Académique Documents
Professionnel Documents
Culture Documents
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
Published by Springer,
P.O. Box 17, 3300 AA Dordrecht, The Netherlands.
www.springer.com
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
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 dif-
ference 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 criti-
cism 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 Lech-
ner, 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
xiii
xiv ABBREVIATIONS AND NOTATIONAL CONVENTIONS
pres present
priv privative
q question particle
real nominal suffix denoting prototypicality
recip reciprocal
refl reflexive
rel relative
s agreement/suppletion type, typical of 1
stat stative
th theme
top topic marker (Japanese)
voc vocative
Kiowa Phonemes and Orthography
Consonants
Vowels
xv
xvi KIOWA PHONEMES AND ORTHOGRAPHY
Vowel diacritics
FRAMEWORK
The current investigation aims to lay the conceptual and basic theoretical
foundations of a morphosemantic theory of number. Until now, morphologi-
cal 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 con-
cerned 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 ex-
act trial, quadral or quintal number, or whether the forms that permit such
readings are really paucals, and hence cardinally inexact. More rarely, mor-
phologists 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, obs-
ervations 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 lan-
guages’ 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
(1) Á∗- OM OM
3a:3a:3s-give.pf
‘They gave it to them’
(2) M ∗-
É OM OM
3D:3a:3s-give.pf
‘They two gave him to them’
(3) Mé- ÓMÓM
3a:3D:3s-give.pf
‘They gave him to them two’
(4) Et- ÓMÓM
3a:3a:3D-give.pf
‘They gave them two to them’
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 syn-
tactic 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).
IP
..
. Inflected Agreeing Verb
DP morphologically significant
number information
D D
6 6
semantically significant
number information
(6) a. Each class is internally coherent, that is, all classmate nouns share
particular semantic characteristics.
b. 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 rela-
tion 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.
(7) Characteristics
First person
Animates and animate-like entities (physically similar to ani-
mates or capable of self-propulsion or determining direction of
motion)
Naturally regarded both as individual and as occurring in col-
lections; 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
others to one or two (tógúl ‘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!Ôˇ ǑdO, ‘tomato’+inv, means ‘one
tomato’ or ‘three or more tomatoes’; áádO, ‘stick’+inv, means ‘one stick’;
tógúúdÓ, ‘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 mem-
bership. 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 mo-
tivates 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 D
depends on content
of Class and Number
⎡Class: Noun ⎤ Number
(±singular) ±singular
⎣(±augmented)⎦ ±augmented
(±group)
The chapter shows further that the relationship between a class’ mnemo-
nic 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 differ-
ent 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 res-
trictive 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 distin-
guishing 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, agree-
ment will reflect that singularity and any suppletive predicates will be in
their singular or singular/dual form:
(9) Tógúl ∅- 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 indi-
cate different numbers. Two examples are (11), where agreement indicates
plurality and suppletion, singularity, and (12), where the reverse holds:
(11) MŹMde t!óMúM gya yáM -
É dôi- et
this shirt :1s:3P-too-big.S
‘This shirt is too big for me’
(12) Tóú ∅- sÓl
house 3S-be set.P
‘There are houses standing’
FRAMEWORK 9
Theoretical arguments are presented that the features that condition sup-
pletion are located on a different head from the features that trigger agree-
ment. 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 treat-
ments 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 phono-
logical 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 crosslin-
guistic 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 the-
ory of number might comprise.
Now, in light of the foregoing exposition, the term ‘morphosemantic’ itself re-
quires clarification, particularly in comparison to terms such as ‘morphosyn-
tactic’ or ‘syntacticosemantic’. ‘Morphosyntactic’ refers to the interaction
10 CHAPTER 1
(13)
Syntax
Morphology Semantics
Phonology Conceptual
Interface
Articulatory
Interface
1.4. Framework
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 fur-
thermore that the phonological structures associated with syntactic feature
bundles are inserted postsyntactically, at the terminal nodes. Let me illus-
trate 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 ex-
act 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 mod-
ules 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 sum-
mary began with semantic features in Chapter 3 and ended with morpho-
logical 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 ex-
pected, 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 de-
termined 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.
12 CHAPTER 1
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 Ko-
daseet (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 flu-
ency 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. How-
ever, 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 13
Kiowa is a rich agreement language with relatively free word order. A basic
(informationally unmarked) order is nonetheless discernible:
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.
Nouns
young men’, áá means ‘two or more trees’ and k!Ôn 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: tógúúdÓ ‘young men’, áádO ‘a tree’ and k!ÔˇǑdO
‘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).
Quantifiers may be pre- or postnominal: 顏ˇde tóú ∼ tóú 顏ˇde ‘this house’,
yŹŹ tóú ∼ tóú yŹŹ ‘two houses’, étté tóú ∼ tóú étté ‘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:
DP modifiers
Consistent with its head finality (default verb finality, postnominal modifi-
cation), Kiowa has postpositions, rather than prepositions:
Similarly, focus modifiers are suffixal. (They differ from quantifiers in being
inseparable from the noun.)
Pronouns
Kiowa has only two pronouns: nÓÓ for all first persons, whether singular, dual
or plural, inclusive or exclusive; and ám for second person, whether singular,
dual or plural. There are no third person pronouns, but deictics can be used
instead, 顏ˇde∼顏ˇgO ‘this’∼‘this.inv’ and óŹde∼óŹgO ‘that’∼‘that.inv’.
Verbs consist of two parts, an agreement prefix (next subsection) and a com-
plex verb. The latter, in which only the root is obligatory, consists of:
(35) A- pénhaa-ou
2s:3s-sugar- pour.imp
‘Pour the sugar’
(36) P!él- heM Mi hétÓ gya-ol- sÓl
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’
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, indi-
rect 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
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 Pho-
netic Alphabet) values (Lagefoged 1993). The palatal affricate, IPA [ts], is
written x, and the alveolo-palatal fricative, IPA [C], sy. The latter represen-
tation 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
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 Won-
derly, Gibson, and Kirk (1954):
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 ar-
bitrariness. 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, mem-
bers of each noun class share semantic characteristics with a natural nexus
to number concepts.
A second insight was provided by Merrifield (1959a), who refined Won-
derly, 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
22 CHAPTER 2
(3) a. To explain how Kiowa noun classes are distinguished on the basis
of number agreement in verbal prefix.
b. To argue that this methodology reveals nine classes and to
present a new nomenclature for these.
c. To show that membership of the nine classes is semantically co-
herent.
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 syn-
tax 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
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 dis-
tinguishes referential cardinalities 1, 2, and 3 or more. In order to keep
morphological and semantic terminology separate, the following notational
convention is adopted:
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. sdp
b. sdi
c. sii
d. idp
e. ids
f. idi
g. sds
h. sss
i. 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 car-
dinality 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, d-
agreement; 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
KIOWA’S NOUN CLASSES 25
The appropriateness of the mnemonic sss cannot be illustrated via the sen-
tence frames used for the previous classes. Nouns in this class are all mass
nouns and, so, are, strictly speaking, uncountable. Nonetheless, the mnemo-
nic 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).
Further motivation for the sss mnemonic comes from mass-to-count con-
version, that is, use of one water to mean ‘one portion / helping of water’.
Though the mechanisms of mass-to-count conversion themselves require clar-
ification, 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.
Second, observe that, in (35), ‘whisky’ has been converted to a count noun.
Notwithstanding, we find s-agreement; d-agreement is unacceptable.
Even when the explicit measure phrase k!Oâl ‘dish, cup’ is included, the verb
shows s-agreement.
sss nouns in virtue of their semantics: all liquids are in this class, as discussed below.
30 CHAPTER 2
See Section 4.7 for discussion of the surprising combination of incorporated s/d ‘drop’
with main verb p ‘lay’ in (ii).
KIOWA’S NOUN CLASSES 31
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 imme-
diately 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.
of unaccusatives, the first person exclusive dual and first person exclusive plural
trigger i-agreement. This is shown below for unaccusatives.8, 9
Compare (44) and (45) with (46) which has i-agreement in virtue of the
inverse-marked tógúúdÓ ‘young men’.
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) Kiowa nouns classes are internally coherent in that there are semantic
properties common to members of a given noun class.
8
Unergatives exhibit object agreement and are classed with transitives.
