INFORMATION TO USERS

INFORMATION TO USERS
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted.
Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from lefl to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back of the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy.
Higher quality 6" x 9" black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.
UMI
A Bell & Howell Infonnation Company
300 North Zeeb Road, Ann Aibor MI 48106-1346 USA
313/761-4700 800/521-0600
BANTU NOMINALIZATION STRUCTURES
by
John Mugane
Copyright © John Mugane
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF LINGUISTICS
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
1997
UMI Number: 9729475
Copyright 1997 by
Mugane, John Muratha
All rights reserved.
UMI Microform 9729475
Copyright 1997, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
2
THE UNIVERSITY OF ARIZONA ®
GRADUATE COLLEGE
As members of the Final Examination Committee, we certify that we have
read the dissertation prepared by
entitled
John Mugane
Bantu Nominalization Structures
and recommend that it be accepted as fulfilling the dissertation
requirement for the Degree of
Doctor of Philosophy
RiLchard T. Oehrle
Date
^J^an Bresnan—,
Darte
3|M/f7
-v^l
Simin Karimi
Date
D. Terence Langendoen /
Date
Date
Final approval and acceptance of this dissertation is contingent upon
the candidate's submission of the final copy of the dissertation to the
Graduate College.
I hereby certify that I have read this dissertation prepared under my
direction and recommend that it be accepted as fulfilling the dissertation
requirement.
D
kfK.Oar^L
Lsser tat" i on D•irpr^nr
—
Harp
'
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an
advanced degree at The University of Arizona and is deposited in the University
Library to be made available to borrowers under the rules of the Library.
Brief quotations from this dissertation are allowable without special permission,
provided that accurate acknowledgement of source is made. Requests for permission
for extended quotation from or reproduction of this manuscript in whole or in part,
may be granted by the copyright holder.
SIGNED:_
4
ACKNOWLEDGEMENTS
This dissertation is a product of my residence at two excellent Departments of
Linguistics: The University of Arizona and Stanford University. I would like to thank my
committee: Dick Oehrle, Joan Bresnan. Terry Langendoen, and Simin Karimi. Dick
Oehrle, Co-Chair, has been an outstanding mentor to me. both when I was in residence in
Tucson and when away in Palo Alto. His vast knowledge and perceptiveness in virtually
all theoretical and empirical matters have made it tremendously profitable to work with him
on Bantu language issues of which little is known about. All this in addition to ensuring
departmental and institutional support for my efforts. Joan Bresnan, Co-Chair, has been a
great mentor and inspiring teacher. I have benefited a lot from her vast knowledge of
linguistics and her engaging theoretical insights. Terry Langendoen has been responsible
for critical improvements ensuring clarity of the ideas presented. Simin Karimi's
suggestions have led to better elaboration of content. All the committee members have
maintained an 'open-door' policy with me which has made communication timely and
progress high-paced.
Thanks also to the linguistics faculty at both The University of Arizona and
Stanford University for making my career as a student fascinating and intellectually
fulfilling. In Arizona. I have learned a great deal from my colleagues especially Jan
Mohammad. Nayla Yateem, Shensheng Zhu, Colleen Fitzgerald, and Chip Gerfen.
Stanford University has provided me an excellent environment for study and
research. I have benefitted a great deal from discussions with Sam Mchombo, Peter Sells,
Chris Culy, Elizabeth Traugott, Whitney Tabor, and Chris Pinon. Invaluable have been the
suggestions by members of the Advanced Syntax Research Group and the Stanford
Syntax Supper Group, especially, Adams Bodomo, Racheal Nordlinger. Maria-Eugenia
Nfno, Hye-Won Choi. Joan Bresnan, Peter Sells, and Tracy King. Also beneficial were the
ideas generated at the Berkeley Linguistics colloquium on 'The Bantu Word' in the Fall of
1993. For my professional development. Will Leben has been instrumental in my
establishing an engaging and promising research agenda. I have benefited from his
mentoring and his tireless efforts to ensure all the necessary departmental and institutional
support.
My parents, my siblings, and especially my nieces and nephews have all done their
share in encouraging me far beyond expectation. They are the best argument I have that all
the effort is wonhwhile. All the shortcomings are solely my own.
5
This dissertation is dedicated with pride and honor to:
my mother, Gathoni,
and
to the memory of my father, Mugane wa Ndegwa
6
TABLE OF CONTENTS
LIST OF FIGURES
13
ABSTRACT
14
CHAPTER 1. PRELIMINARIES
16
1.0
Introduction
16
2.0
The noun phrase
16
3.0
The Data
20
3.1
Noun Classes
20
3.2
Nominal concords
24
3.3
Nouns and noun phrases
26
3.3.1
Root nouns
28
3.3.2
Nominalized forms
29
3.3.3
Linear order
30
4.0
The noun phrase in Bantu
32
5.0
The Issues
35
6.0
5.1
The morphology
35
5.2
The Bantu noun and noun phrase
37
5.3
Categorial ambiguity
38
5.4
Mixed categories
38
5.5
TTie lexicon-syntax divide
39
Theoretical premises
6.1
7.0
Theses
The Lexical-Functional grammar framework
40
41
6.1.1
The c-s
41
6.1.2
The lexicon
43
6.1.3
Thef-s
44
49
7
7.1
8.0
Category types
50
7.1.1
Pure categories
50
7.1.2
Ambiguous Categories
51
7.1.3
Mixed categories
53
7.2
Category level
54
7.3
Tests
56
7.4
Structural Issues
57
7.5
Mapping of information
58
Summary
60
CHAPTER 2. CATEGORIAL AMBIGUITY
63
1.0
Introduction
63
2.0
Syntactic distribution
64
2.1
2.2
2.3
Noun modification
64
2.1.1
Determiners
65
2.1.2
Adjectives
66
2.1.3
Quantifiers
66
2.1.4
Associative Phrases
67
Verbal behav ior
67
2.2.1
Adverbs
68
2.2.2
Emphatic interjections
68
2.2.3
Object prefixation
69
Positional distribution
70
2.3.1
Linear ordering
70
2.3.2
Extraction
72
2.3.3
Event intensification
73
8
3.0
4.0
Morphological properties
74
3.1
Tense and a.spect
75
3.2
Negation
75
3.3
Verbal extensions
76
Lexical restrictions
76
4.1
Auxiliary verbs
77
4.2
Intransitive bases
78
5.0
Summary
79
6.0
Proposal
80
7.0
8.0
6.1
The u-class
82
6.2
Structure-Function association
86
6.3
Summar>'
87
Previous studies
88
7.1
Abney(1987)
88
7.2
Grimshaw (1991)
89
Conclusion
CHAPTER 3.
91
STRUCTURAL ISSUES
93
1.0
Introduction
93
2.0
Facts
94
2.1
The noun and its modifiers
2.1.1 The noun and its complements
2.2
3.0
The verb
Myers (1987)
3.1.1
96
97
Previous Analyses
3.1
94
Problems with Myers (1987)
99
99
99
9
3.2
Carstens (1991)
3.2.1
3.3
Problems with Carstens (1991)
Summar>'
101
102
103
4.0
Proposal on affixes
103
5.0
Conclusion
105
CHAPTER 4.
MIXED CATEGORIES
106
1.0
Introduction
106
2.0
The facts
107
3.0
4.0
2.1
Nominal properties
107
2.2
Verbal properties
110
2.3
Summary
113
Proposal
114
3.1
Word structure
114
3.2
Phrase structure
116
3.4
Summary
125
Conclusion
CHAPTER 5.
126
THE LEXICON/SYNTAX DIVIDE
127
1.0
Introduction
127
2.0
Problems
128
3.0
2.1
Problem 1
128
2.2
Problem 2
129
2.3
Problem 3
129
The facts
3.1
130
Verbal Properties
3.1.1
Stem Valencies
130
130
10
3.2
3.3
3.4
3.1.2
Prefixation
131
3.1.3
Suffixation
132
3.1.4
Modification
133
3.1.4
Summar>'
134
Nominal Properties
135
3.2.1
The Associative Phrase
136
3.2.2
Demonstrative Placement
141
3.2.3
Adjective Placement
146
3.2.4
Summary
148
Phra.sal Properties
149
3.3.1
Phrasal Recursivity
149
3.3.2
Coordination
149
3.3.3
Pronoun Incorporation
153
3.3.5
Summary
155
Conclusion
156
4.0
Issues
158
5.0
Previous Solutions
158
5.1
Sproat (1985)
5.1.1
5.2
Myers (1987)
5.2.1
5.3
Application of Myers (1987) to Grkuyij
Kinyalolo (1991)
5.3.1
5.4
Application of Sproat (1985) to Grkuyu
Application of Kinyalolo (1991) to GFkuyu
B&M(1995)
5.4.1
Application of B&M (1995) to Grkuyu
159
161
162
162
164
168
169
171
5.5
6.0
7.0
Summary
173
Proposal
173
6.2
Solution to Problem 1
176
6.3
Solution to Problem 2
176
6.4
Solution to Problem 3
177
Conclusions
CHAPTER 6.
179
ATTRIBUTIVE PHRASES
181
1.0
Introduction
• 81
2.0
Issues
183
3.0
Root adjectives
185
4.0
3.1
Properties of Root Adjectives
185
3.2
Analysis
187
Participial adjectives
4.1
[Ncl-...-h-u] adjectives
189
4.2
[Ncl-...-u] adjectives
190
4.3
Properties of participial nouns
192
4.3.1
Non-finite participial phrases
193
4.3.2
Finite participial phrases
195
4.3.3
Attributive versus declarative phrases
196
4.3.3.1
The focus particle
196
4.3.3.2
Tonal considerations
197
4.3.3.3
Reduced Relative clauses
197
4.3.4
5.0
189
Analysis
Patientive phra.ses
5.1
Adjectival properties
198
202
202
12
5.2
Analysis
204
6.0
Agentive nominalization
207
7.0
Summary
208
8.0
Previous Studies
211
8.1
9.0
10.0
Application to Grkuyu
Solutions
214
214
9.1
Issue 1
215
9.2
Issue II
215
9.3
Issue III
215
9.4
Issue IV
216
Conclusions
216
CHAPTER 7. CONCLUSION
218
I.O
Introduction
218
2.0
The issues
218
2.1
Tests
219
2.2
The morphology
221
2.3
Infinitive/gerunds
222
2.4
Category and word structure types
223
2.4.1
Root nouns
224
2.4.2
Nominalized verb roots
224
2.4.3
Nominalized verb stems
226
2.5
Phrasal structure
227
2.6
The Lexicon-Syntax Divide
228
2.7
Adjectives
230
REFERENCES
232
13
LIST OF FIGURES
A
a
AGR
AGT
AP
ASP
Assoc
AssocP
Azer
Caus
COMPL
Conj
Dem
DO
fp
fv
GEND
lO
Loc
MA
Ncl
Nstem
NUM
Nzer
O
OP
Pass
PERF
pi
PRED
Pres
Pron
PROXaddr
PROXdist
PROXsp
R
Rel
S
S2
SP
TA
TP
Applicative
agent
Agreement
Agent
Adjectival phrase
Aspect
Associative
Associative phrase
Adjectivalizer
Causative
Complete
Conjunction
Demonstrative
Direct object
Focus particle
Final vowel
Gender
Indirect object
Lx)cative
Manner adverb
Noun class
noun stem
Number
Nominalizer
Object
Object prefix
Passive
Perfect
Plural
Predicate
Present
Pronoun
Proximate to addressee
distant to speaker
Proximate to speaker
Reciprocal
Relative
Subject
Singular
Subject prefix
Temporal Adverb
Tense prefix
14
Abstract
This dissertation studies Bantu nominalizations drawing evidence primarily from
Grkuyu and Bantu languages already in the literature. Ofkuyu provides a very rich system
of deverbal nouns which brings to the fore issues regarding word and phrase composition
of deverbal elements, and the lexical integrity of words. Bantu nominalizations have
received little attention in the literature in works such as Myers (1987). Kinyalolo (1991).
Bresnan and Mchombo (1995). A very striking aspect of nominalized verbs in Grkuyu.
(and Bantu) is that they bear both noun morphology (noun class marking), and verbal
morphology (both inflectional and derivational). Deverbal nouns are many in Bantu
languages and can not be taken to be idiosyncratic elements, without attempting to discover
whether they are subject to principles that explain their large variety and numbers.
In this study it is apparent that deverbal nouns do encapsulate the properties of
nouns and verbs simultaneously. Upon nominalization Grkuyu shows that we get a set
intersection of the properties of N and those of V. These properties are maintained from the
sub-lexical level to the phrasal level. I propose that when the sub-lexical source of these
nominalizations is established, it becomes apparent why deverbal nouns exhibit split
category or mixed category status. This study also employs tests to check distributional and
behavioral properties of all the items under scrutiny. I show that there are N/V ambiguous
elements (infinitive/gerunds) and a family of mixed N/V elements whose category type can
not be uniquely determined. Split and mixed category items challenge conventional
premises of analysis which require every word to be associated to some unique category
type. I have utilized the idea of Extended Heads in Grimshaw (1991) and Bresnan (1996)
and the theory of Lexical-Functional Grammar (LFG) of Bresnan (1982). to account for
the distinction between pure, split, and mixed category elements. Central to the analysis is
15
the association of affixal information on words directly to grammatical function without
first relating the affixes to category and projection.
16
CHAPTER 1
PRELIMINARIES
1.0
Introduction
This dissertation sets out to study the noun phrase of Bantu languages. Formal
studies of the noun phrase are few in the literature, with Lees (1960). Chomsky (1970).
Jackendoff (1977). Hellan (1986). Abney (1987). Giorgi and Longobardi (1991). Lapointe
(1990) being the most representative of works that have been the basis of current thinking.
With regards to Bantu languages a few studies have examined several important aspects of
the noun phrase (Myers (1987). Kinyaiolo (1991). Carstens (1991), Bresnan and
Mchombo (1995)). It is apparent that the structure of the noun phrase taken for granted has
been that provided in the analysis of the earlier mentioned studies.
In this preliminary chapter. I provide a brief review of the noun phrase within
linguistic theory and within Bantu studies. The chapter then provides an outline of the kind
of data this dis.sertation is concerned with and a very brief description of the LexicalFunctional Grammar framework which I adopt in this study. This chapter ends with a
statement of the theses of this dissertation.
2.0
The noun phrase
All the studies mentioned above have made significant contributions to our
understanding of phrase structure. Lees (1960) studied the similarities between noun
phrases (NP) and sentences (S). Upon observing that NP and S are similar in their external
and internal distribution. Lees proposed that the NP must be transformationally derived
from a corresponding S. By external distribution. Lees draws attention to the identical
behavior of subject and object within NP and within S in processes such as passivization.
17
which was considered to involve a transformational process. With regards to internal
structure, the NP and S are similar with respect to binding and control processes.
Examples of binding structures are (1). where reflexive anaphors must have antecedents
which bind them, and (2), where pronouns must be free.
(1)
a. Billk hated himself^ /*himic
b. *himselfk hated Billk /*him)i
(2)
a. Billk's hatred of himself^ /*him|;
b. *his own hatred of Billk /*himi;
In many ways. Lees's dissertation provides the foundational claims upon which
subsequent proposals on the noun phrase have been constructed. Chomsky (1970) for
instance starts off by refining Lees's observations that NPs are transformationally derived
from corresponding S. Chomsky (1970:184) says that the base consists of a categorial
component (which I will assume to be a context free grammar) and a lexicon. In this
'lexicalist* hypothesis, the correspondence between nominalizations with related sentences
is attributed to the base. Chomsky argued that the structural subject - object distinction is
necessary in NPs just as it is in the S. This led to the introduction of intermediate categories
X" in X-bar theory. Chomsky proposed that lexical heads V. N. A, and P are
underspecifications of complex categories involving feature combinations in which nouns
are [+N. -V], verbs [-N. +V]. adjectives [+N. +V], and prepositions [-N. -V]. It is because
of their underlying feature complexity that we get mixed behavior. Jackendoff (1977).
following Chomsky (1970). suggested that independent of the category type, the way the
head defines its projections in structure is constrained by endocentricity. The principle of
endocentricity stipulates that every phrase must have a unique head whose category type is
18
the same as that of the mother node. Also X-bar theor>' has it that every phrase must have
at least one intermediate projection (X') which consists of X and its complements.
On this view, the relation between the two NPs in (3) can be directly captured by a
movement relation within NP, rather than captured indirectly by regarding one NP as the
nominalization of an active sentence and the other NP as the nominalization of a passive
sentence.
(3)
a.
the city's destruction
b.
the destruction of the city
Abney (1987:33-34) points out that the introduction of N' leads to very similar
structures between NP and S which accounts for binding and control facts. He points out
that if c-command is defined as "a c-commands b if neither dominates the other and the
first (branching node after Reinhart (1976)) node dominating a dominates b", the "subjects'
of both NP and S asymetrically c-command the objects, capturing binding and control
facts. Abney follows up on the similarity between the S and the NP and notes that the two
are similar in that they were the 2 cyclic nodes in earlier transformational grammar and
they are the only two categories that freely contain subjects, this despite the fact that N and
V pose the most fundamental opposition in grammar. In a manner that complements
Lees's (1960) study of the similarity between NP and S, Abney provides an inventory of
the differences between NP and S. Despite the observed differences however, he exploits
the similarity and makes the major proposal of his dissertation: that there is an Infl-like
category within the NP (the DP-hypothesis). He proceeds to assign quasi-identical
structures to both NP and S. He bases this proposal on two facts from English: first, a
possessed noun phrase agrees with its subject just as the verb agrees with the subject;
19
second, possessors receive the same case as the subject of the S. Both these analytical
properties point to the existence of AGR in the English noun phrase. The existence of
AGR implies an Infl-like position which it (AGR) occupies.
Giorgi and Longobardi (1991) study the internal structure of the Romance noun
phrase and the theory of empty categories within the "Principles and Parameters"
framework. They stiirt from the ideas found in Chomsky (1970) (stated above) that NPs
such as the city's destruction are transformationally related to the clause the destruction of
the city. They call their approach the "Configurational Hypothesis' which has two premises:
that within the NP there are definite theta and non-theta positions; and that the theta
structure of N parallels that of V.
Lapointe's (1990) study of gerund phrases introduces the idea of mixed category
phrases. He proceeds from Pullum's (1991) observation that gerund phrases exhibit both
external and internal behavioral properties of nouns and verbs. The mixed category status
of these nominal phrases are particularly problematic in maintaining a central X' theoretic
notion, "head of a phrase'. Specifically, due to the mismatches that exist between the
category type of the mother node and the head daughter node, in such a proposal, the
generation of impossible rules (under X-bar theory) such as "NP
—> V. PP" and
"N' —> (QP). V" is left open. Structures which permit mismatches in lexical features
between the mother node and the head daughter node (such as a V that heads an NP) are
not permitted in X' theory. Lapointe proposes the notion of dual lexical category (DLC) of
the type <X/Y>0 in which X and Y are major lexical categories with X determining the
external syntactic properties of the phrase and Y the internal syntactic properties of that
phrase.
20
3.0
The Data
The evidence to be considered comes from a variety of languages documented in the
Hteraiure. The noun is one of the most fascinating elements in Grkuyu grammar'. There are
non-derived as well as derived nouns. Among the derived nouns are nouns with
preprefixes on the one hand and on the other hand nouns that are derived from verbs. All
nouns exist in a complex of gender-type classes which must be reflected in the
morphology of almost all nominal elements including demonstratives, numerals,
quantifiers, associative words, pronouns, subject and object prefixes, possessives. relative
words, and even a set of question words. However, the noun class system does not
determine the morphological shape of prepositions. I begin with a look at the noun class
system.
3.1
Noun Classes
The general property of all Bantu languages is that nouns are divided into classes
which are distinguished morphologically by characteristic prefixes that are affixed to the
noun stem (Meinhof 1932)). These noun classes differ from one Bantu language to
another. In general, a class of nominals is made up of words that have similar prefixes,
which in turn form part of a set of concord elements operating a distinct pattern of
agreement. The numbering of the classes is a means of labelling the different sets of
concord prefixes that operate the grammatical agreement in all given Bantu language (Bleek
1862; Guthrie 1970:79). With respect to the grammatical number of nouns and their
associated prefixes, there are three categories: (i) singular with correlative plurals having
identical lexical meaning as the regular form; (ii) plural with which occur a correlative
' T h e orthography used throughout is the system advocated a s the standard by the United Kikuyu Language
Committee m the early years of the 1900s and used since then by the speat;ers of the language.
T h e vowel f is
equivalent to e in words such as 'ate' and D to O' a s in the word 'goat'. TTie judgements presented throughout are
my o w n and I have found no mstance where my judgements contradict Barlow's grammar. Gfkuyu is spoken in
Kenya by over five million people. Guthrie (1967) classifies Gikuyu a s belonging t o zone E5 I in his overall
classification.
21
singular form; (iii) neutral which has neither correlative plural nor correlative singular
forms. Neutral forms are generally: abstract concepts such as iitugi 'generosity': locatives,
such as haha ~here"(specific place); verbal infinitives/gerunds, such as kuruga
to
cook/cooking". It is apparent then, that Grkuyu has an elaborate system of noun
classification. There are at least 17 noun classes in the language that are classified following
Guthrie's study of Bantu languages (Guthrie. 1948). The list in figure (4) is a general
outline of Grkuyu noun classes arranged in accordance with Meinhof's general guidelines
of Ur-Bantu (Meinhof. 1932). As I have outlined above, the glosses show that some
classes such as 6, 7. and 8 have a whole range of semantic properties associated with them,
including the singular/plural marking properties associated with all other nouns in classes 1
through 14.
clas
s
1
2
3
4
5
6
Prefix
Example
Gloss
muamumfima-
muundu
andu
muti
mftr
ihuti
mahuti
person
people
tree
trees
leaf
leaves
7
ki-
kiTga
8
ci/i-
ciiga
9
10
11
12
13
14
nNrukatuu-
nyiimba
nvumba
ruT
kahn
tuhiT
uugf
Qcuru
15
ku-
16
17
haku-
kuguru
kuruga
haha
guku
Description
Singular of 2
Plural of 1
Singular of 4
Plural of 3 and occasionally of 14
Singular of 6
Plural of 5, 14; the plural pejorative and
augmentative. Also preferred plural form.
organ
Singular of 8; the pejorative and
augmentative singular.
organs
Plural of 7; occasionally plural of
pejorative and augmentative.
house
Singular of 10.
houses
Plural of 9 and 12.
river
Singular of 10.
small boy
Diminutive. Singular of 13.
small boys Diminutive Plural of 12.
intelligence Concepts, where possible plural found in
porridge
6. Also preferred singular form. Contains
a
subclass of tangible objects as well.
Mainly infinitives and gerund. Also other
cookins
miscellaneous objects, e.g body parts.
here
locative - specific place
here
locative - general place
22
The grouping of referential entities into noun classes has clear semantic criteria for
some of the classes, while for others the semantics are quite fuzzy.
As described in Mugane (forthcoming), all nouns in cla.ss 1 are human in Grkuyu
such as mutumia "woman". Class 2 contains the plural forms of class 1 such as atumia
'women'.
Class 3 and its correlative plural class 4 include names for parts of the body, e.g
mutwdnntwe "head/heads'and names of plants, especially trees, e.g muembelmiemhe
'mango tree/mango trees', etc.
Classes 5 and 6, singular and plural respectively, include names of plants.
ikunii/makimu 'mushroom/mushrooms', parts of plants, irhangu/marhanf^u '\eaVleawcs'.
iyembelmaemhe 'mango fruit/ mango fruits', inanimate things such as working tools and
implements, iceembe!mciceenibe 'hoe/hoes'and iiimulmalimu "spear/spears', parts of the
body such as iniuru/maniiiru 'nose/nosts' and names for ocassions such as
igongona/magongona 'ceremony/ceremonies'.
Nouns of classes 7 and 8 are of miscellaneous significance and are generally
impersonal. They include names of languages Gilhungu 'English language'and Kijaruu
'Luo language', anatomical terms, kiara!ciara 'finger/fingers', and names of utensils.
gi cikoiiciko ' spoon/spoons'.
Nouns of classes 9 and 10 do not carry a prefix. These nouns are largely
impersonal including names of animals such as hurialfiuria 'rhino/rhinos'. These classes
also consist of words borrowed from other languages such as ndege/ndege
"aeroplane/aeroplanes'and ngari/ngari 'car/cars'.
The nouns of class 11 consist of names of impersonal objects, parts of the body
such as uthiiilrliiu 'face/faces"and rucwirilnjuiri hair/hairs. The plurals of I 1 are found in
10.
The other classes 12 and 13 are the diminutive singular and plural respectively, for
example, gakaari/iukaari. "little car/little cars'. These are also the classes that denote
endearment. Thus ^akaari may also refer to a car that is seen to be endearing.
Class 14 is the class of abstract concepts such as uugi "knowledge", but also
consists of names for miscellaneous objects such as iicuru 'porridge". The plural for
abstract concepts as well as tangible ones that can be pluralized is found in class 6. mauugr
"many types of knowledge'and macurii 'porridge, in a collective sense for the foregoing
examples.
Class 15 consists of infinitive nouns and gerunds such as kiiriiga "to cook'or
"cooking'as well as names for various body pans such as kiiguru "leg'and gutu "ear'. The
body parts in this cla.ss derive plurals from class 6. so magurit "legs'and matu "ears' are
the resulting plurals for the foregoing examples.
Classes 16 and 17 both refer to location of things with class 16 making reference of
the specific place and class 17 the general place. The specificity of a place is relativized to
the spatial parameters set in the conversation.
Besides those in class 11/10. there are other items whose singular is found in one
class and the plural in a different class. Examples of these are abundant; ibera "guava
fruit'class 5 but mhera "guava fruits'is in class 10 instead of the expected class 6.
It is not the ca.se that a noun class will be identified simply by the appearance of its
prefix. This is e.specially true given that prefixes such as mu- and kit- each mark more than
one noun class in Grkuyu . The prefix mu- marks classes 1 and 3 while ku - marks classes
15 and 17. To be able to differentiate between noun classes with identical prefixes, in
general there are two types of criteria: one semantic and the other grammatical. Nouns
determine the morphology and syntactic behavior of the associative elements, adjectives,
pronouns, demonstratives, etc. I now turn to the nominal concord system.
24
3.2
N'ominai concords
Grkuyu nouns, as is generally the case in all Bantu languages, are classified in
groups not based on the semantic criteria but rather on their concordial properties. The
grammatical behavior and distribution of nouns is a much more reliable criterion than the
semantic one in determining the noun class of any given referential entity. The classes are
distinguished by the occurrence of certain associated sets of concordial affixes that go with
the noun in an utterance. The system of concord is ordered following the noun class
system. A lot of the grammar of Grkuyu and Bantu generally is expres.sed in a system of
agreement concords (5).
