitive prefixes support the same conclusion, with, indeed, several of the segments identified above, identifiable there.

5.4.2.

Transitives

We are now concerned with the following prefixes. (The notation ‘(:3a):’ is dispensed with, in virtue (28).) (65)

Kiowa Transitive Agreement Prefixes x: 1s: 2s: 2i: 3s: 3i: 3a:

3s gya a bá∗ ∅ é∗ á∗

3d nen men bet eM et et

z 3p gyat bat bát∗ gya ét∗ gyá∗

3i dé bé bét é ét et

3a de be bé em ét ém

Singular x 3s:z prefixes were discussed on p. 175. Therefore, here, attention is confined to 1s:z and 2s:z . These two are identical with respect to rhymes and are strongly reminiscent of the y-rhymes of (58), which are included below:


185

(66) singular x singular y non-singular y

3s a á Ó

3d en én ét

z 3p at ań át

3i é Ó Ót

3a e

The points of difference are tone, nasality for z =3p, the vowel for z =3i, and the existence of distinct z =3a forms. However, none poses major difficulties. First, with regard to tone, observe that the singular x -rhymes have low tone throughout, except for z =3i. Here, instead of O, which one might expect, extrapolating from the ditransitive rhymes, we have ´ e. As both vowel and tone are unpredictable, they can be taken as idiosyncratic properties of the vocabulary item (as (67) applies only when there is no y, other statements about [z 3i ] must be contextualized to y, as (59) was): (67)

´ e ⇔ [z 3i ]

Second, with regard to existence of distinct z =3a forms, recall that this agreement category is subject to morphological deletion operations (Section 4.5.2). Therefore, I assume that the distinctive realization of a-agreement depends on these operations not having occurred. (68)

e ⇔ [z 3a ]

Finally, with regard to nasality, I take nasality for z =3d and non-nasality for z =3p as the norm. The former follows immediately from (38), with the latter forming the complement case. Examples are 1s:3d nen and 2s:3d men, as against 1s:3p gyat and 2s:3p bat. The total lack of nasality in the non-singular y row of (66) follows from (59). The nasality of z =3p forms in the middle row of (20) requires additional explanation, however. (69)

[+nasal] ⇔ [z 3p ] / [x −singular] [y 3s ]

The more general condition on nasality, given (38), is: (70)

[+nasal] ⇔ [ d ]

The details of all of these allomorphic variations should not obscure the main point, namely, that many of the regularities and patterns from the ditransitive prefixes carry over in large measure to the transitive system. So saying, let us consider the onsets.

CHAPTER 5

186 (71)

x: 3s gy-a ∅-a

1s: 2s:

3d n-en m-en

z 3p gy-at b-at

3i d-é b-é

3a d-e b-e

The second person onsets are reminiscent of the ditransitive system (see the discussion preceding (64)), and so one is led to: (72)

b- ⇔ [x 2s ]

(The m- in 2s:3d arises standardly by nasalization.) However, observe that 2s:3s has no b-. It, a, consists of the expected rhyme without an onset. So, either there is a zero vocabulary item that realizes x =2s in the context of z =3s, or the features in (72) ordinarily realized by b- have been deleted by a rule, contextualized by z =3s, prior to vocabulary insertion. I am not aware of a general principle to decide between the allomorphy- and the deletion-based solutions, but for concreteness adopt the latter.11

⎡

(73)

| x

⎤ −author ⎢+hearer ⎥ ⎢ ⎥ ⎣+singular ⎦ −augmented

→ ∅ /

[s]

Observe that there is no requirement that z , rather than y, be s. So underspecified, (73) correctly predicts absence of the onset b- from 2s:3s:z and 2s:1s:z , respectively the top and bottom rows of (51). First person onsets are simpler; indeed, they are predictable. Adding g(54) to the rhymes in (66) yields the correct results by standard phonology: gya(t) from g+ia(t) by Glide formation, d(´ e) from g+(´ e) by Dental-velar 11 That allomorphy and deletion can achieve the same surface result may point to a deficiency in the system: how, one might imagine, is the child to decide between them? The only attempt I am aware of to address this redundancy is Trommer’s (1999) formalization of Distributed Morphology, in which all morphological operations are reduced to vocabulary insertion. Two morphological operations deployed above—feature insertion and node (as opposed to mere feature) deletion—are not obviously replicable as vocabulary insertion; nor are non-structure-preserving morphological operations, which may establish or disestablish the adjacency relations taken to be central to suppletion in Chapter 4. Given the inherent interest of Trommer’s proposal and of the issue that motivates it, these criticisms should prompt refinement, rather than abandonment, of the work.


187

switching, nen from g+en by Dental-velar switching and Nasalization. Thus, many of the generalizations over ditransitive prefixes apply to singular x transitive prefixes, and some new generalizations apply equally well to the ditransitive ones. Non-singular x We are now concerned with the following prefixes. (74)

x: 2i: 3i: 3a:

3s bá∗ é∗ á∗

3d bet et et

z 3p bát∗ ét∗ gyá∗

3i bét ét et

3a bé ét ém

Little beyond what has already been said is required here. The most striking fact here is the distribution of ∗. It is restricted to z =s/p, which form a curious natural class, being oppositely specified for [±singular] and [±augmented]. Superficially, we can write: (75)

∗ ⇔ [x −singular] [z ±singular ∓augmented]

However, this has the problem that z =3i is predicted to be tone lowering, as i is of the general form [−singular −augmented +singular/augmented]. To avoid this, it is necessary to look beyond the notational device ‘∗’ to the underlying tonal mechanisms. Following my own analysis (Harbour 2002), the tonal/metrical constituency of tone lowering ∗-prefixes is ϕ)V, that is, a separate foot in the same tonal/metrical domain as the verb. So, (75) really describes the conditions under which a prefix contains a right parenthesis (forms its own foot). For simple (non-∗) high prefixes, the tonal/metrical constituency is ϕ|V or ϕ)|V, that is, a domain separate from the verb. So, if z =3i is realized both by the foot demarcator ‘)’ and additionally by the domain demarcator ‘|’, then it is consistent with (75) but does not tone lower. Moving on to non-tonal segments, consider 3i:z . Recall that non-singular higher arguments block nasalization by d and that they condition a coda when z =3i (59). Here, we see this again, with the 3i:3d, et, being non-nasal and with 3i:3i, ´ et, having a coda. Recall also (60), according to which, if a higher argument is i, then we find the vowel e, instead of the usual z vowel. This applies here too. (We return to 3i:3a, momentarily.)

CHAPTER 5

188

Consider 2i:z . The onset is predictably b-, given (64) and the restriction of (24) to singular. Compare the rhymes of 2i:z and 2s:z : x:

(76)

2s: 2i:

3s a á∗

3d en et

z 3p at át∗

3i é ét

3a e é

Again, effects of (59) are evident in the non-nasality of 2i:3d and the coda on 2i:3i. Further changes concern the tone of 2i:3s/p, which have high tone and ∗, whereas 2s:3s/p simply have low tone. This is predicted by (75). For z =3a, observe that 2i:3a, b´ e, is simply 2s:3a, be, with high tone. Interestingly, a similar relation holds form 3a:3a. It, ´ em, is simply 3s:3a, em, with high tone. This suggests: (77)

If ϕ ⇔ [x +singular F] [z 3a ], then ϕ ⇔ [x −singular F] [z 3a ] ´ ⇔ [x −singular F] [z 3a ]

a. b.

The prefix 3i:3a requires particular comment. It is the only z =3a prefix with a t-coda. Others, by contrast, derive from a regular onset plus a predictable vowel plus predictable tone. For instance, 1s:3a is g+e; 2s:3a is b+e; 2i:3a is b+e+´; for 3s:3a, the form is unexpected, but given it, em, 3a:3a is predictable as em+´. With its coda, 3i:3a ´ et looks quite different. In fact, its form is predictable, if we apply the mechanisms that account for a-agreement’s y-conditioned d∼i alternation. Applied exceptionally to 3i:3a, the result is 3i:3d/i, which, correctly, is ´ et. (78)

The mechanisms of the d∼i alternation apply to 3i:3a.

Observe that, in the preceding paragraph, 3s was treated as the singular correspondent of 3a, for the purposes of applying (77). The z =3p prefixes reinforce this correlation. Consider the relationship between 2s:3p and 2i:3p, bat and b´ at∗, which vary only with respect to tone. Exactly the same relationship holds between 3s:3p and 3a:3p, gya and gy´ a∗. However, for other prefixes, any correlation is less obvious: (79)

x:

3s: 3a:

3s ∅ á∗

3d eM et

z 3p gya gyá∗

3i é et

3a em ém


189

If the onset for x =3a is ∅, then 3a:3d is as expected: ∅+et is et. (Note that 3a, being [−singular], conditions the non-nasal allomorph of z =3d agreement.) Given this, 3a:3i has the expected onset, vowel and coda ∅+e+t, and is surprising only in lacking high tone. Thus, the analysis of the ditransitive prefixes extends naturally to the transitives with minor modifications. Note, again, the crucial reference to features of earlier chapters, for instance in creating a natural, if marked, class of s and p (75) and the analysis of 3i:3a (78). 5.4.3.