9
Kiowa has only two pronouns. NÓÓ, glossed as ‘1’, is used for first person singular, dual
and plural, inclusive and exclusive. Ám, glossed as ‘2’, is used for second person singular,
dual and plural. For third person, deictics are used (Watkins 1984).
KIOWA’S NOUN CLASSES 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 mis-
understandings 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 charac-
teristics of two classes. For instance, rivers are moving bodies of water. Con-
sequently, p!ÓÓ ‘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 McKen-
zie, 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 ženščina ‘woman’, devuška ‘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’, maljčik ‘boy’, tsar j ‘czar’, Karenin ‘male of the Karenin
family’. Despite these robust generalities, mužčina ‘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.
34 CHAPTER 2
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 35
Human animates. Words for humans are all sdi. For example:
Similarly, the sdi class includes designations for humans formed by at-
taching 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’.
10
‘Old’, from which comes ‘elder’, is suppletive for number in Kiowa. See Chapter 4.
36 CHAPTER 2
Inanimates. Although all animates are sdi, not all sdi nouns are an-
imate. One such is hêˇǐii ‘doll’. Its subsumption suggests that animate-
likeness 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:
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.
KIOWA’S NOUN CLASSES 37
However, three terms given by Watkins do not lend themselves to this way
of thinking: k!ÔOsoMǔ ‘whetstone’ [lit.: knife-grind], hôux!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
13
She also includes mOnx!óˇ ‘fingernail’. In my fieldnotes, it is sds, as are Onx!óˇ and
Onk!Ón, both of which mean ‘toenail’.
38 CHAPTER 2
Gloss 1/2 3
neck k!ól k!óttO
toe Onthál OntháttO
tongue dén déMŹMdO
spinal cord pâisen pâiseM MidO
spine gómthoM gómthoM gO
tooth zóMúM zêm
tripe ÓÓbŹMŹM ÓÓbŹMŹMgÓ
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.
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.
KIOWA’S NOUN CLASSES 39
15
Dr McKenzie gives the first form, which is regular given (89), Mrs Kodaseet the second,
which unexpectedly shows zero marking after l.
40 CHAPTER 2
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.
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 is very small, with only eight members (four, according to
previous reports), mostly hair and fruit.16
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.
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 col-
lections 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 bre-
ak 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 p-
final nouns like k!yagóp ‘brain’, class membership is still detectable through Merrifield’s
method of examining the agreement the noun triggers on the verb for referential cardinal-
ities 1, 2, and 3.
18
Mrs Dupoint said: An k!yagóp dé-pÓttO. T!ěǐnéxeiyot!ǎǐmO gO k!yagóp ét-
khOleitOnmO. E-ÓÓt!olǑǑ ‘I eat brains. They cook cook them together with chicken eggs.
44 CHAPTER 2
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.
Like the idi class, the sds class is comparatively small. It subsumes:19
Kiowa has two more classes with symmetric mnemonics, ppp and sss. These
are strongly semantically coherent.
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.
Gloss 1/2/3
ash sOphán
hail, sleet t!én
ice téMŹMgya
rain sép
snow t!ól
eyebutter, sleepy dust táásek!On
snot, mucous 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 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’.
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 speak-
ers. Rather, they appear to be idp nouns, with the inverse-marked noun
corresponding to the minimal part:
Such speakers permit a corresponding dual reading for the granular sense.
However, speakers for whom granular mass nouns are ppp strongly reject
sentences like (70) and (71).20 Instead, they offer syÓnde xóŹséˇǒǔgya ‘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.
48 CHAPTER 2
2.4.6. Summary
The foregoing discussion of the noun classes has shown that they are inter-
nally semantically coherent, that is, that there are properties that members
of each class generally share. These are summarized below.
50 CHAPTER 2
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.
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 Ap-
pendix), k!yᡠá
ˇ hyóp ‘men’ triggers both i- and a-agreement. It relates an
incident in which Mr Bointy was not a participant. The three opening sen-
tences state where the men were and what they were doing—a simple report-
ing of facts, with little scope for empathy. The verbs show i-agreement.
a-agreement in (82) does not motivate an sda class, as, generally, xêˇǐgO
‘horses’ triggers i-agreement, so that xêˇǐgO is best regarded as the inverse-
marked form of the sdi noun xêˇǐ ‘horse’. A typical example is:
when used for adults. a-agreement is all but obligatory for KÓŹgú ‘Kiowas’:
The reflexive ‘oneself’ might be regarded as involving obligatory empathy, rather than as
being in some mysterious sense obligatorily animate plural.
54 CHAPTER 2
Similarly, demonstratives agree for inverse with their noun. This is shown
below for the sdi noun sân ‘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.
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 KÓŹgú,
in contrast to final lows in KÓŹmaa ‘Kiowa woman’, KÓŹtǒgya ‘Kiowa language’).
KIOWA’S NOUN CLASSES 55
And there is a unique l∼p alternation in words built on tól ‘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 tâl
‘skunk’ and idp k!Oâl ‘dish’; -∅ on sdi t!áp ‘deer’, ids k!óp ‘mountain’ and
idi k!yagóp ‘brain’; -dO on idi k!Ôn ‘tomato’, sdi sân ‘child’, and idp són
ˇŹˇ ‘white faced cattle’, idp mÓx!á
‘grass’; -mO on sdi t!óút!á ˇŹˇ ‘paper’, ids t!á
ˇŹˇ
‘sheet’; -op is sdi sǎǎné ‘snake’ and idp mÓnsóúde ‘bracelet’; and so on.
Even the competing -gu∼-̂i do not define classes, as, by lexical coincidence,
56 CHAPTER 2
Themes
This suggests that these thematic nouns have e when non-inverse because
they denote inherent pairs. Consequently, the vowel appears to be inde-
pendent 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.
58 CHAPTER 2
The relationship between themes and the meaning of the nouns to which
they attach shows that themes are not indicative of phonological noun clas-
⎡ ⎤⎡ ⎤
+consonantal −consonantal
(i) [−αback] / ⎣−sonorant ⎦ ⎣+sonorant ⎦
[αback] →
αback
27
The combination b+y standardly simplifies to b; for instance compare (96)’s gyát,
:1p:3p, from d+y. . . , with bát, :2p:3p, from b+y. . . .
KIOWA’S NOUN CLASSES 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!ÓÓgya ‘baby’ and bélkŹtkya ‘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:
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 seman-
tic 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
62 CHAPTER 3
(2) DP
NumberP D
Given Claim #1, this means that features are distributed as in (3).
(3) DP
NumberP D
±singular
±augmented
⎡Class: Noun ⎤ Number
(±singular) ±singular
⎣(±augmented)⎦ ±augmented
(±group)
The features in (3) are not all of the same status. Those on Class and Num-
ber 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 spec-
ification 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 ⎤ Number
(±singular) ±singular
⎣(±augmented)⎦ ±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):
This chapter contains the core ideas of the current study: proposals concern-
ing 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 refer-
ential 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 primi-
tive, 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 expli-
cable. 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 phe-
nomena, 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. Deriva-
tions 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 mnemon-
ically possible classes. It argues that Kiowa almost optimally exploits the
space of possible mnemonics generated by the features at its disposal.
The main claim of this section is that 1, 2 and 3, the values of referen-
tial cardinality, are not atomic. Rather, they are composed of two fea-
tures, [±singular] and [±augmented], to be defined below. The evidence
for this claim comes from the distribution of i-agreement and inverse mark-
ing. (This section recapitulates, in part, arguments from Noyer 1992 and
Harbour 2003a. The analysis of cardinality is essentially Noyer’s.)
The values 1, 2 and 3 tempt one to posit three monovalent number features,
[singular], [dual], [plural]. However, such privative features predict that there
66 CHAPTER 3
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
{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 ex-
plaining 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
. . . . . .
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.
68 CHAPTER 3
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 lat-
tice, partitioning it into [−singular] and [+singular], and [−augmented] and
[+augmented], regions. As a convention, I will define only [+F].5 Then:
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]
(13) x = ∅
[+augmented]
. . . . . .
[−augmented]
[−singular +augmented]
[−singular −augmented]
. . . . . . [+singular −augmented]
Summary
(15) DP
NumberP D
Noun Number
±singular
±augmented
NUMBER FEATURES 71
3.2. 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.