The elements provided in figure (5) are called agreement concords because their
morphological shape is a direct reflection of the properties of the noun they modify or
associate with in some syntactic fashion. From figure (5) we see that pronouns differ from
class to class, that other than the personal pronouns (those related to class 1 and 2). all
pronouns bear the -o ending with the other material being a copy of the morphology of the
noun class in question for all classes except 3. 4. 5, 9. 10. and ]4~. Adjectives have
morphology that is identical to that of the noun class for all clas.ses except for class 8. This
class is not morphologically distinguished from 9 and 10 in adjective formation. As
indicated in figure (5). abstract notion nouns as well as concrete nouns in class 14 (though
forming plurals with class 6 morphology) form adjectives with class 3 morphology-*.
Subject prefixes (SP) and object prefixes (OP) also pattern differently in terms of each
class. The a.s.sociative marker is made by prefixing the SP to the associative morpheme -a.
- l i is not the purpose of this chapter to provide arguments for the type of morphology as opposed to that ot' the
noun classes and I am not aware of any study of the GiTcuyii phonological processes that explains the surface
morphology of pronouns.
•^When class ? marker [i] is affixed to a vowel initial stem, an
that I + ega = rfega (goixJ thing in class 5).
r
is insened at the beginning of the word such
25
while the possessive pronoun is made up of the associative marker plus the possessor
morpheme"*.
Figure (5)
Ncl
example Gloss
I
l^t Sg
you
2nd Sg
Pron
nir
wee
Adj
SP
nd
OP Assoc
nd
Genitive 'mv'
w-akwa
wf
ku
w-aku
mu
w-ake
tu
w-itu
3rd Sg
she/he
we
1st PI
us
ithuf
e
tjj
2nd Pi
you
inyur
mwf
mu
w-anyu
me
mo
w-ao
mu wa
ma a
mu wa
mi ya
rf
na
ma ma
gr
kya
kr
ci
cia
w-akwa
a-akwa
w-akwa
y-akwa
n-akwa
m-akwa
gy-akwa
3rd PI
1. mij
2. a
3. mu
4. mr
5. i
6. ma
7. gr
kr
8. i
ci
9. n
10. N
11. ru
12. ka
13. tu
14. u
15. ku
ku
16. ha
17. ku
they
singer
muini
aini
singers
tree
mutr
mfU"
trees
igego
tooth
magego teeth
gi'tr
chair
kJura
frog
itr
chairs
ciura
frogs
dog
ngui
ngui
dogs
njQr
river
little boy
kahiT
tuhiT
little boys
uuru
evil
kurfa
to eat
to see
kuona
here
haha
auku
here
V
0
we
cio
mu
a
miJ
mr
i
ma
gr
kf
N
yo
cio
ruo
ko
tuo
guo
kuo
n
N
ru
ka
tu
mu
ka
u
a
u
r
n
ma
gr
kr
i
ci
r
ci
ru
ka
tu
u
ku
ho
ku
ha
ku
ha
ku
0
guo
yo
no
mo
kyo
mr
ci
ru
ka
tu
wo
ku
ku
ha
ku
ci-akwa
ya
cia
rwa
ka
twa
wa
kwa
y-akwa
ci-akwa
rw-akwa
g-akwa
tw-akwa
w-akwa
gw-akwa
ha
gwa
ha-kwa
2w -akwa
In addition to the concordial elements in figure (5) there is Figure (6) as well. The
table in figure (6) shows the morphological types of demonstratives, relative words,
examples of numerals, and quantifier words in the language. Demonstratives are of three
kinds each with singular and plural forms. The ones glossedr/7«5 refer to things in
proximity to the speaker, thatj refers to those in proximity to the addressee and things
•*The other prevalcni case nf possessor morphemes arc locative ones. Thus in Figure (5) we may also use -kuo
(general place) a n d or -ho (specific place) to indicate that the entity rel'ered to occupies the said location.
26
mentioned previously in a conversation, and thatj marked with long vowel are those that
are non-proximal to both speaker and addressee-''. Relative words -na "that" or "which"
morphologically resemble that2. Quantifiers are a closed class with members such as
othelantwe "all/some'. These use subject prefix morphology as do the numerals.
Figure (6)
Nclass this
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
3.3
mu
a
mu
mr
i
ma
of
i
n
N
ru
ka
tu
u
ku
ha
ku
uyu
aya
uyu
fno
rfn
maya
afkr
ici
Tno
ici
rCiru
gaka
tiitu
uyu
giiku
haha
gOku
that!
that2
ucio
acio
ijcio
Tyo
rfu
macio
kJu
icio
ivo
icio
ruu
kau
tuu
ucio
kuu
hau
kQu
uurfa
aarfa
uurfa
una
riTrfa
maana
kiTrfa
iiria
iTrfa
iiria
ruuria
kaarfa
tuuria
Qurfa
kuurfa
haaria
kuurfa
Rel
'that'
uria
aria
urfa
frfa
rina
maria
kfrfa
iria
frfa
iria
rurfa
kan'a
turfa
urfa
kuria
haria
kuria
Num
I &3
umwe
atatu
umwe
ftatu
rim we
matatu
kfmwe
ithatu
fmwe
ithatu
rOmwe
kamwe
tutatu
umwe
kumwe
hamwe
kumwe
Quant
all
wothe
othe
wothe
yothe
riothe
mothe
gfothe
ciothe
yothe
ciothe
ruoihe
gothe
tuothe
wothe
guothe
hothe
guothe
Quant
some
amwe
fmwe
mamwe
imwe
imwe
tumwe
umwe
kumwe
hamwe
kumwe
Nouns and noun phrases
While the nominal system is clearly organized in a complex system, the architecture
is buttressed by a complex of deverbal words which bear various functions in the noun
phrase. This class of deverbal words avails another class of referential entities. Nouns bear
different kinds of morphological shapes in Grkuyu as is illustrated in (7) below. Broadly
speaking, two types of nouns can be distinguished: root, or underived. nouns and derived,
primarily deverbal, nouns. Derived nouns can be thought of as existing in two forms; one
uses the preprefixes and the other involves suffixation. Preprefixes consist solely of the
•^The vnwel ihai indicates non-proximaie things may be lengthened arbitrarils. The longer the vowel is made, the
iireater the distance indicated.
27
diminutive ka- and lu- (class 12/13) . the augmentative kC- (class 7) as well as collectives
ma- (class 6). These noun class markers are used as preprefixes mainly when the stem of
the noun in question is monomorphemic. The derived nouns may also involve suffixation
on nouns. In Grkuyu nouns derived by suffixation are many and quite spectacular, as will
be seen in section 3.3.2. One noun suffix -im is remarkable in that it can be suffixed to any
noun, derived or underived. with the effect of changing the meaning from a referential
entity to a location. The types of nouns distinguished morphologically are on the left branch
of figure (7). which includes all the nouns under the node labelled "referential entity". The
referential entities are combined with nominal modifiers to produce complex noun phra.ses.
28
Figure (5)
NOUN PHRASE
REFERENTIAL ENTITY
MODIFIERS OF ENTITY
Numerals / \ \
Demonstratives \ \
Pronouns
UNDERIVED NOUN
mij-ndu
people
\
Quantifiers
Associatives
DERIVED NOUN
etc.
PREPREFIXES
SUFFIXATION
ROOT NOUNS
tu-mu-ndu
little people
VERBAL BASE
POSTPOSITIONAL
-inr
NOMINALIZATION
atlinlon
manner
agentive /
reflexive
occassion
X^result
etc.
\ locative
patientive
Nominal modifiers include quantifiers, numerals, pronouns such as the possessive, and the
demonstrative. In what follows examples are provided for each noun type and. whenever
possible, its properties in terms of grammatical behavior and distribution. For the latter,
properties of nominal modifiers will also be made explicit.
3.3.1
Root nouns
Root nouns are made up with the affixation of the noun class to stems. Apart from
the noun class prefix, these nouns are not further decomposable into simpler components.
29
The result is a referential expression. Examples of root nouns in Grkuyu are provided in
(8). Two morphological processes are possible with root nouns. They can be locativized by
(m'-suffixation (8b) and (8d) to bear the meaning "on/at/in/by" or take preprefixes as in (8c)
and (8d).
(8)
3.3.2
a.
mu-tf
3-tree
"tree"
b.
mu-tf-inf
3-tree-LOC
"on/at/in/by the tree"
c.
ka-mu-tf
12-3-tree
"small tree"
d.
ka-mu-tf--inr
12-3-tree-LOC
"on/at/in/by the small tree'
Nominalized forms
Derived nouns are those whose base is verbal. These are formed by the
nominalization of their stems by a number of prefixes and suffixes. They differ from root
nouns and nouns derived from other nouns in that they are formed by prefixing and
suffixing simultaneously to verbal stems. Examples include phrases such as (9a). studied
in chapter 5. patientive deverbal adjectives (9b) and participial adjectives (9c). discussed in
chapter 4. infinitive/gerund constructions (9d) which are dealt with in chapter 2. Other
forms which fall in the category of deverbal nominal elements include manner
nominalization (90. locativization (90 and occasionalization (9g).
agentive deverbal nouns
(9)
a.
mu-thi7nj-i
1 -slaughter-Nzer
"goat-slaughterer"
mburi
1 Ogoat
30
patienlive deverbal adjectives
b.
mburi
thiTnj-e
9goat
3-slaughter-Azer
"slaughtered goat'
pjirticipial deverbal adjectives
c.
mu-ndu
mu-thom-u
1 -person 1 -read-Azer
lit. 'a read-person'
'an educated person'
infinitive/gerund
d.
ku-rug-a
ucuru
15-cook-fv
11 porridge
"cooking/to cook porridge"
manner nominalization
e.
mu-thom-ere
15-read-Nzer
"manner of reading'
locativization
f.
i-kur-fro
5-pluck-Nzer
"location of plucking'
occasionalization
g.
3.3.3
i-tuur-a
5-live-fv
"residence"
Linear order
The pragmatically neutral linear order of modifiers is noun, demonstrative,
possessive pronoun, quantifier and. finally, the adjective (Barlow 1951) as shown in
example (10) and its elements in terms of category in (10").
31
(10)
(10')
nyungu id
ciake
lOpot
lODem 10-3sgPron
"all these red pots of hers/his'
N
Dem
Pro
ciothe ndune
lOall lOred
Q
A
Root and derived nouns of the type described above have a number of distributional
properties in Grkuyu. First, as in other languages, they form noun phrases either by
themselves or in combination with nominal modifiers. Nouns can be premodified or
postmodified by demonstratives as exemplified by (11). (11a) is considered the
pragmatically neutral order. (1 lb) is an order that is derived under what is considered to be
stylistic permutation geared to some pragmatic effect such as focusing or clefting.
(11)
a.
nyungu Tno
9pot
9 Dem
'this pot'
b.
Tno
nyungu
9 Dem
9pot
"this pot'
Now consider the modifiers that the anaphoric NPs take provided below. It is clear
that all the anaphoric elements can occur without the head as seen in (12). In fact all
nominal elements including the head are optional.
(12)
a.
ici
ciake
ciothe
lODem 10Assoc-3sgPron lOall
"all these red (things) of hers/his"
b.
iria
ciake
ciothe
lORel
10Assoc-3sgPron
lOall
"all these red (things) of hers/his'
c.
ci-ake
lOAssoc-3sg-Pron
"all her/his red ones'
d.
ciothe ndune
lOall
lOred
"all red ones'
ciothe ndune
lOall lOred
ndune
lOred
ndune
lOred
32
Should a modifier that occurs to the left of the modifier in linear precedence be
permuted to the right of it. then the comma intonation must be employed as indicated in
(13). I assume that the comma intonation employed in this manner indicates appositiveness
and that the modifier to the right of the comma is outside the noun phrase. When the
quantifier phrase in (13a) is permuted to the right as in (13b). the comma effect results. The
permutations of the pos.sessive pronoun in (13c) and the adjective in (13d) produce a
similar effect. This comma effect indicates that a nominal modifier is actually outside the
noun phra.se.
(13)
a.
ici
ciake
ciothe ndune
lODem 10Assoc-3sgPron lOall lOred
"all these red things of her's/his'
b.
ici
ciake
ndune,
lODem 10Assoc-3sgPron lOred
"these red (things) of hers/his, all'
c.
ici
ciothe
ndune, ci-ake
lODem lOall
lOred
10Assoc-3sg-Pron
'all these red ones of her's/his'
d.
ici
ndune, ciothe
ci-ake
lODem lOred
lOall
10Assoc-3sg-Pron
'all these red ones of her's/his'
ciothe
lOall
With all these forms it should be noted that the comma is employed to mark the
point at which the linear order has been violated. Elements to the right of a comma that are
well ordered in terms of precedence are not marked with the commas as seen in (13d).
4.0
The noun phrase in Bantu
The noun phra.se in Bantu provides a rich area for the study of the morphology-
syntax interface. As indicated in section 3. the noun is the governor and distributor of
concord/agreement marking in Bantu. As a staning point, any study of the Bantu word
structure must evaluate what the status and contribution of the noun class marker is. This is
not easy. Mufwene (1980) finds that noun class markers exhibit both inflectional and
derivational behavior. As will be further elaborated in this study. Mchombo (1978), Sproat
(1985), Myers (1987). Kinyaiolo (1991), Carstens (1991). Bresnan and Mchombo (1995).
have all advanced different proposals on the status of the noun class marker. That the noun
class marker is so central in the study of the Bantu noun phrase structure is not at all
suprising as any preliminary study of any Bantu language will reveal. This is because the
assumptions made on its status have broad consequences on central issues in current
linguistic theorizing including (but not limited to); nominalization; adjectivization; the
lexical integrity of words (Chomsky. (1970): Bresnan and Mchombo. (1995) and the
references they cite) versus word formation in the syntax (as found in Baker (1985)); and.
as argued in chapter 3. on noun phrase structure itself.
With regard to specific aspects studied by these researchers, the ones at issue in this
dissertation include the status of the noun class, the identity of the nominalizing morpheme,
the lexicality or phrasality of synthetic compounds (agentive nominalizations). as well as
the structure of the noun phrase. The main proposals about the status of the noun class
marker and the nominalizer are made by Mchombo (1978). Sproat (1985), Myers (1987),
Kinyaiolo (1991) and Bresnan and Mchombo (1995). A lexical account for synthetic
compounds is proposed by Bresnan and Mchombo (1995) for Chichewa synthetic
compounds while Kinyaiolo (1991) proposes a phrasal account for Kilega ones. Carstens
(1991) makes proposals about noun phrase structure. As can be inferred below (and
repeatedly throughout this study) none of the studies cited above have agreed on a given
solution.
Sproat (1985). a.ssuming a structure suggested by Mchombo (1978), proposes that
the noun class marker in Chichewa is a phrasal element which selects N or V. For him
class prefixes are not the agentive nominalizers equivalent to -er in English for words like
34
"teacher', because -er in EngHsh (unlike class markers in Chichewa) link up to the external
theta-role of the verb. This issue is addressed in chapter 5.
Myers (1987) studies Chishona and analyzes the noun class prefix as an element
which takes NP and VP complements. Following Fortune (1955. 1985). he identifies the
nominal modifiers as NPs because they seem to possess all the syntactic attributes of a
noun phrase except the missing head noun.
Like NPs. nominal modifiers can function as
subjects and objects of sentences, as well as complements of prepositional phrases. Since
these nominal elements take modifier complements, such as intensifiers. degree phrases
etc.. Myers, citing Fonune (1955. 1985). calls them qualificative nouns. While Lapointe
(1990) points out the challenge posed by gerunds to X-bar notion of "head of a phra.se". the
problem introduced by these "qualificative nouns' is that they are headless or. as Myers
(1987) proposes, they are headed by heads which are phonologically null. Like an overt
head, the phonologically null head determines the number and cla.ss of its modifiers and of
the NP it heads according to Myers (1987). This issue and some of Myers's claims are
taken up in chapter 3.
Carstens (1991) proposes an analysis of noun class prefixes and noun phra.se
structure in Bantu drawing data primarily from Kiswahili. According to her. noun class
prefixes are added to nouns as number morphology by gender-specific redundancy rules.
To Carstens. Number is a functional head which selects NP complements. Kiswahili Ninitial order within NP is as a result of noun raising and NPs are covert determiner phrases
with phonologically empty heads to which nouns raise obligatorily through number
projection. Chapter 3 of this dissertation examines some aspects of Carstens's proposal.
Kinyalolo (1991) argues that the rule for synthetic compound formation operates
on NPs rather than on DPs.
He proposes that the rule of synthetic compound formation
because the rule makes reference to NPs and not to DPs contrary to Sproat (1985) who
argues that only
are input to the rule and Abney (1987).who says synthetic-
compounding makes reference to DPs. On the question of the status of the noun class
prefix. Kinyalolo (1991) assumes that "the morphological process of affixation is nothing
more than adjunction of an X® to a Y® . The demerits of an analysis of noun class
markers as X® are discussed in Bresnan and Mchombo (1995).
Bresnan and Mchombo (1995) present several arguments against deriving synthetic
compounds in the syntax. TTiey test noun class marker for lexical integrity effects. They
test for the extractibility of noun class marker, its conjoinability. its gapping ability, whether
it displays inbound anaphoric islandhood. and whether it is phrasally recursive. They
conclude that noun class markers (excluding the locatives) respect the lexical integrity of
words and can not be products of phrasal formation.
Tlie studies mentioned in the previous section and this one provide the background
on which this dissertation stands. Many of the proposals in those studies are brought to
question in this study as outlined in the next section.
5.0
The Issues
The issues under scrutiny in these study have to do with the morphology of nouns
and nominal elements, the Bantu noun and noun phrase, categorial ambiguity, mixed
category items, and the lexical integrity of words. These issues together unravel some of
the core properties regarding the internal organization of the Bantu noun phrase. As will be
clear throughout, the assumptions made with regards to a particular issue either in this
study or in the literature, have consequences on all the subsequent issues. In what follows I
provide a brief note on each issue.
5.1
The morphology
Morphological marking of nouns and is similar to that of adjectives as seen in (14).
In (14). the prefixal morphology of the adjective is mostly identical in shape to that of
36
nouns. The similarity is indicated by the material in bold on nouns (.column 1) and
adjectives (column 4). The literature offers conflicting reports on the question of the kind of
information contributed by the noun class marker. Mufwene (1980) found that noun class
markers have a status ambiguous between inflection and derivation. Bresnan and
Mchombo (1995) have argued that with the exception of the locative classes. Chichewa
noun class markers are lexical elements which are only manipulable by word formation
processes while Myers (1987), Kinyalolo (1991), treat noun class markers as phrasal
elements.
1
Noun class
3
2
Gloss
example
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
mumi
aini
mun"
mTtf
igego
mageao
mu
a
tnu
mf
i
ma
gf
i
n
n
ru
ka
tu
u
ku
ha
ku
gftr
itr
ngui
ngui
ruuf
kahiT
tuhiT
uuru
kuruga
haha
guka
singer
singers
tree
trees
tooth
teeth
chair
chairs
dog
dogs
river
little boy
little boys
evil
to cook
herehere-.
4
Adjective
-nene 'big'
munene
anene
munene
mfnene
inene
manene
kfnene
nene
nene
nene
runene
kanene
tunene
munene
kunene
hanene
kunene
In contrast to the noun and adjective, the morphology of modifiers (demonstratives,
associatives. and quantifiers) is identical to that found on verbs (familiarly called the subject
prefix within verbs) as (15) indicates. Since I assume that the difference in morphological
marking (between (14) and (15)) is not accidental, the i.ssue here is why nominal elements
in columns 2 - 7 are distinguished morphologically from columns 2 and 4 in (14) above.
37
(15)
1
Class
ist Sg
2
SP
nd
3
OP
nd
2nd Sg
wf
5
Assoc
wa
6
Dem
uyu
ku
wee
wa
uyu
wothe
mu
wa
ucio
wothe
ist Pi
e
tij
we
tu
ithuf
a
aya
ithuothe
2nd PI
mwr
mu
inyuf
a
aya
inyuothe
3rd PI
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
me
mo
othe
a
ma
u
r
ri
ma
a
wa
a
wa
ya
na
ma
kya
cia
ya
cia
rwa
ka
twa
wa
kwa
ha
gwa
aya
0
a
ij
1
n
ma
gf
i
T
ci
ru
ka
tu
u
ku
ha
ku
0
we
0
guo
yo
no
mo
kyo
cio
yo
cio
ruo
ko
tuo
guo
kuo
ho
ku
uurfa
aaria
uuna
nrfa
runa
maana
kiTria
iiria
uria
iiria
ruurfa
kaaria
tuuna
uuna
kuuna
haana
kuuna
wothe
othe
wothe
yothe
nothe
mothe
arothe
ciothe
yothe
ciothe
ruothe
gothe
tuothe
wothe
guothe
hothe
guothe
3rdSg
5.2
7
all
wothe
4
Pron
nir
P
i
r
ci
rii
ka
tu
ii
ku
ha
ku
The Bantu noun and noun phrase
With regards to the noun, we have to identify what the category changing morpheme
is among the candidate affixes (which include the noun class marker). By so doing we will
be able to show whether word structure is head-initial or head-final^. If the noun class
marker is subject to the rules of phrasal formation (Myers. 1987), then words are fonned
in the syntax as Baker (1985) suggests. However, if the noun class marker is only
manipulable by lexical processes of word formation, then lexical integrity is maintained
(Bresnan and Mchombo. 1995). It is therefore in looking at structural issues that we are
^^Thc terms 'hcad-tnitial' and 'head-final' are used here to refer to the category determining a f f i x .
38
able to evaluate the status of the noun class marker, whether it is a lexical or phrasal
element. This is particularly crucial in the nominaiized verbs which this study looks into
extensively.
Another issue involving structure is with regards to how best ambiguous category
items and mixed category phrases can be represented.
5.3
Categorial ambiguity
Gfkuyu infinitive/gerunds are ambiguous between the category N and category V.
This ambiguity is demonstrated by the distributional and behavioral properties of (18)
versus (18"). The contrastive behavior between infinitive/gerunds with complements
exemplified by (18) and those without (18') suggests that the former (those with
complements) are of category V and those without belong to category N.
(18)
kuria
mfanga
1 Seat
4cassava
'to eat cassave / eating cassava'
(18')
kurfa
ISeat
"to eat/eating"
5.4
Mixed categories
How to analyze mixed category items and how such items figure in the syntax of
noun phrases is the main challenge here. Unlike infinitive/gerunds which display
ambiguous categorial status, agentive nominals such as (19) present us with examples of
mixed category items which display the properties associated with elements of category N
and of category V.
(19)
mu-thunj-i
mburi
Islaughter-Nzer
lOgoat
"slaughterer of goats'
39
Other cases of mixed N and V elements are exemplified by (20a) and (20b) which are part
of the system of attributive expressions in Grkuyu. It is the study of expressions such as
these that provide insights into the countervailing mechanisms in "impoverished" Bantu
adjective systems.
(20)
a.
njogoo
mak-u
9rooster 9worry-Azer
lit. 'a worried rooster'
"a rooster that is worried'
b.
irio
ngarang-e
lOfood lOfry-Azer
"fried food"
As in the cases of ambiguous categories, the implications of mixed category items to the
internal organization of the noun phrase is central to this study.
5.5
The lexicon-syntax divide
The Grkuyu agentive nominalization bearing [mu-...-i] morphology (21) is
complicated by the existence of its double [mu-...-a] in (22).
Both [mu-...-i] and
[mij-...-a] express similar meanings but behave quite distinctly from each other in terms of
their categorial behavior and syntactic distribution. Both display a number of characteristics
that are a.ssociated with products of lexical rules of word formation seen in Bresnan &
Mchombo (henceforth B&M) (1995) for Chichewa. However. [mu-...-a] morphology
present something of a paradox in that they exhibit lexical properties while still maintaining
some phrasal characteristics. The loose compounding exhibited by [mu-...-a] compounds
immediately raises questions pertaining to the principle of the lexical integrity of words
proposed in various fashions by Chomsky. (1970); Wasow and Roeper (1972). Lapointe.
(1980); B&M. (1995); Sells. (1995). and others. The issue of lexical integrity is studied in
the light of (21) and (22).
40
(21)
mu-endi
andu
1 lover
2people
•people lover"
(22)
mu-enda andu
1 lover
2people
•people lover"
Having outlined the issues. I now provide a brief description of the Lexical-Functional
Grammar framework (Bresnan. 1982; 1995) and Grimshavv"s (1991) theory of Extended
Projections, both of which I will be using throughout this study.
6.0
Theoretical premises
This dissertation relies heavily on Lexical-Functional Grammar (LFG) expounded
in Bresnan (1982; 1995) and Grimshaw's (1991) theory of Extended Projections in its
exposition. In its design the theory of LFG provides a framework that allows for base
generation of elements instead of structural derivation by movement mechanisms such as
those proposed by Baker (1985). In addition to the nonderivational nature of Bantu words,
the lexical status of words inspires the lexical integrity principle, which states that
morphological words are subject to lexical rules and are not manipulable by syntactic rules
of phra-se formation (Bresnan and Mchombo. 1995). Adopting a lexical view of Bantu
words has support from many studies including Kiparsky (1982); Selkirk (1982); Di
Sciullo and Williams (1987); Cho and Sells (1995); Bresnan and Mchombo (1995); etc.
These have provided very compelling evidence that the lexical integrity of words is crucial
in grammatical theory. The Extended Head theory of Grimshaw (1991) has proved very
useful in modelling facts about Bantu mixed category items. As such it is worthwhile to
introduce it early in the study. I begin with a brief description of LFG in section 6.1 and
then proceed with a sketch of Grimshaw"s (1991) theory of Extended heads.
41
6.1
The Lexical-Functional grammar framework
Lexical functional grammar contends that the syntactic information of phrases and
clauses is represented in parallel levels, the most researched being the c-structure (c-s). fstructure (f-s). The theory permits the possibility that other levels of representation may be
in existence, an example being a level of semantic representation suggested by Kaplan
(1987). Halvorsen and Kaplan (1988), Dalrymple et. al, and Andrews and Manning
(1993). Dalrymple (1993), by way of contrast, assumes that semantic encoding is in the
form of a lambda calculus. Other levels could conceivably be of discourse and
phonological representation. The proposal of the c-s and f-s level may be said to be
motivated by two generalizations known as the principle of variability and the principle of
universality. The principle of variability states that expression structures (c-s) vary accross
languages and that they (expression structures) encode such properties as word order,
substitutability. and phonological phrasing. The principle of universality observes that the
internal structure
(f-s) of language is largely invariant and abstracts away from expression structures (c-s).
The internal structure captures phenomena such as reflexivization. binding, agreement, case
government (Bresnan 1995).
6.1.1
The c-s
The c-s is the level of configurational representation. This level represents relations
that are expressed by phrase structure rules, such as those relations pertaining to linear
order,
precedence, dominance as well as constituency. Phrase structure rules have
annotations accompanying them as indicated in (23) for the S. NP. and VP rules. The
formally annotated phrase structure rules need not be stipulated, but can be derived from
general principles of structure-function as.sociation (see below).
42
(23) a.
s
NP
VP
(TSUBJ) = i
b-
NP _>
N
T =i
c.