Intransitives

We now turn to the last prefixes in the system, namely, the intransitive. (80)

x: 1s: 2s: 2i: 3s: 3i: 3a:

z =∅ a em ba ∅ e á

Given the preceding discussion, there is little to surprise us here. 2s:∅, em, is a special realization of the features common to 2s intransitive and 2s indirect object agreement (p. 181). The correlation between intransitive and indirect object agreement is suggestive also for 1s:∅ and 3s:∅, neither of which was realized by an onset when the values of y. Here, again, we find them onsetless. This correlation cannot be extended, however, to non-singular intransitives. If it could, 3i:∅ would have onset b-, as any:3i:z prefixes have. Nonetheless, the non-singular prefixes are reminiscent of other prefixes. Specifically, for non-singular x, x :∅ is predictable from x :3s. (81)

x:

2i: 3i: 3a:

z

∅ ba e á

3s bá∗ é∗ a´∗

If the x :∅ forms are taken as basic, then addition of ∗ yields x :3s. (Recall that the metrical reality behind ∗ entails that, if a prefix is ∗, then it has

CHAPTER 5

190

high tone, p. 187.) This suggests that the x :∅ forms are, for non-singular prefixes, the basic vocabulary items for these persons. (82)

a. b. c.

ba ⇔ [ 2i ] e ⇔ [ 3i ] ´ a ⇔ [ 3a ]

If this reasoning is more or less correct, then the only additional explanation required for intransitive prefixes is the occurrence of a for 1s:∅. (83)

a ⇔ [ 1s ]

5.5.

Conclusion

The point of the foregoing micromorphological argumentation is that it places us in a position to give a vocabulary list with allomorphic variants which map prefix structures that have been acted on by morphological rules into phonological segments that yield familiar prefixes by regular phonology. At this point, though we are clearly nearing a solution, I lay the matter to rest. As emphasized above, the analysis of the agreement prefixes is relevant to the current investigation as a means of justifying the existence of the features and values argued for herein. By showing that complex allomorphy conditions and morphological operations are statable precisely in these terms, that aim has been achieved above. I leave a complete and final analysis, down to the level of a full and fully compositional vocabulary list, for future work. 5.6.

Appendix

Although the Kiowa prefix system has been the object of extensive investigation for nearly a century now, one aspect of it has escaped previous attention: the syncretic behavior of ∅:3y:3a. Ordinarily, y-agreement triggers the d∼i alternation for z =3a. For ∅:3y:3a, however, the alternation does not occur. Furthermore, this is (I tentatively claim) the only argument combination where it is significant whether z =3a arises from an animate plural or from a reflexive. Let us distinguish these cases as ∅:3y:3a and ∅:3y:refl. For animate plurals, we apparently have the syncretism ∅:3y:3a ≡ 3y:3a; that is, for third persons, action on and possession of an animate plural results in identical agreement. For ∅:3y:refl, matters are more complex: note the odd switching of s for p and p for s.

THE AGREEMENT PREFIX x:

(84) ∅:

:z

:refl

191 y

3s 3p:3a

3d ∅:3d:3i

3p 3s:3a

3i ∅:3i:3i

As these facts have only newly come to light, they require further investigation. However, in the context of the current chapter, which is merely a detailed sketch, not a full solution, this open-ended aspect is not detrimental.


Chapter 6 Conclusions and Consequences Despite its crosslinguistic frequency, noun classification is commonly conceived of as an arcane and arbitrary feature of language, something possessed of little less arbitrariness than the Saussurean sign itself. The preceding chapters, investigating one such system, that of Kiowa, have shown that, far from being an isolated fact about individual lexical items, the mechanisms of noun classification are deeply embedded in the grammar, exhibiting effects in the semantics, syntax, and morphology. In this final chapter, I wish to consider what Kiowa suggests for the treatment noun classification more generally.

6.1.

Noun classification

At first glance, the Kiowa-Tanoan inverse is an exotic morpheme that switches or toggles number between opposing values. However, we have now seen that the inverse does no such thing as toggle. Rather, it is simply a response to feature conflicts that arise on D from the interaction of Class and Number. With the veil of exoticism lifted, it is useful to ask to what extent Kiowa’s noun class system resembles those of other languages, such as the gender and declension class systems of Indo-European, or the gender and number systems of Bantu. Kiowa, it emerges, is like Bantu, but distinct from IndoEuropean. To show this, I consider an example of each system, Sanskrit for Indo-European, Kiswahili for Bantu, and show that it is possible to combine the core characteristics of Kiowa and Sanskrit (Section 6.1.1), but that it is impossible to do likewise for Kiowa and Kiswahili (Section 6.1.2). This shows that Kiowa and Kiswahili represent manifestations of the same system. 193

CHAPTER 6

194 6.1.1.

Gender and declension class

An important distinction inherited by generative grammar from traditional grammar is that between gender and declension class (see, e.g., Aronoff 1994). Though both are (mostly) arbitrary properties of individual lexical items, they are crucially different. Gender ‘propagates’ throughout the syntax, whereas declension class is confined to the individual noun itself. For instance, consider three Sanskrit noun phrases (based on Coulson 1992): (1)

(2)

(3)

r¯aj¯a k¯antah. king.nom.s beloved.masc.nom.s ‘beloved king’

-an class

a¯tm¯a k¯antam self.nom.s beloved.neut.nom.s ‘beloved self’

-an class

suhr.t k¯antah. friend.nom.s beloved.masc.nom.s ‘beloved friend’

consonant-final class

The notion of gender is illustrated by the adjectives, that of declension class by the nouns. The nouns r¯ ajan ‘king’ and ¯ atman ‘soul’ belong both to the -an declension class. Accordingly, in the nominative singular, they end in ¯ a. By contrast, suhr.d ‘friend’ belongs to the consonant-final declension class, and so, in the nominative singular, ends in zero, with the consonant devoicing. Thus, from the point of view of case endings, ‘king’ and ‘soul’ form a natural class to the exclusion of ‘friend’. However, when we turn to the adjectives, we find that ‘king’ and ‘friend’ form a natural class to the exclusion of ‘self’. For the former, ‘beloved’ ends in -ah., but, for the latter, it ends in -am. According to the traditional designations, ‘king’ and ‘friend’ are ‘masculine’, ‘self’ ‘neuter’. Clearly, these genders are independent of declension class, as the two classificatory schemes crosscut.1 Declension class determines only the form of the morpheme that 1

The fourth possibility, neuter gender, consonant-final declension, is manas ‘mind’. Nominative singular on the noun is zero and on the adjective -am; by regular phonology: (i)

manah. k¯antam mind.nom.s beloved.neut.nom.s ‘beloved mind’

consonant-final class

CONCLUSIONS AND CONSEQUENCES

195

realizes nominative singular, and other case-number combinations, on the noun itself, such as ¯ a for the -an declension class, zero for the consonantfinal declension class. Gender determines the form of the morphemes that realize nominative singular, and other case-number combinations, on agreeing elements beyond the noun, such as -ah. for masculine, -am for neuter.2 With this distinction in mind, we can now ask: Does Kiowa have a gender or declension class system? Kiowa does not have a gender or declension class system. The easiest way to see this is by sketching languages, Kiowa and Kiowa , that combine Kiowa’s noun class system with a declension class and a gender system, respectively. To sketch Kiowa , let us first simplify Kiowa by removing all allomorphy of the inverse marker (Section 2.6.1), replacing them with two allomorphs -inv and -erse. We now define two declension classes, the -inv declension class and the -erse declension class, as follows. Take all the nouns in Harrington (1928) and assign them to the -inv class. All other nouns are assigned to the -erse class. This information is represented on the vocabulary entries. For example, Harrington gives the Kiowa for ‘stick’ but not ‘walking stick’. Consequently, ‘stick’ belongs to the -inv and ‘walking stick’ to the -erse declension class in Kiowa . The vocabulary items are: √ (4) stick ⇔ á a´-inv class √ (5) walking stick ⇔ t!optéM aa-erse class The superscripts are ‘class diacritics’ (cf., Harris 1991) that condition the realization of feature conflict on D, that is, the inverse marker: (6)

[−F +F] ⇔ -inv / [ N-inv class ] [ D ⇔ -erse / [ N-erse class ] [ D

] ]

Note that the assignment of nouns to declension classes affects only the form of inv the inverse suffix on nouns. It does not affect the noun classes them2

As an example of other things the endings of which are determined by gender, consider a past participle, such as pratibuddha ‘awoke’. In the masculine singular nominative, pratibuddhah. may be combined with the masculine nouns above to yield R¯ aj¯ a/Suhr.t k¯ antah. pratibuddhah. ‘The beloved king/friend awoke’; and in the neuter singular nom¯ inative, pratibuddham may be combined with the neuter nouns to yield Atm¯ a/Manah. k¯ antam pratibuddham ‘The beloved soul/mind awoke’.