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:
ber and the right column gives the corresponding agreement type(s).
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 Num-
ber. This happens with [−singular −augmented] / [−singular +augmented],
which, according to (16), yields d-agreement for 2 / p-agreement for 3. Con-
sequently, 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 generaliza-
tion 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:
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.
(21) DP
NumberP D
(22) DP
NumberP D
u singular
u augmented
Class: Noun Number
(±singular) ±singular
(±augmented) ±augmented
(23) Valuation of D
Uninterpretable number on D is valued by a computation over Num-
ber and Class. The features of both are replicated on D.
(24) DP
NumberP D
u singular
u augmented
Class: Noun Number
−augmented +singular
−augmented
(25) DP
NumberP D
+singular
−augmented
Class: Noun Number
−augmented +singular
−augmented
(26) DP
NumberP D
u singular
u augmented
Class: Noun Number
−augmented −singular
−augmented
(27) DP
NumberP D
−singular
−augmented
Class: Noun Number
−augmented −singular
−augmented
(28) DP
NumberP D
u singular
u augmented
Class: Noun Number
−augmented −singular
+augmented
(29) DP
NumberP ⎡ D ⎤
−singular
⎣−augmented⎦
Class: Noun Number +augmented
−augmented −singular
+augmented
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 uninterpretabil-
ity. 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. Depend-
ing 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
(31) DP
NumberP D
dÓ
Class: fish Number
ÓÓpŹˇŹˇ ∅
This section applies the mechanisms of Section 3.2.2 to the Class features of
Section 3.2.1 to derive other class mnemonics:
(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 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
αsingular
βaugmented
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.
(35) DP
NumberP D
u singular
u augmented
Class: Noun Number
−singular +singular
−augmented
(36) DP
NumberP ⎡ D ⎤
−singular
⎣+singular ⎦
Class: Noun Number −augmented
−singular +singular
−augmented
(37) DP
NumberP D
−singular
−augmented
Class: Noun Number
−singular −singular
−augmented
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 D
−singular
+augmented
Class: Noun Number
−singular −singular
+augmented
The noun will not be inverse marked and will trigger p-agreement. This
derives the idp class.
(39) DP
NumberP D
−singular
−augmented
Class: Noun Number
−singular −singular
−augmented −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
NumberP ⎡ D ⎤
−singular
⎣+singular ⎦
Class: Noun Number −augmented
−singular +singular
−augmented −augmented
(41) DP
NumberP ⎡ D ⎤
−singular
⎣−augmented⎦
Class: Noun Number +augmented
−singular −singular
−augmented +augmented
Given the conflicts, both nouns will be inverse marked and trigger i-agreement.
This derives the idi class.
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,
82 CHAPTER 3
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 un-
accusatives, the first person exclusive dual and first person exclusive plural
trigger i-agreement.
Compare (43) and (44) with (45) which has i-agreement in virtue of the
inverse-marked tógúúdÓ ‘young men’.8
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-.
Syncretism #3. For indirect and direct objects, the first person in-
clusive 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
(51) A- x!óŹgyá
1s-fall.s/d.pf
‘I fell’
(52) E- x!óŹgyá
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.: dét 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.
84 CHAPTER 3
(53) E- k!úŹgyá
1i-fall.p.pf
‘They and I fell’
Given the mechanisms argued for above, we can account for these syn-
cretisms 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 agree-
ment 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.
10
[−author −hearer] corresponds to third person. However, as here, one may also repre-
sent 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 ⎡ D ⎤
u author
⎢u hearer ⎥
⎢ ⎥
⎡Class: I ⎤ Number ⎣u singular ⎦
+author +singular u augmented
⎣−hearer ⎦ −augmented
+singular
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.
(56) DP
NumberP ⎡ D ⎤
+author
⎢−hearer ⎥
⎢ ⎥
⎡Class: I ⎤ Number ⎣+singular ⎦
+author +singular −augmented
⎣−hearer ⎦ −augmented
+singular
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:
86 CHAPTER 3
(57) DP
NumberP ⎡ D ⎤
+author
⎢−hearer ⎥
⎢ ⎥
⎡Class: I ⎤ Number ⎢−singular ⎥
⎢ ⎥
+author −singular ⎣+singular ⎦
⎣−hearer ⎦ ±augmented ±augmented
+singular
(58) DP
NumberP ⎡ D ⎤
+author
⎢+hearer ⎥
⎢ ⎥
Class: I, you⎤ Number ⎢−singular ⎥
⎡ ⎢ ⎥
+author −singular ⎣+singular ⎦
⎢+hearer ⎥ −augmented −augmented
⎢ ⎥
⎣+singular ⎦
−augmented
(59) DP
NumberP ⎡ D ⎤
+author
⎢+hearer ⎥
⎢ ⎥
Class: I, you⎤ ⎢−singular ⎥
⎡ Number ⎢ ⎥
+author −singular ⎢+singular ⎥
⎢+hearer ⎥ ⎢ ⎥
⎢ ⎥ +augmented ⎣−augmented⎦
⎣+singular ⎦ +augmented
−augmented
Nominative and Non-Nominative (see Adger and Harbour 2007 for an anal-
ysis 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] → ∅ /
⎡ ⎤
Non-nom.
(61) ∅ / ⎣ +author ⎦
[αhearer] →
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)
⎡ ⎤ Nominative(58)
⎡ ⎤ Nominative(59)
⎡ ⎤
+author +author +author
⎢−hearer ⎥ ⎢+hearer ⎥ ⎢+hearer ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢−singular ⎥ ⎢−singular ⎥ ⎢−singular ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣+singular ⎦ ⎣+singular ⎦ ⎢+singular ⎥
⎢ ⎥
±augmented −augmented ⎣−augmented⎦
+augmented
(64) Non-nominative(57)
⎡ ⎤ Non-nominative(58)
⎡ ⎤ Non-nominative(59)
⎡ ⎤
+author +author +author
⎢−hearer ⎥ ⎢+hearer ⎥ ⎢+hearer ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢−singular ⎥ ⎢−singular ⎥ ⎢−singular ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣+singular ⎦ ⎣+singular ⎦ ⎢+singular ⎥
⎢ ⎥
±augmented −augmented ⎣−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, constitut-
ing 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 nÓÓ
1 and ám 2 are never inverse marked. I assume that this is a specially
conditioned form of zero 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 num-
ber 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:
(67) DP
NumberP ⎡ D ⎤
−singular
⎣−augmented⎦
NumberP Adjective +augmented
(68) DP
NumberP D
bO
NumberP Adjective
k!óm
Class: man Number
ˇ á
k!yá ˇ hẐˇǐ ∅
ˇ á
Standard phonology yields k!yá ˇ hŹˇk!ǒǔbO.
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,
90 CHAPTER 3
顏ˇ, and distal, óŹ. They are found, for instance, in the locatives, 顏ˇhOO ‘here’
and óŹhyOO ‘there’. Demonstratives comprise a deictic root and a number-
sensitive suffix, -gO if modifying an inverse-marked noun, -de otherwise.
For thalŹŹ ‘boy’, an sdi noun, the inverse-marked demonstratives occur when
referential cardinality is 3. For an idp noun, say, they occur for 1.
(72) DemonstrativeP
DP Demonstrative
u singular
u augmented
NumberP ⎡ D ⎤
−singular
⎣−augmented⎦
Class: boy Number +augmented
−augmented −singular
+augmented
(73) DemonstrativeP
DP Demonstrative
⎡ ⎤
−singular
⎣−augmented⎦
+augmented
NumberP ⎡ D ⎤
−singular
⎣−augmented⎦
Class: boy Number +augmented
−augmented −singular
+augmented
(74) DemonstrativeP
DP Demonstrative
gO
NumberP D
op
Class: boy Number
thalii ∅
Inserting óŹ, the distal deictic, we have thalyóp óŹgO. Variant word orders
are derived by syntactic means irrelevant here. The non-inverse-marked form,
DP+óŹde, 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:
The simplest way to derive common number marking between relative clauses
92 CHAPTER 3
and deictically modified DPs is to suppose that the two are structurally iden-
tical 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 decom-
pose 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 ex-
planation 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
They also form part of third person kin possession (sometimes with -de
devoiced; see Watkins 1984 for possession in general):
In (80), both the possessed noun and the possessor suffix are inverse marked.