VP
V
T =i
T =i
NP
(TOBJ ) = i
In the examples T refers to "the f-structure of the node above me' while i refers to "the
f-structure of this node". Consequently. (T SUBJ) =sL reads that "the subject f-structure of
the mother node is the same as the f-structure of this node" while the annotation
(T OBJ) =vl reads that 'the object f-structure of the mother node is the same as the
f-structure of this node". These annotations on Phrase structure rules express the f-s
relations between the correspondents of the c-s nodes. Annotated c-s rules produce
annotated c-structures. The Swahili sentence in (24). has the annotated c-s provided in (25).
(24)
Juma
a-na-som-a
Juma
1 S-Pres-read-fv
•Juma is reading a book'
kitabu
7book
(25)
(TSUBJ) = ^
NP
T=i
N
T =i
V
(TOBJ) = i
NP
Juma
anasoma
is reading
T =i
N
kitabu
a book
43
LFG assumes an X-bar theory in which Functional projections (FP) are
distinguished from lexical projections (XP). These result from structural endocentricity
which involves universal structure function association principles (Kroeger (1991); King
(1992); Bresnan (1995)) listed in (26).
(26)
a.
C-structure heads annotated with T=vl are f-structure heads.
b.
Sisters of
c.
Specifiers of lexical categories are a subclass of adjuncts or absent.
d.
Complements of lexical categories are (non discourse functions)
c-heads are co-f-heads.
argument functions.
e.
Constituents adjoined to maximal projections are non-argument functions.
Also among LFG assumptions is the idea that all c-s nodes are optional, unless required by
general principles such as completeness and coherence (Bresnan 1995).
6.1.2
The lexicon
The lexicon consists of elements with particular information which encodes the
attribute and the value for the attribute. For the terminal elements in (25b) above, we have
the following lexical entries.
(27)
a.
Juma
N:
(T PRED) = 'Juma'
(T GEND) = 1
b.
anasoma
V;
(T PRED) = 'read <SUBJ. OBJ>"
((T SUBJ PRED) = 'pro')
(T SUBJ GEND) = 1
(T TENSE) = PRESENT
c.
kitabu
N;
(T PRED) = 'book'
(T GEND) = 7
44
6.1.3
The f-s
The f-s level carries a different kind of information from the c-s. This level is
represented as an attribute-value matrix and carries information regarding grammatical
relations and predicate argument relations (Kaplan and Bresnan (1982), Dalrymple
(1993)). For every c-s node there is a corresponding f-s. The f-s is subject to completeness,
coherence, and consistency. These are well-formedness conditions. Completeness requires
that the f-s must contain all governable grammatical functions required by the PRED. In
(14b) for instance, the verb soma "read" requires a SUBJ and an OBJ to satisfy its
subcategorization requirements. The PRED for the outermost f-s being •read '<(TsUBJ)>'
and the PRED of the innermost f-s being •read<TOBJ>'. Coherence specifies that the f-s
must not contain any grammatical functions not required by the PRED. In terms of the c-s
provided in (25) and the lexical entries in (27b). consistency requires the absence of either
Juma or the PRED "pro" in the lexical specification of the inflected verb (27b). We achieve
this effect by making the PRED "pro' feature available only when the subject ywma is not
present. Consistency requires that each attribute must have a unique value even though
values may be identical for different attributes. In (27) for instance, the attribute GEND has
the same value for Juma in (27a) as it does for soma
"read" in (27b). However, by
consistency, we can not have the value of GEND being both class 1 and 9.
Each c-s node is annotated. The relation between the c-s and f-s is specified by
statements called functional descriptions. The c-s in (25) and the lexical entries given in
(27) may be instantiated as in the c-s in (28). In (28). fp refers to the f-s corresponding to
the c-s node marked n.
45
(28)
S
(TSUBJ) = if
NP
(TOBJ) = i .
MD
'O
NP
(TPRED) = 'Juma'
(TTENSE) = PRES
(tSUBJGEND)= 1
((T SUBJ PRED) = 'pro' )
(TPRED) = 'read<TSUBJ. TOBJ >
(TGEND)=1
Juma
anasoma
is reading
(T PRED) = 'book'
(TGEND) = 7
kitabu
a book
Following the algorithm explained by Kaplan and Bresnan (1982). the combination
of the annotations (25b), the lexical entries (27) and the instantiation (28), are solved as
indicated by the functional correspondence statements provided in (29a) through (29h).
(29)
a.
b.
(fl SUBJ) = f2
(by the (T SUBJ) = i equation on the NP daughter of S)
f2 = f3
(by the T = i equation on the N daughter of the NP rule)
c.
(f3 PRED) = "Juma"
(by the lexical equation (T PRED) = "Juma")
d.
(f3 GEND) = 1
(by the lexical equation ( TGEND) = I)
e.
fl = f4
(by the T = i- equation on the
VP daughter of S rule)
f.
f4 = f5
(by the T = i equation on the V daughter of VP rule)
g.
(f5 PRED) = 'read < SUBJ. OBJ >'
(by the lexical equation (T PRED) = 'read <TSUBJ. TOBJ>')
46
h.
((f5 SUBJ PRED) = "pro")
(by the lexical equation ((T SUBJ PRED) = 'pro'))
i.
(f5 SUBJ GEND) = 1
(by the lexical equation (T SUBJ GEND) = 1)
j.
(f5 TENSE) = PRES
(by the lexical equation (T TENSE) = PRESENT)
k.
(f4 0BJ) = f6
(by the equation (T OBJ) = -i equation on the NP daughter of VP rule)
1.
f6 = f7
(by the T = i equation on the N daughter of the NP rule)
m.
(f7 PRED) = "book'
(by the lexical equation (T PRED) = 'book')
n.
(fy GEND) = 7
(by the lexical equation (T GEND) = 7)
The result of solving the annotations with their corresponding instantiations is a full
grammatical syntactic structure provided in (30). where the arrows indicate the mapping of
the c-structure to the f-structure.
(30)
SUBJ
NP
PRED 'Juma'
,
GEND
1
YE
--TOED
NR
Juma
anasoma
is reading
kitabu
a book
'read < SUBJ, OBJ >'
'
TENSE PRES
•
^
Q9J
PRED 'book'
^; GEND 7
47
6.2
Grimshaw (1991) Extended Projection
In the theory of extended projection. N and D have the same categorial features, but
are distinguished by their lexical versus functional status. This also applies to the relation
between V and I. In this theory. I and D are of the same category as their lexical
counterparts V and N. I and D are however distinguished from V and N by their functional
status. I and D have the functional feature {FI } while V and N are designated as { F q }. The
{F} value of a node plays a role in the formation of extended projections, it is not a binary
feature nor does it interact with categorial features. The extended projection is a category
consisting of triple specification; category, level and function. (31) shows the verbal
projections with their triple specifications. Verbal categories consist of {V. V. VP. 1. I'.
IP. C. C. CP}, these are identical to each other in terms of lexical specifications <+V. -N>.
but differ in level of projection (indicated as 0. 1. or 2) and functional level (labelled fQ. f].
or f2).
(31)
CP <[tNl 2.f2>
C < n , l l.f2>
V <[tN]- o-fo>
Nominal categories include {N. N". NP. D. D". DP}as indicated in (32). They too are
marked with triple feature specifications.
48
(32)
DP<[.^]. 2,f,>
<[;N]. i-fi>
^'<[;U '-fo
n<[;N ]• o-fo>
It is by requiring that the F values of X and Y be shared that the usual (unextended)
definition (33) of head and projection is obtained.
(33)
X is the perfect head of Y. and Y is the perfect projection of X iff;
(a)
Y dominates X
(b)
Y and X share all categorial features
(c)
all nodes intervening between X and Y share all categorial features.
(d)
the F value of Y is the same as the F value of X.
By this definition CP is a projection C and C as it shares both categorial and functional
features with them as illustrated in (31) but not of I, I', and IP with which it only shares
categorial features. The same reasoning applies to the nominal projections in (32). The
extended head notion differs from the perfect head in that the F value of the head is lower
than the F value of its projections as defined in (34).
(34)
X is the extended head of Y, and Y is an extended projection of X iff:
(a)
Y dominates X.
(b)
Y and X share all categorial features.
(c)
all nodes intervening between X and Y share all categorial features
49
(d)
If X and Y are not in tlie same perfect projection, the F value
of Y is higher than the F value of X
Extended head and extended projections (Grimshaw. 1991)
In this theory of extended projections. (34d) prevents a DP from being directly dominated
by a DP. an NP dominated by an NP. IP by IP. etc. from forming extended projections. In
the definition of extended head and extended projection, only categorial features need be
mentioned. TTiere is no requirement that members of a single projection share their lexical
or functional features.
7.0
Theses
In this dis.sertation I have assembled facts on nominal and verbal properties of
elements in GrkQyu. Several important aspects are considered at length, including issues on
category types and distinctions; tests for category type and category status; lexicalization
effects; structural representation; and a study of countervailing mechanisms in
impoverished Bantu root adjective systems.
In matters relating to category types, this study unravels a spectrum consisting of
pure .N' and V categories; ambiguous/split N/V categories; and mixed .N/V categories.
Also revealed with regards to N - V contrasts is a distinction between lexical and phra.sal
categories. With regards to lexicalization, I argue that the analysis of agentive nominals has
proved recalcitrant because there are confounding effects that are caused by lexicalization.
The Extended Head approach described in section 6.2 above provides a mechanism for
adequately representing some of the phenomenon studied here. Concerning the rarity of
root adjectives in Bantu. 1 show how functional information may be represented
independent of category types and c-structure considerations so as to permit mixed
categories to bear a modificational function. In what follows. I present the main results.
50
7.1
Category types
As pointed out above, this study identifies three types of categories: pure N and
pure V categories; ambiguous N - V ones; and mixed N/V categories. A study of each of
these brings about new evidence on infinitive/gerunds and deverbal nominalizations which
have so far gone unnoticed. The results obtained here have not been anticipated by previous
studies.
7.1.1
Pure categories
Pure nouns can be straightforwardly distinguished from pure verbs in Bantu
languages. The easiest way to contrast N properties from V ones (in Bantu) is by their
morphological marking and syntactic distribution. Verbs are typically marked with
concordial morphology and can be extended and inflected by affixation. Nouns bear noun
class morphology (which is distinguishable from concordial morphology) and do not
permit any affixation other than the preprefixes. Syntactically, verbs are also subject to
very different syntactic distribution from nouns, and do display the most fundamental
opposition in the grammar (Abney, 1987).
With regards to word structure considerations, verbs differ from nouns by their sublexical ability to bear inflectional and derivational affixation. Thus derivational and
inflectional affixation is a property uniquely associated with Bantu verbs. In verb
formation, the verb stem which is the locus of affixation, combines with a subject concord
prefix as in (35). The verb stem always ends with a final vowel.
51
(35 )
V
Subject
Final vowel
stem
Bantu noun formation involves the combination of noun class markers with noun roots as
in (36). Since it is the case that noun stems are restricted to occur only with certain noun
class markers, it is descriptively accurate to say that noun roots select their gender classes'^.
Noun class markers are part of word formation, which serves to provide the semantic
classification (albeit a fuzzy one in synchronic Bantu) of entities in gender type classes.
(36)
N
Noun class
N.root
7.1.2
Ambiguous Categories
Grkuyii infinitive/gerunds are ambiguous between the category N and category V.
The distributional and behavioral properties clearly display that we get contrastive behavior
between infinitive/gerunds with complements (gerund [+NPobj]) exemplified by (37) and
those without (gerund [-NPobj]) (37') suggest that the former (those with complements)
are of category V and those without belong to category N.
(37)
kuna
mfanga
15eat
4cassava
"to eat cassave / eating cassava"
^This behavioral propeny is in contrast to that of adjectives which can occur with any noun class
marker m Bantu (Moshi. 1992).
52
(37')
kuna
ISeat
'to eat/eating'
With gerund [+NPobj] roots ail derivational (Ext*) and (restricted) inflectional affixation
is applicable to the verb stem (38).
(38 )
V
V
stem
Verbs are categories which will permit extraction of their thematic objects, allow for object
prefixation when the root is transitive (or transitivized by derivational affixation), like
regular VPs, gerund [+NPobj] can be fronted without object prefixation, they cannot be
extracted by relativization and by nf-clefting. It is because gerund [+NPobj] are verbal that
they do not allow use of associative phrases but can be formed with auxiliaries, be negated
and admit tense and aspect affixation. They also take typical verbal modification because
they are verbal.
With gerund [-NPobjl the verb root is nominalized (by -a affixation). The root is
verbal but its combination with -a yields a noun stem (Nstem), the Nstem combines with
the class marker
- to form a noun as illustrated by (39). Infiectional and derivational
affixation takes place at the stem level and not at the verb root. This explains why
nominalizing the verb root rules out any extension possibilities having to do with object
prefixation, marking of negation and tense and aspect affixation.
53
(39)
N
stem
It is because they are nouns that gerunds [-NPobj] have the behavior and distribution of
noun phrases. They can be extracted when they occur as VP complements, they partcipate
in associative phrase formation and are modified like typical nouns. Semanticaily empty
verbs such as auxiliaries are not permitted because nouns must have semantic content.
The sub-lexical level reveals that infinitive/gerunds are ambiguous category items.
Gerund [+NPobjl
of category V and gerund [-NPobj] belong to category N. I attribute
the disjunction of N and V properties exhibited by infinitive/gerunds to their different word
structures captured at the sub-lexical level.
7.1.3
Mixed categories
While infinitive/gerunds are categorially ambiguous dsiplaying a di.sjunction of N and
V properties, there is a class of deverbal nouns which are categorially mixed encapsulating
the intersection of N and V properties. The deverbal nouns of interest are. agentive with
[mij-...-i] morphology (40), participial marked with [mu-...-u] (41), and patientive bearing
[mu-..-e] morphology (42).
(40)
mu-thunj -i
mburi
Islaughter-Nzer lOgoat
"slaughterer of goats"
(41)
mu-ndu
mu-mak -u
1-person
Iworry-Nzer
"a worried person'
54
(42)
mu-ana mu-end -e
1-child l-love-Nzer
"a loved child"
These deverbal nouns share a number of properties which point to the word structure in
(43). The formation of mixed categor>' items involves the nominalization of the verb stem
and the combination of noun class markers with deverbal noun stems to form nouns.
(43)
N
^jtem
-i
-u
-e
Mixed category items combine the lexical and phrasal characteristics of gerund [+NPobj]
and gerund [-NPobj]-
Like gerund [+NPobj]' deverbal nouns are very rich
morphologically as is evidenced by the affixation that they admit on their stems. Unlike
gerund [+NPobj] types however, deverbal noun formation involves the combination of the
noun class marker with a nominalized verb stem to form a noun. The only sub-lexical
difference between mixed category items and the gerund [-NPobj] is that in the latter, the
infinitive class marker combines with a nominalized verb root to form a noun.
7.2
Category level
Agentive nouns are of various kinds. There are those that are fully lexicalized as
words, there are those with an intermediate status and exhibit loose compounding, then
there are those that are fully phrasal. The lexicalized agentive nouns exemplified by (44) do
not permit phrasal recursivity. they are anaphoric islands, no element within such nouns is
substitutable. be elidable or conjoinable. (44) is a word which simply projects an N as in
(45).
55
(44)
mutigain
"deceased person who left some offspring'
(45)
N
I
mutiga-in
'rainbow'
In contrast to (44). agentive elements such as (46) allow for some phrasal
recursivity, yet disallow substitutability and interpolation of anything between the
head and the thematic complement. These exhibit loose compounding as illustrated
by (47) where the complement to the head is phrasal (a DP).
(46)
mij-end-a andu
I want-Nzer 2people
"one who loves people"
(47)
mu -enda
DP
A
andu
'one who loves people'
Agentive expressions such as (48) permit all kinds of syntactic manipulations
because they are totally phrasal. Their structure is as in (49) involving phrasal
categories.
(48)
mii-end-i andu
1 want-Nzer 2people
"one who loves people"
56
(49)
mu-end-i
DP
A
andu
'one who loves people'
I conclude that the Bresnan and Mchombo (1995) study of Chichewa deals with
lexical elements (synthetic compounds) which are expressions of the type provided
by (44). By way of contrast, the study done by Kinyaloio (1991) on Kilega
encounters data involving phrasal categories such as that exemplified by (48). It is
not suprising therefore, that the two studies come up with opposite results, the
former in support of lexical integrity and the latter in violation of it. Grkijyu has
both types of data including that in (46) which exhibits loose compounding.
Evidence from Grkuyu enables us to resolve the issue by showing that both the
results of Bresnan and Mchombo (1995) and Kinyaloio (1991) can be maintained
once the lexical domain is distinguished from the phrasal one.
7.3
Tests
TTie procedural strategy in this study is the use of tests as the basis for the proposals
made. I have introduced distributional and behavioral evidence in identifying the category
type, category level, the lexical and phrasal status. These empirical tests reveal that mixed
categories have properties intersecting those of nouns and verbs. Infinitive/gerunds
(gerund [+NPobj] ^irid gerund [-NPobj]) ^re ambiguous between category N and V (a
virtual disjunction of N and V properties). Many of the tests introduced as the study
progresses may be relevant only for Grkuyu and perhaps other Bantu languages. The
agenda in doing this is to attempt to assemble proofs for claims made of Bantu languages
without parochial inclinations towards the more studied languages.
57
7.4
Structural Issues
The word structure proposals provided in section 7.1. have a direct influence on the
phrase structure. The most salient thing about the structures I suggest is that they lead to
some adjustments on the general theory of heads and also to modifications of Grimshaw's
(1991) theory of Extended Heads. The structure we assume for mixed category items is
(53) in which a deverbal N is the extended head of VP such that the N and the VP are
mapped onto the same f-structure in a manner made precise in chapter 4 and 6.
(53)
DP
-
_
[mu-..V..-i]
-e
-u
In (53). N. a lexical category head is an extended projection of VP. This proposal
contradicts the notion that a lexical category
(N in the case of (53)) and a lexical
projection XP (VP in the case of (53)) can not be co-heads. This prevents the mapping of
Y onto the same f-structure as the XP instead of the expected relationship between lexical
heads and their lexical sisters (that of head and complement) as proposed by Bresnan
1982. The extended head theory is violated in that according to Grimshaw. a nominal
category, that is one with <+N. -V> cafegorial features, can not be the extended head or the
extended projection of a category with <-N. +'V> features. The proposal presented in
chapter 4 suggests adjustments to Grimshaw (1991) so as to permit (53).
A very impressive result from the grammar of mixed categories presented in this study
is that mixed category properties of [mu-...-i]. [mu-...-e]. and [mu-...-u] words splits the
structure of the noun phrase into two domains; the concordial one (DP) and the lexical one
58
(NP, VP) as illustrated in (54). All nominal modifiers occur in the concordial domain, while all
verbal modifiers occur in the lexical domain. The concordial domain holds categories whose
members consist of closed class items (determiners, pronouns, quantifiers, etc.) and the
lexical level is the domain where open class items (verbs, nouns, and their projections) occur.
(54)
DP
concordial elements
lexical elements
7.5
Mapping of information
This study sets to answer questions regarding how languages which lack or have a
paucity of certain category items make up for those categories. Specifically for Bantu, there
are very few elements which can be classified as adjectives (Dixon, 1977), which makes
Bantu attribution intriguing. How is attribution done in Bantu languages or does the paucity
of adjectives conversely curtail attribution in these languages? In answer to this question,
we observe that empirical facts show that mixed category items are used for attribution
compensating for the impoverished root adjective system. 1 conclude that the grammatical
design of
LFG which allows for the mapping of information from morphological
components to the f-structure (say by pronominal affixation (Bresnan and Mchombo
1987) to functions independent of category and c-structure. we are also able to associate
mixed N/V elements to the modifier function in the f-structure without requiring that there
be a category adjective and its projections. I have proposed (55) with the functional
annotation
59
(T MOD) = i on the NP as the structure involving attributive modifiers within noun
phrases.
(T MOD) =i
mundu
person
muthomu
studied
(TOBJ) = i
DP
thiomi
languages
'a person who has studied languages'
lit. 'a language studied person'
Notice that in the f-structure for (55) provided in (56), it is the NP with an extended head
that bears a modifier function in the f-structure. This possibility in the theory of parallel
(independent) levels of representation of grammatical information advanced by LFG
permits us to characterize how Bantu languages use mixed category items to compensate
for the paucity of root adjectives observed by Barlow (1951), Dixon (1977) and elsewhere.
This is a very fascinating result.
60
(56) -
HEAD
GEND
SUBJ
1
OEND
'P^°' ^
1
OBJ PRED 'languages'
GEND 10
ASPECT
COMPL
PRED 'study < SUBJ. OBJ >'
Having sketched out the main proposals of this dissertation. I provide a summary for each
chapter in the section below.
8.0
Summary
Chapter 2 explores issues pertaining to categorial ambiguity of infinitives/gerunds. It
is shown that gerunds/infinitives are categorially ambiguous between N and V. I list the
properties as.sociated with gerund [+NPobjJ (those with complements) and gerund [NPobj] (those without complements). The word formation level is identified as the locus at
which the two types are distinguished. The gerund [+NPobj] type is verbal hence its
potential for inflectional and derivational affixation. Gerund [-NPobj] words are formed by
the nominal ization of the verb root. The verb root is a level deeper than the verb base where
affixation takes place. It follows then, that gerund [-NPobj] words can never permit
inflectional or derivational affixation because they are not sufficiently rich in the sub-lexical
level.
Chapter 3 is a descriptive study of nouns and their modifiers alongside verbs and
their complements. The main issues examined here have to do with morphological
marking and its implications in word and phrase structure. While augmenting chapter 2.
this chapter evaluates Myers (1987) and Carstens (1991), claims about the status of the
noun class marker. I argue here that the distinctions between the grammatical categories
considered here is best captured at the sub-lexical level.
Chapter 4 examines mixed category items. The various proposals on how to
analyze mixed category items; one way suggested by Chomsky (1970) is to have complex
categories so we could have categories [+N, -V]. [-N. +V], [+N.+V], and so on. If
behavior is characterized by feature-descriptions, then we can get mixed behavior. Lapointe
(1990) suggests dual category phrases of the type <n O /v O >. Another way is through
lexical/phrasal split, with verbal properties or nominal properties characterized in a way that
is sensitive to the lexical versus phrasal distinction. A third alternative (which is argued for
in this study) is through word formation which involves the nominalization of verb stems
permitting the deverbal noun to inherit (partially) properties of V. I propose that once the
notion of Extended Head (Bresnan. 1996) is introduced in the grammar of noun phrases,
an explanation for the behavior of mixed category items is available.
Chapter 5 identifies and studies some problems all having to do with deverbal
nouns of the agentive type in Grkuyii. These problems are: First, the identity of the
nominalizing morpheme and the status of the noun class marker; second, the
lexical/phrasal ambiguity displayed by agentive deverbal phrases; third, the variation in
morphological shape among different kinds of agentive nouns and; fourth, the whole issue
of the lexical integrity of words. It is shown that derivational word structure is affix-final
while phrase structure is head initial for Grkuyii. It is also posited that there is a descriptive
generalization to be gained by considering the nominalizing material as suffixal and the
class marker as a lexical element which is manipulable only by word formation processes.
It is shown that there is a lexical/syntactical ambiguity exhibited by agentive phrases in that,
despite the fact that agentive deverbal words in Orkuyu are products of syntactic formation
(see Kinyalolo (1991) for Kilega). they also behave in a manner which is typical of
products of lexical formation (see Bresnan &Mchombo (1995) for Chichewa). Faced by
the dilemma that neither a purely lexical account nor a purely syntactic analysis can explain
all the facts in Grkuyu, it is proposed that a notion of permeability within the lexiconsyntax divide itself best captures Grkuyu facts. Given that there are lexical integrity effects
which are clearly manifest, it is concluded that the principle of lexical integrity is needed in
grammatical theory. With regards to the different morphological shapes, it is pointed out
that the variant shapes differentiate the lexicalized forms from those forms which are
phrasal.
Chapter 6 examines Grkuyii atrributive phrases pursuing themes of mixed
categories, and morphological marking. Here I assemble facts regarding participial nouns,
and patient nouns and conclude that the analysis provided in chapter 4 for agent nouns
using the idea of extended heads (Bresnan 1996) suffices for these nouns as well. This
chapter unravels some countervailing mechanisms in the 'impoverished' Bantu root
adjective systems.
I recapitulate the major points of this dissertation in Chapter 7.
63
CHAPTER 2
CATEGORIAL AMBIGUITY
1.0
Introduction
The categor>' distinction between V and N has been said to be pose the greatest
opposition in grammar (Abney. 1987). Elements that qualify to have the designation 'verb'
have a very distinguishable array of distributional and behavioral properties from nouns. In
Bantu language studies, verbs are described as bearers of inflectional morphology such as
tense and aspect, and can be extended with derivational morphology with various
syntactically interesting effects on the valency of the ba.se verb. In addition, verbs can be
negated, be modified with adverbs of various kinds. The usual syntactic distinction
between nouns and verbs has it that verbs act as consumers (functors) of nouns to form
larger verbal structures. Nouns are subject to a very different set of properties, they
function as arguments of verbs in the syntax, their typical modifiers include determiners,
adjectives, and quantifiers. In semantic terms nouns denote entities while verbs denote
eventualities (Bach. 1981).
In this chapter, I show that while the distinction between root N and root V is clear
in Bantu, infinitive/gerunds are split between the two categories. I present a number of
distributional and behavioral properties of Grkuyu infinitive/gerunds to demonstrate that
they exhibit a clear ca.se of categorial ambiguity between N and V. This conclusion is
reached upon observing that there exists conirastive behavior between infinitive/gerunds
with complements (henceforth gerund [+NPobj]) exemplified by (I) and those without
(gerund [-NPobj] henceforth) in {1"). There is preponderant evidence which suggests that
gerund [+NPobj] belong to category V and gerund [-NPobj] belong to category N.
64
(1)
kuna
mfanga
1 Seat
4cassava
'to eat cassave / eating cassava'
(1 •)
kuna
15eat
"to eaL'eating"
I attribute the recalcitrant nature of infinitive/gerunds in theoretical analysis to this
shift indicating that the categoriai ambiguity they exhibit need not be captured
simultaneously, by allowing dual or mixed categor>' lexical projections as attempted by
Pullum (1991), Lapointe (1990). The category shift I demonstrate argues for two separate
representations for the infinitive/gerund structures. While obviating the dilemma of having
a mismatch between the category of the head with that of the mother which is a flagrant
violation of X-bar theory (Jackendoff 1977). this proposal exposes the nature of the lexical
processes of word formation for elements that display categoriai ambiguity.
In what follows. I employ a series of tests on both gerund [-NPobj]
gerund
[+NPobj] to check for their categoriai statuses. I look at their syntactic distribution, their
morphological properties and their lexical restrictions. While it is the ca.se that I do most of
the exposition using transitive bases, the implications of using intransititve bases is
discussed in section 4.2.