CHAPTER 6

196

selves, such as sdi, idp; nor does it affect the form of i-agreement on the verb. The possibility of combining Kiowa’s noun class system with declension classes that crosscut them, Kiowa , shows that Kiowa noun classification is distinct from declension classes.3 By way of illustration, consider the idp nouns ‘stick’ and ‘walking stick’ in the sentence frame ‘It’s a ’. Kiowa (7) and (8) are analogous to Sanskrit (1) and (3), the -an and consonant-final masculine nouns. (7)

(8)

´ a- inv e- dÓÓ A´ stick-inv 3i-be ‘It’s a stick’ T!optéM aaerse e- dÓÓ walking stick-inv 3i-be ‘It’s a walking stick’

Kiowa , -inv class

Kiowa , -erse class

To sketch Kiowa , we define two genders, inv and erse, by assigning all nouns in Harrington (1928) the gender inv and all others erse, where the phonological reflexes of these genders are, say, creaky voice and lip rounding, respectively. That is, if N, a Kiowa noun, has gender inv in Kiowa , then, wherever N triggers the inverse marking or agreement form ψ in Kiowa, it will trigger ψ + creaky voice, ψ , in Kiowa . For example: (9)

´ adO déhÓÓgya [A´ eét] -gO stick.inv 3i.inv -big.s-inv.inv 1s:3i.inv -get.pf ‘I got a stick that is big’

Kiowa

And if N has gender erse in Kiowa , then, wherever N triggers the inverse marking or agreement form ψ in Kiowa, it will trigger ψ + rounding, ψ, in Kiowa . For example:

(10)

déhÓÓgya [T!optéM aadO eét] -gO walking stick.inv 3i.inv -big.s-inv.inv 1s:3i.inv -get.pf ‘I got a walking stick that is big’

Kiowa

Observe that in both (9) and (10), the inverse marking on the noun is exactly as it would be in Kiowa and, moreover, that this marking is unrevealing of 3

Note that the claim here is that Kiowa is a logically possible language. I make no claim as to whether Kiowa , or any other of the grammatical hybrids below, are acquirable.

CONCLUSIONS AND CONSEQUENCES

197

gender, as in (1) and (2).4 Combining Kiowa’s noun class system with a gender system that crosscuts these classes, Kiowa shows that Kiowa noun classification is distinct from gender classification.

6.1.2.

Gender-number systems

A system more similar to Kiowa’s than those considered above is that of Bantu, or, more precisely, Kiswahili on Carstens’ (1991) analysis.5 It is typical of Bantu languages in possessing a large number of noun classes. Each is associated with a class-particular nominal prefix and with ‘a distinctive pattern of agreement borne by modifiers and arguments of the noun, and on auxiliaries and predicates in the relevant syntactic relations to it’ (pp. 2–3). Some classes and examples are provided below (cf., Carstens 1991, p. 3). (11)

Class 1 2 3 4 5 6 7 8 9 10 11 .. .

Example m-tu wa-tu m-ti mi-ti gari ma-gari ki-atu vi-atu n-yumba n-yumba u-bao .. .

Gloss person people tree trees car cars shoe shoes house houses board .. .

4 The syntactic mechanisms that give rise to the different forms of agreement and inverse marking in Kiowa are, I assume, the same as in Sanskrit, and other languages, in which adjectives, verbs and so on agree for gender and number. This would be implemented by having D bear number and gender features simultaneously, rather as it bore person and number features in the treatment of sii syncretisms. Thus, in the preceding examples, D would bear [−F +F inv -gender] or [−F +F erse-gender]. Other gender-number combinations are easily representable: for example, for an sdi noun of gender inv and referential cardinality 2, D would be [−singular −augmented inv -gender]. 5 The language is more commonly called ‘Swahili’ in English. I follow Carstens’ terminology; see her footnote 1, p. 1.

CHAPTER 6

198 (12)

Mtoto huyu wangu mzuri 1.child 1.this 1.my 1.good ‘This my good child has fallen down’

ame-anguka 1.agr-pf- fall

(13)

Watoto hawa wangu wazuri wa- me-anguka 2.child 2.this 2.my 2.good 2.agr-pf- fall ‘These my good children have fallen down’

(14)

Mti huu wangu mzuri 3.tree 3.this 3.my 3.good ‘This my good tree has fallen down’

(15)

Miti hii yangu mizuri ime-anguka 4.tree 4.this 4.my 4.good 4.agr-pf- fall ‘These my good trees have fallen down’

ume-anguka 3.agr-pf- fall

The traditional views in Bantu and Kiowa linguistics make the two languages’ noun classification systems seem quite dissimilar. Bantuists have considered ‘Class, an amalgam of number and gender, . . . to be a lexical property of Bantu nouns and/or their prefixes’ (p. 6). Generalizations about Classes are then of the form ‘If a noun is singular in 1, it is plural in Class 2’, or (cf., p. 28) ‘If a Class 11 noun triggers Class 3 agreement in the singular, it triggers Class 10 agreement in the plural’. Traditional Kiowa descriptions, such as Wonderly, Gibson, and Kirk (1954), view referential cardinality as part of the inherent meaning of the noun; for instance, t´ og´ ul means ‘one or two young men’, with the inverse suffix giving the inverse number. (Such approaches, of necessity, deemphasize mass nouns and pluralia tantum.) Carstens, however, gives an analysis of the morphology and syntax of Kiswahili DPs that solves several problems in traditional Bantu linguistics and at the same time makes Kiswahili more similar to Kiowa, on the analysis given above. Specifically, she claims that there is no primitive notion of Class that amalgamates gender and number. Rather, these belong to distinct projections in the syntax. Modifying her structure slightly, to emphasize commonalities between our analyses, mtu ‘a person’ has the structure:6 6

The differences are (a) that Carstens writes ‘Group A’ where I have ‘Gender A’—the extra term does not benefit the discussion here; (b) that the projections in Carstens’ tree are labelled category-neutrally—the functional structure of the DP is dealt with later in her dissertation; (c) that the content of Number is not [+singular −augmented]. The last, for uniformity with Kiowa, is the only controversial alteration. I believe that the content of Number varies crosslinguistically; in Kiswahili, it is probably just [±singular].

CONCLUSIONS AND CONSEQUENCES (16)

199

DP

D ⎤ +singular ⎣−augmented⎦ Gender A

NumberP

⎡

m

Number Class: person


Gender A

tu On Carstens’ analysis, prefixes like m- are the realization of gender and singular∼plural alternations such as m-∼wa- are contextual allomorphy:7 (17)

[Gender A] ⇔ m- / [D +singular ⇔ wa- / [D −singular

] ]

The point of contact between Carstens’ analysis of Bantu and the current analysis of Kiowa is that, in both cases, what has been traditionally treated as class-specific morphology is analyzed instead as the realization of features on D. Moreover, these features come to be on D by copying from lower heads, Class and Number. Kiowa and Kiswahili differ, however, in that the classifying features of Kiswahili are not number features. Consequently, copying Class and Number onto D cannot result in feature conflict. A further point of contact between Kiowa and Kiswahili is that both are compatible with a declension class system. This was shown above for Kiowa and similar facts are found in Kiswahili. Animate nouns provide one example of this. All animates trigger, on their modifiers, arguments and predicates, 1-agreement forms in the singular and 2-agreement forms in the plural. However, animate nouns themselves do not always bear the 1∼2 prefixes m-∼wa-. For instance, vifaru ‘rhinos’ displays the Class 8 prefix vi-. However, it triggers Class 2 agreement on its modifiers and predicates, as in: (18)

7

Vifaru wawili wa- na- pigana 8.rhino 2.two 2.agr-pres-fight.recip ‘Two rhinos are fighting’

Alternatively, m-/wa- could realize [±singular Gender A] on D.

CHAPTER 6

200

These facts are reminiscent of the Sanskrit cases treated above, where marking on the noun is not a reliable predictor of agreement forms elsewhere. Carstens’ analysis is that all animates have Gender A (i.e., are Class 1 when singular, Class 2 when plural), but that some, such as ‘rhino’, are lexically marked as conditioning prefixes from different classes on themselves. So, the noun imposes a declension class on its own prefix, above and beyond any morphological forms that it imposes elsewhere in the sentence. Note further that the Kiowa and Kiswahili systems are not wholly mutually compatible, as expected if they are different manifestations of the same mechanisms. Given that D in Kiowa does not always have number features that reflect the content of Number, there can be no language Kiowa in which nouns bear inverse morphology for certain referential cardinalities and where agreement forms of nouns’ modifiers, arguments and predicates consistently reflect referential cardinality. Thus, Kiowa’s noun classification system is akin to that of Kiswahili and so is a gender-number system. 6.2.

Parting comments

This investigation has had both empirical and theoretical aims. Empirically, it has sought to provide an analysis of a variety of seemingly unrelated phenomena in Kiowa grammar: the distribution of agreement types and inverse marking, the semantic characteristics of noun classes, the structure of adjectivally or deictically modified DPs, the inventory of noun classes, the motivation and nature of agreement∼suppletion mismatches, the nature of Kiowa’s agreement prefixes. Theoretically however, it has suggested a line of inquiry—research into a unified morphosemantic theory of number—that strikes me as important and interesting, and has offered some elements of what such a theory of number might comprise. If this program of research is accepted by others, then the immediate goal is to find other empirical domains, such as the nature of distributivity, or associative plurals, or pluractionality, in which semanticists and morphologists have equal stakes. By thus widening the domain of inquiry of both morphologists and semanticists, a broader range of data is brought to bear on core problems than has until now been considered relevant. Affording the total theory of Universal Grammar one theory of number, the morphosemantic, rather than two, the morphological and the semantic, we make it ontologically more constrained, descriptively more adequate and explanatorily more powerful. This, I believe, is what linguistics should be about.