Further modification with ‘only’ leads, optionally, to triple inverse marking:
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 in-
stances 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
(82) DP
NumberP D
u singular
u augmented
Class: Noun Number
(±singular) ±singular
(±augmented) ±augmented
With the underlying classificatory features of five (and a half) classes uncov-
ered, 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 posi-
tive 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 inde-
pendently 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
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 97
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 inter-
feres with this correlation. The classes under consideration in this section,
and in the second part of the chapter generally, also interfere with the cor-
relation. 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.
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:
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:
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:
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].
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 ⎡ D ⎤
−singular
⎣+augmented⎦
Class: Noun Number +group
+group −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:
arising via the mechanisms discussed by Noyer 1998, Harbour 2003a, a di-
gression profitably avoided here. See Section 4.5.1 for crucial evidence that
the representation of [±augmented] in these cases is indeed different from
others.)
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
(96) DP
NumberP D
u singular
u augmented
Class: Noun Number
+group ±singular
−augmented
(97) DP
NumberP D
±singular
−augmented
Class: Noun Number
+group ±singular
−augmented
This is all but identical to an sdp DP for referential cardinalities 1/2. s/d-
agreement result, as desired.
For ids nouns of referential cardinality 1/2, Class is [−singular +group]:
(98) DP
NumberP D
u singular
u augmented
Class:
Noun
Number
−singular ±singular
+group −augmented
(99) DP
NumberP D
u singular
u augmented
Class: Noun Number
+group −singular
+augmented
(100) DP
NumberP D
+singular
−augmented
Class: Noun Number
+group −singular
+augmented
(101) DP
NumberP D
+singular
−augmented
Class:
Noun
Number
−singular −singular
+group +augmented
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 D
u singular
u augmented
Class: Noun Number
+augmented ±singular
−group ±augmented
NUMBER FEATURES 105
(103) DP
NumberP D
−singular
+augmented
Class: Noun Number
+augmented ±singular
−group ±augmented
3.4.3. Summary
We can now expand the list of noun classes and classificatory features.
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 con-
stant agreement for both granular and non-granular nouns. If we abandoned
the assumption, the combination [+augmented +group] would ensure con-
stant s-agreement for non-granular mass nouns; addition of [−group] to Class
would ensure constant p-agreement for granular mass nouns. The second as-
sumption 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 expos-
itory 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⎤ D
−singular +singular
⎣+augmented⎦ −augmented
+group
NUMBER FEATURES 107
(107) DP
Class: Noun D
−singular −singular
+augmented +augmented
3.5.1. Conjunction
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.
108 CHAPTER 3
(108) DP
NumberP D
u singular
u augmented
⎡Class: andP ⎤ Number
(±singular) ±singular
⎣(±augmented)⎦ ±augmented
(±group)
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
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 coher-
ent. In this chapter, the feature system that underlies the classification has
been examined. The results are that [±singular] and [±augmented], the fea-
tures that compose referential cardinality, are also used for noun classification
along with a third feature, [±group], which qualifies type of augmentation.
(110) DP
NumberP D
u singular
u augmented
⎡Class: Noun ⎤ Number
(±singular) ±singular
⎣(±augmented)⎦ ±augmented
(±group)
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
One can imagine how such reasoning would continue, but it is rather super-
ficial, 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] ⎬
(112) [−singular] [−augmented] [−group]
⎩ ⎭⎩ ⎭⎩ ⎭
∅ ∅ ∅
112 CHAPTER 3
Only three of the resulting mnemonics are new: sip, sis and iip, correspond-
ing 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, ob-
serve 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 liq-
uids at once.
NUMBER FEATURES 113
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 seman-
tic properties of nouns, not with mnemonics, which underdetermine class fea-
tures, but with the class features themselves directly. For instance, footwear
might be sdp not because it is default, but because it is [−group]; and sub-
classes 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.
3.8.1. Cardinality
I now present four simple proofs justifying (9) and the exclusion of [+singular
+augmented].
3.8.2. Grouphood
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.
∗ (A, B) ≡ ¬A → B (3.1)
¬ ∗ (A, B) ≡ A → ¬B (3.2)
| ¬ ∗ (A, B) | = 0 iff | A → ¬B | = 0
iff | A | = | B | = 1 (3.3)
| ¬ ∗ (A, B) | = 0 iff | ∗(A, B) | = 1
So, by (3.1):
| ¬ ∗ (A, B) | = 0 iff | ¬A → B | = 1 (3.4)
| A | = | B | = 1 iff | ¬A → B | = 1
Several predicates in Kiowa are sensitive to the number of their inner argu-
ment; 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 cardinal-
ity 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 bear-
ing 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 allo-
morphy 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
118 CHAPTER 4
4.1. Suppletion
Kiowa’s number-sensitive predicates fall into two classes. Those in (1) display
an S∼D/P opposition, those in (2), S/D∼P.
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 -gyá.
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 tóúdé ‘shoe’. In (3), referential
cardinality is 1 and we find s-agreement and the s-form of the predicate, êl.
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
120 CHAPTER 4
argument. (Tóúdé ‘shoe’ loses its -de when modified; Section 2.6.2.)
(12) a. E- k!ÓÓ
1/3I-be lying.S/D
‘It is lying’ or ‘We two are lying’
b. NÓ- k!úl
:1s:3I-be lying.P
‘Mine are lying’
(13) a. YáM - dôi- et
:1s:3P-too-big.S
‘It is too big for me’
b. ∅- SÓl
3S-be sitting.P
‘It is sitting’
4.1.2. Clarification
relevant to distinguishing suppletion from forms that are related but by non-
obvious means. Only the definition of suppletion is central to what follows
and even this may be skipped by readers familiar with the concept.
‘Suppletion’ has been used to cover three phenomena that probably deserve
explanation by different theoretical mechanisms. English examples are:
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.
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 āi∼æ alternation of satisfy∼satisfaction (pp. 201−202). See also
Yang (1999, 2002) for more recent discussion of consequences of distinguish-
ing go∼went from sing∼sang alternations.
Now consider -ed∼∅, which I regard as allomorphy:
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.
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.
Rightwards spreading of high tone onto -i∼-ǐ and shortening of the root vowel
are phonologically regular (Watkins 1984, Harbour 2002). Second, note the
alternation (y)i∼ya:
By orthographic convention, V́V̀ is written V̂V. So, the desired yâi 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):
Observe the lowering of the high vowel uu in ‘hit’ and ‘lay.p’ in (25). Gen-
eralizing lowering to all high vowels, we derive ‘die.pf’ from hẐˇǐ-m:
Recall that velars become dental before e (Section 2.6.2). This switching,
together with processes affecting hẐˇǐ ‘die’, apply to kŹˇŹˇ ‘pull’:
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
4.2.1. Assumptions
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
(31) DP
.. D
.
u singular
VP u augmented
NumberP V
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 pre-
dictions 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 127
4.2.2. 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
(33) DP
NumberP D
αsingular
βaugmented
Class: Noun Number
∅ αsingular
βaugmented
6
In (35), as in suppletive vocabulary entries below, both forms have conditioning con-
texts, [+singular] for êl, [−singular] for bẐn. 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
128 CHAPTER 4
√
(35) big ⇔ êl / [+singular]
⇔ bẐn / [−singular]
(36) VP
NumberP V
bẐn
Class: shoe Number
∅ −singular
−augmented
(37) DP
NumberP D
−singular
−augmented
Class: shoe Number
∅ −singular
−augmented
(40) VP
NumberP V
k!úl
Class: shoe Number
∅ −singular
+augmented
(41) DP
NumberP D
−singular
+augmented
Class: shoe Number
∅ −singular
+augmented
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 ⇔ ól / [−augmented]
⇔ p!él / [+augmented]
130 CHAPTER 4
(44) VP
NumberP V
ót
NumberP Adjective
bẐn
Class: shoe Number
tóú- −singular
−augmented
4.2.3. Summary
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.
For 2, the predicate is still in its s/d-form but bears d-agreement and the
noun is not inverse marked.
For 3, the noun is again inverse marked and triggers i-agreement. Again, the
predicate, now in its p-form, still reflects referential cardinality.