2.0
Syntactic distribution
Both
properties.
gerund [-NPobj])
gerund [+NPobj] have a number of syntactic
This section examines their modification potential, their behavior under
clefting, their positional distribution with respect to canonical placement of nouns and verbs
as well as linear order considerations. I demonstrate that with respect to these tests, gerund
[-NPobj] exhibit a behavior (that of pure N) that is the exact opposite of that brought out by
gerund [+NPobj] (that of pure V).
65
2.1
Noun modiflcation
Modification of syntactic elements is governed by their categorial status. Modifiers
of nouns include a set of determiners (demonstratives, possessive pronouns), adjectives,
quantifiers and numerals.
2.1.1
Determiners
Determiner elements are a small closed set of elements in Bantu mainly consisting
of possessive pronouns, and demonstratives. Possessive pronouns are not permitted with
gerund [+NPobj] (2) in contrast to (2') which shows that gerund [-NPobjl permits such
modification which is typical of root nouns as provided in (2").
(2)
*kuna
mfanga
gwake
15eat
4cassava 15-3sgPron
'his/her eating (of) cassava'
(2")
kuria
gwake
15eat
15-3sgPron
'her/his eating'
(2')
njugu
yake
lOpeanut 15-3sgPron
'her/his peanut"
Likewise, demonstratives combine only with root nouns (3") and gerund [-NPobjl (3') but
not gerund [+NPobj]
(3) reveals.
(3)
*kuna mfanga
kuu
Kamau a-ra-n-a
15eat 4ca.ssava
15Dem Kamau 3sgS-Pres-eat-fv
Lit. 'that cassava eating that Kamau is eating "
'that eating of cassava that Kamau is engaged in'
(3")
kuna kuu
Kamau a-ra-ri-a
15eat
iSDem Kamau 3sgS-Pres-eat-fv
Lit. 'that eating that Kamau is eating'
'that eating that Kamau is engaged in"
(3")
njugu
fyo
Kamau a-ra-rf-a
lOpeanut lODe'm Kamau 3sgS-Pres-eat-fv
'that peanut that Kamau is eating"
66
A typical property of the Bantu noun phrase is that the demonstrative can occur on either
side of the noun it modifies. This is the case with root nouns such as njiigu "peanut" in (4")
and gerund [-NPobj] as seen in (4') but not with the gerund [+NPobj] types (4).
(4)
*kuu
kuna
mfanga
nf-gu-gii-ku-rwar-i-a
15Dem 15eat
4cassava FP-15-Fut-2sgO-sick-A-fv
"that eating of cassava will make you sick'
(4")
kOu
kuna
gwaku
ru-gu-gu-ku-rwar-i-a
15Dem 15eat
i5-2sgPron FP-15-Fut-2sgO-sick-A-fv
'that eating you are engaging in will make you sick"
(4")
fyo
njugu
yaku
nf-r-gu-ku-rwar-i-a
lODem lOpeanut
10-2sgPron FP-IO-Fui-2sgO-sick-A-fv
"that peanut of yours will make you sick"
When the distribution of determiners is considered, gerund [-NPobj] behaves like the
canonical noun and gerund [+NPobj] behaves like something else
2.1.2
Adjectives
Adjectives modify nouns while adverbs typically modify verbs. That gerund
[+NPobj] iire verbs is evident in the fact that they can not be modified with adjectives (5).
TTie reverse is the case with the gerunds [-NPobj] types (5').
(5)
*kuna
mianga
kwega
1 Seat
4cassava 1 Sgood
"good cassave eating'
(5")
kurfa
gwake
15eat
15-3sgPron
"her/his good eating"
2.1.3
Quantifiers
kwega
ISgood
Like the determiner class, quantifiers constitute of a very restricted number of
elements which include -orhe 'all', -nin e 'some', -ingi' 'many', and all numerals. These
elements can only modify gerunds [-NPobj] types (6') but not gerunds [+NPobj] forms
(6). Thus gerunds [-NPobj] patterns with nouns.
67
(6)
*guthoma
ibuku
15read
5book
"all book, reading'
guothe
15-all
(6')
guthoma
kuu
Isread
ISDem
"all that reading'
guothe
1 5-all
2.1.4
Associative Phrases
The behavior and distribution of associative phrases also indicates that gerunds [-
NPobj] are nouns. Associative phrases feature prominently in the grammar of Bantu noun
phrases permitting the arbitrary association of any two nouns subject to the semantic
plausibility of the relation that can obtain between any two nouns. The gerunds [-NPobj]
types permit such associations as seen in (7') and (8') but the use of associatives with
gerunds [-nNPobj] is disallowed as indicated by (7) and (8).
(7)
*kurfa
mfanga
gwa
ciana
ni-gu-gu-ci-rwar-i-a
15eat
4cassava 15 Assoc lOchild
FP-15-Fut-lOO-sick-A-fv
"the children's eating of cassava will make them ill'
(7')
kuria
15eat
gwaku
I5-2sgPron
kwa
I5Asso
mfanga
4cassava
nf-gu-gii-ku-rwar-i-a
FP-15-Fut-2sgO-sick-A-fv
c
•your eating of cassava will make you ill'
(8)
*guthoma
ibuku
kwa
magegania
15read
5book
15 Assoc astonishing
"an astonishing reading of a book"
(8')
guthoma
kuu
kwa
ISread
15Dem
15 Assoc
"that astonishing reading'
2.2
magegania
astonishing
Verbal behavior
Modifiers of verbs include adverbs of various kinds and emphatic interjections. As
in the above cases the distinction between gerund [+NPobj] and gerund [-NPobj] 's also
evident in modification involving adverbs, emphatic interjections, and object prefixation.
68
2.2.1
Adverbs
Like regular finite verbs in (9) and (10). the gerund [+NPobj] can be modified by
the use of adverbs of various kinds. In (9'). gerund [+NPobj] is modified with the adverb
of extent and in (10') by the manner adverb. Adverbial modification is not possible with
gerund [-NPobj]
seen in (9") and (10").
(9)
a-ra-ri-a
mfanga
muno
3sgS-Pres-eat-fv 4cassava
a lot
"she/he is eating cassava a lot'
(10)
a-ra-n-a
mfanga
kahora
3sgS-Pres-eat-fv 4cassava
slowly
"she/he is eating cassava slowly"
(9')
kuna
mTanga
muno
1 Seal
4cassava a lot
"to eat cassave / eating cassava a lot"
(10')
kuna
mfanga
kahora
15eat
4cassava slowly
"to eat cassave / eating cassava slowly'
(9")
*kuna
gwake
ISeat
15-3sgPron
"her/his eating a lot'
muno
a lot
(10")
*kuna
gwake
ISeat
15-3sgPron
'her/his eating slowly'
kuu
kahora
15Dem slowly
2.2.2
Emphatic interjections
In Grkuyu there are special purpose interjections (Intj) which emphasize the
completeness or totality of state. Examples of Grkuyij emphatic interjections are evident in
(11). (12). and (13).
(11)
nyau
nF-ya-kir-a
9cat
FP-9S-quiet-fv
"the cat is dead silent'
ki
Intj
69
(12)
nf-a-ku-end-et-e
FP-3sgS-2sgO-like-Perf-fv
"he/she really likes you"
tu
Intj
(13)
ru-a-ku-end-et-e
FP-3sgS-2sgO-like-Perf-fv
"she/he really likes you'
ma
Intj
Emphatic interjections only go with verbs and not with nouns. That pattern is maintained
with respect to the distinction which exists between the gerund [+NPobj] (14) and the
gerund [-NPobj] (14').
(14)
kuiyura
ISfill
maf
6water
ki
Intj
"to be filled with water totally"
(14')
*kuiyura kiiu
ISfili
15Dem
"that total filling"
2.2.3
Object prefixation
ki
Intj
It is evident that gerund [+NPobj] types ( 15) are of category V because they permit
object prefixation (16).
(15)
kurfa
mTanga
15-4eat
4cassava
"to eat cassava / eating cassava"
(16)
ku-mr-na
15-40-eat
"to eat/eating it"
That the object prefix is really an incorporated pronoun is shown by the fact that
infinitive/gerunds with an object prefix do not permit modification with determiners (16")
which is consistent with the behavior of infinitive/gerunds with noun phrase complements
discus.sed in the foregoing sections.
70
(16")
2.3
*ku-mr-na
gwake
15-40-eai
15-3sgPron
"her/his eating it'
Positional distribution
Gerund [+NPobj] ^ind gerund [-NPobj]
exhibit their N/V split character in their
positional distribution. In what follows I e.xamine their distribution with respect to; linear
ordering, extraction possibilities of both gerund [+NPobj] ^ind gerund [-NPobj].
the
possibility of event intensification.
2.3.1
Linear ordering
In Bantu. DP fronting is done under different conditions from that involving verb
phrases. If it is correct to conclude that gerund [+NPobj] 's a verb phrase and that gerund
[-NPobj] are nouns, each one of them may be expected to undergo the relevant kind of
fronting. Thus in (17) and (17') both forms do occur in postverbal positions, which is a
position from which fronting of both NP and VP is possible.
(17)
(17')
Wanja
ni-a-re -end-a
Wanja
FP-3sgS-Pres -want -fv
"Wanja wants to read the bible'
Wanja
ni-a-re-end-a
Wanja
FP-3sgS-Pres-want-fv
'Wanja wants to read'
guthoma
Tsread
kirfkanfro
7bible
giithoma
ISread
Even when they occur in the postverbal position, infinitive/gerunds with complement noun
phrases are not noun phrases, because they do not trigger object prefixation (in bold) in the
matrix verb as seen in (18) for Grkuyu.
(18)
*Wanja nf-a-ra -ku-end-et-e
Wanja
FP-3sgS-Pres-150-like-PERF-fv
"Wanja likes it' (to read/reading the bible)
It has been demonstrated in (Mchombo, 1995) that in Bantu object prefixation induces the
possibility of free word order in the sentential level. The prediction with regards to
infinitive/gerunds is that free word order would not be possible since they do not permit
71
object prefixation. It turns out that with infinitive/gerunds, free word order exists in the
absence of the object prefix.
(i) fronting of gerund and its complement yields good forms (19).
(19)
Wanja
nf-a-re-end-a
Wanja
FP-3sgS-Pres-want-fv
"Wanja wants to read the bible"
(20)
a.
guthoma
b.
Wanja,
kTnkanrro.
guthoma
guthoma
Tsread
kfnkanfro
7bible
Wanja nf-a-re-end-a
kJrfkanfro
nf-a-re-end-a
(ii) Object prefixation with infinitive with its complement
When the object prefix -ku - is used in the matrix verb, (20a) and (20b) become
unacceptable as shown in (20'a) for (20a) and (20b') for (20b).
(20')
a.
*guthoma
b.
*Wanja.
kfrfkanfro
guthoma
Wanja nf-a-ra -ku-end-et-e
kfrfkanfro nf-a-ra -ku -end-et-e
Note however that the facts in (20) and (20') are reversed when the gerund [-NPobj] 's
considered . Unlike the case of gerund [-i-NPobj]. we do not get alternative ordering
without object prefixes (21). Rather, as expected with sentential noun phrase objects,
gerund [-NPobj]
permit multiple word orders with object prefixation as provided in
(21-).
(21
(21")
a.
*gijthoma
Wanja ni-a-re-end-a
b.
*Wanja. guthoma
a.
guthoma
Wanja nf-a -ku -end-et-e
b.
Wanja.
gQthoma
nf-a-re-end-a
nf-a -ku-end-et-e
Thus, with regards to the behavior and distribution of the object prefix, there is a very clear
split between gerund [-i-NPobj]
gerund [-NPobj]- Infinitives with complements are
purely verbal while those without are purely nominal, a virtual category shift.
72
2.3.2
Extraction
In Bantu, noun phrase complements of verbs can be extracted by clefting and
relativization. while verb phrase complements can not. Extraction of noun phrase objects in
Grkuyu is done by m' -clefting and relativization. If the proposal that the gerund [+NPobj]
is a verb phrase is correct, it may be expected that they will allow m' -clefting and
relativization. Tlie gerund [-NPobj] would permit such extraction. In (22) and (23) we have
the gerund [-i-NPobj] ^ind the gerund [-NPobjJ respectively appearing as the thematic
objects of the verb enda 'like'. Extraction of the thematic object by m -clefting and
relativization is permitted with gerund [-NPobj] types (23') and (23") and not with the
gerund [+NPobj] types (22') and (22").
(22)
Wanja
e-end-et-e
guthoma KFmathai
Wanja
3sgS-like-PERF-fv
ISstudy
Maasai
"Wanja likes to study the Maasai language'
(23)
Wanja
e-end-et-e
guthoma
Wanja
3sgS-like-PERF-fv
ISstudy
"Wanja likes to study/studying"
ni -clefting
(22') *Nr guthoma Kfmathai Wanja e-end-et-e
FP
Tsstudy
Maasai
Wanja 3sgS-like-PERF-fv
"It is studying the Maasai language that Wanja likes'
(23')
NT
guthoma Wanja e-end-et-e
FP
ISstudy
Wanja 3sgS-like-PERF-fv
"It is studying that Wanja likes'
Relativization
(22") *guthoma
Krmathai kuria Wanja e-end-et-e
ISstudy
Maasai
ISRel Wanja 3sgS-like-PERF-fv
"The studying of the Maasai language that Wanja likes'
(23")
guthoma
kuna Wanja e-end-et-e
ISstudy
ISRel Wanja 3sgS-like-PERF-fv
"The studying of the Maasai language that Wanja likes'
73
TTiese extraction tests thus reinforce the claim that gerund [-NPobjJ types are nouns
and that gerund [+NPobj]
something different. That gerund [+NPobj] behave like
verbs, is evident in (24') where a regular verb phrase such as agi'tengera 'while running'
can not be m' clefted. TTiis indicates that m' clefting is not used in the fronting of VPs. just
as in the cases involving gerund [+NPobj]
(24)
Wanja
a-ra-ri-a
a-gf-tenger-ag-a
Wanja
3sgS-Pres-eat-fv
3sgS-Consec-run-Hab-fv
"Wanja is eating (while) running'
(24')
*nf a-gf-tenger-ag-a
FP 3sgS-Consec-run-Hab-fv
"It is cassava Wanja is eating'
2.3.3
Event intensification
Wanja
Wanja
a-ra-rf-a
3sgS-Pres-eat-fv
In GTkijyu there exists a strategy for emphasizing the intensity of events and
processes by the use of infinitive/gerunds. The strategy involves using the infinitive/gerund
such as kun'a 'to eat/eating' in (26) which is identical to the verb of the higher clause rf'eat'. The effect is emphatic intensification of the described event. Interestingly this process
is sensitive to the inclusion or exclusion of the complement noun phrase. Hence as seen in
(25) in contra.st to (26). intensification of the verb can only be done with gerund [-NPobj]^(25)
*Wanja a-ra-n-a
kurfa
Wanja
3sgS-Pres-eat-fv
15eat
"Wanja is really eating cassava'
mfanga
4cassava
*^Thesc examples are somewhat reminiscent of English gerund conjunction for intensification effect m
(1).
( I)
a. He's practicing and practicing
b. *Hc's practicmg and practicing the recital piece
This IS m contrast to verbal conjunction for the same effect which is acceptablc as in (1").
( r)
a.
b.
He practiced and practiced
He practiced and practiced the recital piece
74
(26)
Wanja
a-ra-n-a
Wanja
3sgS-Pres-eat-fv
"Wanja is really eating'
kuna
15eat
m clefting is applicable only to (26) as indicated by (26').
(26')
Nr
kuna
Wanja a-ra-n-a
FP
15eat
Wanja 3sgS-Pres-eat-fv
"It is the case Wanja is eating'
The fact that kiirra 'to eat/eating' in (22') is not a thematic object of the verb of the higher
clause (as it is the case that intransitive verbs can also be used in this manner) is irrelevant.
The significant point is that m clefting is categorially restricted to elements of category N
and those include gerund [-NPobj]-
In this section, I have shown that the distribution of the gerund [+NPobj]
different from that of gerund [-NPobj]- "Hie former is of category V and the latter bears the
behavioral properties associated with elements of category N.
3.0
Morphological properties
In Bantu languages, noun class markers are affixed to noun stems while prefixation
and suffixation is done to verb stems. Verbal prefixation includes tense/modality, reflexive,
subject and object prefixation. Suffixation consists of a number of valency-affecting
operators and aspect markers. It is therefore a deceptively simple task to distinguish
between verbs and nouns by simply checking their morphological extensions.
Unfortunately with infinitive/gerunds this issue is not straightforwardly resolved. This is
because in addition to the category shift this chapter documents, ku- marks both the noun
(class 15) and also occurs as the concordial marker on the verb. It turns out however that
gerund [+NPobj] can be distinguished from gerund [-NPobj] by checking ten.se/aspect
marking, negation, and derivational morphology.
75
3.1
Tense and aspect
That gerund [+NPobj]
verbal is indicated by their ability to take aspectual (27)
and tense markers (28). The term 'infinitive' which means that a word or phrase is
unmarked for tense is clearly not useful in the discussion of Bantu words with class 15
prefixes.
(27)
Kamau a-hot-a
gu-ciar -ag-a
Kamau 3sgS-able-fv
15-bear-Hab-fv
'Kamau might give birth to twins'
(28)
Kamau a-hot-a
gti -ga-ciar-a
Kamau 3sgS-able-fv
15-Fut-bear-fv
'Kamau might give birth to twins'
mahatha
twins
mahatha
twins
Gerund [-t-NPobj] permit future tense marking on their stems (28), habitual aspect
marking (27). but not Perfect (29). or Complete (30).
(29)
*gQ-taar-ut-e
andu
I5-advise-PERF-fv
2people
'to have advised people'
(30)
*gu-taar-ir-e
andu
15-advise-COMPL-fv 2f)eople
'to have completed advising people'
It follows then that though gerund [-f-NPobj] types are very restricted in terms of their
inflectional markings, it is clear that they nonetheless behave like verbs, unlike gerund [NPobj] ones. More evidence of this split is observed in the marking of negation.
3.2
Negation
Swahiii infinitive/gerunds can be negated by introducing the morpheme -to - into
the prestem morphology of the base verb (31). Negation is generally marked in verbal
morphology in Bantu but not in nominal elements, so evidence that the infinitive/gerund
word is a verb is compelling.
76
(31)
Ku-to-chez-a
ngoma
15-Neg-play-fv
lOdrum
"not playing/to play daims"
Gerund [-NPobj] 'ike regular nouns, can not be marked with the negation prefix (31').
(31")
3.3
*ku-to-chez-a
kwake
15-Neg-play-fv
15-2sgPron
"her/his not playing"
Verbal extensions
Like verbs, gerund [+NPobj] permit affixation of derivational morphology -
applicativization (32a). causativization (32b).
(32)
a.
gu-taar-ir-a
miindu
ciana
15-advise-A-fv 1 person
Schild
'advising children for a person'
b.
gu-taar-ith-ya
mundu
ciana
15-advise-C-fv 1 person
Schild
'help advise children for a person'
Extension with derivational suffixes is a property that is unquestionably verbal in Bantu
languages. Unlike gerunds [-NPobj] types and nouns more generally, extended forms can
not be modified with nominal modifiers.
(33)
a.
*gCi-taar-rr-a
kwake
15-advise-A-fv 15-2sgPron
'your advising for'
b.
*gu-taar-ith-i-a
kwake
15-advise-C-C-fv
15-2sgPron
'your causing to cause to advise'
4.0
Lexical restrictions
In terms of lexical information that is pertinent to the distinctions I am drawing
between infinitive/gerunds, there are two interesting issues that come to the fore. One
relates to semantically empty verbs such as auxiliaries and the other relates to intransitive
bases where the distinction between gerund [-NPobj] ^nd gerund [-NPobj] does not apply.
77
In considering the function of nouns as arguments in the syntax, it must be the case
that for something to be a noun it must name something (an entity, individual etc.). In
effect, it has to be semantically contentful. If this assumption is made, then it may be
expected that auxiliary type verbs (and other semantically bleached verbs) would prohibit
the formation of gerund [-NPobjl type which behave like nouns. It would however be
possible to form gerund [+NPobj] types with auxiliary verbs. Another issue pertaining to
the lexical specifications involves infinitive/gerunds formed with intransitive verb stems. It
may be expected that such forms do not exhibit a categorial shift between N and V. I
examine each of these issues in turn.
4.1
Auxiliary verbs
Can aiLxiliary verbs noniinalize?
Since all gerund [-NPobj]
nouns, they must be semantically contentful
disallowing the use of auxiliary-like verbs. While this is indeed the case in the Swahili
examples in (34'). no such restriction is imposed on gerund [+NPobjJ cases (34).
(34)
ku-w-a
mzazi
ni kazi
gumu
15-be-fv 1 healer is lOwork lOhard
"to be a parent is hard work'
(34')
*ku-w-a kwake
ni
15-be-fv 15-2sgPron
is
"her/his being is hard work'
kazi
lOwork
gumu
lOhard
By this behavior, infinitive/gerunds are verbs.
Can infinitive/(gerunds be complements to auxiliary verbs?
As pointed out by Bresnan and Mchombo 1995:233 for Chichewa. Grkuyu and
Swahili have a class of auxiliary-like verbs which take only infinitival complements. One
such verb is hota "be able' in Grkuyu and weza 'be able' in Kiswahili. Note that gerund
[+NPobj]
the one that occurs in that position (35) and (36) and not the gerund [-NPobj]-
This is equivalent to English expressions such as "John is able to speak French'. No noun
78
phrase can occur in this position as (35') and (36') demonstrate. Thus, the Grkuyu and
Swahili examples demonstrate that the relevant distinction between gerund [+NPobj]
gerund [-NPobj] 's that between N and V.
Grkuyu
(35)
Kamau a-hot-a
15-bear-fv
Kamau 3sgS-able-fv
'Kamau might give birth to twins'
(35')
gu-ciar-a
* Kamau a-hot-a
15-bear-fv
Kamau 3sgS-able-fv
'Kamau might give good birth'
Swahili
(36)
Mzee
a-na-wez-a
1 elder
3sgS-Pres-able-fv
'the elder can recite poetry'
(36')
4.2
ku-imb-a
15-sing-fv
*Mzee a-na-wez-a
ku-imb-a
I elder 3sgS-Pres-able-fv
15-sing-fv
'the elder can (give) good recitation'
mahatha
twins
kwega
15good
mashairi
6poem
kuzuri
15200d
Intransitive bases
Another lexical property of relevance to the discussion of category shift between N
and V is infinitive/gerunds with intransitive bases. The question arises whether the
proposal that gerund [-NPobjJ is of category N implies that all infinitives/gerunds based on
intransitive verb stems are consequently nouns. Categorial shift between N and V can be
demonstrated by the same tests outlined above. Depending on the type of modifications
that employed above, intransitive infinitive/gerunds will behave consistently. As evidenced
in (37) and (38). intransitive infinitive/gerunds can be of category N or of category V
depending on how they are modified which encapsulates categorial shift between the two
categories.
(37)
gu-kir-a
kiju
15-quiet
15Dem
"that silence of yours"
2waku
15-2sgPron
79
(38)
gu-kir-a
muno
I5-quiet
a lot
'to be very silent'
(39)
gu-kir-a
ki
15-quiet
Intj
"to be totally silent'
It is interesting to note that transitive stems such as -taar- 'advice' in (40) when rendered
intransitive by -an- sufFixation as in (40') exhibit the mixed patterns observed of the regular
intransitives as provided in (41) and (42).
(40)
gu-taar-a
ciana
15-advise-fv
lOchild
'to advise/advising children'
(40')
gu-taar-an-a
15-advise-R-fv
'to advise/advising each other'
(41)
gu-taar-an-a
kuu
kwanyu
15-advise-R-fv
15Dem
15-2plPron
"that advising each other of yours'
(42)
gCi-taar-an-a
muno
15-advi.se-R-fv
a lot
"to each other too much'
5.0
Summary
In studying the behavioral and distributional propenies of gerund infinitives. I have
arrived at results that indicate very clearly that we are dealing with elements that are
different in category status. By the propenies in (43), it may be concluded that gerund
[-i-NPobj] types are of category V.
(43)
a.
permit extraction of thematic objects
b.
allow object prefixation
c.
permit free word order without object prefix
d.
disallow relativization
80
e.
cannot be nf-clefted
f.
disallow use of associative phrases
g.
can be formed with auxiliaries
h.
permit negation (Swahili)
i.
very restricted use of tense and aspect
j.
take verbal modification (adverbs, intensifiers. emphatic interjections)
Properties opposite of those in (43) have been exhibited by the gerund [-NPobj] types (43')
by which I have concluded that they have N status.
(43')
a.
do not have thematic objects
b.
disallow object prefixation
c.
permit free word order only with prefix
d.
permit relativization
e.
can undergo m'- clefted
f.
permit use of associative phrases
g.
restricted from semantically empty verbs
h.
disallow negation
i.
tense and aspect disallowed
j.
take nominal modification (adjectives, determiners, possesive pronouns etc.)
It is ver>' clear then that there is a shift in category which permits the distinctions made in
(43) and (43') to exist. I proceed to propose reasons for the categoriai shift from V to N.
6.0
Proposal
Infinitive/gerunds suggest that there is a category split between gerund [+NPobjJ
and gerund [-NPobj]- The former are of category V and the latter of category N. Evidence
comes from behavioral and distributional properties typical of verbs and nouns. The head
of the gerund [+NPobj] bears the structure indicated in (44). The verbal properties of
81
marking inflectional and derivational morphology, and of disallowing nominal modifiers
but permitting adverbial modification provide convincing evidence that (44) is the right
structure. The base is verbal which combines with -a to produce a verb stem which
combines with fiii - to form the
(44)
category.
V
gij
item
stem
^CX3l
Ext*
Words with a sub-lexical structure as in (44) project the category V. The verb
phrase permits extraction of its thematic objects (43a), and allows for object prefixation
when the root is transitive (or transitivized by derivational affixation). Like regular VPs.
gerund [+NPobj] can be fronted without object prefixation (43c), they cannot extracted by
relativization (43d) or by nr-clefting (43e). It is because gerunds [+NPobj] are verbal that
they do not allow the use of associative phrases (43f) but can be formed with auxiliaries
(43g). be negated (43h) and admit tense and aspect affixation (43i). They also take typical
verbal modification (43j) because they are verbal as proposed in (44).
The behavior and distribution of the gerund [-NPobj] suggests that it has the
structure (44"). There is no dispute that the base is verbal but its combination with -a
yields an Nstem. which combines with the class marker gu - to form a noun. Inflectional
and derivational affixation takes place at the stem level and not at the verb root. This
explains why nominalizing the verb root rules out any extension possibilities having to do
with object prefixation (43'b). marking of negation (43'h). and tense and aspect affixation
(43'i).
82
(44')
N
stem
This structure shows that the gerund [-NPobjl 'S a noun and like other nouns in the
language, it participates in free word order distribution triggered by object prefixation
(43 'c). It is because gerund [-NPobj] elements are nouns that their extraction by
relativization and nf-clefting (43 'd) and (43 'e). Since gerund [-NPobj] words are nouns, the
associative phrase is permitted (43 'f) as well as typical noun modification (43j).