Appendix A A Hunting Story This appendix presents the short text quoted in Section 2.5. It was recorded from Mr Vincent Bointy in August 2001, at Anadarko, Oklahoma and transcribed (with some amendments) under his supervision. It illustrates several phenomena discussed in the preceding chapters, including the use of numberconditioned suppletive predicates. The accuracy of the glossing owes much to Laurel Watkins’ comments on an earlier draft. The following points explain aspects of the glossing, such as policy decisions and tricky cases. For discussion of individual grammatical points, see Watkins (1984) or the more specific references cited below. 1. Kiowa employs a switch reference system for clause linking (Watkins 1993). The switch reference markers are simply glossed as conj. As a rule of thumb, gO (and its derivative forms, such as gig´ O) indicates continuity of reference, whereas nO (and its derivative forms, such as neg´ O) indicates discontinuity of reference. (One might, on these grounds, have expected nO instead of gO in the sentence after Ky´ ap ´ a-´ˇZ´ˇZbaahel ‘They went on hunting in the other direction’. If this expectation is correct, it is possible that the process of transcription, with its frequent interruptions and repetitions, disrupted the contintuity of the discourse and, so, of reference tracking.) 2. Morphemes’ surface tones depend on their position in the word and their wordmate morphemes (Harbour 2002). Glosses below reflect surface, not underlying tones. Hence the tonal variation between dÓˇÓˇ∼dOˇ Oˇ and mˆ ei∼mei in: 201

APPENDIX A

202 (a)

gya-tháp-dÓˇÓˇ-mˆ ei 3p- dry- be- impf.hsy ‘They were dry’

(b)

gya-h´ oM On-dOˇ Oˇ-mei 3p- path-be- impf.hsy ‘There was a path’

3. As Mr Bointy did not participate in the events described, hearsay forms are used throughout the narrative (though not in recreation of participants’ speech). 4. Two distinct morphemes are glossed as fut: tOO and t!OO, used for verbs with external arguments and verbs without, respectively. E.g.: (a)

éMthÓtté-tOO 3s:1s-shoot- fut ‘He will shoot me’

(b)

ba- thóu ´gyâi-t!OO 1p.in-pass.pf- fut ‘We will pass’

5. The agreement prefix in En-th´ Ott´ etOO ‘He will shoot them [two]’ is unexpected, as 3s:3d is eˇ , not en. Watkins (p.c.), who has also encountered such forms, suggests that n is a phonologically conditioned variant of the pure nasal. 6. Agreement in Gya-g´ uu ´ d´ O´ Oˆ Oˇ OˇmO ‘They were singing hunting songs’ is puzzling. Gya is either 3s:3p or 3p. However, 3s:3p fails to capture the plurality of the agent, and 3p, which is intransitive, fails to match the transitivity of the verb root. Possibly, I misheard the verb, which should be detransitive, Gya-g´ uu ´ d´ O´ Oˆ Oˇ Oˇma; the prefix 3p then makes sense and an impersonal translation is appropriate ‘There was hunt song singing going on’ (cf., German ‘Es wurde(n Jagdlieder) gesungen’). Alternatively, singular agents, as in 3s:3p, may be a grammatical default (cf., the paragraph beginning ‘They got ready to sleep’); this remains to be investigated. 7. Square brackets around parts of the translation indicate elements not strictly part of the Kiowa original but which are implied by it and aid in reading the English. ´ uu 8. U´ ´´Z is a narrative element seeming to indicate (traversal of a) distance. It may be related to the distal ´ o´Z, as in the demonstratives ´ o´Zde ‘that’, ´ o´ZgO ‘that.inv’. It is not glossed below. 9. Several different morphemes—em, eˇ ˇi, pa—are glossed as loc. Their differences are discussed by Watkins (1984).

APPENDIX A

203

Íˇ´Zˇheˇ iˇtegya Ph´ a ˇ´ a ˇ o k!y´ a ˇ´ a ˇ hy´ op e-´ˇZ´ˇZa ˇa ˇ hel ´ a´ adom. Sy´ Ond´ e gya-h´ o ˇ OndOˇOˇmei d´ e em t´ e´Z e-khOl´ e´Za ˇa ˇ hel. GO e-x´ andehel ´ OgO ´ a´ adO e-bˆ ondOˇ Oˇmei d´ eeˇ ˇi gO h´ o ˇ On d´ omgy´ a´ OphOO gya-x´ ou ´dˆ ei. NO ’gO t!´ op´ a ˇ´ a ˇ hˆ elde t´ o ˇu ´ ˇ nˆ ei: ‘X´ a´ a! ´ OgO á ádO e-bôndOOde dómgyá ba-thóu´gyâit!OO. Bát-mOOhol, hÓˇÓˇ b´ a-salOm. ´ OÓkO, ááboˇ uˇ dOdethai khóˇ u´ˇ gya t!ámde mOOkyáˇiˇik!OO. BátmOOkholdOOpe. D´ omb´ e ba-khˆ utdet!OO. H´ ay´ atto th´ a´Zde khyˆ aigunt!OO.’

A Hunting Story Three men were hunting in the woods. They were all going along a narrow path when they arrived where the path lay under a bent tree. The one in front said: ‘Friends! We’re going to go under that bent tree. Let’s get ready and load the guns. There, up in the bent tree, there’s a black panther lying in wait. Let’s be ready: we’re going to pass underneath, and maybe he’ll pounce on [us].’

PháM áM o k!yáM áM hyóp e- ´MZ´MZ- aM aM - hel á a´- dom. SyÓn- dé gya-h´ oM On-dOM OM three man.inv 3i-hunt-come-hsy tree-loc small-adv 3p- path- bemei dé- em impf.hsy nom-loc

té´Z e- khOlé´Z- aM aM - hel. GO e- xánde-hel ÓgO all 3i-together-come-hsy conj 3i-reach- hsy rel

a´ a´dO e- bôn- dOM OM -mei dé- eM Mi gO h´ oM On dómgyá ÓphOO gyatree.inv 3i-bend-be- impf.hsy nom-loc conj path under there 3pxóu ´-dêi. lay- impf.hsy O´ gO á a´dO rel tree.inv

NO ’gO ∅- t!óp-áM áM - hêl- de ∅- t´ oM u ´M - nêi: ‘Xáá! conj rel 3s-first-come-hsy-nom 3s-say-impf.hsy friends

e- bôn- dOO-de dómgyá ba- thóu ´gyâi-t!OO. Bát3i-bend-be- nom under 1p.in-pass.pf- fut 1p.in:3p´ÓkO, á a´- boM uM dO- de- thai kh´ mOOhol, hÓM ÓM bá∗salOm. O oM u ´M gya ready.imp gun 1p.in:3s-load.imp there tree-bend.inv-nom-above black

t!ámde ∅- mOOkyá-MiMi- k!OO. Bát∗mOOkhol- dOOpe. Dómbé bapanther 3s-ready- hunt-lie 1p.in:3p-ready.inc-be.imp under 1p.inkhˆ utdet!OO. Háyátto thá´Z- de ∅- khyâi- gun- t!OO.’ pop/flee out.detr-fut maybe on top-adv 3s-stretch-jump-fut

204

APPENDIX A

NO th´ Op ´ a-khˆ utdehel. T!Om´ a ˇ´ a ˇ d´ e gˆ om gya-b´ o ˇu ´ ˇ hˆ el nO h´ eˇ´ˇZ dÓˇÓˇmˆ ei. ´ T´ ekhop ´ a-donhel n´ e h´ On ´ a-thOˇOˇmOO. Em-x´ a´Zhel: ‘Hagya t!´ amde ob´ ´ O´Zk!OO?’ ´ uu Ky´ ap ´ a-´ˇZ´ˇZbaahel gO p´ a´ a pˆ eˇ ˇi h´ olhel. U´ ´´Z ´ OgO th´ o ˇu ´ ˇ Otkh´ Ongy´ a d´ eem, th´ o ˇu ´ ˇ k!´ O´ Opa, ´ O´ZhyOO ´ em-k!´ uu ´ hel. K!´Z´Zˆ elxeihel ´ O´ZhyOO d´ epeidou h´ On h´ Ond´ e hˆ OigO ´ Omd´ et!OOdei. ´ O´ZhyOO pêˇ ˇi á-piˇˇiOmhel gO á-hanhel. So they passed [quickly] to the other side. The one at the head looked back and it was gone. They looked for it all over but didn’t find it. ‘Was it really there?’ they asked themselves. They went on hunting in the other direction and one of them killed a turkey. Well, way over where the water came to an end, against a bank, that’s where they camped. They built a big fire so that nothing would come near and there they cooked the turkey and ate it.

NO thÓp a´- khˆ utdehel. ∅- T!Om-áM áM - dé gôm gya- b´ oM u ´M - hêl conj beyond 3a-pop/flee out-hsy 3s-first- come-nom back 3s:3p-look-hsy nO héM´MZ ∅- dÓM ÓM-mêi. Tékhop á∗- donhel né hÓn á∗conj gone 3s-be- impf.hsy everywhere 3a:3s-seek-hsy but neg 3a:3s´ thOM OM -mOO. Emxá´Z- hel: ‘Hagya t!ámde ∅- óbÓ´Z- k!OO?’ find- neg 3a:3a-ask-hsy possibly panther 3s-really-lie ´ uu Kyáp a´- ´MZ´MZ- baa- hel gO pá a´ pêMMi ∅hól-hel. U´ ´´Z ÓgO other way 3a-hunt-come-hsy conj some turkey 3s:3s-kill-hsy . . . rel th´ oM u ´M ∅- OtkhÓn-gyá dé- em, th´ oM u ´M - k!ÓÓ- pa, Ó´ZhyOO ém- k!´ uu ´water 3s-end- detr nom-loc water-bank-loc there 3a:3a-camp.s/d∅- K!´Z´Z- êl- xeihel Ó´ZhyOO dé- peido hÓn hÓndé 3s-wood-big.s-lay.s/d-hsy there nom-because neg something ´´ZhyOO pêMMi á∗- piMMi- Om- hel gO hÔigO ∅- Óm-dé- t!OO- dei. O near 3s-do- detr-fut-hsy there turkey 3a:3s-food-make-hsy conj hel. hsy