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.)
For sdi nouns, such as báou ‘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 indi-
rect object (possessor), is éˇ-ét. In the second clause, Number is [−singular
+augmented], which conditions the d/p-form of ‘small’:
√
(50) small ⇔ syÓn / [+singular]
⇔ syân / [−singular]
Class and Number have opposite specifications of [±augmented]. So, the
possessed noun, báougO, is inverse marked and triggers i-agreement. The
resulting verb, with agreement for the possessor, is gÓ-syân, which shows
number-conditioned suppletion despite number-neutral i-agreement.
In contrast to (49), consider (51), with an idp noun.
(51) MŹMgO
É pŹMáádO e- ét, né óŹde 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, 顏ˇgO, and noun, pŹˇáádO, are inverse
marked and the verb bears i-agreement. The resulting e-ét 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 trig-
gers p-agreement via D. These, with (50), yield gya-syân.
Consider, lastly, an appositively used s∼d/p adjective with the idi noun
álOO ‘apple’. Since we are only concerned with the DP, I simplify the discus-
sion 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
AGREEMENT AND SUPPLETION 133
(55) DP
NumberP ⎡ D ⎤
−singular
⎣+singular ⎦
NumberP Adjective −augmented
(56) DP
NumberP D
tO
NumberP Adjective
êl
Class: apple Number
álOO +singular
−augmented
(57) DP
NumberP D
−singular
−augmented
NumberP Adjective
(58) DP
NumberP D
∅
NumberP Adjective
bẐn
Class: apple Number
álOO +singular
−augmented
(59) DP
NumberP ⎡ D ⎤
−singular
⎣−augmented⎦
NumberP 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 bẐn for ‘big’.
136 CHAPTER 4
(60) DP
NumberP D
dO
NumberP Adjective
bẐn
Class: apple Number
álOO +singular
−augmented
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 ⎡ D ⎤
+author
⎢−singular ⎥
⎢ ⎥
⎢+singular ⎥
⎡Class: I (you)⎤ Number ⎢ ⎥
+author −singular ⎢±augmented ⎥
⎢ ⎥
⎢+singular ⎥ ±augmented ⎣(+hearer) ⎦
⎢ ⎥
⎣(±hearer) ⎦ (−augmented)
(−augmented)
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) a. E- x!óŹgyá
1i-fall.s/d.pf
‘She and I fell’
b. E- k!úŹgyá
1i-fall.p.pf
‘They and I fell’
√
(63) a. lay ⇔ x!óú / [−augmented]
⇔ k!úú / [+augmented]
b. detr ⇔ -Źgyá
(64) a. Ba-thóúya
2p- move.s/d.impf
‘You and I are moving around’ (Watkins, p.c.)
b. Ba-zéMŹMma
2p- move.p.impf
‘You, I and (s)he/they are moving around’ (Watkins, p.c.)
√
(65) wander ⇔ thóú / [−augmented]
⇔ z顏ˇ / [+augmented]
Summary
4.4.1. Collectives
(66) DP
NumberP D
+singular
−augmented
Class: Noun Number
+group −singular
+augmented
In the first two sentences, agreement and suppletion both transparently re-
flect 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 ⇔ xéŹ / [−augmented]
⇔ sÓÓ / [+augmented]
b. stat ⇔ -l
(71) VP
NumberP V
sÓl
Class: Noun Number
+group −singular
+augmented
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.
140 CHAPTER 4
In each case, ‘hair’ triggers s-agreement. However, the s-form of the predi-
cates is unacceptable and the d/p-form must be used. Again, this is exactly
as expected given (66) and the vocabulary items:
√
(77) long ⇔ kŹˇŹˇnŹŹ / [+singular]
⇔ kyó ˇŹˇ / [−singular]
√
(78) short ⇔ xáádóú / [+singular]
⇔ xéŹ / [−singular]
Thus, nouns that trigger s-agreement when referential cardinality is 3
trigger suppletion as expected.
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 Num-
ber (Section 3.5), then suppletion in such cases cannot depend on Num-
ber. 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.
AGREEMENT AND SUPPLETION 141
Evidence for how sss mass nouns behave with [±singular]-sensitive pred-
icates is harder to come by, owing to the meanings of these predicates. Sen-
tences 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 ÓlhǑǑsyan ‘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: syân 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.)
Pluralia tantum nouns, such as hólda ‘dress’, khÓÓdé ‘trousers’, tóú ‘teepee’,
ˇkút ‘book, letter’, and t!ó
tó ˇ ú
ˇ gya ‘shirt’, are ppp, [+augmented −group].
Their behavior with respect to suppletion only partly conforms to the gener-
alization that suppletion depends on the feature content of Number, or Class,
when Number is absent: [±singular]-sensitive predicates obey the generaliza-
tion, [±augmented]-sensitive ones do not, displaying the p-form throughout.
Consider, first, a [±singular]-sensitive predicate.
(95) NumberP
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 145
Animate plurals share this suppletive pattern with reflexives. For in-
stance, there are three expressions built on suppletive predicates and de-
manding reflexive agreement: hóút!al ‘part ways’ (travel-sever.s/d), hóúol
‘turn off’ (travel-drop.s/d), and hónx!oigya ‘come late’ (last-land.s/d):8
8
Hónx!oigya ‘come late’ is a dative reflexive, with the reflexive triggering, generally,
i-agreement. See pp. 146ff.
146 CHAPTER 4
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 *hóúthaa and *hónk!uigya, are explicitly rejected. Consistently, if the
subject is singular, the predicate appears in s/d-form, as in:
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 supple-
tion 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 ⎡ D ⎤
−singular
⎢+singular ⎥
⎢ ⎥
Class: Animate ⎣−augmented⎦
Number
−singular +singular +empathic
+empathic −augmented
4.6. Conclusion
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 sem-
antically 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:
The glosses suggest that the meanings of each s∼d/p adverb pair are iden-
tical. 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.
AGREEMENT AND SUPPLETION 151
To be sure, there are cases where the members of each pair are inter-
changeable. (Dôi ‘too’ in (113) is irrelevant; compare (123)−(124).)
(125) emphasizes repeated gain of lots of money; I surmise that étté (126)
is more stative, implying constant wealth.
A similar difference arises for other pairs in (111). For instance, for
‘small’, the [−singular] adverb syânde means ‘a little at a time’, but the
[+singular] adverb syÓnde simply ‘a little’.
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).
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 sec-
ondary 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 rel-
ative clauses (cf., the same suggestion for demonstratives in Section 3.2.4).
Their meaning would then be something like ‘being big’, and, being agree-
mentless, 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):
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 pre-
fix 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 argu-
ments, 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 par-
ticipants in as little as a single vowel, the question is how so much meaning
gets into so little sound.
157
158 CHAPTER 5
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 vocab-
ulary list and explicit derivation of the entire inventory (see also Appendix
5.6). Such delimitation is reasonable within the confines of the current inves-
tigation 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
Section 5.1 presents the full inventory of prefixes and explains the slightly
modified prefix notation used throughout the chapter. In addition, it summa-
rizes 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, pre-
dictable 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 methodolog-
ical and theoretical points of the previous paragraph apply again here.