Semantically empty verbs such as auxiliaries are not permitted (43 'g) because nouns must
have semantic content.
It is clear then that the distinction between gerund [+NPobj] and gerund [-NPobj]
that between V and N respectively and the best way to capture it is at the stem level. The
Bantu verb stem has been shown to be the locus of a number of important processes in the
literature (Mchombo. 1993; 1995), the most salient among them being derivational and
inflectional suffixation possible in (43). It is because of the categorial split that occurs due
to the nominalization of a verb root at the sub-lexical level that disallows morphological
extensions on Nstems for (44').
The categorial split that is documented in this chapter is not entirely unprecedented.
When the abstract concept class 14 m- is considered, we get behavioral properties that
suggest that the category of the stem makes crucial differences in the syntax.
6.1
The u-class
As pointed out in chapter 1. u- words belong to class 14 which is the class of
abstract concepts such as iiugi' 'knowledge', but also consists of names for miscellaneous
83
objects such as ucuru 'porridge'. The plural for abstract concepts as well as tangible ones
that can be pluraiized is found in class 6. maiiugr 'many types of knowledge'and macuru
•porridge, in a collective sense' for the foregoing examples. Naturally, it is built extensively
with verb bases as exemplified by (45).
(45)
a.
u-tungat-i
14-serve-Nzer
'service'
b.
u-thom-i
14-read-Nzer
'reading'
c.
u-rug-i
14-cook-Nzer
'cooking'
Given that there are valency requirements with all verbs it is reasonable to expect the forms
in (45) to be able to have their thematic objects expressed. This is however, not the case as
(45') indicates.
(45')
a
*u-tungat-i
kanitha
14-serve9church
'church service'
b.
*Q-thom-i
14-read
'book reading'
c.
*u-rug-i
irio
14-cook
1 Ofood
'food cooking'
mabuku
6book
Since (45) are permitted and (45') are disallowed, m - words are expected to exhibit the
patterns observed with gerund [-NPobj] types and not that of gerund [+NPobj] ones. This
turns out to be right because as (46) indicates, all il- words can be modified with nominal
elements. As pointed out in chapter I. m - words pattern with the class 3 concord system.
84
(46)
a.
u-tungat-i
mwega
14-serve3good
'good service'
b.
u-tungat-i
14-serve'that service'
c.
u-tungat-i
wake
14-serve3sgPron
'his/her service'
d.
u-tungat-i
wa
14-serve3 Assoc
'church service'
kanitha
9church
e.
u-tungai-i
uria
14-serve3Rel
'service which is good'
mwega
3good
Qcio
3Dem
Further, it- words are never modified with adverbials (47a) or the emphatic interjections
(47b). Infact, their verb status is totally inaccessible to common verbal processes such as
reduplication (47c). object prefixation (47d), and even valency changing affixation such as
that induced by causativization (47e).
(47)
a
*u-tungati
wega
14-serve
well
'servicing well'
b.
*u-tungati
14-serve
'real service'
c.
*u-tungatungati
14-serve
'service'
d.
*u-mf-tungati
14-30-serve
'service of it'
e.
*u-tungat-ith-i-a
14-serve-C-C-fv
'cause to cause to serve'
tu
Intj
85
It is important to pay attention to the kinds of suffxal vowels that are found on m words as illustrated by (48). In these examples we see that suffixal vowels are
-a. -u. -e.
-o. These suffixes are among those described by Mugane (1996) as nominalizing affixes in
Grkuyu grammar'^.
(48)
a.
u-thak-a
14-be pretty'beauty'
b.
Ci-nen-e
14-big'bigness'
c.
Q-tung -u
15-advice-R-fv
'bigness'
d.
ij-tungat-i
14-serve'service'
'^I (Jo not think that these tmal vowels are simply a case of vowel harmony. Rather these final vowels are the
nommalizing morphemes. Support for this is seen in the examples below ( 4 8 ' ) where the suffixal vowel is not
identical to the root vowel. This does not rule out the possibility of utilizing feature specification theories such a s
Radical Underspecification (Archangeli. 1988) t o determine the underlying features that determine the quality o f
these surface vowels. Such a task is however orthogonal t o the concerns of this study.
(4S')
a.
u-cg-a
14-good-Nzer
'goodness'
bi
u-tun-c
14-rcd-Nzer
'redness'
c.
u-mat-u
14-densc-Nzer
'density'
d.
u-ther-1
l4-l|ght-N/cr
'liahi'
Li-ath-<)
l-Haw-N/er
'law'
86
e.
u-tuur -o
14-live'living'
It is not surprising that one of the nominalizing vowels is -a in (48a). Thus in the
infinitive/gerund constructions, the -a in gerund [-NPobjl is likely not the conventional
suffixal vowel commonly associated with direct imperatives but the nominalizer
morpheme which upon combination with the verb root (Vroot) converts it into an noun
stem (Nstem) as indicated by (49).
. ] — ]
\iasc
Nstem
TTie conventional status of -a as a final vowel in verbs is correct for gerund [+NPobjJ (50).
(50)
Tjasc
6.2
*siem
Structure-Function association
In terms of structure-function association, there are the usual distinctions between
noun phra.ses and verb phrases as indicated in (51) for gerund [+NPobj]- The V is both the
lexical and functional head of the VP. Depending on the transitivity of the head V.
complements to the head can be expressed. Sisters to lexical heads are subcategorized
arguments to the head as provided by universal principles of structure-function association
discussed in chapter 1.
87
(51)
PRED 'eating/to eat <OBJ >'
kuria
eatinglio eat
mi ansa
cassava
PRED 'cassava'
OBJ i GEND 4
'eating/to eat cassava'
The gerund [-NPobj] types have the conventional structure of noun phrases (52). In this
structure gerund [-NPobj] project
and their maximal projection is NP. The information
contained in regular Bantu nouns typically includes a PRED value and the noun class. In
Bantu as in other language families nouns do not take complements which explains why
the gerund [-NPobj] types do not permit object prefixation. among the other behavioral
properties outlined in this chapter.
(53)
NP
T=i
N
PRED 'eating'
GEND 15
kuria
eating
6.3
Summary
To summarize this section. I have shown that the gerund [+NPobj] can be
distinguished from gerund [-NPobj]- The shift in category from V to N happens at the
stem level where the Vbase in gerund [-i-NPobj] projects a V stem while the gerund
[-NPobj] though having a V base shifts to an N stem. Once this distinction is made, all the
contrasts in behavioral and distributional properties between the two types of
88
infiniiive/gerunds follow. I have also extended the argument by briefly drawing a parallel
between gerund [-NPobj]
u- words in order to show that it is not strange at all for
valency bearing elements such as verbs to be converted to nouns and thereby losing all
their predication properties.
7.0
Previous studies
While previous studies on gerunds are numerous, none of them provides any
satisfactory account for the facts assembled in this chapter. In this section. 1 will brietly
explain why I do not assume the conclusions arrived at by Abney (1987) and why in.spite
of the fact that I use Grimshaw's (1991) theory of extended projections. 1 do not assume
her theory of gerunds.
7.1
Abney (1987)
Abney (1987) points out that verbal gerunds in English have the external distribution of
NPs and the internal structure of VPs. As NPs their subjects appear as possessives while
as VPs they take adverbials. As a consequence, gerunds switch between category N (54a)
and V (54b).
(54)
a. His writing of the essav did not help clarify matters
b. 1 resent the cat's eating mice in front of me
For (54) Abney proposes that gerunds are VP complements to D. which makes it possible
to account for the nominal/verbal properties and to indicate at what juncture in phrase
structure the switch occurs without violating the basic principles of X-bar theory. The
category D is viewed by Abney as being the noun phrase equivalent of inflectional
projections of verb phrases within sentences. This is because both gerunds and verb
phrases have subjects. In their internal distribution, gerunds have subjects which appear as
89
possessives. they take adverbiais and assign case as verbs do. Externally, gerunds take
adjectives etc. To Abney. these behavioral and distributional properties suggests that the
structure of gerunds is as in (55).
(55)
DP
D
VP
Given the data examined in the foregoing sections. Abney's theory applied to Grkuyu does
not work because there is no need to capture the verbal and the nominal properties
simultaneously. Rather, the evidence from Grkuyu suggests that the nominal set of
gerunds has the behavior and distribution of root nouns and must be characterized as such
in the constituent structure. The verbal ones are VPs and should be represented as such.
The disjunction of properties between gerunds with complements and those without
obviates the necessity of (55) and requires that data on gerunds be construed differently.
The proposal (argued for in this chapter) that there is need for the separate characterization
of nominal gerunds from verbal one finds support in the fact that gerund properties are
uniquely nominal or uniquely verbal, but never a mixed. Thus no cases are attested where
there are verbal case-markings with adjectival modifiers simultaneously - a puzzle
Grimshaw (1991) notes, though her proposal is inadequate in the light of Grkuyu (and
Bantu) data.
7.2
Grimshaw (1991)
Even though I adopt Grimshaw's theory of extended projections in chapters 4. 5.
and 6.1 do not adopt her theory of infinitive/gerunds. In her theory of extended projections
Grimshaw (1991) considers Abney's (1987) idea that gerunds are VP complements to D
90
lo be illegitimate. This is because in her theory (explained in chapter 1) V and D belong to
different syntactic categories, and hence, the VP and the DP cannot form an extended
projection. Grimshaw then points at the categorial hypothesis (that V and D are involved)
as the source of the problem. She says that inspite of exhibiting both nominal and verbal
behavior, the morphology of the nominal gerund is identical to that of the verbal gerund.
She proposes then that ing head of the gerund is itself N/V neutral. This means that the V
that is headed by -ing and the VP that is headed by the V are N/V neutral and not really V
and VP. An N/V neutral element is either verbal or nominal. For Grimshaw the V is
[+substantive. -N] and N is [+substantive. +N], which necessitates that the presumably
N/V neutral -ing is simply [4-substantive] and unspecified for the feature [N]. Thus.
Grimshaw refines the definition of extended projection to require that the categories be
non-distinct rather than identical, which means that the gerund "VP" can be a complement
to D without violating her assumptions. Still, as she points out, there is a problem -to
explain why gerund properties are uniquely nominal or uniquely verbal, and never mixed.
Grimshaw assumes that the answer to this problem lies with "the characteristics of
neutralization".
Turning to GHcuyu (and Bantu) facts and the analysis presented in this chapter.
Grimshaw's (1991) proposal of category neutrality finds no support. Bantu languages
being morphologically explicit do allow us to see processes not visible in morphologically
impoverished languages such as English. The fact of the matter is that we have sub-lexical
morphology which clearly indicates whether an element is a nominalized verb or a pure
verb. I have shown that nominalized verb roots (gerund [-NPobj]) unlike pure verbs (and
gerund [-i-NPobj]) do not permit inflectional and derivational affixation. Thus gerund
[-NPobj] iire unambiguously of category N and gerund [-t-NPobj])
articulately, of
category V. When word formation processes are carefully considered there is no point at
91
which there is categorial neutrality or any need for extended projections in gerund word and
phrase construction. Rather, the categorial ambiguity exhibited by gerunds is a consequence
of general properties of Bantu nominalization structures.
8.0
Conclusion
In this chapter, it has been shown that infinitive/gerunds exhibit genuine categorial
ambiguity. I have assembled a number of syntactic, lexical, and morphological properties
associated with gerund [+NPobj]
gerund [-NPobj]- I h^ive identified the word
formation level as the locus at which the two types may be distinguished. The gerund
[+NPobj] type is verbal hence its potential for inflectional and derivational affixation
among other properties listed in (43). Gerund [-NPobj] words are formed by the
nominalization of the verb root. The verb root is a level deeper than the verb stem where
affixation takes place. It follows then, that gerund [-NPobj] words can never permit
inflectional or derivational affixation. I have pointed to empirical evidence to show that -a is
a possible nominalizing suffix along with
-e. -u. and -o. In this analysis the noun class
marker combines with a noun stem to make nouns. I have also briefly looked at the uclass marker to show that no noun class marker combines with a verb stem to make a
noun. As such. I have concluded that noun class markers are not the category determining
elements of the words they are part of. but rather the elements which encode information
about gender class and number (Bresnan and Mchombo. 1995). Evidence in this chapter
thus indicates that the category determining material in CFkuyu word structure is suffixal.
I have also argued that it is not necessary to posit (as does Grimshaw (1991)
category neutral N/V elements or even extended projections in gerund word and phrase
formation. Rather, the category split or ambiguity exhibited by gerunds is a consequence of
a nominalization process. Once the attendant word structures are distinguished, the
category of gerund [-NPobj] can be uniquely determined as N and that of gerund [+NPobj]
as V.
93
CHAPTER 3
STRUCTURAL ISSUES
1.0
Introduction
In chapter 2. we have chronicled the properties of pure V and pure N. and in the
process observed that infinitive/gerunds are ambiguous between these two categories.
Maintaining Abney's (1987) assumption that the noun and the verb pose the most
fundamental opposition in grammar'®, in the present chapter we investigate some panicular
aspects of N and V. One of the most salient features of nominal modifiers is that their
prefixal morphology is identical to that of the subject prefix which occurs on verbs. The
noun bears the noun class affix which is morphologically distinguished from the marking
of modifiers. The status of affixes with regards to whether they are morphological
formatives or whether they are phrasally derived has been a major issue is current Bantu
studies. This chapter will evaluate some of the claims made about affixes using Grkuyu
facts. The analyses of the Bantu noun class currently available in the literature in works
such as Myers (1987). Carstens (1991), and Kinyalolo (1991) offer inadequate proposals
concerning Bantu affixes. In fact there have been conflicting reports on the question of the
kind of information contributed by the noun class marker. Mufwene (1980) found that
noun class markers have a status ambiguous between inflection and derivation. Bresnan
and Mchombo (1995) have argued that with the exception of the locative classes. Chichewa
noun class markers are lexical elements which are only manipulable by word formation
processes while Myers (1987). Kinyalolo (1991). treat noun class markers as phrasal
elements.
" ' D i c k Oehrle int'orms me that he tlrst heard o t ' t h i s assumption from Roman Jakobson who a d m i t t e d that he
c o u l d n ' t decide whether the N V opposition was most fundamental o r the opposition between c o n s o n a n t and
vowel.
94
This chapter also elaborates on the properties of noun phrases and verb phrases in
terms of their complements and modification. For our purposes here, we will assume that
the structure of the Bantu verb phrase proposed in current literature (Bresnan & Mchombo
(1987, 1995). Alsina (1993). Mchombo (1993. 1995). etc..) forms an adequate
background. As such our discussion of the verb phrase will be brief.
2.0
Facts
In this section we take a brief look at some properties of the Bantu noun and verb.
We will first look into the linear order of nouns and their modifiers and the morphology of
nouns, their specifiers and modifiers. We will then consider verbs and their complements.
2.1
The noun and its modiflers
The pragmatically unmarked linear order of elements in the GiTcuyu noun phra.se is
the noun (N), demonstrative (Dem), possessive pronoun, quantifiers (QP). adjectives
(AP). and the associative phrase (XPassoc) as in(l).
(1)
nyungu ici
ciake
ciothe ndune
lOpot
lODem 10-3sgPron lOall lOred
"all these red pots of the village'
cia
lOAssoc
gfcagi
7village
Of the.se elements, only APs and XPassoc can be stacked by iteration.
In chapter 1 section 3.3.3. it is demonstrated that when the canonical order of
elements is violated, a comma intonation occurs. Tne comma intonation is taken to mark
appositivity. indicating that any elements that occur to the right of a comma are outside the
noun phra.se. The point at which a comma is put marks the point at which the linear order
has been violated. Elements to the right of a comma that conform to the unmarked linear
order are not internally separated by further commas.
Also as pointed out in chapter I section 5.1. elements of the noun phrase bear one
of two types of morphological markings. Consider the forms indicated in (2) and (3).
95
1
Noun class
1 mu
2 a
3 mu
4 mf
5 i
6 ma
1 gi"
8 i
9 n
10 n
11 ru
12 ka
13 tu
14 u
IS ku
16 ha
17 ku
2
example
muini
aini
mutf
mTtr
igego
magego
gftr
itr
ngui
ngui
ruor
kahu
tuhiT
uuru
kuruga
haha
guku
3
Gloss
singer
singers
tree
trees
tooth
teeth
chair
chairs
dog
dogs
river
little boy
little boys
evil
to cook
herehere-
4
Adjective -nene 'big'
munene
anene
munene
mTnene
Inene
manene
krnene
nene
nene
nene
runene
kanene
tunene
munene
kunene
hanene
kunene
Nouns and adjectives bear noun class morphology (2) which is in contrast to the
morphology of modifiers such as demonstratives, associatives, and quantifiers. The
morphology of nominal modifiers is identical to that found on verbs (familiarly called the
subject prefix within verbs) as (3) indicates. The question here is why nominal elements in
columns 2 and 4 in (2) are distinguished morphologically from columns 2. 3. and 4 in (3).
96
3
Dem
Qyu
4
all
wothe
waku
uyu
wothe
wake
ucio
wothe
1
Class
ist S2
2
Pron
wakwa
2ndSg
3rci Sg
ist Pi
witu
aya
ithuothe
2nd PI
wanyu
aya
inyuothe
3rd PI
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
wao
aya
othe
wake
wao
wake
yake
rfake
make
giake
ciake
yake
ciake
ruake
sake
twake
wake
gwake
hake
gwake
uurfa
aarfa
ijurfa
una
riTria
maaria
kiTria
iiria
TTrfa
iiria
ruurfa
kaarfa
tuurfa
Qurfa
kuuria
haarfa
kuuria
wothe
othe
wothe
yothe
rTothe
mothe
gfothe
ciothe
yothe
ciothe
ruothe
gothe
tuothe
wothe
guothe
hothe
guothe
2.1.1 The noun and its complements
Underived nouns can have complements as seen in (4a) and (4b). Complements of
pure nouns are associative phrases. In these examples the noun to the right of the
associative is the thematic argument of the one to the left.
(4)
a.
mbica
cia
muhiki
lOpicture lOAssocl bride
"pictures of a bride'
b.
mbuthu ya
Qcuru
9gourd 9 Assoc 14porridge
'gourd of porridge"
97
That nouns take associative phrases provides further proof of the distinctions made in
chapter 2 with regards to infinitive/gerunds. Gerund [-NPobj] cases are nouns hence they
take associative phrase complements as in (5) but Gerund [+NPobj] ^re verbs and can not
take associative phrase complements (but do take direct complements) as seen in (5').
(5)
guthoma
kuu
kwa
15read
15Dem
15 Assoc
"that astonishing reading"
(5")
*guthoma
ibuku
kwa
magegania
I5read
5book
15 Assoc astonishing
'an astonishing reading of a book'
2.2
magegania
astonishing
The verb
In contrast to the noun, verbs can take complements of various kinds (VP. S. NP)
cross linguistically. Most important among the differences between verbs and nouns is
that verbs have selectional restrictions governed in part by their valencies (some are
classified in verb classes based on aspectual semantics etc.). Typically base intransitives
unlike simple transitives do not take complements. In Bantu languages, valency increasing
morphologies (such as the causative, the applicative) and valency reducing derivational
affixes (the reciprocal, the stative) all apply to verbs. In Gfkijyii. verbs of any valency can
mark passivization (Mugane (1996), in addition to the more conventional Bantu verbal
pretlxal morphology such as reflexivization, and tense marking.
Structurally, Bantu verb phrases are head initial endoceniric structures (6b). As is
the case with other language families. Bantu verbs take clausal complements which include
NP. VP. andS'i.
" T h e structure of V P is complex and there has yet t o be a convincing case of the presence of an IP projection.
Carstcns and Kinvalolo (1991) have analyzed compound tense construction to be projecting an I P but motivation
for an IP projection in Swahili is not conclusive. In fact, there are several tests that can be performed t o
determine whether kuwa + V is a discontinuous morpheme o r whether the relation between kuwa and V is
syntactic. Contini-Morava (1991)
proposes that k u u a is a lexical item with the selectional properties.
.Addressing the syntax of CTs. Carstcns and Kinyalolo (1991) argue that C T constructions are fcttua-suppon
(Wald. 1973) constructions with k u a a inserted to support the tense morpheme. Here we take the position
98
ma-ra-kiy-a
njohi
2S-Tns-brew-fv
9beer
"thev are brewins beer"
(6)
b.
In addition to the pronominal object prefixation (OP in column 3 of (7)) already introduced
in chapter 2. verbs bear morphology that concords (SP) with their subject arguments as
indicated in (7) by column 2.
1
Class
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
2
SP
ua
u1nma
21"ciTcirijkatuukijhaku-
3
OP
-mu-ma-Q-\-
-n-ma-gi-ci-F-ci-ru-ka-tu-u-ku-ha-ka-
proposed in Mugane (1994) that kuwa has non-trivial semantic content. There are several tests that suppon the
idea that kiiwa (like fully Hedged V) is complement taking. These include; negation processes of kuwa and its
V P complement, agreement, adverb placement. S e e also Demuth & Gruber (1994) and the references cited there
for a discussion of the C T complement phrase analyzed aas clausal complement. There may be reason (still
unexplored I to think that kuti a is diachronically a verb that is synchronically being reanalyzed a s an auxiliary as
VIcWhorter (1992) proposes regarding the Swahili copula - m - .
99
3.0
Previous Analyses
Some studies have analyzed the structure of Bantu NPs and VPs. Baker (1985)
proposes ways of utilizing morphology in the syntax disregarding lexical integrity, Myers
(1987) following Fortune (1955. 1985). studies Shona word structure. Carstens and
Kinyalolo (1991). study the Swahili IP. Kinyalolo (1991) looks at Kilega phrase structure.
Carstens (1991) studies the Swahili noun phrase and Mchombo (1978. 1993). look at
word structure as autonomous of syntactic manipulation. Of these studies, we will look at
the proposals in Myers (1987) and Carstens (1991). and how they analyze the Bantu noun
class.
3.1
Myers (1987)
Myers studies Chishona morphology. He analyzes the noun class prefix as an
element whose selectional requirements indicate that it may take NP complements as
exemplified by (8).
(9)
N'
NP
This structure should allow recursive nesting involving the noun class which is not
possible. This structure is also at odds with our suggestion that the noun class marker is
pan of the structure of the word and not of the phrase as (9) indicates.
3.1.1
Problems with Myers (1987)
The problem with Myers's analysis of the noun class marker is that it does not
respect the lexical integrity of words (Bresnan and Mchombo (1995)) in that the noun class
prefix selects NP to project an N'. Responding to Myers's claims. Bresnan and Mchombo
100
study Chichewa and draw the conclusions in (10). (10) indicates that Chichewa noun class
markers (except locative classes) respect lexical integrity of words.
(10)
a.
noun class prefixes (exluding the locative classes) are
bound morphemes.
b.'
following (a), noun classes (excluding locatives)
can not be conjoined nor undergo ellipsis, among the
other results of lexical integrity tests.
These properties hold for Grkuyu as well and the class of locatives has been replaced by a
system in which affixing -ini on any noun has the effect of indicating location. Employing
some of the tests indicated in (10b) yields the expected results. (11a) illustrates the
conjunction of two words and in (11 b) where the conjunction of stems under the .scope of
one noun class marker is impossible.
(11)
a.
[ma-rigu
na
ma-embe]
6banana
Conj 6mango
'bananas and mangoes'
b.
*ma-[rigij na
-embe]
6banana
Conj 6mango
'bananas and mangoes'
Contrasting (11) with (12). the associative -a is a phrasal element unlike the subject prefix.
Thus we can conjoin NPs under the scope of the associative.
(12)
a
[mburi
na
ng'ombe]
2 Assoc lOgoat
Conj lOcow
'those associated with goats and cows'
In chapter 2, we proposed that the distinction between V and a deverbal noun is
explained in the sub-lexical level. We provided (13) as the structure of a gerund [-NPobj]
where the noun class marker (Ncl) combines with a noun stem (Nstem) to form a noun.
101
(13)
N
stem
Similarly, in the case of root nouns, we have (13') which indicates that the Ncl combines
with a root noun to form a noun'-. By (13'), (9) is bad becau.se it is an attempt to have the
noun class taking a clausal element as Myers (1987) suggests.
Ncl
3.2
Carstens (1991)
Carstens concentrates on the structure of the noun phrase. She analyzes Swahili NPs
as DPs with empty
projects #P.
to which nouns raise. Number is a functional category #. which
#P is a complement to D^.
Nouns raise to # from NP positions to collect
number features and then rise further to D® in order to account for the linear order [Npronoun-X] as shown in (14). The figure (14) is adopted from Carstens. 1991:108. In
her analysis, the noun cla.ss marker is an allomorph of the #®.
' ~ l n 11)') Wt.' draw a thicker line between the Nstem and N to indicate that the Nstem is head of the whole word
since the usual assumption is that the head daughter of a complex constituent has the same category features a s
the constituent itself and the same or fewer bar-levels docs not decide whether noun class or Nstem is the head.
102
D
#P
kikombe
cup 4
Pron
changu
mine
tmax
A9ent
NP
theme
kikombe hiki changu
7cup
7this 7my
'this my cup'
3.2.1
Problems with Carstens (1991)
The problem with Carstens's proposal applied to Grkuyii is in relation to the
formation of nouns with preprefixes. The 'stacked' gender markers or preprefixes require
phrase structure functional projections (#P) inside of the morphology, violating lexical
integrity (Bresnan and Mchombo 1994). If the formation of nouns with one prefix such as
mu-iiiriru 'flute' (15a) has the phrasal source shown in (14). then for nouns with
preprefixes. such as (15b) ka-mu-turiru 'small flute', require an additional phrasal .source.
(15)
a. mu-turiru
3-flute
'a flute'
103
b. ka-[mu-turiru]
12-3-flute
'a small flute'
If the analysis in (14) is the correct one for (15). Bresnan and Mchombo have argued that
we should be able to find both outer and inner concords alternating freely. From (16) we
see that only outer concords are permitted. Carstens (1991) offers no e.xplanation why data
such as (16b) are consistently ruled out across Bantu languages (Bresnan and Mchombo
(1995); Chichewa, Kiswahili).
(16) a.
ka-mu-turiru
ga-cong'i
12-3-flute
12-ugly
'a small ugly flute'
b. *ka-[mQ-turiru mu-cong'i]
12-3-flute
3-ugly
3.3
Summary
To conclude this section, it is notewonhy that neither Myers (1987) nor Carstens
(1991) recognize that AP gets noun class morphology while -ote (quantifier) and
demonstratives get verbal SP morphology (the three elements hypothesized to be base
generated as adjuncts in Carstens' analysis) as the grammars of Swahili (Ashton (1947).
Mbaabu (1985) reveal. While details could vary from one Bantu language to another, the
morphology of words is inaccessible at the level of constituent structure, but it (the
morphology) does contribute significantly to the functional structure in the syntax
(Bresnan, 1995).