a´∗- hanhel. 3a:3s-eat up-hsy

APPENDIX A

205

GO ’igO ´ em-hˆ eˇ ˇitesOOhel ph´Z´Z´ O´ Ok!obe. Neg´ O p´ a´ a t´ o ˇu ´ ˇ nˆ ei: ‘S´ Ot ap´ e´Zgyaxan, u ú ú ´´Z, gya-g´ uu ´ d´ O´ Oˆ OˇOˇmO gO ´ eˇ´ˇZgO d´ o ˇu ´ ˇ gya a-h´ e´Zba gO ´ eˇ´ˇZhOdedOOgya gya-hˆ aapO: “HOyO hOO, . . . ” ’ ´ An-ph´ att´ ehelxeˇ ˇi, m´ a´ ade h´ Ond´ e´ an-x´ ath´ a´ adehel gO pˆ aagO-dOOgya ´ u´Zde ´ hˆ aapehel. U´ an-d´ O´ Op´ o ˇu ´ ˇ´ a ˇ´ a ˇ hˆ el. M´ amdexo m´ O´ O em-d´ O´ Ot´ O´ OyˆZi. Khodˆ eide m´ am khˆ utdehel. . . Gya-deˇ ˇimOOh´ olhel. K!´ O´ Opa gya-h´Znhel. ´ Oˇ ÓˇgO gya-thélhel, ÓˇÓˇgO h´ e´Zbet!OO d´ eotte gya-h´Znhel. Then they sat down around the fire to tell stories. One said: ‘I just thought of [this time], way back, they were singing hunting songs and I went down into the center and I started to sing this song: “HOyO hOO, ...” ’ When he finished, something upstream gave a shout and started to sing the same song. And the sound of singing got closer. It was like the singing was coming from above. Suddenly, it flew right over them. . . They got ready to sleep. They dug into the bank. Each one opened up [a hole], each one dug just enough to get into. GO conj

[he]gÓ ém- hêMMitesOO-hel then 3a:3a-tell stories-sit- hsy

∅- t´ oM u ´M - nêi: 3s-say-impf.hsy

‘SÓt just

ph´Z´Z-ÓÓk!obe. fire- around

a- pé´Zgya- xan, 1s-thought-arrive.pf

u ú ú ´´Z, ...

NegÓ conj.then

pá a´ some

gya-g´ uu ´dÓÓ3p- good shot-song-

ÔM OM mO gO éM´MZgO d´ oM u ´M gya a- hé´Zb- a gO éM´MZhO-de- dOOgya do.impf conj there down 1s-enter-intr.pf conj this- nom-song ´ gya- hâap-O: “HOyO hOO, . . . ” ’ Anphátté-hel- xeM Mi, mááde 1s:3s-raise- tr.pf . . . . . . :3s:3p-finish- hsy-when upstream hÓndé ánxáthá a´de-hel gO pâ agO-dOOgya ∅hâape-hel. something :3s:3p-war cry-shout- hsy conj same- song 3s:3s-raise- hsy ´ u´Zde ánU´ dÓÓ- p´ oM u ´M - áM áM - hêl. Mám- de- xo mÓÓ em- dÓÓ- tÓÓthere :3s:3p-sing-sound-come-hsy above-nom-instead like 3s:3a-sing-actyˆZi. Khodêide mám ∅- khˆ utdehel. . . Gya-deM Mi- mOOhól-hel. impf.hsy suddenly above 3s-pop/flee out-hsy 3p- sleep-ready- hsy ´ K!ÓÓ- pa gya- h´Zn-hel. OM ÓM gO gya- thélhel, ÓM ÓM gO ∅- hé´Zbe-t!OO bank-loc 3s:3p-dig- hsy self 3s:3p-open up-hsy self 3s-enter- fut dé- otte gya- h´Zn-hel. nom-only 3s:3p-dig- hsy

206

APPENDIX A

´ Zdeˇ gya-th´ S´ Ot ´ em-deˇ ˇis´ O´ Ogy´ a´ a gO h´ Ond´ e´ an-x´ ou ´´ a ˇ´ a ˇ p´ o ˇu ´ ˇ hˆ el. A´ apdÓˇÓˇmˆ ei gO gya-ph´ otkh´ Otk´ OdOˇOˇmei. Pah´ˇZ´ˇZ thˆ OOxo ´ an-x´ ou ´´ a ˇ´ a ˇ p´ o ˇu ´ ˇ nˆ ei. PhÓˇÓˇhel gO, klk klk klk, hÓˇÓˇgOt ´ e-s´ al´ Omhˆ el. T´ e´Z p´ a´ axokO ´ a-´ Onhˆ el: ‘Hagya n´ O´ O´ eˇ -th´ Ott´ etOO, hagya o ´´Zde en-th´ Ott´ etOO?’ Khodˆ eide ´ an-x´ at´ Onhel gO m´ am gya-´ a´Zhel gO gya-ph´Z´Zth´ a ˇ´ a ˇ holhel. NO sOph´ ol ´ a-bo ˇu ˇ . B´ ethOO sOph´ ol bˆ oudOˇOˇmei. M´ On x!´ ol eˇ -kh´ u´ZbOOdo m´ O´ O´ an-x´ ou ´´ a ˇ´ a ˇ p´ o ˇu ´ ˇ nˆ ei. They had just sat down to sleep when they heard something crawling. The leaves were dry and had become all crackly. Clearly, the crawling sound was coming from over there. It stopped and, click click click, loaded a gun. They all thought the same thing: ‘Is he going to shoot at me or the other two?’ Suddenly, he gave a shout, flew up and showered embers down on them. And [in the light] they saw an owl. It had been an owl all along. I guess he’d been dragging his wings and that’s why it had sounded like someone crawling.

´ ZdeM SÓt ém- deM Mi- sÓÓgyá a´ gO hÓndé ánxóu ´- áM áM - p´ oM u ´M - hêl. A´ just 3a:3a-sleep-sit.pf conj something :3s:3p-crawl-come-sound-hsy leaf gya-tháp-dÓMÓM-mêi gO gya-phótkhÓtkÓ-dOM OM -mei. Pah´MZ´MZ thÔO3p- dry- be- impf.hsy conj 3p- cracklybe- impf.hsy clearly beyondxo ánxóu ´- áM áM - p´ oM u ´M - nêi. ∅- PhÓMÓM-hel gO, klk klk klk, instead :3s:3p-crawl-come-sound-impf.hsy 3s-stop- hsy conj [clicks] hÓM ÓM gOt ésálÓm-hêl. Té´Z pááxokO a´- Ón- hêl: ‘Hagya nÓÓ éMgun.inv 3s:3i-load- hsy all alike 3a-think-hsy possibly 1 3s:1sthÓtté-tOO, hagya ó´Zde en- thÓtté-tOO?’ Khodêide ánxátÓnshoot- fut possibly that 3s:3d-shoot- fut suddenly :3s:3p-war cry-soundhel gO mám gya- á´Zhel gO gya- ph´Z´Z-tháM áM -hol-hel. NO sOphól hsy conj above 3s:3p-go off-hsy conj 3s:3p-fire- beat-kill-hsy conj owl a´∗- boM uM . BéthOO sOphól ∅- bôudOM OM -mei. MÓn x!ól 3a:3s-see.pf unbeknownst owl 3s-long time-be- impf.hsy perhaps wing eM kh´ u´Z- bOO- do mÓÓ ánxóu ´- áM áM - p´ oM u ´M - nêi. 3s:3d-drag-bring-because like :3s:3p-crawl-come-sound-impf.hsy

Bibliography Adger, David and Harbour, Daniel. 2007. Syntax and Syncretisms of the Person-Case Constraint. Syntax 10:1ff. Adger, David, Béjar, Susana, and Harbour, Daniel. 2001. Allomorphy: Adjacency and Agree, paper presented at the 24th GLOW colloqium, Braga. Aronoff, Mark. 1994. Morphology by Itself: Stems and Inflectional Classes. Cambridge, MA: MIT Press. Béjar, Susana. 2007. Conditions on Phi Agree. In Daniel Harbour, David Adger and Susana Béjar, eds., Phi Theory: Phi Features Across Interfaces and Modules. Oxford: Oxford University Press. Bonet, Eulàlia. 1991. Morphology after syntax: Pronominal clitics in Romance. Ph.D. thesis, MIT, Cambridge, MA. Boyd, Maurice. 1983. Kiowa Voices: Myths, Legends and Folktales. Fort Worth, TX: Texas Christian University Press. Carstens, Vicki May. 1991. The morphology and syntax of determiner phrases in Kiswahili. Ph.D. thesis, University of California, Los Angeles. Chierchia, Gennaro. 1998. Plurality of mass nouns and the notion of ‘semantic parameter’. In Susan Rothstein, ed., Events and Grammar , 53–103, Dordrecht: Kluwer. Chomsky, Noam and Halle, Morris. 1968/1991. The Sound Pattern of English. Cambridge: MIT Press, paperback edition. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka, eds., Step by Step: Essays on Minimalist Syntax in Honour of Howard Lasnik , 89–115, Cambridge, MA: MIT Press. 207