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 161
x :y : z
∅ 3s 3d 3p 3i 3a
1s:(3a:) a gya nen gyat dé de
1ex:(3a:) e é∗ et ét∗ ét ét
1in:(3a:) ba bá∗ bet bát∗ bét bé
2s:(3a:) em a men bat bé be
2d:(3a:) ma má∗ mén mán∗ mén∗ mé
2i:(3a:) ba bá∗ bet bát∗ bét bé
3s:(3a:) ∅ ∅ eM gya é em
3d:(3a:) eM éM∗ én én∗ én én
3i:(3a:) e é∗ et ét∗ ét ét
3a:(3a:) á á∗ et gyá∗ et ém
∅/2s/3s:1s: éM éM né yáM nÓ né
2d:1s: mâa∗ mâa∗ ménêi∗ mánẐi∗ mÓnÔO∗ d∼i
2i:1s: bâa∗ bâa∗ bédêi∗ bágẐi∗ bÓdÔO∗ d∼i
3d:1s: êM Mi∗ êMiM∗ éMnêi∗ éMnẐi∗ éMnÔO∗ d∼i
3i:1s: êi∗ êi∗ édêi∗ égẐi∗ édÔO∗ d∼i
3a:1s: âa∗ âa∗ dêi∗ gyâa∗ dÔO∗ d∼i
any:1d/p: dÓ dÓ dét gyát dÓt d∼i
∅/1s:2s: em gyá nén yán gÓ d∼i
other :2s: gO gÓ dét gyát gÓt d∼i
any:2d: mÓ mÓ mén mán mÓn d∼i
any:2i: bÓ bÓ bét bát bÓt d∼i
1s:3s: gyá nén yán gÓ d∼i
1ex:3s: êi∗ édêi∗ égẐi∗ édÔO∗ d∼i
1in:3s: bâa∗ bédêi∗ bágẐi∗ bÓdÔO∗ d∼i
∅/2s/3s:3s á én án Ó d∼i
any:3d: mé mén mén mén d∼i
any:3i: bé bét bét bét d∼i
2d:3s: mâa∗ ménêi∗ mánẐi∗ mÓnÔO∗ d∼i
2i:3s: bâa∗ bédêi∗ bágẐi∗ bÓdÔO∗ d∼i
3d:3s: êMiM∗ éMnêi∗ éMnẐi∗ éMnÔO∗ d∼i
3i:3s: êi∗ édêi∗ égẐi∗ édÔO∗ d∼i
3a:3s: âa∗ dêi∗ gyâa∗ dÔO∗ d∼i
162 CHAPTER 5
The table requires the following comments: (a) Cells representing impos-
sible 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’ can-
not 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’. Con-
sequently, 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 impossi-
ble argument combination in Kiowa.) (f) Third plural inanimate intransitive
agreement, 3p:∅, absent from the table, is gya.
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 163
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 V
x y z
(11)
Prefix
⎡ x ⎤ ⎡ y ⎤ z
−singular −singular +singular
⎢+singular ⎥ ⎢+singular ⎥ −augmented
⎢ ⎥ ⎢ ⎥
⎣−augmented⎦ ⎣−augmented⎦
+empathic +empathic
Phonology
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] αF βG γH βG γH
Feature insertion inserts the (contextually) unmarked value of a fea-
ture. 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 165
5.2.1. ‘any’
⎡ y ⎤ z
−author +singular
⎢+hearer ⎥ −augmented
⎢ ⎥
⎣−singular ⎦
−augmented
Or the agent could be first person exclusive, any number, or third person,
any number:
⎡ x ⎤ ⎡ y ⎤ z
(+author) −author +singular
⎢(−hearer) ⎥ ⎢+hearer ⎥ −augmented
⎢ ⎥ ⎢ ⎥
⎣ ±singular ⎦ ⎣−singular ⎦
±augmented −augmented
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.
(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).
THE AGREEMENT PREFIX 167
(+author) ⎡ y ⎤
(−hearer) −author
(24) ± singular
→ ∅ / [ ] ⎢+hearer
⎢
⎥
⎥
⎣+singular ⎦
± augmented
−augmented
(25)
⎡ x ⎤ ⎡ y ⎤ z
(+author) −author ±singular
⎢(−hearer) ⎥ ⎢+hearer ⎥ ±augmented
⎢ ⎥ ⎢ ⎥
⎣ ±singular ⎦ ⎣+singular ⎦
±augmented −augmented
(26)
x ⎡ y ⎤ z
−author ±singular
⎢+hearer ⎥ ±augmented
⎢ ⎥
⎣+singular ⎦
−augmented
(27)
x ⎡ y ⎤ z
−singular −author ±singular
⎢+hearer ⎥ ±augmented
⎢ ⎥
⎣+singular ⎦
−augmented
Recall (Sections 2.5 and 4.5.2) that transitive agreement prefixes are system-
atically 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:
168 CHAPTER 5
|
⎡ y ⎤
(28) −singular → ∅
⎢+singular ⎥
⎢ ⎥
⎣−augmented ⎦
+empathetic
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 ]
The effect of (29), syncretism of first person with third person, arises else-
where and again requires person deletion, though under a different condi-
tioning 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 169
(31) x :y : z
∅ 3s 3d 3p 3i 3a
2d:1s: mâa∗ mâa∗ ménêi∗ mánẐi∗ mÓnÔO∗ d∼i
2d:3s: mâa∗ ménêi∗ mánẐi∗ mÓnÔO∗ d∼i
2i:1s: bâa∗ bâa∗ bédêi∗ bágẐi∗ bÓdÔO∗ d∼i
2i:3s: bâa∗ bédêi∗ bágẐi∗ bÓdÔO∗ d∼i
3d:1s: êMiM∗ êMiM∗ éMnêi∗ éMnẐi∗ éMnÔO∗ d∼i
3d:3s: êMiM∗ éMnêi∗ éMnẐi∗ éMnÔO∗ d∼i
i:1s: êi∗ êi∗ édêi∗ égẐi∗ édÔO∗ d∼i
i:3s: êi∗ édêi∗ égẐi∗ édÔO∗ d∼i
a:1s: âa∗ âa∗ dêi∗ gyâa∗ dÔO∗ d∼i
a:3s: âa∗ dêi∗ gyâa∗ dÔO∗ 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 (éˇ) and ∅:3s:3s (á) 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 Pān.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.
170 CHAPTER 5
+author
(32) → ∅ / [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
x :y : z
∅ 3s 3d 3p 3i 3a
1s:(a:) a gya nen gyat dé de
2s:(a:) em a men bat bé be
2d:(a:) ma má∗ mén mán∗ mén∗ mé
2i:(a:) ba bá∗ bet bát∗ bét bé
3s:(a:) ∅ ∅ eM gya é em
3d:(a:) eM éM∗ én én∗ én én
i:(a:) e é∗ et ét∗ ét ét
a:(a:) á á∗ et gyá∗ et ém
∅/2s/3s:1s: éM éM né yáM nÓ né
2d:1/3s: mâa∗ mâa∗ ménêi∗ mánẐi∗ mÓnÔO∗ d∼i
2i:1/3s: bâa∗ bâa∗ bédêi∗ bágẐi∗ bÓdÔO∗ d∼i
3d:1/3s: êMiM∗ êMiM∗ éMnêi∗ éMnẐi∗ éMnÔO∗ d∼i
i:1/3s: êi∗ êi∗ édêi∗ égẐi∗ édÔO∗ d∼i
a:1/3s: âa∗ âa∗ dêi∗ gyâa∗ dÔO∗ d∼i
any:1d/p: dÓ dÓ dét gyát dÓt d∼i
∅/1s:2s: em gyá nén yán gÓ d∼i
other :2s: gO gÓ dét gyát gÓt d∼i
any:2d: mÓ mÓ mén mán mÓn d∼i
any:2i: bÓ bÓ bét bát bÓt d∼i
1s:3s: gyá nén yán gÓ d∼i
∅/2s/3s:3s á én án Ó d∼i
any:3d: mé mén mén mén d∼i
any:i: bé bét bét bét d∼i
THE AGREEMENT PREFIX 171
Duals
Prefixes where x /y=3d are predictable from prefixes where x /y=3i, by nasal-
ization 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 : z
3s 3d 3p 3i 3a
any:3d: mé mén mén mén mén
any:3i: bé bét bét bét bét
172 CHAPTER 5
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).
(35) x :y : z
∅ 3s 3d 3p 3i 3a
3d:1s: êMiM∗ êMiM∗ éMnêi∗ éMnẐi∗ éMnÔO∗ éMnêi∗
3i:1s: êi∗ êi∗ édêi∗ égẐi∗ édÔO∗ édêi∗
(36) x: z
∅ 3s 3d 3p 3i 3a
3d: eM éM∗ én én∗ én én
3i: e é∗ et ét∗ ét ét
(37) x :y : z
∅ 3s 3d 3p 3i 3a
any:2d: mÓ mÓ mén mán mÓn mén
any:2i: bÓ bÓ bét bát bÓt bét
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
(40) x :y : z
∅ 3s
2i:1/3s: bâa∗ bâa∗
1/3i:1/3s: êi∗ êi∗
3a:1/3s: âa∗ 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:
(42) x :y : z
3s 3d 3p 3i
3i: é∗ ét ét∗ ét
3i:1/3s: êi∗ édêi∗ égẐi∗ 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 êi to
4
For some speakers, x :3s prefixes are used here. This follows immediately from (32).