4.0
Proposal on affixes
With regards to the split in morphological marking between nouns and their
modifiers, we have seen that their the subject prefix has the same information within the
104
verb phrase as it does within the noun phrase. Thus the subject prefix which occurs on
nominal modifiers and verbs encodes agreement information. This leaves unexplained the
status of the noun class marker and the information it contributes. Here. I will take the
position arrived at in Bresnan and Mchombo (1995) that noun class markers are part of
word formation, which serves to provide the semantic classification (albeit a fuzzy one in
synchronic Bantu) of entities in gender type classes. I contend then that the categorydetermining material of the word is affix-final as in (17a) for verbs and (17b) for nouns.
Within verbs, the verb stem is the locus of derivational affixation, as such verbs will permit
such affixation. Since in (17b) there is a noun stem and not a verb stem, derivational
affixation is not possible within root nouns. Similarly for nominal modifiers, the quantifier
stem (Qstem) though combining with the subject prefix to form a quantifier word (I7c). it
can not permit derivational suffixation.
(17)
a.
V
Subject
V
stem
b.
N
Noun cla.ss
M
'
TOOl
Q
c.
Subject Prefix
\
Item
105
By word structure considerations, verbs differ from nouns and nominal modifiers by the
former's (verb's) affixational possibilities which is revealed by their sub-lexical properties.
TTius derivational and inflectional affixation is a property of verbs in Bantu.
5.0
Conclusion
In this chapter we have looked at root nouns and pure verbs in terms of their
modificational properties and morphological marking. We have seen that one of the most
salient features of nominal modifiers is that their prefixal morphology is identical to that of
the subject prefix which occurs on verbs. The noun, by way of contrast, bears the noun
class affix which is morphologically distinguished from the marking on modifiers. We
have proposed that affixes are elements which are utilizable only in word formation and as
such are not phrasally derived. This chapter has also evaluated some of the claims made
about Bantu affixes using Grkuyu facts. In this regard we have looked at work by Myers
(1987), and Carstens (1991). Concerning the contrast between pure nouns and their
modifiers on the one hand and verbs on the other, we have attributed their contrasting
behavior to their sub-lexical structure. Specifically, elements with verb stems should be
able to be inflected and extended, while elements with other types of stems such as those
present in nouns and nominal modifiers do not.
106
CHAPTER 4
MIXED CATEGORIES
1.0
Introduction
In chapter 2, I have shown that infinitive/gerunds display categorial ambiguity
when their behavioral and distributional properties are examined. I attributed the ambiguity
to their word structure. In the present chapter I study agentive nominalized words and their
structural projections. In doing this, it is apparent that agentive nominals are clear ca.ses of
mixed category items which have the properties associated with elements of category N
and of category V. The categorial behavior and syntactic distribution of expressions such
as (1) show that they exhibit mixed category status which literally displays quite
systematically a combination of the properties displayed by the gerund [+NPobj] and the
gerund [-NPobjl studied in chapter 2.
(1)
mu-thiTnj-i
mburi
I slaughter-Nzer 1 Ogoat
"slaughterer of goats"
In order to explain its behavior I will begin with the word structure of nominalized words
whose morphology is [mu-...-i]. I claim that the mixed category behavior of [mu-...-i]
words stems from their word structure itself and the lexical information that it bears. I then
proceed to study the syntax of phrases invoving [mu-...-i] words. In terms of the phrasal
category that [mu-...-i] words project, the Extended Head projection of Bresnan (1996),
helps us explain [mu-...-i]'s syntactic behavior and distribution. The leading question in this
chapter then, is how to capture the mixed category status of [mu-...-i] both within words
and within the phrase. I will examine certain facts about [mu-...-i] words and then proceed
to look into issues on word and phrase structure.
107
2.0
The facts
In considering the facts, I will begin by providing evidence of mixed category
behavior of {mu-...-i] words. I begin with nominal properties and then proceed to verbal
properties.
2.1
Nominal properties
In this section I will provide evidence that [mu-...-i] words behave like nouns. To do
this I present information on positional distribution and the modificational properties.
Beginning with positional distribution. [mu-...-i] phra.ses can be extracted using m' -clefting
and relativization. In chapter 2. I observed that noun phrase complements of verb phrases
can be extracted by clefting and relativization. while verb phra.se complements (of VPs) can
not. Extraction of [mu-...-i] object phrase muthfTnji mhiiri 'goat slaughterer'in (2) by h T
-clefting is exemplified by (3) and by relativization as in (4).
(2)
Wanja
e-end-et-e
mu-thiTnj-i
Wanja
3sgS-like-PERF-fv
Islaughter-Nzer
"Wanja likes a slaughterer of goats'
mburi
lOgoat
(3)
Nr
mu-thiTnj-i
mburi
Wanja e-end-et-e
FP
1 slaughter-Nzer lOgoat Wanja 3sgS-like-PERF-fv
"It is a slaughterer of goats that Wanja likes"
(4)
mu-thiTnj-i
mburi
una
Wanja e-end-et-e
1 slaughter-Nzer lOgoat
IRel Wanja 3sgS-like-PERF-fv
"The slaughterer of goats that Wanja likes'
TTie.se extraction tests suggest that [mu-...-i] words are nominal. This is not the ca.se
with other types of phrasal complements of the same verb encia 'want' in (2'). Extraction of
verb phrases in GTkuyu can not be done by m' -clefting (3') nor relativization of sentential
complements (4').
(2')
Wanja
a-re-end-a
u-thnnj-e
Wanja
3sgS-Pres-want-fv 2.sgS-slaughter-Nzer
"Wanja wants you to slaughter goats'
mburi
lOgoat
108
(3')
*Nr u-thnnj-e
mburi
Wanja a-re-end-a
FP
2sgS-slaughter-Nzer
lOgoat Wanja 3sgS-Pres-wani-fv
"Ii is for you to slaughter goats that Wanja wants'
(4')
*ij-thiTnj-e
mburi
una
Wanja a-re-end-a
2sgS-slaughter-Nzer
lOgoat
IRel Wanja 3sgS-Pres-want-fv
'TTiat she slaughters goats the way in which Wanja wants'
Similarly, with other sentential complements of the verb enda 'like' in (2"). neither the
extraction by m -clefting (3") nor relativization (4") is possible.
(2")
Wanja
a-re-end-a
au" a-ga-thir
Wanja
3sgS-Pres-want-fv that 3sgS-Fui-go
lit. "Wanja wants that she will go abroad'
'Wanja wants (now) to go abroad (in the future)
riiraya
11 abroad
(3")
*nf
atf
a-ga-thir
ruraya
Wanja a-re-end-a
FP
that 3sgS-Fut-go I 1 abroad Wanja 3sgS-Pres-want-fv
'Wanja wants (now) to go abroad (in the future)
(4")
*air a-ga-thir
ruraya
una
Wanja
that 3sgS-Fut-go 11 abroad IRel Wanja
'Wanja wants (now) to go abroad (in the future)
a-re-end-a
3sgS-Pres-want-fv
These extraction tests further attest to the claim that [mu-...-i] expression behave
like nouns.
Another property which reveals the nature of [mij-...-i] words is that they alternate
with incorporated pronouns within the verb as shown by (6) for (5).
(5)
Wanja
nr-a-ra-on-a
mQ-thiTnj-i
Wanja
FP-3sgS-Pres-want-fv
Islaughter-Nzer
"Wanja is seeing the one who slaughters goats'
(6)
Wanja
nr-a-ra-mu-on-et-e
Wanja
FP-3sgS-Pres-10-like-PERF-fv
"Wanja had seen him'
mburi
lOgoat
109
Moreover, with object prefixation multiple word orders are possible (Mchombo (1995). I
expect that with [mu-...-i] expressions allowing object prefixation. free word order would
be possible as in (7) and bad when the prefix is absent (7').
(7)
(7')
a.
mu-thiTnj-i
mburi.
b.
Wanja.
a.
*mu-thiTnj-i
b.
*Wanja. mu-thunj-i
Wanja nf-a-ra-mu-on-a
mu-thunj-i
mburi.
mburi
nf-a-ra-mu-on-a
Wanja nf-a-ra-on-a
mburi
nF-a-ra-on-a
Thus, with regards to the behavior and distribution of the object prefix, [[mu-.-.-i)
expressions exhibit the distributional properties of elements identified as category N.
Further nominal behavior of [mu-...-i] is demonstrated by the fact that it can be modified
with demonstratives (8a). possessive pronouns (8b). adjectives, quantifiers (8c). adjectives
(8d) and associative phrases (8e).
a.
mu-thunj-i
mburi
1 slaughter-Nzer 1 Ogoat
"this goat-slaughterer'
uyu
1 Dem
b.
mu-thi7nj-i
mburi
1 slaughter-Nzer 1 Ogoat
'our goat-slaughterer'
witu
Ipl-Pron
c.
a-thiTnj-i
mburi
2slaughter-Nzer 1 Ogoat
"all goat-slaughterers'
othe
2all
d.
a-thiTnj-i
mburi
2slaughter-Nzer 1 Ogoat
"big goat-slaughterers'
anene
2bi2
e.
wa
mu-thiTnj-i
mburi
1 Assoc
1 slaughter-Nzer 1 Ogoat
'2oat slaughterer of the village
(8)
7village
In addition to the nominal properties in this section. [mu-...-i] words have verbal propenies
as well.
110
2.2
Verbal properties
In chapter 2. it was indicated that verbs can be modified not just by adverbs but also
by emphatic interjections. Despite their nominal properties. [mu-...-i) phrases participate in
verb-like distribution. They can be modified with adverbs (9) and by emphatic interjections
(10).
(9)
a-thiTnj-i
mburi
uuru
2slaughter-Nzer lOgoat
badly
"those who slaughter goats badly'
(10)
mu-kir-i
ki
2-love-Nzer
Intj
"one who is very silent
It' agent nouns are verbal, those with transitive stems may be expected to permit object
prefixation and they do as indicated by (11 b) from (1 la).
(11)
a.
mu-thiTnj-i
mburi
I slaughter-Nzer
1 Ogoat
"slaughterer of goals'
b.
mu-i-thfinj-i
1 -1OO-slaughter-Nzer
"slaughterer of them'
That the object prefix is really an incorporated pronoun is shown by the fact that the
cooccurrence of the object noun phrase with object prefixation is disallowed as shown by
(12). This is a condition that obtains in Grkijyij where the object prefix and the object noun
phra.se are in complementary distribution.
(12)
*mQ-i-thiTnj-i
mburi
1-IOO-slaughter-Nzer
1 Ogoat
"slaughterer of (them) goats'
A striking a.spect of deverbal nouns with incorporated pronominal objects is that they can
only be modified as verbs (13) and not as nouns (14)
Ill
(13)
(14)
a.
mu-i-thiinj-i
biu
I -1 OO-slaughter-Nzer completely
"slaughterer of them completely"
b.
mu-i-thiTnj-i
tu
Islaughter-Nzer Intj
"a real slaughterer of them"
a
uyu
*mu-i-thi7nj-i
I bem
1 -1 OO-slaughter-Nzer
"this slaughterer of them"
b.
*mu-i-thiTr)j-i
waku
1 slaughter-Nzer 15-2sgPron
•your slaughterer of them'
Despite the verbal behavior in (13), agent nouns do not mark tense/aspectymodality.
This property is unlike that displayed by verbal infinitive/gerunds in chapter 2 which do
permit very restricted tense and aspect marking. Thus "future" tense marking in agent
nouns induces the ungrammaticality in (15) as does "habitual" aspect marking in (16).
(15)
*mu -ga-thiTnj-i
mburi
1-Fut-slaughter-Nzer
lOgoat
"slaughterer of goats (in the future)"
(16)
*mu-thiTnj -ag-i
mburi
1-slaughter-Hab-Nzer
lOgoat
"slaughterer of goats (habitual)"
Likewise marking of negation is not permitted as seen in (17).
(17)
^mu-ti-thnnj-i
mburi
1-Neg-slaughter-Nzer
lOgoat
'one who does not slaughter goats"
In Grkuyu. reduplication is a property of verbs and as such the reduplicated forms in (18)
are expected to be modifiable with adverbs as in (19) and not with nominal elements as is
shown to be possible by (19'). Reduplication of verb-stems induces the meaning "to do a
little (more)" to the base meaning of the verb.
112
(18)
mu-thiTnjathiTnj-i
mburi
1 siaughter/slaughter-Nzer 1 Ogoat
"one who slaughterers goats a little"
(19)
mu-thunjathiTnj-i
mburi
kahora
I slaughter/slaughter-Nzer
1 Ogoat slowly
"one who slaughterers goats slowly a little'
(19')
*mu-thiTnjathiTnj-i
mburi
1 slaughter/slaughter-Nzer 1 Ogoat
'this slaughterer of goats a little'
uyii
IDem
Like verbs. [mu-...-i] phrases permit affixation of derivational morphology applicativization (20). causativization (21), and -an- affixation (22).
(20)
mu-thiTnj-fr-i
andu
mburi
Islaughter-A-Nzer 2person
lOgoat
"one who slaughteres goats for people'
(21)
mu-thiTnj-ith-i-a
andu
mburi
Islaughter-C-C-Nzer
2person
1 Ogoat
'this slaughterer of goats for people'
(22)
mu-thiTnj-an-i
Islaughter-R-Nzer
'one who does surgery"
As expected. (20). (21). and (22) can be modified with adverbs as seen in (20'). (21'). and
(22').
(20')
mu-thiTnj-ir-i
andu
mburi
wega
1 slaughter-A-Nzer 2person
1 Ogoat
well
'one who slaughters goats for people well"
(21')
mu-thi7nj-ith-i-a
andu
mburi
wega
Islaughter-C-Nzer 2person
1 Ogoat
well
"one who causes people to slaughterer goats well"
(22')
mu-thunj-an-i
wega
Islaughter-R-Nzer well
one who does surgery- well"
Suprisingly however, [mu-...-i] phrases with derivational extension can be modified with
nominal modifiers as (20"). (21"). (22") indicate.
113
(20")
mu-thnnj-ir-i
andu
mburi
Islaughter-A-Nzer 2person
lOgoat
"this slaughterer of goats for people"
(21")
mu-thiTnj-ith-i-a
andu
mburi
uyu
Isiaughter-C-Nzer 2person
lOgoat
IDem
"this one who causes people to slaughterer goats"
(22")
mu-thflnj-an-i
Islaughter-R-Nzer
"this slaughterer'
2.3
uyu
IDem
uyii
IDem
Summary
Properties of [mu-.-.-i] words are provided in (23). These properties indicate that
[mu-...-i] words exhibit mixed category behavior, the virtual combination of the behavior
of the gerund [-NPobjJ kind and gerund [+NPobj] types discussed in chapter 2.
(23)
a.
permit extraction as thematic objects
b.
allow object prefixation
c.
do not permit free word order without object prefix
d.
allow relativization
e.
can be m -clefted
f.
allow use of associative phrases
a
c
can not be formed with auxiliaries
h.
do not permit negation
i.
no tense/aspect/aspect marking is permitted
J-
take verbal modification (adverbs, intensifiers. emphatic interjections)
J'-
take nominal modification (adjectives, determiners, possesive pronouns etc
1 14
3.0
Proposal
In this section. I address issues regarding [mu-...-i] word structure with a view to
proposing how the mixed properties exhibited by [mu-...-i] words can be represented so
that the properties in (23) fall out.
3.1
Word structure
How can the mixed properties of lnni-...-i] be captured lexically?
It is theoretically possible that the categories associated with [mu-...-i] phrases, or
their morphologically interesting first words, are simply idiosyncratic. That is. it might be
the case that they are neither nominal nor verbal, but simply something else. This
theoretical possibility fails to provide any insight into the distribution of nominal and verbal
properties displayed by [mu-...-i] phrases. So the challenge is to account for the mixture of
properties that are observed. There are various proposals in the literature; one way
suggested by Chomsky (1970) is to have complex categories so we could have categories
[+N. -V]. [-N. +V], [+N.+V]. and so on. If behavior is characterized by featuredescriptions. then we can get mixed behavior. Lapointe (1990) exploits this idea and
suggests dual category phrases. Another way is through lexical/phrasal split, with verbal
properties or nominal properties characterized in a way that is sensitive to the lexical versus
phrasal distinction. A third alternative (which I argue for) is through word formation
which involves the nominalization of verb stems permitting the deverbal noun to inherit
(partially) properties of V.
The first step towards a partial solution to this puzzle is to sort out what word
structure is like. Specifically, the identity of the category determining morpheme. If we
assume that it is the noun class marker, word structure has to be head-initial. If we decide it
is the suffix as in (24). then the noun structure would have to be head-final. The direction
one takes has important consequences on the lexical or phrasal status of the noun class
115
marker. In (24), the verb stem (Vstem) combines with a nominalizer (Nzer) to form a
noun stem (Nstem) and the noun class (Ncl) in turn combines with a noun stem to form a
noun (N). By this proposal, [mu-...-i] words can have derivational extensions because
nominalization makes reference to the extended stem (Vstem).
(24)
N
-Stem
Their mixed category behavior of [mu-...-i] words is attributable to their verbal and
nominal composition. This is explained in the next section.
What is the status of the category determining affix?
That the category changing affix is suffixal as in (24) is deducible from (25) where
the prefix can be held constant and one still gets different semantic categories by varying
the suffixal material . Evidence that -i is a nominalizer is revealed by the fact that it is the
suffixal material that changes as various types of nominalizations are constructed.
Furthermore the vowel is -i is not a possible verb final vowel in Grkuyu. (25) provides a
sample of Gfkuyu nominalizations.
(25)
a. thilnja
b. thiTnja
c. thfTnja
c. iTra
d. cokia
e. gwata
f thuurd
O ikuruka
c
h. ruga
i. getha
'slaughter*
'slaughter"
'slaughter"
'tell one-self
'return'
'hold'
'hate'
'descend'
'cook'
'harvest'
mu-thnnj-i
mO-thflnj-a
mu-thilnj-e
mu-iTr-o
ma-cok-io
n-gwat-Fro
ru-thuur-o
mu-ikuruk-o
mu-rug-Fre
i-geth-a
'slaughterer'
'slaughterer'
'slaughtered'
'self-deception'
'reply'
'place o f holding'
'hatred'
'descent'
'cooking manner'
'harvesting occasion'
< Agentive >
< Agentive >
<Patientive>
<Reflexive >
< Result >
<Locative >
< State >
<Movement >
< Manner >
<Occasion>
116
Myers assumes that it is the class marker that nominalizes the V by heading a NP in the
syntax. This assumption does not explain the role of the deverbal nominalizing suffixes
seen in Grkuyu.
Sproat (1985) suggests that the noun class combines with lexical categories to form
lexical categories. The noun class marker in (26) is an N which takes a V complement to
form an N.
(26)
Again, just as pointed out above. (26) can not explain some of the final vowels, such as -i.
and -u (discussed further in chapters 5 and 6) that appear on nominalized forms never
occur on verbs. Sproat's proposal can therefore not explain the Grkuyu facts.
3.2
Phrase structure
How can the mixed properties of [mu-...-i] be captured syntactically?
To explain how [mu-...-i] words are able to take verbal and nominal modification
poses immediate problems for the idea that every phrase must have a unique head
(endocentricity). In an endocentric pattern, X® is the c-head of X'. and X' is the c-head of
X". A phrase and its c-head belong to the same category type N. V. P. The principle of
endocentricity guarantees that the c-head of every phrase is a catcgor>' of the same type but
of a lesser level of structural complexity. Given the mixed behavior of [mu-...-i] words,
how can endocentricity be satisfied? The notion of the Extended Head Projection (Bresnan.
1996) defined in (27) addresses precisely this problem.
(27)
X is an extended head of Y if X corresponds to the same f-structure as
Y. X is of the same category type as Y and every node other than Y
that dominates X also dominates Y (Bresnan. 1996).
117
In (28) I is the extended head of VP. According to (27), I is an inflectionally defined lexical
element such as a finite verb which is related to arguments within its phrasal co-head at the
level of f-structure. According to Bresnan (1996) the notion of extended heads is possible
only when the co-heads are of the same category type. In Bresnan (1995). I and VP belong
to the same class, following Grimshaw (1991).
(28)
IP
VP = Y
Verb = X
A logical solution to the mixed behavior is to have the head of [mu-...-i] words be N which
projects an NP and also serves as the extended head of VP as provided in (29). This is not
strange because [mu-...-i] bears a large number of verbal qualities.
(29)
DP
[mu-..V..-i]
What kinds of conditions permit this to obtain? One of the universal principles of
endocentric structure-function association states that complements of functional categories
are f-structure co-heads (Bresnan, 1982) as is illustrated by (30). The
is both the
c-structure head of F and an f-structure co-head. XP is an f-structure co-head but not a
118
c-head of F.
(30) captures the idea that the relations of
categories to their complements is not that of
predicator to argument.
The situation in (29) is contrary to (30) since N is a lexical category' Y® and not an
pO as required by (30). The question then is when does (31) obtain where the relation
between a lexical category
and an XP is that of co-heads where the Y is mapped into
the same f-structure as the XP instead of the expected relationship between lexical heads
and their lexical sisters. While (30) applies to the usual notion of head, the extended head
definition permits (31) so long as Y^ and XP are of the same category.
(31)
Y'
The structure of the noun phrase headed by [mu-...-i] would be as in (32).
(32)
DP
[mu-..V..-i]
(TOBJ)=i
DP
119
In (32) concordiai elements occur within the DP while noun class marked elements
and lexical categories (N and V) occur within the extended head projection. The mixed
category properties of [mu-...-i] words splits the structure of the noun phrase into two
domains; the concordiai one and the lexical one as illustrated by (33b) for (33a).
(33) a. a-thiTnji
mburi aya
2S-slaughterer
lOgoat 2Dem
'these goat-slaughterers'
b.
DP
concordiai elements
lexical elements
T=i
T=i
VP
a-thTT nji
slaughterers
aya
these
(TOBJ) =i
DP
A
mburi
goats
Given (32). the mixed properties of [mu-...-i] words listed in (23) can now be
explained. Like noun phrases, [mu-...-i] phrases can be extracted as thematic object (23a).
they undergo extraction by relativization (23d) and nf-clefting (23e). [mu-...-i] phrases also
permit the use of associative phrases (23f) which are a part of the grammar of noun
phrases in Bantu. They also require that the root be semantically contentful (23g), thus
[mii-...-i] words can not be formed with auxiliaries. Also like elements which are nominal,
they do not permit marking of negation (23h). tense and aspect (23i). Also as indicated in
(23j') [mu-...-i] words are modified with typical nominal modifiers (demonstratives.
120
pronuons, adjectives). Noun phrase modification takes place in the concordial domain
hypothesized in (33b).
The verbal behavior of [mQ-...-i] words is explained by the fact that the verb stem
in (24) allows for the presence of V properties and hence the verbal projection VP in (33b).
Object prefixation which is a characteristic verbal behavior is permitted (23b). Verb
phrases can be fronted without object prefixation in the matrix verb (23c). and [mu-...-i]
words can be modified with adverbs, intensifiers and emphatic interjections (23j) just like
typical verb phra.ses. Verb phra.se modification takes place in the lexical domain propo.sed
in (33b).
3.3 Base versus structural generation
The idea of extended heads assummed here looks suspiciously like head to head
movement. A head to head movement analysis would propose that the verb stem is moved
from its V position as head of VP to the N position where it collects nominal features and
heads the NP. This kind of proposal derives words by syntactic derivation.
Head
movement violates lexical integrity which requires that the processes attendant to word
formation are different from procsses involving the syntactic formation of phrases (B&M.
1995).
It is important to realize however that this is a simulation rather than actual head
movement. There are a number of reasons why movement is undesirable. First, structural
generation approaches encounter serious empirical and theoretical problem.s. second
nothing prevents both positions from being lexically filled as is evidenced by a variety of
typologically unrelated languages.
3.3.1
Against head movement
Structural base generation is preferable to derivational accounts involving
movement as is shown by Bresnan (1994a) study of category mismatches. There are real
paradoxes if movement is involved. In typical movement configurations, there is no
121
categor>' type shift between the source and the target of movement, in (34) from (Bresnan
1994a), there is a mismatch in category type between the apparently moved (by
topicalization) ihat complement (34a) and the position from which they are supposed to be
moved (34b). The target position is of category CP yet the source category is an NP.
(34)
a.
That he was tired he didnt think about
b.
*he did not think about that he was tired
Passivization and raising shows that the supposedly moved category (the that clause) in
(35a) can not occur in the place it is supposed to have been moved from (35b)
(35)
a.
that languages are learnahle is captured by this theory
b.
*this theory captures that languages are learnable
According to Bresnan (1994a), for expressions such as (35), the that complement is basegenerated in topicalized position as a CP adjoined to the IP in c-structure. The fact that
auxiliary inversion with the CP is not possible as indicated by (36) is evidence for a subject
external position of the CP in English (Koster 1978), that is it can not be in the SPEC of IP
position.
(36)
*Is that languages are learnable captured by this theory
Since there is a constraint that the realization of subject and object functions is nominal
(Bresnan (1994b). the CP can not occupy the subject NP position as a sister of I' in (37).
122
(37)
that lanauaaes are leamable
I(Aux)
captured
theory
The general principles of structure-function correspondence define the position of the
adjoined CP as that of topic of focus function and indicates that the CP corresponds to a
single functional-structure nucleus (38). This as Bresnan (1995 lecture notes) points out
corresponds to Grimshaw's (1991) notion of "extended projection'.
(38)
IP
^"that..."
SUBJ
I(Aux)
IS
According to Bresnan (1995. Lecture notes) a gap is represented by the omission of the
subject NP from its usual c-structure position as a sister to I". In this framework, the
extraction configuration which involves a displaced constituent with a gap arises without
movement from general principles.
Other evidence against movement comes from Bresnan and Kanerva (1989) study
of locative inversion in Chichewa. In locative inversion structures, the inverted subject is
123
the thematic subject, the syntactic object, and the presentational focus in discourse.
Bresnan and Kanerva (1989) propose that locative inversion properties involving
nonextractibility, agreement, word order, lexical restrictions on invertibility. and
nonpassivizability are better explained by parallel levels of grammatical structure rather
than by derivation through mechanisms such as movement.
Further evidence of
the undesirability of movement based approaches comes from non-configurational
languages such as Warlpiri (Bresnan 1995a). In these languages, there is no evidence for
groups of words that are contiguous (phrases), so as to allow substitutability and the
interpolation of parentheticals.
In the extended head structure (29). both the VP head position and the NP position
can be tilled. In English, the auxiliary and the main verb can both be mapped onto the same
f-s. In Grimshaw's (1991) theory of extended heads. I is an extended head of VP and (39)
indicates that both positions could be filled. This rules out the necessity of viewing (29)
above as instantiating head movement.
Aux
V
According to Nordlinger (1995). in Wambaya. a Northern Australian language,
tense and imperative mood is marked simultaneously on two different heads in one clau.se
as in (40). This challenges movement-based frameworks that each piece of functional
information present in a clause corresponds to only one node in the constituent structure, a
situation clearly in conflict with (41). These kinds of examples (concludes Nordlinger)
124
show that there is no one to one correspondence between morphosyntactic categories and
syntactic positions.