208

BIBLIOGRAPHY Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz, ed., Ken Hale: A Life in Language, 1–52, Cambridge, MA: MIT Press. Conklin, Harold. 1962. Lexicographic treatment of folk taxonomies. In F. W. Householder and S. Saporta, eds., Problems in Lexicography, Publication of the Indiana University Research Center in Anthropology, Folklore, and Linguistics 21, 119–141, Bloomington, IN. Corbett, Greville. 2000. Number . Cambridge: Cambridge University Press. Coulson, Michael. 1992. Sanskrit: An Introduction to the Classical Language. London: Hodder and Stoughton, second edition, revised by Richard Gombrich and James Benson. Cysouw, Michael. 2003. The Paradigmatic Structure of Person Marking. Oxford: Oxford University Press. Dryer, Matthew S. 2005. Coding of nominal plurality. In M. Haspelmath, M. Dryer, D. Gil, and B. Comrie, eds., World Atlas of Language Structures, 138–139, Oxford: Oxford University Press. Ellis, Clyde. 1996. To Change Them Forever: Indian Education at the Rainy Mountain Boarding School, 1893–1920 . Norman and London: University of Oklahoma Press. Everett, Daniel. 1986. Pirahã. In D. C. Derbyshire and G. K. Pullum, eds., Handbook of Amazonian Languages I , 200–325, Berlin: Mouton de Gruyter. Foley, William. 1986. The Papuan Languages of New Guinea. Cambridge: Cambridge University Press. Foley, William. 1991. The Yimas Language of New Guinea. Stanford, CA: CSLI Publications. Hale, Kenneth. 1962. Jemez and Kiowa correspondences in reference to Kiowa-Tanoan. International Journal of American Linguistics 28:1– 8. Hale, Kenneth. 1967. Toward a reconstruction of Kiowa-Tanoan phonology. International Journal of American Linguistics 33:112–120. Hale, Kenneth. 1973. Person marking in Walbiri. In S. R. Anderson and P. Kiparsky, eds., A Festschrift for Morris Halle, 308–344, New York: Holt, Rinehart, and Winston, Inc.

BIBLIOGRAPHY

209

Hale, Kenneth. 1975. Navajo linguistics: Part III, ms. MIT. Hale, Kenneth. 1997. Some observations on the contribution of local languages to linguistic science. Lingua 100:71–89. Halle, Morris and Marantz, Alec. 1993. Distributed Morphology and the pieces of inflection. In Ken Hale and Samuel Jay Keyser, eds., The View from Building 20 , 111–176, Cambridge, MA: MIT Press. Halle, Morris and Vaux, Bert. 1998. Theoretical aspects of IndoEuropean morphology: the nominal declensions of Latin and Armenian. In Craig Melchert, Jay Jasanoff, and Lisi Oliver, eds., Mr Curad: Studies in Honor of Calvert Watkins, Innsbruck: Institut f¨ ur Sprachwissenschaft der Universität Innsbruck. Halle, Morris. 2005. Palatalization/velar softening: What it is and what it tells us about the nature of language. Linguistic Inquiry 36:23–42. Harbour, Daniel. 2002. On the metrical nature of Kiowa tone, ms. MIT. Harbour, Daniel. 2003a. The Kiowa case for feature insertion. Natural Language and Linguistic Theory 21:543–578. Harbour, Daniel. 2003b. Some outstanding problems of Yimas. Transactions of the Philological Society 101:125–136. Harbour, Daniel. 2004. Argument structure and morphological richness. Lecture notes, Center for Advanced Studies in Theoretical Linguistics, Universitätet i Tromsø. Harbour, Daniel. 2006a. Number: The morphological use of semantic atoms. Queen Mary’s OPAL #5. (Google: QMOPALs). Harbour, Daniel. 2006b. On the unity of ‘number’ in morphology and semantics. In Ryo Otoguro, Gergana Popova, and Andrew Spencer, eds., Essex Research Reports in Linguistics 47 (Proceedings of the York–Essex Morphology Meeting 2), 21–30, University of Essex: Department of Language and Linguistics. Harbour, Daniel. 2006c. Valence and atomic number. Queen Mary’s, OPAL #8. (Google: QMOPALs). Harbour, Daniel and Guoladdle, Carrie. In preparation. Carrie’s Kiowa correspondence course, ms. QMUL and Riverside Indian School. Harrington, John P. 1910. On phonetic and lexic resemblances between Kiowan and Tanoan. American Anthropologist 12:119–123.

210

BIBLIOGRAPHY Harrington, John P. 1928. Vocabulary of the Kiowa Language. Washington, DC: Government Printing Office. Harris, Jim. 1991. The exponence of gender in Spanish. Linguistic Inquiry 22:27–62. Heath, Jeffrey. 1998. Pragmatic skewing in 1↔2 pronominal combinations in Native American languages. International Journal of American Linguistics 64:83–104. Kayne, Richard. 1994. The Antisymmetry of Syntax . Cambridge, MA: MIT Press. Kihm, Alain. 2002. What’s in a noun: Noun classes, gender, and nounness. www.llf.cnrs.fr/fr/kihm. Kihm, Alain. 2005. Noun class, gender, and the Lexicon–Syntax– Morphology interfaces: A comparative study of Niger-Congo and Romance languages. In G. Cinque and R. S. Kayne, eds., The Oxford Handbook of Comparative Syntax , 459–512, Oxford: Oxford University Press. Krifka, M. 1992. Thematic relations as links between nominal reference and temporal constitution. In Ivan Sag and Anna Szabolcsi, eds., Lexical matters, 29–53, Stanford, CA: CSLI. Kroskrity, Paul V. 1993. Language, History, and Identity: Ethnolinguistic Studies of the Arizona Tewa. Tucson and London: University of Arizona Press. Lagefoged, Peter. 1993. A Course in Phonetics. Fort Worth: Harcourt Brace College Publishers, third edition. Lasersohn, Peter. 1995. Plurality, Conjunction and Events. Dordrecht: Kluwer. Link, Godehard. 1983. The logical analysis of plurals and mass terms: A lattice-theoretical approach. In R. Bäuerle, C. Schwarze, and A. von Stechow, eds., Meaning, Use and Interpretation of Language, 302– 323, Berlin: Walter de Gruyter. Marantz, Alec. 1997. No escape from syntax: don’t try morphological analysis in the privacy of your own lexicon. In A. Dimitriadis, L. Siegel, C. Surek-Clark, and A. Williams, eds., PWPL 4.2, Proceedings of the 21st Annual Penn Linguistics Colloquium, 201–225,

BIBLIOGRAPHY

211

Pennsylvania: University of Pennsylvania Working Papers in Linguistics. Matthews, Peter. 1972. Huave verb morphology: Some comments from a non-tagmemic viewpoint. International Journal of American Linguistics 38:96–118. McGinnis, Martha J. 2003. On the universal asymmetry between first and second person, ms. University of Calgary. McKay, Graham. 1978. Pronominal person and number categories in Rembarrnga and Djeebbana. Oceanic Linguistics 17:27–37. McKay, Graham. 1979. Gender and the category unit augmented. Oceanic Linguistics 18:203–210. McKenzie, Parker. n.d.a. Kiowa terms for birds and reptiles, ms. Oklahoma Historical Society. McKenzie, Parker. n.d.b. Kiowa terms for mammals, ms. Oklahoma Historical Society. Merrifield, William R. 1959a. Classification of Kiowa nouns. International Journal of American Linguistics 25:269–271. Merrifield, William R. 1959b. The Kiowa verb prefix. International Journal of American Linguistics 25:168–176. Merrill, William L., Hanson, Marian Kaulaity, Green, Candace S., and Reuss, Frederick J. 1997. A Guide to the Kiowa Collections at the Smithsonian Institution. Washington, D.C.: Smithsonian Institution Press. Mooney, James. 1898/1979. Calendar History of the Kiowa Indians. Washington, D.C.: Smithsonian Institution Press. Noyer, Rolf. 1992. Features, Positions and Affixes in Autonomous Morphological Structure. Cambridge, MA: MITWPL. Noyer, Rolf. 1998. Impoverishment theory and morphosyntactic markedness. In D. Brentari, S. Lapointe and P. Farrell, eds., Morphology and Its Relation to Phonology and Syntax , 264–285, Palo Alto, CA: CSLI. Ojeda, Almerindo E. 1998. The semantics of collectives and distributives in Papago. Natural Language Semantics 6:245–270.

212

BIBLIOGRAPHY Ouhalla, Jamal. 2005. Agreement features, agreement and antiagreement. Natural Language and Linguistic Inquiry 23:655–686. Pesetsky, David and Torrego, Esther. 2001. T to C movement: causes and consequences. In Michael Kenstowicz, ed., Ken Hale: a Life in Language, 355–426, Cambridge, MA: MIT Press. Ritter, Elizabeth. 1991. Two functional categories in noun phrases: evidence from Modern Hebrew. In Susan Rothstein, ed., Syntax and Semantics 26 , 37–62, New York: Academic Press. Ritter, Elizabeth. 1993. Where’s gender? Linguistic Inquiry 24:795–803. Schwarzschild, Roger. 1996. Pluralities. Dordrecht: Kluwer. Sportiche, Dominique. 1997. Reconstruction and constituent structure, paper presented at MIT; available at http://www.linguistics.ucla. edu/people/sportich/papers/mittalk97.pdf. Sprott, Robert W. 1992. Jemez syntax. Ph.D. thesis, University of Chicago. Takahashi, Junichi. 1984. Case marking in Kiowa: A study in the organization of meaning. Ph.D. thesis, City University of New York. Trager, Edith Crowell. 1960. The Kiowa language: A grammatical study. Ph.D. thesis, University of Pennsylvania. Trommer, Jochen. 1999. Morphology consuming syntax’ resources: Generation and parsing in a minimalist version of distributed morphology. In Proceedings of the ESSLI Workshop on Resource Logic and Minimalist Grammars, Utrecht. Vergnaud, Jean-Roger. 1974. French relative clauses. Ph.D. thesis, MIT. Watkins, Laurel (with the assistance of Parker McKenzie). 1984. A Grammar of Kiowa. Lincoln: University of Nebraska Press. Watkins, Laurel. 1990. Noun phrase versus zero in Kiowa discourse. International Journal of American Linguistics 56:410–426. Watkins, Laurel. 1993. The discourse functions of Kiowa switchreference. International Journal of American Linguistics 59:137–164. Winter, Yoad. 2001. Flexibility Principles in Boolean Semantics: Coordination, Plurality, and Scope in Natural Language. Cambridge MA: MIT Press.