5
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
174 CHAPTER 5
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 ı̇i to 3i:3p yields 3i:1/3s:3p, once again voicing
the alveolar, and switching it to velar before i/y. Lastly, addition of ÔO to
3i:3i and voicing of the alveolar yields 3i:1/3s:3i.
Now, the identity of the added vowels, êi, Ẑi and ÔO, is not at all mys-
terious. 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:
(44) x :y : z
3s 3d 3p 3i
2i: bá∗ bet bát∗ bét
2i:1/3s: bâa∗ bédêi∗ bágẐi∗ bÓdÔO∗
The z =3s form follows from lengthening of the vowel, and the 3d, 3p and 3i
forms from respective addition of êi, Ẑi and ÔO. The minor difference concerns
the first vowel in 2i:1/3s:3i, which is Ó, not é, as in 2i:3i. This difference is
discussed below.6 (Again, the 2d forms are predictable by nasalization.)
Finally, consider:
(45) x :y : z
3s 3d 3p 3i
3a: á∗ et gyá∗ et
3a:1/3s: âa∗ dêi∗ gyâa∗ 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, *édêi and *édÔO. In these cases, the
initial vowel is not inserted, or, if inserted, does not survive the phonology.
Summarizing these patterns, we have:
(48) 3s 3d 3p 3i 3a
:∅ ∅ eM gya e á
3s: ∅ eM gya é 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
(i) x :y : z
3d 3p 3i
2i:1/3s: bét.êi∗ bát.Ẑi∗ bét.ÔO∗
1/3i:1/3s: ét.êi∗ ét.Ẑi∗ ét.ÔO∗
7
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.
176 CHAPTER 5
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, é, 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 á, but 3s:3a is em. I
am unsure how far to pursue this resemblance.
Summary
x :y : z
∅ 3s 3d 3p 3i 3a
1s:(a:) a gya nen gyat dé de
2s:(a:) em a men bat bé be
2i:(a:) ba bá∗ bet bát∗ bét bé
3s:(a:) ∅ ∅ eM gya é em
i:(a:) e é∗ et ét∗ ét ét
a:(a:) á á∗ et gyá∗ et ém
∅/2s/3s:1s: éM éM né yáM nÓ né
any:1d/p: dÓ dÓ dét gyát dÓt d∼i
∅/1s:2s: em gyá nén yán gÓ d∼i
other :2s: gO gÓ dét gyát gÓt d∼i
any:2i: bÓ bÓ bét bát bÓt d∼i
1s:3s: gyá nén yán gÓ d∼i
∅/2s/3s:3s á én án Ó d∼i
any:i: bé bét bét bét d∼i
THE AGREEMENT PREFIX 177
5.4.1. Ditransitives
x :y : z
∅ 3s 3d 3p 3i 3a
∅/2s/3s:1s: éM éM né yáM nÓ né
any:1d/p: dÓ dÓ dét gyát dÓt d∼i
∅/1s:2s: em gyá nén yán gÓ d∼i
other :2s: gO gÓ dét gyát gÓt d∼i
any:2i: bÓ bÓ bét bát bÓt d∼i
1s:3s: gyá nén yán gÓ d∼i
∅/2s/3s:3s á én án Ó d∼i
any:i: bé bét bét bét d∼i
Let us first consider y=singular prefixes (from which other :2s:z is ex-
cluded on principled grounds, as explained below).
Singular y
(51) x :y : z
∅ 3s 3d 3p 3i
∅/2s/3s:1s: éM éM né yáM nÓ
∅/1s:2s: em gyá nén yán gÓ
1s:3s: gyá nén yán gÓ
∅/2s/3s:3s: á én án Ó
178 CHAPTER 5
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. á ⇔ [y 3s ] [z 3s ]
b. én ⇔ [y 3s ] [z 3d ]
c. án ⇔ [y 3s ] [z 3p ]
d. Ó ⇔ [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+á → gyá by Glide insertion; g+én → nén by
Dental-velar switching and Nasalization; g+án → yán by Glide insertion,
Nasalization, Engma-deletion; and g+Ó → gÓ 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:
⎧ ⎫ ⎧ ⎫
⎪
⎪ ⎡ x ⎤ y ⎪
⎪ ⎪ ⎪
⎪⎡ y ⎤⎪⎪
⎪
⎪ ⎪
⎪ ⎪ ⎪
⎪
⎨ +author ⎬ ⎨ −author ⎬
g- ⇔ ⎢ −hearer ⎥ +singular ∩ ⎢+hearer ⎥
⎪ ⎢ ⎥ ⎢ ⎥
⎪
⎪ ⎣+singular ⎦ −augmented ⎪ ⎪
⎪
⎪
⎪
⎪⎣+singular ⎦⎪ ⎪
⎪
⎪
⎩ ⎪
⎭ ⎩ ⎪ ⎪
⎭
−augmented −augmented
(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.
THE AGREEMENT PREFIX 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 . How-
ever, 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
The distribution of codas here is different from that in the other rows of
(51) in that Ón (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:
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 mor-
phology. 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, yᡠderives 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.
180 CHAPTER 5
Non-singular y
(57) x :y : z
∅ 3s 3d 3p 3i
any:1d/p: dÓ dÓ dét gyát dÓt
other :2s: gO gÓ dét gyát dÓt
any:2i: bÓ bÓ bét bát bÓt
any:i: bé bét bét 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 -Ót
(except that the vowel for any:3i:3i is e); in the 3p column, the rhyme is -át
(except that the vowel for any:3i:3p is e again); in the 3d column, the rhyme
is -ét; in the 3s column, the rhyme is -Ó (except that the vowel for any:3i:3p is
e again); in the ∅ column, the rhyme is -Ó (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 ).
(58) z
3s 3d 3p 3i
singular y á én án Ó
non-singular y Ó ét át 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
THE AGREEMENT PREFIX 181
allomorphy of z -agreement.
A potential problem for these allomorphy relations is that other :2s:z con-
ditions 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 ϕ-
structure (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.
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 : z
∅ 3s
2i:1/3s: bâa∗ bâa∗
i:1/3s: êi∗ êi∗
a:1/3s: âa∗ âa∗
any:1d/p: dÓ dÓ
any:2i: bÓ bÓ
∅/2s/3s:1s: éM éM
1s:2s: em gyá
other :2s: gO 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.
THE AGREEMENT PREFIX 183
(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) 2 3i
x b- ∅-
y b- 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 át. 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’, yáŹpÓ
‘rope’, yátkyá ‘eight’: if the underlying form were ya. . . , as it is in yál ‘hopefully’, et
cetera, then ∅:3s:3s would be yán. 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 paabyói ‘brother.inv’.) (1s:3s:3p and ∅/2s/3s:2s:3p yán have
initial y-, in contrast to ∅/2s/3s:3s:3p án, owing to opaque phonological processes. Yán
derives from g+ián by Glide formation, gyán, Nasalization, Nyán, and Engma-deletion,
yán. By contrast, in ∅/2s/3s:3s:3p án, there is never a consonant before ián, so that
Vowel-in-hiatus deletion applies, yielding án.)
184 CHAPTER 5
Summary
5.4.2. Transitives
We are now concerned with the following prefixes. (The notation ‘(:3a):’ is
dispensed with, in virtue (28).)
x: z
3s 3d 3p 3i 3a
1s: gya nen gyat dé de
2s: a men bat bé be
2i: bá∗ bet bát∗ bét bé
3s: ∅ eM gya é em
3i: é∗ et ét∗ ét ét
3a: á∗ et gyá∗ et ém
Singular x
(66) z
3s 3d 3p 3i 3a
singular x a en at é e
singular y á én án Ó
non-singular y Ó ét át Ót
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 é. 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) é ⇔ [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.
(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.
186 CHAPTER 5
(71) x: z
3s 3d 3p 3i 3a
1s: gy-a n-en gy-at d-é d-e
2s: ∅-a m-en b-at b-é 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 ]
Non-singular x
(74) x: z
3s 3d 3p 3i 3a
2i: bá∗ bet bát∗ bét bé
3i: é∗ et ét∗ ét ét
3a: á∗ et gyá∗ et 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:
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, ét, 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.)