(40)
Dagumalba
gun -u
hit -Fut
3:sg:masc:A -Fut
"He will hit him/her/it'
(41)
S
T
^
AUX
(TTM)=FUT
dagumaj-ba
(TTM)=FUT
g^^.u
Another example where both the extended head and the element for which it is an
extended head are lexically filled is Finnish. The study of Finnish split inflection by Nino
(1996) shows that the negative auxiliary e- *NEG' and the participial main verb puhu"speak" in (42a) and (42b) are morphologically marked for number agreement with the
subject. This indicates that both I and VP are lexically filled with the identical grammatical
elements which are mapped onto the same f-structure component. In (43) there is no
motivation for head movement since the relevant grammatical features are already present
in both I and V. Nino (1996) concludes that words apply to the syntax fully formed and
not as sublexical elements, which explains why I and V bear NEG features and obviates
the necessity of head movement from V to I in pursuit of NEG features. The Split
inflection supports lexical views of grammatical architectures based on the lexical integrity
principle.
(42)
a.
e -n
puhu -nut
NEG-ISG speak-PPART.SG
"I did not speak"
125
b.
(43)
e -mme
puhu-nee -t
NEC-1 PL
speak-PPART-^
"We did not speak'
r
IP
VP
ISGl
en
3.4
V
fSG
puhunut
i
PRED
i
SUBJ
'speak'
PRED 'pro'
PERS 1
•-NUM SG
TENSE
PAST
Summary
To capture the mixed behavior of (mu-...-i] at the phrase level. 1 have utilized the
notion of the Extended Head projection of Bresnan (1996), allowing [mu-...-i) to be a
special type of N which heads an NP and is also an extended head for VP. A nominalized
verb can be u.sed as an extended head only if the ncminalization makes reference to the
verb stem and is therefore able to bear derivational affixation. This explains why the
analysis for [mu-...-i] words is different from gerund [-NP] (chapter 2) where it is the
Verb root which is nominalized. Unlike verb stems, verb roots are not extended. The [mu...-i] word is a special type of N because its ability to be used as an extended head is relative
to the depth of nominalization. The richer the verb is in terms of its affixation potential, the
greater its likelihood to be u.sed as an extended head. [mu-...-i] words are very rich
morphologically, they can mark object prefixes, reflexive prefix, be extended with multiple
derivational affixes. When such morphologically rich verb forms are nominalized as is the
case with [mu-...-i] it is not suprising that they would portray the mixed V nd N behavior.
The information contained in such rich stems is very difficult to suppress upon
nominalization. hence the mixed behavior.
4.0
Conclusion
In this chapter I have studied the categorial behavior and syntactic distribution of
[mQ-...-i] phrases. I have shown that they exhibit mixed category status which
systematically picks out a combination of the properties displayed by the gerund [+NPobj]
and the gerund [-NPobj] studied in chapter 2. I have proposed the word structure of [mu...-i] which indicates that derivational affixation precedes nominalization. The claimed here
is that the mixed category behavior of [mu-...-i] words stems from their word structure
itself and the fact that [mtj-...-i] words act as extended heads. By introducing the notion of
extended heads of Bresnan (1996), I have not only been able to capture the mixed
properties of the [mu-...-i] word and its projection but also have discovered that the [mu-...i] phrase induces two domains; the concordial one where all nominal modifiers are located,
and the lexical domain where N and V projections are localized. These domains result from
the mixed category status of [mu-...-i] words.
127
CHAPTER 5
THE LEXICON/SYNTAX DIVIDE
1.0
Introduction
The Grkuyu agentive nominalization bearing [mu-...-i] morphology which has been
studied in the preceding chapter is further complicated by the existence of its double
[mu-...-al.
Both [mu-...-a] and [mu-...-i] express similar meanings but behave quite
distinctly from each other. It is shown here that these two types of agentive deverbal
nominals, express similar semantics, but differ radically in terms of their categorial
behavior and syntactic distribution. We present evidence to show that agentive deverbal
words in GrkuyiJ are products of syntactic formation just as argued by Kinyalolo (1991)
for Kilega. However, Grkuyu displays a number of characteristics that are associated with
products of lexical rules of word formation seen in Bresnan &Mchombo (henceforth
B&M) (1995) for Chichewa. We see that one type of Grkuyii agentive nouns marked with
[mu-...-a] morphology present something of a paradox in that they exhibit lexical
properties while still maintaining some phrasal characteristics. We see that the loose
compounding exhibited by [mu-...-a] words immediately raises questions pertaining to the
principle of the lexical integrity of words proposed in various fashions by Chomsky,
(1970); Wasow and Roeper (1972), Lapointe, (1980); B&M, (1995); Sells. (1995), and
others.
This chapter shows that in order to explain the attested facts of agentive nominals,
we have to recognize a three way split in Grkuyu nominalization. One, a lexicalized form;
two. phrases that are in the process of becoming lexicalized; and three, elements that are
phrasal. It is proposed that this split is demonstrative of a lexicalization process involving
phrases. The first kind is frozen and has become a word, while the second type shows that
there exists a constrained and localized phrasal recursivity in the lexicalizing forms, and the
128
third type is open to all kinds of syntactic manipulations. The process of the lexicalization
of phrases is seen to be quasi-innovative and context dependent as as seen in Clark and
Clark (1979) for English. Our analysis points to a middle ground' in the lexical-syntactic
divide stated by the notion of lexical integrity. Grkuyu agent nominals suggests a
permeable notion of lexical integrity in place of the hard and fast separation between the
lexical module and the syntactic one suggested by B&M (1995). The chapter observes that
the violation of the lexical integrity principle by [mu-...-aJ words, is not all over the map
(which would make no sen.se of lexical integrity) but rather highly constrained. Lexical
integrity is thus maintained, contrary to Baker (1988) and Lieber (1992) among others.
2.0
Problems
Grkuyu and Swahili agentive nominalizations have two morphological forms. In
Grkuyu agentive nominalization produces agent nouns, one bearing the [mu-...-a]
morphology exemplified by (1) and the other. [mu-...-i] morphology as shown in (2).
(1 )
mu-end -a
1 love-Nzer
'people lover'
(2)
mu-end -i
andCi
1 love-Nzer 2people
'people lover'
2.1
andu
2people
Problem 1
The first problem is with regards to how the status of [mQ-...-a] words in Grkuyii
can be captured given that they display lexical/phrasal ambiguity.
Specifically, what is the
status of a and 3 in (3)? Do [mu-...-a] elements project XP or an X^?
If compounding
is done under N. then there arises the question whether we have a process involving the
'-^Tabor 11994) addresses issues pertaining to the notion of 'in-betweenness' using diachronic data for synchronic
representation of English.
He notes that it is sometimes helpful to choose a representation that varies
continuously with lime. It is along such lines that the idea of a middle-ground is used in this paper.
129
compounding of phrases within N. This means that the status of P (the complement
argument) as to whether it is sublexical X"' ( part of word formation), lexical
(a
syntactic atom), or an X"^^^ must be resolved.
(3)
a
andu
2.2
Problem 2
The second problem is with regards to why there exists two types of morphological
forms
in GrkijyiJ. [mQ-...-al and [mu-...-i]. The issue here is how we can distinguish
between the different morphological types.
2.3
Problem 3
Accompanying the problems above is the tenability of the lexical integrity principle
itself. The lexical integrity principle stipulates that the internal structure of words is not
transparent to syntactic processes. According to B&M 1995, this restriction follows from
the fact that words are built out of different structural elements and by different principles
of composition than syntactic phrases. For our purposes, this implies that phrases cannot
be syntactically attached to words. Yet such an attachment is precisely what we find with
the formation of [mQ-...-a] expressions, apparently calling to question the lexicon/syntax
divide.
In what follows, we substantiate the problems we have stated for Grkuyu.
begin with a detailed look at the facts.
We
130
3.0
The facts
As stated above. (I) and (2) reveal both nominal and verbal properties as well as
lexical and phrasal behavior. That they are verbal is clearly indicated by their complement
taking property. That they are nominal is quite evident in the fact that (1) and (2) are
expressions naming entities. We will now provide evidence to make these properties
explicit. In section 3.1 we study the verbal properties of agent nouns, then in section 3.2
we look at their modificationai properties, after which we will consider their phrasal
properties in section 3.3.
3.1
Verbal Properties
Grkuyu like all other Bantu languages permits the inflection of its stem with
prefixes as well as suffixes. Prefixes include pronominal elements; the subject prefix, the
object prefixas well as the reflexive prefixes. These prefixes can occur in cenain
nominalized forms. Prefixes encoding negation, tense, aspectuality. and modality do not
appear in Grkuyu agentive nominalized forms.
Suffixes which can occur in
nominalizations include the reciprocal -an-, the causative -ith-, and the applicative
Agent nouns appear to be sensitive to stem valencies, to which we now turn.
3.1.1
Stem Valencies
As seen in (4a) intransitive stems as well as triadic stems (4c) cannot participate in
[mu-...-a] formation. Perhaps (4a) (the case of intransitives) fails because [mu-...-a] must
have a complement and (4c) (ditransitive cases) due to some restriction to the number of
complement arguments.
(4)
a.
*mu-nr-a
1 -cry-Nzer
'one who cries'
'•^See B&M (1987) for a discussion on (he properties of subjeci and objcci prefixes.
131
b.
c.
mu-end-a
andu
1 want-Nzer
2people
'one who loves people'
*mu-tum-a
marua thukuru
1 send-Nzer
Sletter 9school
'sender of a letter to a school'
With the [mQ-...-i] type stems, all types of valencies are usable as seen in (4'). There is
absolutely no restriction in satisfying the valencies of stems here.
(4')
a.
mu-nr-i
1 -cr>'-Nzer
'one who cries'
b.
c.
mu-end-i
andu
1 want-Nzer
2people
'one who loves people'
mu-tum-i
marua thukuru
1 send-Nzer
Sletter 9school
'sender of a letter to a school'
In Bantu languages, transitive stems such as (4b) and (4'b). (4'c) would also permit object
preflxation and subject reflexivization where appropriate.
3.1.2
Preflxation
Given (4b) we would expect that (5b) should be able to permit object preflxation as
do all transitive verbs. This is however not the case as indicated in (5) for [mu-.-.-a]
expressions. The behavior seen in (5) indicates that perhaps [mu-...-a] is not a verb but
may be a noun, since it does not permit object preflxation (5a) or allow reflexivization (5b)
of the stem.
(5)
a. *mu-me-end-a
1 -3plO-want-Nzer
'one who likes them'
132
b. *mu-r-end-a
1 -Rfl-wanl-Nzer
'one who likes him/herself
Object prefixation (5'a) and reflexivization (5'b) are perfectly grammatical with the
[mtj-.-.-i) types.
(5')
a. mu-me-end-i
1 -3plO-want-Nzer
'one who wants them'
b. mi3-r-end-i
1 -Rfl-want-Nzer
'one who wants them'
Perhaps the suffixal -a in the [mu-...-al types, is not the usual final vowel (verbalizer) in
Bantu but rather a nominalizing suffix which precedes all affixation to the verbal stem. If
this is the case, we would expect to find that suffixation should be disallowed for
[mQ-...-a] type but allowed for the [mu-...-ij type.
3.1.3
Suffixation
The expectation that suffixation will be disallowed in [mu-...-a] but will be
permitted in [mu-...-i] is correct as seen in (6) for [mu-...-a] and (6') for mu-...-i]. Even
though the [mu-...-a] word has a complement, it does not permit argument absorption by
reciprocalization (6a) nor valency increase (addition of arguments) by applicativization
(6b)'-'^ but the [mu-...-i] type does permit both processes as seen in (6').
(6)
a. *mu-end-er-a
andu
mbiya
1-like-A-Nzer
2people lOmoney
'One who likes people for money'
b. *mu-end-an-a
1 -want-Rec-Nzer
'^'Rcc' stands tor reciprocal even tliough the gloss has nothing to do with reciprocity. This is a quite unfamihar
use oC the reciprocal and is discussed in Mugane (1996a).
133
(6')
a. mu-end-er-i
andu
mbiya
l-like-A-Nzer
2people lOmoney
'One who likes people for money'
b. mu-end-an-i
1 -want-Rec-Nzer
'one who loves (some unspecified people)'
Indeed the [mu-...-i] type of phrase has verbal content, the simultaneous affixation of the
applicative and -an provides the expected result. The predicted result is that the valency
reducing -an and the valency increasing applicative cause no surface change to the number
of arguments to the stem becau.se the simuitaneous addition of one argument and the
absorption of one has the effect of cancelling out as seen in (7).
(7)
mu-thnnj-an-fr-i
mburi
l-slaughter-R assoc-A-Nzer lOgoat
'One who associatively slaughters goats on behalf of someone/another'
3.1.4
Modiflcation
In chapter 4, we saw that adverbs and emphatic interjections are used for verb
phrase modification. If [mu-...-a] expressions are verb phrases, they should permit such
modification but that is not the case as seen in (8) and (9).
adverb modification
(8)
*mu-thiTnj-a
mbQri
uuru
1 slaughter-Nzer 1 Ogoat
badly
"one who slaughter goats badly'
emphatic interjection
(9)
*mu-thiTnj-a
mbiiri
tu
1 slaughter-Nzer 1 Ogoat
Intj
"one who really slaughter goats badly"
As discussed in chapter 4. [mu-...-i] phrases can be modified with adverbs (8') and
emphatic interjections (9') just like verb phrases do.
134
adverb modification
(8')
mu-thnnj-i
mburi
uuru
2slaughter-Nzer lOgoat
badly
•those who slaughter goats badly'
emphatic interjection
(9')
mu-thiTnj-i
mburi
tu
Islaughter-Nzer lOgoat
Intj
"one who really slaughter goats badly'
3.1.4
Summary
In this section we have shown that there are systematic differences between the
[mu-.-.-a] type of agent noun from the [mu-...-i] kind. The [mu-...-a] type has no detectable
verbal propenies other than the fact that it is sensitive to the valency of the stem (stem must
be transitive) which allows [mO-.-.-a] elements to be formed with complement arguments
as the second member of the expression. We have seen that because the
[mu-...-a]
types disallow any kind of affixation, they behave like root nouns. Since they are clearly
deverbal by composition, we have proposed that affixation is disallowed because
nominalization (affixing -a ) precedes all affixation. In the [mu-...-i] phrases, where all
verbal behavior is permitted, nominalization (affixing -i) follows all affixation.
The behavior and distribution of [mu-...-a) types is reminiscent of that displayed by
gerund [-NPobj] studied in chapter 2. The word structure (10) explains why we get
nominal modification and no inflectional morphology with [mu-...-a] types. As with the
gerund [-NPobj]. there is no dispute about the base of [mu-...-a]. The verb ba.se combines
with -a to create an Nstem and the Nstem combines with the class marker mH - to form a
noun.
135
(10)
N
stem
With [mQ-...-i] types where nominalization follows all affixation, we have the structure
(10') argued for in chapter 4.
In what follows we proceed to test the quite plausible hypothesis that on the one
hand, [mu-...-a] words are actually of category N (lexical) and on the other hand.
[mu-...-i] phrases are NPs (phrasal elements) hence their participation in phrasal behavior.
In doing this the identity of P in (3) will be determined.
3,2
Nominal Properties
Since agent deverbal nouns name entities, it is expected that they can function as
subjects and objects of sentences, as well as be oblique objects. Examples with agent
phrases as subjects are given in (11). as objects in (12). and as oblique objects in (13).
(11)
a.
mu-end-a
andu
a-ra-rug-a
1 want-Nzer
2peopIe 3sgS-TNS-cook-fv
'one who loves people is cooking food'
irio
lOfood
136
b.
(12)
(13)
mu-end-i
andu
a-ra-nag-a
1 wam-Nzer
2people 3sgS-TNS-cook-fv
'one who loves people is cooking food'
irio
lOfood
a.
Kamau
a-ra-ror-a
mu-end-a
Kamau 3sgS-TNS-look-fv I want-Nzer
"Kamau is looking at the people-lover'
andu
2people
b.
Kamau
a-ra-ror-a
mu-end-i
Kamau 3sgS-TNS-look-fv 1 want-Nzer
"Kamau is looking at the people-lover'
andO
2people
a.
Kamau
a-ra-in-a
na mu-end-a
Kamau 3sgS-TNS-sing-fv with 1 want-Nzer
"Kamau is singing with the people-lover'
andu
2people
b.
Kamau
a-ra-in-a
na
mu-end-i
Kamau 3sgS-TNS-sing-fv with 1 want-Nzer
"Kamau is singing with the people-lover'
andu
2people
Clearly then, both types of deverbal nouns have the distribution of regular noun phrases in
the language. In what follows, we examine the lexical nature of the agent nouns. It is
expected that if these nominalized forms are true compounds, they will behave as words.
Specifically, the head noun and its complement should be inseparable and be modifiable by
adjectives only as a unit. With regards to the separability of the head noun from its
complement, we consider the associative phrase construction and demonstrative placement
facts. In the case of the modificational potential we will consider adjective placement.
3.2.1
The Associative Phrase
As reported in Mugane (forthcoming), associative phrases have various functions
in the grammar of noun phrases. They permit any two nouns or more to be associated with
each other, and the meaning arrived at is determined by the semantic plausibility of the
relation that can hold between the attendant nouns. The associative word takes a DP
complement. If the [mu-...-a] and its complement are products of lexical process of word
137
formation (a word and not a phrase), we would expect that (14a) and (14b) are the only
ways to use the [mu-...-a] phrases to form associative phrases, that is, the head [mu-...-al
and its complement would be inseparable units.
(14)
a.
mu-end -a andu
wa
GFcuka'^
1 love-Nzer 2people 1 Assoc Nairobi
'A Nairobi person who loves people'
b.
mwana
wa
mu-end -a
1 child
1 Assoc 1 love-Nzer
'A child of people-lover'
andu
2people
The attendant structure for (14a) apparently suggests a structure such as (15).
(15)
0
A
muenda andu
people - i
as.soc
wa
of
DP
A
Grcuka
Nairobi
'a Nairobi person who loves people'
If [mQ-...-a] items are words, it should be impossible to take the complement out of them
and form an associative phrase with it creating constructions such as (16). In (16), we see
that we can take the complement of [mu-...-a] and form an associative phrase with it. This
indicates that mu-end-a "lover' takes a phrasal complement.
A possible structure of (16)
is (17) where the N takes a DP complement. The element P is therefore an XP. We will
leave what [mu-...-a] items project marked with an a for the time being.
(16)
mu-end -a andu a Gfcuka
1 love-Nzer 2people 2Assoc Nairobi
'A person who loves people of Nairobi'
"^Gfcuka is the GrkuvO word for the city ot' Nairobi. Its literal translation is "the place where people deboard' as it
IS where people from all rural areas in Kenya end their journey.
138
DP(B)
XP
muenda
lover
iSOC
NP
Xassoc
andu
people
a
of
Nairobi
'lover of Nairobi people'
Assuming the analysis in chapter 4. we would not expect concordial elements within the
[mu-...-a] phrase, if they are NPs. This is because all concordial elements (of which the
associative is one) are outside the NP'"'. The inadmissibility of (18) emanates from the
same reason intransitive verb stems cannot form [mu-...-a] independent words, namely
that [mu-...-a| deverbal nouns must have direct complements. Associative phrases such as
\va andu "he/she associated with people' do not form complements for deverbal nouns.
(18)
*inu-end -a wa andu
1 love-Nzer 1 Assoc 2people
'A lover of people'
With regard to [mu-...-i] phrases, we note that like [mu-...-a], they too can function
as a unit in associative phrase formation.
(14')
a.
mu-end -i andu
wa
GFcuka
1 love-Nzer 2people 1 Assoc
Nairobi
'A Nairobi person who loves people'
'^The reading can be grammatical only in the case where niwenda 'lover' is a proper noun, in which ease he she
is someone belonging to ant/u 'peopic" hence mwenda docs not take the complement andu and in this sense is not
pan Q ( our discussion here.
muenda wa andu
muenda I Assoc 2people
"muenda of people'
139
mwana
wa
mu-end -i andu
1 child
1 Assoc 1 love-Nzer 2people
'A child of people-lover'
b.
The structure for (14'a) is as iri (15') where it is indicated that [mu-...-i] words are of
category N and they take DP complements and project NP which accompanied by the
demonstrative projects a DP. making the formation of an associate phrase possible.
tssoc
Gfcuka
Nairobi
muendi
lover
A
andu
people
'a person of Nairobi who loves people'
As with any DP. the associative can be employed to link up one DP to another as seen in
the structure (17') for (16'). Just like the [mu-...-a] type, we have evidence that the
complement of the deverbal noun has phrasal status.
(16')
mu-end-i andu a Gfcuka
1 love-Nzer 2people 2Assoc Nairobi
'A person who loves people of Nairobi'
140
(17')
T=X
T=i
N
VP
muendi
lover
(./.)
ssoc
andu
people
assoc
DP
A
GTcuka
Nairobi
'lover of Nairobi people'
The next thing then is to provide evidence for the phrasal status of the deverbal noun
muendi iover'. We did show above that [mu-...-i] types permit the whole range of
behavioral properties expected of Bantu verb forms which have phrasal reflexes such as
object prefixation, reflexivization. It is due to the phrasal status [mu-...-i] in constructions
such as (18'a) that we get the structure (18'b) where niii-end-i 'lover' being a DP can be
associated to another DP andU 'people'.
(18')
a.
mu-end-i wa andu
1 love-Nzer 1 Assoc 2people
'A lover of people'
141
b.
assoc
assoc
muendi
lover
andu
people
'a lover of people'
3.2.2
Demonstrative Placement
The phrasal status of [mu-...-a]. that we have proposed above, permits the
modification of the complement as in (19) where the complement to muencla 'lover' is a
DP as shown in (20).
(19)
mu-end-a [andii aya]
1 love-Nzer 2people 2Dem
'A lover of these people'
(20)
muenda
lover
N
aya
these
andu
people
'lover of these people'
The head noun ([mu-...-a]) can be modified with a demonstrative as well. The
demonstrative is in a higher position as shown in (22) for (21).
142
(21)
[mu-end-a andu] uyu
1 love-Nzer 2people 2Dem
'This lover of people'
(22)
I
I
D'
D
DP
^
N
uyu
muenda
lover
'this lover of people'
The implication of the structures we have provided so far is that there is nothing to stop
expressions such as (23) which is ungrammatical. Demonstrative fronting is generally
permitted under focusing in Grkuyu. In (23), the demonstrative cannot be fronted as in
(24) yet the thing the head noun is compounded with is a phrase. This failure is quite
mysterious and we hypothesize that the identity of a in (22) and elsewhere is likely to be
N. which prevents demonstrative fronting within the N, even though the complement is an
Xmax
Suffice it to say that focusing is a phrasal property and not a lexical one. The
relation between the head noun and the complement is very tight, such that no manipulation
is allowed in that area.
(23)
*mu-end-a aya andu
1 love-Nzer 2Dem 2people
'A lover of these people'
143
(24)
DP
N
D'
muenda
lover
focus
aya
these
NP
N
andu
people
'lover of these people'
Not mysterious, though, is the unacceptability of (25) in which case the demonstrative
modifying the deverbal noun cannot occur inside of the NP because it occupies a higher
structural position as shown by (20) above.
(25)
*mu-end-a
uyu andu
1 love-Nzer 2Dem 2people
Hiis lover of people'
By the syntactic analysis we have proposed, the demonstrative modifying the complement
cannot be fronted outside of the [mu-...-a] expression as given in (26) but the
demonstrative modifying the head can be fronted under focusing as seen in (27)
(26)
*aya mu-end-a andu
2Dem I love-Nzer 2people
'A lover of these people'
(27)
uyu [mu-end-a andu]
2Dem I love-Nzer 2people
This lover of people'
Also expected is the linear order in which the respective demonstratives must occur should
they occur together. As seen in (28), contrasted with (29), the demonstrative modifying
the complement must precede that which modifies the head. This is the expected result
144
since the iiyu 'this one' referring to the head is higher structurally than aya 'these ones'
referring to the complement as can be deduced from (22).
(28)
mu-end-a
andu
aya, uyu
1 love-Nzer 2people 2Dem. 1 Dem
'this lover of these people'
(29)
*mij-end-a
andu
uyu aya,
1 love-Nzer 2people 1 Dem. 2Dem
'this lover of these people'
With regards to the [mu-...-i] phrases, the syntactic analysis we have proposed for them
allows for the modification of the complement just as with [mu-...-a] types. This is shown
in (19') where the complement argument andu "people" a DP is sort of glommed to
miiendi 'lover' which according to our proposal above is a NP as indicated in (20').
mu-end-i [andu aya]
1 love-Nzer 2people 2Dem
'A lover of these people'
(19')
(20')
muendi
lover
/
(..'.)
andu
people
'lover of these people'
145
[mu-...-i] phrases behave in a manner similar to [mu-...-a] ones with respect to modifying
the deverbal noun with the demonstrative higher in the structure as exemplified by (21').
[mu-end-i
andu]
uyu
1 love-Nzer 2people 2Dem
'This lover of people'
(21')
With [mu-...-i] phrases however, expressions such as (23') where the demonstrative is
fronted under focusing, are permitted, in contrast to (23). In (23'). the demonstrative can
be fronted as expected because the complement within which fronting happens is a DP and
the [mu-...-il phrase is an NP. Thus what fails mysteriously with [mu-...-a] expressions
works with Imu-...-i] ones as seen in (23') and its structure in (24').
mu-end-i aya andu
1 love-Nzer 2Dem 2people
'A lover of these people'
(23')
(24')
NP
muendi
lover
[)p
D'
D
NP
aya
these
N
andu
people
'lover of these people'
The unacceptability of (25') follows from the same reason given for (25) where the
demonstrative modifying the deverbal noun is not expected inside of the NP because it
occupies a higher structural position.
146
(25')
*mu-end-i uyu andu
1 iove-Nzer 2Dem 2people
This lover of people'
The expressions in (26') and (27') follow for [mu—i] phrases for the same reasons given
for [mu-...-a]types.
(26')
*aya mu-end-i
andu
2Dem 1 love-Nzer 2people
'A lover of these people'
(27')
uyu [mii-end-i
andu]
2Dem 1 love-Nzer 2people
'This lover of people'
As expected from the structures we have proposed, the linear order in which the respective
demonstratives must occur, should they occur together, is the same for [mii-...-i] phrases
as is the case for [mu-...-a] ones.
As suggested for the [mQ-...-a] cases the relative
structural positions for the demonstrative in the [mO-.-.-i] cases explains the data in (28')
contrasted with (29').
(28')
mu-end-i andu
aya, uyu
1 love-Nzer 2people 2Dem, 1 Dem
'this lover of these people'
(29')
*mu-end-i andiJ
uyu
aya,
I love-Nzer 2people I Dem, 2Dem
'this lover of these people'
We now turn to the modification potential of [mu-...-a] and [mu-...-i].