BIBLIOGRAPHY

213

Wonderly, William, Gibson, Lorna F. and Kirk, Paul L. 1954. Number in Kiowa: Nouns, demonstratives and adjectives. International Journal of American Linguistics 20:1–7. Yang, Charles. 1999. Words, rules, and competitions, ms. MIT. Yang, Charles. 2002. Knowledge and Learning in Natural Language. Oxford: Oxford University Press. Yumitani, Yukihiro. 1998. A phonology and morphology of Jemez Towa. Ph.D. thesis, University of Kansas.

Index [±augmented], 5, 61, 67, 101, 110 definition, 63, 69 [±group], 5, 61, 100–102, 110, 115 and suppletion, 138–144 definition, 63, 98–99 [±singular], 5, 61, 67, 70, 101, 110 definition, 63, 68–70

ClassP, 73, 93, 99, 105, 106, 109 interpretation, 96, 99 collections of collections, 44, 99, 139 collectivity, 36, 41–44, 50, 95, 97, 101, 138–140 of action, 42, 45 conjunction, 28–29, 107–109, 201

abstract nouns, 48 adjacency, 122, 125, 126, 186 adjectives, 16, 89, 129, 132–136, 141 adverbs, 17, 150–154 agreement prefixes, 9, 18–19, 157– 191, 202 notation, 19, 162 table of, 161 agreement types, 23, 71, 72 Aktionart semantics of, 63, 68, 99 Aktionsart semantics of, 69 allomorphy, definition, 122 animacy, 34–38, 50, 94, 145–149, 162, 167, 190 Arapesh, 33, 63

d∼i alternation, 108, 146, 188, 190 default agreement, 77, 179 deletion, 164–167, 175, 186 demonstratives, 16, 89–92, 148, 202 Distributed Morphology, 10–12, 186 DP structure, 62, 100, 109, 126, 144 dual, see referential cardinality empathy, 51–53, 145–149 English, 6, 12, 31, 36, 43, 46, 69, 93, 95, 121, 144, 155 feature insertion, 106, 164, 167, 179, 181, 186 features conflicting, 62, 76, 77 lattice semantics, 67–70 negation, 68 semantic composition, 70 first person, 31–32, 34, 50, 81–88, 94, 136–137, 165, 168–169, 182

Bantu, 193 body parts, 37, 41–45, 95, 97 case marking, 15 Catalan, 159 214

INDEX footwear, 49, 94, 113 gender, 4, 56, 193–200 Georgian, 169 Hopi, 119 idi, 27, 42–44, 80–136 idp, 26, 38–42, 79–80, 95, 130, 132 ids, 26, 38–42, 100, 103–104, 108, 130 Ilocano, 69 implements, 36, 39, 42, 45, 50, 95 impoverishment, 87, 88, 164, 168– 170, 179, 181 incorporation, 17–19, 154 individual-level predication, 118, 150 inverse allophony, 54–56 and suppletion, 131–137 feature specification, 76 marking, 6, 15 syntactic locus, 62, 73, 89, 93 Italian, 43 Jemez, 112 Kiowa-Tanoan, 13, 111, 193 Kiswahili, 193 Latin, 179 liquids, 45–46 bodies of, 45 markedness, 68 mass nouns, 28–31, 43 and suppletion, 140–142 granular, 47–48, 50, 100, 106 non-granular, 45–46, 50, 106 micromorphology, 160, 190

215 Minimalism, 10–12, 73–74 mnemonics, 24, 78–88, 109 naturalness, 8, 94–96, 102 paucity, 8, 110–113 morphosemantic, 9, 99, 160, 200 motion, independent, 34, 36, 94 noun class mnemonics, 24 noun classification arbitrariness, 33 default, 33, 49, 50, 94 features, 72 non-uniqueness, 33 semantic coherence, 4, 32–50 subclasses, 34, 44, 113 number features distribution, 7, 62, 73, 88–93, 144 syntactic flow, 4 only, 92–93 Person-Case Constraint, 158, 171 person features, 84 phonology, 19–20, 54–58, 123–125, 163–164, 172, 178–181, 183, 187, 201–202 plural, see referential cardinality pluralia tantum, 46, 50, 98, 99, 101, 104, 142–144 possession, 92 ppp, 27, 46–48, 100, 104, 106, 130 PRO, 154 pronouns, 15, 17, 32, 68, 69, 88, 92 referential cardinality and mass nouns, 31 definition, 22

INDEX

216 feature composition, 67, 70, 71, 96, 113–114 notation, 23 reflexives, 53, 145–149, 162, 167, 190 relative clauses, 16, 54, 91, 148, 154 Rio Grande Tewa, 155 rule ordering, 163, 169, 183 Russian, 33 Sanskrit, 193–195 Saussurean sign, 12, 193 sdi, 25, 34–38, 50–53, 74–78, 94, 107, 130, 132, 136–137 sdp, 25, 78–77, 81, 94, 127–130 sds, 27, 100, 102–104, 130, 138 second person, 86, 168, 172, 182, 183, 186 shape inductivity, 45, 50 sii, 31, 34, 81–88, 94, 168 singular, see referential cardinality sisterhood, 122, 144 sss, 28–31, 45–46, 106, 130 stage-level predication, 118, 150 suppletion, 8–9, 117–155 definition, 121 mismatches, 130 Swahili, see Kiswahili switch reference, 201 Tewa, 125 thematic nouns, 56–59, 120 toggle, 193 tone(s), 3, 4, 7, 19–20, 54–55, 124–125, 132, 157, 163, 172–173, 175–177, 179–182, 185, 187–190, 201 trial, 77

uninterpretability, 62, 73, 76–78, 110, 126 valuation, 74 vegetation, 38–42, 50, 95, 97 word order, 14–16 Y-model, 10 Yimas, 53, 169

STUDIES IN NATURAL LANGUAGE AND LINGUISTIC THEORY

Managing Editors Marcel den Dikken, City University of New York Liliane Haegeman, University of Lille Joan Maling, Brandeis University Publications 1. L. Burzio: Italian Syntax. A Government-binding Approach. 1986 ISBN Hb 90-277-2014-2; Pb 90-277-2015-0 2. W.D. Davies: Choctaw Verb Agreement and Universal Grammar. 1986 ISBN Hb 90-277-2065-7; Pb 90-277-2142-4 ´ Kiss: Configurationality in Hungarian. 1987 3. K. E. ISBN Hb 90-277-1907-1; Pb 90-277-2456-3 4. D. Pulleyblank: Tone in Lexical Phonology. 1986 ISBN Hb 90-277-2123-8; Pb 90-277-2124-6 5. L. Hellan and K. K. Christensen: Topics in Scandinavian Syntax. 1986 ISBN Hb 90-277-2166-1; Pb 90-277-2167-X 6. K. P. Mohanan: The Theory of Lexical Phonology. 1986 ISBN Hb 90-277-2226-9; Pb 90-277-2227-7 7. J. L. Aissen: Tzotzil Clause Structure. 1987 ISBN Hb 90-277-2365-6; Pb 90-277-2441-5 8. T. Gunji: Japanese Phrase Structure Grammar. A Unification-based Approach 1987. ISBN1-55608-020-4 9. W. U. Wurzel: Inflectional Morphology and Naturalness. 1989 ISBN Hb 1-55608-025-5; Pb 1-55608-026-3 10. C. Neidle: The Role of Case in Russian Syntax. 1988 ISBN1-55608-042-5 11. C. Lefebvre and P. Muysken: Mixed Categories. Nominalizations in Quechua 1988 ISBN Hb 1-55608-050-6; Pb 1-55608-051-4 12. K. Michelson: A Comparative Study of Lake-Iroquoian Accent. 1988 ISBN1-55608-054-9 13. K. Zagona: Verb Phrase Syntax. A Parametric Study of English and Spanish 1988 ISBN Hb 1-55608-064-6; Pb 1-55608-065-4 14. R. Hendrick: Anaphora in Celtic and Universal Grammar. 1988 ISBN1-55608-066-2 15. O. Jaeggli and K.J. Safir (eds.): The Null Subject Parameter. 1989 ISBN Hb 1-55608-086-7; Pb 1-55608-087-5