188 CHAPTER 5
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 :
(76) x: z
3s 3d 3p 3i 3a
2s: a en at é e
2i: á∗ et át∗ ét 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, bé, is simply 2s:3a, be, with high tone.
Interestingly, a similar relation holds form 3a:3a. It, ém, is simply 3s:3a,
em, with high tone. This suggests:
The prefix 3i:3a requires particular comment. It is the only z =3a pre-
fix 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 ét 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 ét.
(79) x: z
3s 3d 3p 3i 3a
3s: ∅ eM gya é em
3a: á∗ et gyá∗ et ém
THE AGREEMENT PREFIX 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 agree-
ment.) 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: z =∅
1s: a
2s: em
2i: ba
3s: ∅
3i: e
3a: 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: z
∅ 3s
2i: ba bá∗
3i: e é∗
3a: á á∗
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
190 CHAPTER 5
high tone, p. 187.) This suggests that the x :∅ forms are, for non-singular
prefixes, the basic vocabulary items for these persons.
(82) a. ba ⇔ [ 2i ]
b. e ⇔ [ 3i ]
c. á ⇔ [ 3a ]
If this reasoning is more or less correct, then the only additional expla-
nation required for intransitive prefixes is the occurrence of a for 1s:∅.
(83) a ⇔ [ 1s ]
5.5. Conclusion
5.6. Appendix
Although the Kiowa prefix system has been the object of extensive investiga-
tion 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 191
(84) x: :z y
3s 3d 3p 3i
∅: :refl 3p:3a ∅:3d:3i 3s:3a ∅:3i:3i
As these facts have only newly come to light, they require further inves-
tigation. 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.
This page intentionally blank
Chapter 6
193
194 CHAPTER 6
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 ⇔ áá-inv class
√
(5) walking stick ⇔ t!optéMaa-erse class
The superscripts are ‘class diacritics’ (cf., Harris 1991) that condition the
realization of feature conflict on D, that is, the inverse marker:
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.
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:
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:
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.
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 combina-
tions 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’ termi-
nology; see her footnote 1, p. 1.
198 CHAPTER 6
The traditional views in Bantu and Kiowa linguistics make the two lan-
guages’ 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, tógúl 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 199
(16) DP
⎡ D ⎤ NumberP
+singular
⎣−augmented⎦
Gender A Number Class: person
+singular Gender A
m −augmented
tu
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:
7
Alternatively, m-/wa- could realize [±singular Gender A] on D.
200 CHAPTER 6
These facts are reminiscent of the Sanskrit cases treated above, where mark-
ing 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 mutu-
ally 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.
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 tran-
scribed (with some amendments) under his supervision. It illustrates several
phenomena discussed in the preceding chapters, including the use of number-
conditioned 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 deci-
sions and tricky cases. For discussion of individual grammatical points, see
Watkins (1984) or the more specific references cited below.
201
202 APPENDIX A
͡Źˇhěǐtegya
Phá ˇ á
ˇ o k!yá
ˇ á
ˇ hyóp e-ŹˇŹˇǎǎhel áádom. SyÓndé gya-hó ˇOndǑǑmei déem
téŹ e-khOléŹǎǎhel. GO e-xándehel ÓgO áádO e-bôndǑǑmei déěǐ gO
ˇOn dómgyá ÓphOO gya-xóúdêi. NO ’gO t!ópá
hó ˇ á
ˇ hêlde tó
ˇ ú
ˇ nêi: ‘Xáá!
ÓgO áádO e-bôndOOde dómgyá ba-thóúgyâit!OO. Bát-mOOhol, hÓˇÓˇ
bá-salOm. ÓÓkO, áábǒǔdOdethai khó ˇ ú
ˇ gya t!ámde mOOkyáǐǐk!OO. Bát-
mOOkholdOOpe. Dómbéba-khûtdet!OO. Háyátto tháŹde khyâigunt!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].’
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 á- khûtde- hel. ∅- T!Om-áM áM - dé gôm gya- bóMúM - hêl
conj beyond 3a-pop/flee out-hsy 3s-first- come-nom back 3s:3p-look-hsy
nO héMŹM ∅- dÓM ÓM-mêi. Tékhop á∗- don- hel né hÓn á∗-
conj gone 3s-be- impf.hsy everywhere 3a:3s-seek-hsy but neg 3a:3s-
thOM OM -mOO. Ém- xáŹ- hel: ‘Hagya t!ámde ∅- óbÓŹ- k!OO?’
find- neg 3a:3a-ask-hsy possibly panther 3s-really-lie
ˇ á
SÓt ém-děǐsÓÓgyáá gO hÓndé án-xóúá ˇ pó
ˇ ú
ˇ hêl. ÁŹdě gya-thápdÓˇÓˇ-
mêi gO gya-phótkhÓtkÓdǑǑmei. PahŹˇŹˇ thÔOxo án-xóúá ˇ á
ˇ pó
ˇ ú
ˇ nêi.
PhÓˇÓˇhel gO, klk klk klk, hÓˇÓˇgOt é-sálÓmhêl. TéŹ pááxokO á-Ónhêl:
‘Hagya nÓÓ éˇ-thÓttétOO, hagya óŹde en-thÓttétOO?’
Khodêide án-xátÓnhel gO mám gya-áŹhel gO gya-phŹŹthá ˇ á
ˇ holhel.
NO sOphól á-bǒǔ. BéthOO sOphól bôudǑǑmei. MÓn x!ól ě-khúŹbOOdo
ˇ á
mÓÓ án-xóúá ˇ pó
ˇ ú
ˇ nêi.
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.
SÓt ém- deM Mi- sÓÓgyáá gO hÓndé án- xóú- áM áM - póMúM - hêl. ÁŹdeM
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. PahŹMŹM thÔO-
3p- dry- be- impf.hsy conj 3p- crackly- be- impf.hsy clearly beyond-
xo án- xóú- áM áM - póMúM - nêi. ∅- PhÓMÓM-hel gO, klk klk klk,
instead :3s:3p-crawl-come-sound-impf.hsy 3s-stop- hsy conj [clicks]
hÓMÓMgOt é- sálÓm-hêl. TéŹ pááxokO á- Ón- hêl: ‘Hagya nÓÓ éM-
gun.inv 3s:3i-load- hsy all alike 3a-think-hsy possibly 1 3s:1s-
thÓtté-tOO, hagya óŹde en- thÓtté-tOO?’ Khodêide án- xát- Ón-
shoot- fut possibly that 3s:3d-shoot- fut suddenly :3s:3p-war cry-sound-
hel gO mám gya- áŹ- hel gO gya- phŹŹ-tháM áM -hol-hel. NO sOphól
hsy conj above 3s:3p-go off-hsy conj 3s:3p-fire- beat-kill-hsy conj owl
á∗- boM uM . BéthOO sOphól ∅- bôu- dOM OM -mei. MÓn x!ól
3a:3s-see.pf unbeknownst owl 3s-long time-be- impf.hsy perhaps wing
eM - khúŹ- bOO- do mÓÓ án- xóú- áM áM - póMúM - 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 In-
terfaces 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 An-
geles.
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
[±augmented], 5, 61, 67, 101, 110 ClassP, 73, 93, 99, 105, 106, 109
definition, 63, 69 interpretation, 96, 99
[±group], 5, 61, 100–102, 110, 115 collections of collections, 44, 99, 139
and suppletion, 138–144 collectivity, 36, 41–44, 50, 95, 97,
definition, 63, 98–99 101, 138–140
[±singular], 5, 61, 67, 70, 101, 110 of action, 42, 45
definition, 63, 68–70 conjunction, 28–29, 107–109, 201
214
INDEX 215
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
STUDIES IN NATURAL LANGUAGE AND LINGUISTIC THEORY
Managing Editors
Publications
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 Move-
ment 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 Ital-
ian. 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 Varia-
tion 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
STUDIES IN NATURAL LANGUAGE AND LINGUISTIC THEORY
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
64. P. Štekauer 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. Mišeska Tomić: The Balkan Sprachbund Morpho-syntactic Features. 2006
ISBN1-4020-4487-9
68. K. É. Kiss, Event Structure and the Left Periphery: Studies on Hungarian. 2006
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