16. H. Lasnik: Essays on Anaphora. 1989 ISBN Hb 1-55608-090-5; Pb 1-55608-091-3 17. S. Steele: Agreement and Anti-Agreement. A Syntax of Luiseño. 1990 ISBN 0-7923-0260-5 18. E. Pearce: Parameters in Old French Syntax. Infinitival Complements. 1990 ISBN Hb 0-7923-0432-2; Pb 0-7923-0433-0 19. Y. A. Li: Order and Constituency in Mandarin Chinese. 1990 ISBN 0-7923-0500-0 20. H. Lasnik: Essays on Restrictiveness and Learnability. 1990 ISBN0-7923-0628-7; Pb 0-7923-0629-5 21. M. J. Speas: Phrase Structure in Natural Language. 1990 ISBN0-7923-0755-0; Pb 0-7923-0866-2 22. H. Haider and K. Netter (eds.): Representation and Derivation in the Theory of Grammar. 1991 ISBN 0-7923-1150-7 23. J. Simpson: Warlpiri Morpho-Syntax. A Lexicalist Approach. 1991 ISBN 0-7923-1292-9 24. C. Georgopoulos: Syntactic Variables. Resumptive Pronouns and A’ Binding in Palauan. 1991 ISBN 0-7923-1293-7 25. K. Leffel and D. Bouchard (eds.): Views on Phrase Structure. 1991 ISBN 0-7923-1295-3 26. C. Tellier: Licensing Theory and French Parasitic Gaps. 1991 ISBN0-7923-1311-9; Pb 0-7923-1323-2 27. S.-Y. Kuroda: Japanese Syntax and Semantics. Collected Papers. 1992 ISBN0-7923-1390-9; Pb 0-7923-1391-7 28. I. Roberts: Verbs and Diachronic Syntax. A Comparative History of English and French. 1992 ISBN 0-7923-1705-X 29. A. Fassi Fehri: Issues in the Structure of Arabic Clauses and Words. 1993 ISBN 0-7923-2082-4 30. M. Bittner: Case, Scope, and Binding. 1994 ISBN 0-7923-2649-0 31. H. Haider, S. Olsen and S. Vikner (eds.): Studies in Comparative Germanic Syntax. 1995 ISBN 0-7923-3280-6 32. N. Duffield: Particles and Projections in Irish Syntax. 1995 ISBN0-7923-3550-3; Pb 0-7923-3674-7 33. J. Rooryck and L. Zaring (eds.): Phrase Structure and the Lexicon. 1996 ISBN 0-7923-3745-X 34. J. Bayer: Directionality and Logical Form. On the Scope of Focusing Particles and Whin-situ. 1996 ISBN 0-7923-3752-2 35. R. Freidin (ed.): Current Issues in Comparative Grammar. 1996 ISBN 0-7923-3778-6; Pb 0-7923-3779-4 36. C.-T. J. Huang and Y.-H. A. Li (eds.): New Horizons in Chinese Linguistics 1996 ISBN 0-7923-3867-7; Pb 0-7923-3868-5 37. A. Watanabe: Case Absorption and WH-Agreement. 1996 ISBN 0-7923-4203-8


38. H. Thráinsson, S.D. Epstein and S. Peter (eds.): Studies in Comparative Germanic Syntax. Volume II. 1996 ISBN 0-7923-4215-1 39. C. J. W. Zwart: Morphosyntax of Verb Movement. A Minimalist Approach to the Syntax of Dutch. 1997 ISBN0-7923-4263-1; Pb 0-7923-4264-X 40. T. Siloni: Noun Phrases and Nominalizations. The Syntax of DPs. 1997 ISBN 0-7923-4608-4 41. B. S. Vance: Syntactic Change in Medieval French. 1997 ISBN0-7923-4669-6 42. G. Müller: Incomplete Category Fronting. A Derivational Approach to Remnant Movement in German. 1998 ISBN 0-7923-4837-0 Studies in Natural Language and Linguistic Theory 43. A. Alexiadou, G. Horrocks and M. Stavrou (eds.): Studies in Greek Syntax 1998 ISBN 0-7923-5290-4 44. R. Sybesma: The Mandarin VP. 1999 ISBN 0-7923-5462-1 45. K. Johnson and I. Roberts (eds.): Beyond Principles and Parameters. Essays in Memory of Osvaldo Jaeggli. 1999 ISBN 0-7923-5501-6 46. R. M. Bhatt: Verb Movement and the Syntax of Kashmiri. 1999 ISBN 0-7923-6033-8 47. A. Neeleman and F. Weerman: Flexible Syntax. A Theory of Case and Arguments 1999 ISBN 0-7923-6058-3 48. C. Gerfen: Phonology and Phonetics in Coatzospan Mixtec. 1999 ISBN 0-7923-6034-6 49. I. Paul, V. Phillips and L. Travis (eds.): Formal Issues in Austronesian Linguistics. 2000 ISBN 0-7923-6068-0 50. M. Frascarelli: The Syntax-Phonology Interface in Focus and Topic Constructions in Italian. 2000 ISBN 0-7923-6240-3 51. I. Landau: Elements of Control. Structure and Meaning in Infinitival Constructions. 2000 ISBN 0-7923-6620-4 52. W. D. Davies and S. Dubinsky (eds.): Objects and other Subjects. Grammatical Functions, Functional Categories and Configurationality. 2001 ISBN 1-4020-0064-2; Pb 1-4020-0065-0 53. J. Ouhalla and U. Shlonsky (eds.): Themes in Arabic and Hebrew Syntax. 2002 ISBN 1-4020-0536-9; Pb 1-4020-0537-7 54. E. Haeberli: Features, Categories and the Syntax of A-Positions. Cross-Linguistic Variation in the Germanic Languages. 2002 ISBN 1-4020-0854-6; Pb 1-4020-0855-4 55. J. McDonough: The Navajo Sound System. 2003 ISBN 1-4020-1351-5; Pb 1-4020-1352-3 56. D.E. Holt (ed.): Optimality Theory and Language Change. 2003 ISBN 1-4020-1469-4; Pb 1-4020-1470-8 57. J. Camacho: The Structure of Coordination. Conjunction and Agreement Phenomena in Spanish and Other Languages. 2003 ISBN 1-4020-1510-0; Pb 1-4020-1511-9 58. I. Toivonen: Non-Projecting Words. A Case Study of Swedish Particles. 2003 ISBN 1-4020-1531-3; Pb 1-4020-1532-1


59. D. Adger, C. de Cat and G. Tsoulas (eds). Peripheries. Syntactic Edges and their Effects. 2004 ISBN 1-4020-1908-4; Pb 1-4020-1909-2 60. C. Goria: Subject Clitics in the Northern Italian Dialects. A Comparative Study Based on the Minimalist Program and Optimality Theory. 2004 ISBN 1-4020-2736-2; Pb 1-4020-2737-0 61. V. Dayal and A. Mahajan (eds.): Clause Structure in South Asian Languages 2004 ISBN 1-4020-2717-6; Pb 1-4020-2718-4 62. P. Kempchinsky and R. Slabakova (eds.): Aspectual Inquiries. 2005 ISBN 1-4020-3032-0; Pb 1-4020-3035-5 63. M. Arad: Roots and Patterns. Hebrew Morpho-syntax. 2005 ISBN 1-4020-3243-9; Pb 1-4020-3245-5 ˇ 64. P. Stekauer and R. Lieber (eds.): Handbook of Word-Formation. 2005 ISBN 1-4020-3595-0; Pb 1-4020-3597-7 65. A. Johns, D. Massam and J. Ndayiragije (eds.): Ergativity. Emerging Issues 2006 ISBN Hb 1-4020-4186-1; Pb 1-4020-4187-X 66. F. Lee: Remnant Raising and VSO Clausal Architecture. A Case Study from San Lucas Quiavini Zapotec. 2006 ISBN1-4020-4300-7 67. O. Miˇseska Tomić: The Balkan Sprachbund Morpho-syntactic Features. 2006 ISBN1-4020-4487-9 ´ Kiss, Event Structure and the Left Periphery: Studies on Hungarian. 2006 68. K. E. ISBN 1-4020-4753-3 69. D. Harbour, Morphosemantic Number: From Kiowa Noun Classes to UG Number Features. 2007/2008 ISBN Hb 978-1-4020-5037-4; Pb 978-1-4020-5039-8

springer.com

Morphosemantic Number: From Kiowa Noun Classes to UG Number Features

From number theory to physics

Number: From Ahmes to Cantor

From Number Theory to Physics

From Number Theory to Physics

Introduction to Number Theory

Invitation to number theory

Introduction to number theory

Introduction to Number Theory

Number Story: From Counting to Cryptography

Number Story: From Counting to Cryptography

Introduction to Number TheoryChapters4,7,8,9

Introduction to Number Theory

Introduction to Number Theory

Invitation to number theory

Number Story: From Counting to Cryptography

Hot Number

Number Theory

Hot Number

Number Theory

Number Theory

Hot Number

Number Theory

Number Theory

Number Play

Number Play

Number theory

Feigenbaum Number

Number Play

Number fields

Number Theory

Morphosemantic Number: From Kiowa Noun Classes to UG Number Features

From number theory to physics

Number: From Ahmes to Cantor

From Number Theory to Physics

From Number Theory to Physics

Introduction to Number Theory

Invitation to number theory

Introduction to number theory

Introduction to Number Theory

Number Story: From Counting to Cryptography

Number Story: From Counting to Cryptography

Introduction to Number TheoryChapters4,7,8,9

Introduction to Number Theory

Introduction to Number Theory

Invitation to number theory

Number Story: From Counting to Cryptography

Hot Number

Number Theory

Hot Number

Number Theory

Number Theory

Hot Number

Number Theory

Number Theory

Number Play

Number Play

Number theory

Feigenbaum Number

Number Play

Number fields

Number Theory

Recommend Documents