3.2.3
Adjective Placement
In terms of linear order considerations, adjectives in Grkuyu occur to the right of
demonstratives. In terms of our analysis that the complement of [mQ-...-a] is a phrase, we
expect that the complement may be modified as shown in (30) and the whole expression as
in (31). Following the structure we have assumed for the noun phrase, this linear order
147
position of adjectives in noun phrases predicts that we cannot modify the phrase as in (32)
where the adjectival phrase occurs within NP. because APs are higher in the structure. We
also predict that when we simultaneously modify both the complement and the head of the
expression, the adjective modifying the complement must linearly precede that modifying
the entire expression as seen in (33).
(30)
mu-end-a
andu
a-nyinyi
1 love-Nzer 2people 2young
'lover of young people'
(31)
mu-end-a
andu
mu-nyinyi
1 love-Nzer 2people 1 young
'a young lover of people'
(32)
*mi3-end-a mu-nyinyi andu
a-nyinyi
I love-Nzer I young
2people 2young
'a young lover of young people'
(33)
mu-end-a
andu
a-nyinyi,
1 love-Nzer 1 young
2young
'a young lover of young people'
mu-nyinyi
2people
With [mu-...-i] phrases, the facts (given in (30') to (33') are identical to those presented for
the [mu-...-a] ca.ses and for the same reasons.
(30')
mu-end-i
andu
a-nyinyi
1 love-Nzer 2people 2young
'lover of young people'
(31')
mu-end-i
andu
mu-nyinyi
1 love-Nzer 2people 1 young
'a young lover of people'
(32')
a-nyinyi
*mu-end-i mu-nyinyi andu
1 love-Nzer 1 young
2people 2young
'a young lover of young people'
(33')
mu-end-i
andu
a-nyinyi,
1 love-Nzer 2people 2young
'a young lover of young people'
mu-nyinyi
1 young
148
3.2.4
Summary
In this section we have seen that agent nouns in Grkuyu are of two kinds; the
[mu-...-a] type and the [mu-...-i] ones. We have seen that [mu-...-i] types are phrasal and
that is why we have all the modificational posibilities that have been sketched above. These
modificational possibilities have motivated the structures we have proposed for [mu-...-a]
words (34) and [mLi-...-i] ones (34').
(34)
N
DP
A
[mu-..V..-a]
(34')
[mu-..V..-i)
We have seen that modificational possibilities are sensitive to the relative structural position
of modifiers. We have also suggested that [mu-...-a] deverbal nouns must have direct
complements"^. This requirement eliminates the possibility of [mu-...-a] formation with
intransitive stems. For the surprising case presented in (23) we will provide evidence in
support of a process involving the lexicalization of phrases in section 7.0. I now proceed to
look for information regarding the identity of a. Employing lexical integrity tests will help
us to do this in addition to making explicit the lexical/syntactic nature of Grkuyu agent
noun expressions.
"^Notc thai ihis behavior o f lmu-...-a| types is like that o f verbal inllnitives discussed in chapter 2. Both
and gerund [+NPobjl
similar suftlxal morphology.
149
3.3
Phrasal Properties
In employing lexical integrity tests, we will assume as do B&M (1995) that
because words are products of lexical formation, they are anaphoric islands and are neither
phrasally recursive nor conjoinable. If a is an X® element, then we expect that it will
exhibit these properties.
3.3.1
Phrasal Recursivity
With respect to the complement noun, both types of deverbal nouns show
productive recursivity which is only characteristic of phrasal constituents as those in (35)
and (35'). (35) shows that the complement of the head deverbal noun can be modified quite
liberally, suggesting that what we have with [mQ-...-a] words is loose compounding
between the head and the complement. (35') is the expected behavior since [mu-...-i] are
products of syntactic formation.
(35)
mu-end-a
[andu
aya
othe
anene
a
gfchagi]
1 love-Nzer 2people 2Dem 2all
2big 2Assoc 7village
'lover of all these big village people'
(35')
mO-end-i
[andu
aya
othe
anene
a
gfchagi]
1 love-Nzer 2people 2Dem 2all
2big 2Assoc Tvillage
'lover of all these big village people'
3.3.2
Coordination
With [mu-...-a] items, neither the head noun nor the complement noun can be
conjoined as seen in (36) and (37). In (36a). we see that two [mu-...-a] expressions can be
conjoined, but in (37b). the deverbal nouns cannot be conjoined so as to take one
complement. In (36c). a deverbal noun cannot be conjoined with a compound, as is
expected from the fact that [mu-...-a] words must have complements. In (37). we see that
the complement does not allow conjuction. Given the fact that the complement in
150
[mu-...-a] is phrasal, we would expect coordination to be possible. (37) is therefore quite
surprising.
(36)
(37)
a
[mu-thnnj-a
mburi] na [mu-rug-a
nguku]
1 slaughter-Nzer 9goat and
1 cook-Nzer 9chicken
'goat slaughterer and cook of chickens'
b.
* [mQ-thimj-a
na mu-rug-a]
ngukfl
1 slaughter-Nzer and 1 cook-Nzer 9chicken
c.
* mu-thilnj-a
na [mu-rug-a nguku]
1 slaughter-Nzer and 1 cook-Nzer 9chicken
* mu-thilnj-a
[mburi na ngukCi]
1 slaughter-Nzer 9goat and 9chicken
It is expected then that if conjunction happens at the DP level, that [mu-...-i] phrases can be
conjoined which is indeed the case as seen in (36') and (37'). (36'a) shows that [mu-...-i]
phrases are conjoinable. (36'b) indicates that [mu-...-i] words can not be conjoined so as to
share a single complement. Unlike [mu-...-a] cases, [mu-...-i] words are conjoinable as
seen in (36'c). Also unlike the [mu-...-a] cases, the complement permits conjunction with
like phrases as is expected of phrases in the syntax as provided in (37').
(36')
(37')
a.
[mu-thiTnj-i
mburi] na [mu-rug-i
ngflku]
1 slaughter-Nzer 9goat and 1 cook-Nzer 9chicken
'goat slaughterer and cook of chickens'
b.
*[mu-thrTnj-i
na mu-rug-i]
nguku
1 slaughter-Nzer and 1 cook-Nzer 9chick
'a slaughterer and cook of chicken'
c.
mu-thnnj-i
na [mii-rug-i
nguku]
1 slaughter-Nzer and 1 cook-Nzer 9chicken
'a slaughterer and cook of chicken'
mu-thiTnj-i
[mbiiri na nguku]
1 slaughter-Nzer 9goat and 9chicken
'a slaughterer of goat and chicken'
151
With regards to the structure we have proposed for these agent nouns, we see that (36'b)
fails because conjunction does not take place between NPs but between DPs as shown in
(38) for (36c). The semantics derived by the bracketting in (36b) are that the slaughterer
and the cook are the same individual, while the reading that speakers of the language gel is
one where two different individuals are implied which is the reading derived from the
bracketting in (36c)
(38)
DP
DP
A_
mu-thiinj-i
slaughterer
CONJ
mu-rug-i
cook
nauku
chickens
'slaughterer and cook of chickens'
(38) is surpported by the fact that the demonstrative has scope over only one of the
conjucts as seen in (39a). In (39a) and (39b), the demonstrative has scope over NP and all
its constituents but not over conjuncts. Thus, in (39b)the demonstrative does not have
scope over the first conjunct. The attempt to conjoin phrases under the scope of one
demonstrative as indicated by the bracketing in (39c) fails.
(39)
a.
[[mu-thunj-i ]
uyu] na [[mi]-rug-i
nguku]
1 slaughter-Nzer IDem and 1 cook-Nzer lOchicken
'this slaughterer and that cook of chickens'
uurfa]
IDem
'''When the meaning intended is with reference to the same individual, then other material such niin^Cnowc 'also
is' must be used together with the conjunction na 'and' as exemplitled by .
[muKhilnj-i
na
niinar nowe
I ACT slaughter-Nzer
and also is
'a slaughterer as well as cook of chickeri'
mu-rug-i|
ngukii
lAGTcook-Nzer 9chickcn
152
b.
mu-ihunj-i
na
[mu-mg-i
nguku
1 slaughter-Nzer and
1 cook-Nzer lOchicken
'a slaughterer and this cook of chickens'
uyu ]
IDem
c. *[[mLi-thrTnj-i
na [mu-rug-i
nguku]] uyu]
I slaughter-Nzer and 1 cook-Nzer lOchicken IDem
'this slaughterer and cook of chickens'
The question that remains is whether the conjunction within the complement phrase in (37')
are NPs or DPs. Here demonstrative scope is revealing. Clearly, (40a) in contrast to (40b)
shows that demonstrative scope does not range over conjuncts. Assumming the analysis
in Mugane (1995). that phonologically null demonstratives mark indefiniteness. then
conjunction such as that in (37') is between DPs (with null demonstratives) and not NPs as
shown in (41). (41) predicts that each of the conjuncts can take its own demonstrative, be
modified as well as have its own local associative phrases as illustrated in section 3.2
above.
(40)
a.
mu-thilnj-i
mburi na [nguku
ici]
1 slaughter-Nzer lOgoatand lOchicken lODem
'a slaughterer of goats and these chickens'
b.
*mu-thrrnj-i
[[mbijri na nguku ]
ici]
1 slaughter-Nzer lOgoat and lOchicken lODem
'a slaughterer of these goats and chickens'
153
(41)
mu-thiinj-i
slauifhterer
DP
DP
A
mburi
goals
CONJ
DP
A
nguku
chickens
'slaughterer of goats and chickens'
By the conjunction test, we have seen that the relevant domain for coordination is the DP
and not N or NP. a fact which we have assumed explains the conjoinabiiity of [mu-...-i]
phrases and the inadmissibility of it for [mu-...-a] types. This distinction among N. NP,
and DP is further evidenced in the test involving pronoun incorporation.
3.3.3
Pronoun Incorporation
To bring out the wordhood of [mu-...-a], we note that studies of Bantu pronouns
(Bergvall (1986), B&M (1987)) have argued that object prefixation involves a type of
pronominal incorporation of the object argument. Object arguments in the syntax of Bantu
sentences have been analysed as being NPs. We would expect that if [mu-...-a] items are
of category NP and not N, they should not show anaphoric islandhood (Postal 1969;
Ward. Sproat, and McKoon 1991). If they are NPs. we should be able to get the subject
prefix being indexed to mu-thiTnj-a 'slaughterer' as in (42) and the complement argument
of the [mu-...-a] mburi 'goat' occurring as an object prefix within an inflected verb as in
(43). But we cannot do that as seen in (42) and (43) for independent reasons. We
proposed earlier that [mu-...-a] words must have direct complements as seen in (4). We
154
might expect however that because the complement of [mu-...-a] is phrasal it can at least
occur as a pronoun if and only if [mu-...-a] expressions are of category NP. This is
however not the case as seen in (44). We may conclude that (44) tells us that [mQ-...-a]
expressions involve a type of compounding to form an N because were it an NP. then (44)
would be grammatical, and not display anaphoric islandhood.
(42)
*mu -thrTnj-a
a-car-ag-ya
1 slaughter-Nzer 3sgS-search-ASP-fv
'A slaughterer searches for goats'
(43)
*mu-thrrnj-a
nf-a -cpO-car-ag-ya
I slaughter-Nzer FP-3sg-100-search-ASP-fv
'A slaughterer searches for them'
(44)
mburi
lOgoat
*mij-thiTnj-a
mburi
nf-a-ci-car-ag-ya
1 slaughter-Nzer lOgoat
FP-3sg-100-search-ASP-fv
'A goat-slaughterer searches for them (goats)'
Following Bresnan and Mchombo (1995) (who test for inbound anaphoric islands in
Chichewa synthetic compounds), we expect that anaphoric and deictic uses of pronouns
should occur within the phrasal complement of the deverbal word but that is not the case
as seen in (45).
(45)
*mu-thnnj-a
Tyo
1 slaughter-Nzer 9Dem
'that slaughterer'
^InGrkuvu objeci NPs can not cooccur with the object prefix in the intlected verb as seen in (1) below, (see
Barlosv (I960), Bergvall (I9S6). However, if the object argument is fronted then it may cooccur with the objeci
prefix as shown in (2) below.
1
11
i2)
'Kamau
nf-a - c i -car-ag-ya
IKamau
FP-3sg-IOO-search-ASP-fv
'Kamau searches tor (them) goats'
mburi
lOgoats
Kamau
mburi
nf-a - c i -car-ag-ya
IKamau
lOgoats
FP-3sg-IOO-search-ASP-fv
'Kamau searches for (them) goats'
155
With [mi]-...-i] phrases however, we do get pronominalization which fails with
[mu-...-a] types indicating that these items are phrasal as seen in (42'). Specifically.
[mQ-...-i] and its complement constitute an NP and this explains why (42'). (43'). and (44')
are all good. Thus with [mu-...-i] we do get outbound anaphoricity (B&M. 1995).
(42')
mu-thnnj-i
a-car-ag-ya
1 slaughter-Nzer 3sgS-search-ASP-fv
'A slaughterer searches for goals'
(43')
mu-thiTnj-i
nf-a -ci-car-ag-ya
I slaughter-N'zer FP-3sg-iOO-search-ASP-fv
'A slaughterer searches for them'
(44')
mu-thunj-i
mbOri
nr-a-ci-car-ag-ya
1 slaughter-Nzer lOgoat
FP-3sg-100-search-ASP-fv
'A goat-slaughterer searches for them (goats)'
mburi
lOgoat
Deictic and anphoric use of the pronoun is also possible with [mu-...-i] phrases as seen in
(45').
(45')
3.3.5
mu-thiTnj-i
1 slaughter-Nzer
'that slaughterer*
Tyo
9Dem
Summary
Throughout this section we have proposed that [mu-...-a] elements exhibit phrasal
recursivity in their complements even though they are compounded under N. The identity
of a is the X® category N as indicated in (46).
(46)
[mu-..V..-aj
156
It is because a is a word that conjunction of complements under the scope of the head N is
not possible.
In contrast to [mu-...-a] types, conjunction is permitted in [mu-...-i] phrases
because [mu-...-i] is phrasal, and it is at the phrasal level that conjunction is possible as
seen in (42') for deverbal elements and (43') for the complements. We have also seen that
only objects which are of category NP can be pronominalized. This requirement has ruled
out pronominal elements referring to the [mil-...-a] to be affixed to the inflected verb, [mu...-a] words being of category N cannot be pronominalized, while [mu-...-i] ones being of
category NP can be pronominalized. hence we get an object prefix which stands for [mu...-i] objects.
3.4
Conclusion
The study of Gfkuyu agentive nominalizations, has revealed a number of
interesting properties. We have shown that there are two types of morphological shapes
with regard to agentive nominalization; that agentive deverbal nominals have nominal and
verbal properties, as well as lexical and phrasal ones. Though having similar semantics,
[mu-...-a] and [mu-...-i] phrases display quite different properties.
In terms of compounding, we see that [mu-...-a] involves a phrasal projection, DP.
inside a lexical projection N as illustrated in (46) above. [mu-...-i] phrases on the other
hand are products of phrasal formation as illustrated in (47), whereby the deverbal noun
has a DP complement and the mother node is a phrasal projection (NP).
(47)
N
[mu-..V..-i]
DP
A
157
Table (48) provides an overview of the two types of agent nouns. Regarding the
verbal properties we see that Grkuyu [mu-...-a] forms display none at all, thus providing
further evidence that the compound is frozen as an N not admitting manipulations of its
stem. The [mu-...-i] types on the other hand admit the whole range of verbal properties
(prefixation of the object and the reflexive morpheme, as well as the suffixation of the
applicative, the reciprocal, and the causative).
With respect to their modificational behavior. [mu-...-a] compounds reveal loose
compounding. On the one hand they display phrasal recursivity within the complement,
but disallow complement conjunction, pronoun incorporation (of the complement), and
extraction out of the complement by relativization. [mu-...-i] compounds on the other hand
permit all these syntactic manipulations.
(48)
Property
Valency restriction
Prefixation
Suffixation
Head and
Complement separable
Complement
modification
Complement Phrasal
recursivity
Coordination
[mu-...-a]
head noun
complement noun
Pronoun incorporation
[mu-...-i]
V
X
X
V
X
V
X
V
V
V
V
V
X
X
X
V
V
V
Clearly, several empirical and theoretical issues arise out of these observations. In
what follows we provide a list some of them.
158
4.0
Issues
Here we will pay attention to the questions that motivated the study in this chapter;
the identity of the nominalizer; the status of the noun class marker; the analysis of agent
nouns; and the lexical integrity principle.
One of the questions that arises is whether nominalizing morphology is prefixal
(Mufwene. (1980). Myers (1987). Kinyalolo (1991) or suffixal. Grkuyu has the possibility
that it is the suffixal morpheme. Thus we are left with a puzzle about what the exact identity
of the nominalizing morpheme is.
Second, how do Grkuyu and Swahili facts we have observed compare with other
studies such as (Mchombo (1978). Carstens (1991)); Shona (Myers (1987). Sproat
(1985)); Kilega (Kinyalolo (1991). and Bresnan & Mchombo (1995). Specifically, why
are Shona (Myers, (1987)) and Chichewa so restricted in synthetic compound formation
Bresnan & Mchombo (1995) Grkuyu and Kilega (Kinyalolo (1991) so liberal and
Kiswahili hanging in the middle way. The question, then, is whether there is an analysis
out there that can capture the phenomena of Bantu agent nominals.
Third, assuming that lexical integrity of words (Chomsky (1970). Aronoff (1976)
Lapointe (1980). Bresnan & Mchombo (1995). among others) makes any sense, there are
clearly no rules of phrase structure which can permit phrases within words. The question
here then, is whether the properties both in behavior and distribution of agent nouns
provide a knock down argument against lexical integrity.
5.0
Previous Solutions
Several proposals have been advanced, none of which can explain the Grkuyu and
Kiswahili data we have considered in the sections above. The main propo.sals have come
from Sproat (1985). Myers (1987). Kinyalolo (1991) and B&M (1995). B&M (1995)
have provided arguments in favor of a lexical account, which argues that synthetic
159
compounds are outputs of lexical formation for Chichewa. Sproat (1985). Kinyalolo
(1991) present accounts which propose that synthetic compounds are products of syntactic
formation in Chichewa and Kilega respectively. In what follows. I will consider each
proposal briefly, paying attention to what each of these studies says about the status of the
noun class prefix, the identity of the nominalizing morpheme, the distributional properties
of agent nouns as well as their structure.
5.1
Sproat (1985)
Sproat looks at data from Chichewa such as (49).
According to him. in (49a) and
(49b). the verb stem is followed by an internal argument, while in (49c) it is followed by a
preposition. In (49d). the internal argument is modified by an adjective but the output is
somewhat odd.
Sproat (1985:119))
(49)
a
chi-lim-a
mpunga
class-dig
rice
'rice cultivator"
b.
m-pala
matabwa
class-scrape
wood
'wood scraper' (=carpenter)
c.
m-kala
class-sit
'rock sitter'
d.
'?
pa
on
m-kusa
agalu
class-catch
dog
'black dog catcher'
mwala
rock
akuda
black
Sproat proposes that the class prefix chi- and m- do not play the function that -er does for
English in words like 'teacher', because -er in English links up to the external theta-role of
the verb, whereas Chichewa chi- and m- do not.
160
For Sproat (1985). the Ncl selects some projection in V in the syntax, a structure
suggested earlier by Mchombo (1978). His structure of synthetic compounds involves
lexical categories exclusively (50). As in English, the sister of V cannot be an
in
Chichewa. which disallows (49d).
N
(50)
matabwa
' wood scraper' (= carpenter)
Contrasting (49b) and (49d) with (51c), Sproat proposes that the sister of V cannot be an
Xmax in synthetic compounds in both Chichewa and English. (51c) indicates that the
noun complement may not take a specifier, otherwise the article would be included in (51c)
since the Bronx comes with an inherent determiner.
(51)
a.
the Bronx
b.
a Bronx hater
c.
*a [the Bronx] hater
Sproat following Keyser and Roeper (1984) (who stipulated that the compounding rule
operates on
rather than
proposes that only
are input to the rule of
synthetic compound formation. This explains the contrast between (52) and (53).
(52)
a. the driver of that green car
b. *that green car driver
161
(53)
a. he drives that
b. *that driver.
For (49c) Sproat argues that pa mwala is not an
since pa occurs there in order to
assign case where the verb cannot do so. He then says that the case assigning mechanism
in Chichewa permits (49). a mechanism which is absent in English thus explaining the
ungrammaticality of *book on shelves putter" in the latter language.
5.1.1
Application of Sproat (1985) to GiTiuyu
We agree with Sproat that noun class marker in Chichewa is not equivalent to -er in
words like "preacher' in English. We disagree however with Sproat's analysis that the
noun class prefix is the nominalizer. which is evident in the structure he proposes for
synthetic compounds given in (50). Sproat's structure of synthetic compounds cannot
explain Grkuyu facts which overwhelmingly support the view that the agent nouns of the
[mu-...-i] kind are phrasal.
Secondly, the whole discussion of the ungrammaticality of (51c) is irrelevant for
Grkuyu because nothing prevents the existence of such expressions in the language. This
is clearly demonstrated for Grkuyu by examples such as (54). Examples such as (54)
have precisely what is ruled out by (51c). a demonstrative ana "those' (which apart from
providing deictic properties is also used for definiteness just like English the ) occurring
with the complement andii "people' while the deverbal noun muendi 'lover' without a
demonstrative gets an indefinite interpretation as can be read from the gloss.
(54)
mu-end-i [andu arfa]
1 love-Nzer 2people 2Dem
'A lover of those people'''a lover of the people'
With regards to the contrast which exists in English between (52) and (53). the
discusions as to why (52a) and (53a) are good as opposed to the respective (b) examples is
moot because as shown in (55) and (56). they are grammatical in Grkuyu.
162
(55)
mu-thiTnj-i
[mo
mburi njeke]
1 slaughter-Nzer 9Dem 9goat 9thin
'this thin goat slaughterer' (the goat is thin)
(56)
mu-thiTnj-i
Tyo
1 slaughter-Nzer 9Dem
'that slaughterer' (that refers to 'goat')
5.2
Myers (1987)
Myers studies Chishona morphology. He analyzes the noun class prefix as an
element which takes NP and VP complements as exemplified by (57). For Myers, the
final vowel of the nominalized form is analyzed as the usual Bantu final vowel -a for verbs.
In Shona. the morphology of the final vowel does not change in the examples Myers
provides.
(57)
NP
mu
vanhu
The analysis provided by Myers (1987) argued that the -a in Shona is the usual
final vowel (verbalizer).
5.2.1
Application of Myers (1987) to Grkuyu
Myers's proposed structure of synthetic compounds can not explain GiTcuyu facts.
For us the nominalizer is not the noun class but rather the suffixal vowel. Different final
163
vowels have different semantic categories-'. In GTkuyu the final vowel is -/ only occurs
with deverbal nouns and never with vowels. There is no evidence that the suffixal -a on
synthetic compounds is the Bantu stative final vowel. Evidence that the -a and -i are
nominalizers is revealed by the fact that other nominalization effect morphological change
on suffixal material and not on the noun class. Thus for GrkQyu we get (58a) which is the
nominalization of completed action that can only be done with primary objects only; (58b)
is the nominalization of a stative act which involves subject arguments only, while (58c) is
nominalization of verbs to derive the meaning of "act of --ing'.
(58)
a.
thiTnj- STEM
mbQri thiTnj-e
lOgoat 1 Oslaughter-A
'goats that have been slaughtered'
b.
gij-
STEM
muti magu-u
3tree 3fall-A
'a tree which has fallen'
c.
uniir- STEM
mu-unur-o
3-peel-A
'act of peeling'
Consider ( ! ) below. In ( l a ) we have the final vowel -a which marks stativeimperative mood on verbs, in ( l b )
the final vowel indicates the subjunctive mood, and in (Ic) -i is the agentive nominalizer.
(1)
X
thiTnj-a
slaughter-fv
b.
ma-thiTnj-c
3pl-slaughter-t'v
'that they slaughter'
c.
mu-thiTnj-i
I AGT-slaughter-Nzer"
'slaughterer'
164
Myers assumes that it is the class marker that nominalizes the V by heading an NP in the
syntax. This a.ssumption leaves unexplained the role of the ever varying suffixes with each
nominalization type abserved in Grkuyu.
The problem inherent in Myers' proposal that the noun class prefix is an NP
selecting element which then projects N' is shown to be problematic for Bantu in general
by B & M (1995).
5.3
Kinyalolo (1991)
Kinyalolo (1991) says that synthetic compound formation is preponderant in
Kilega and their meanings very transparent. He proposes that given the data in (59).
synthetic compounds are uncontroversially syntactic units rather than words formed in the
lexicon.
With regards to (59), Kinyalolo points out that the verb bears a class prefix
rather than an agreement marker. Secondly, the verb is obligatorily followed by both an
internal argument as well as an adjunct, just as it should in a sentence. In (59a), (59b). and
(59c). the verb is followed by an NP complement. In (59c) the verb may be modified. In
(59d). the verb is followed by a time adjunct, an element which hangs off VP. Thirdly, the
verb may appear with the applied morpheme -il- to indicate location where the event takes
place as seen in (59b) and (590- Following Kratzer (1988) (who proposed that stage-level
predicates have an extra argument position for events or spatial temporal location).
Kinyalolo proposes that the morpheme -il- selects a VP (an event) rather than V. Fourth.
the verb may bear the applied affix in double object constructions such as (59g). He points
out that double object constructions have been analyzed to involve VP and not V in the
literature (Larson. (1988). Marantz. (1990), among many others).
Kinyalolo (1991:218)
(59)
a.
ki-li-a
mitutu
7-eat-fv 4ripe banana plantain
'eater(of bananas/ ripe plantains)'
165
b.
ki-li-l-a
(.mitutu)
mu-luuzi
7-eat-fv
4banana/ripe plantain 18-11 river
'eater(of banans/ ripe plantains) in the river'
c.
ki-bing-a (mbwa
zi-keke)
7-hunt-fv lOdog lOagr-small
'hunter (of small dogs)'
d.
ki-nyam-a mu-muusi
7-sleep-fv 18-3daytime
'sleeper (in the daytime)'
e.
ki-enda n-gendo
7-go-fv lOjoumey
'a globe trotter"
f.
ki-il-il-il-u-a
na
busi
mu-lubanda
7-fall-appl-appl-pass-fv with 14day 18-11 forest
'one who is usually out in the forest when night fails'
g.
ki-ek-el-a
basambi
ilya
7-carry-appl-fv 2sick person Sfood
'a carrier of food to sick persons'
Kinyalolo does point out that despite the evidence that they are phrasal, synthetic
compounds are subject to a number of constraints. The first of such constraints is that the
NP cannot be pronominal, hence it cannot be represented as an object prefix.
Constructions such as (60) cannot be derived from the pronominalization of mbwa zi-keke
'small dogs' which is the complement argument in (59c).
(60)
*ki -zi -bing-a
(mbwa zi-keke)
7-10-hunt-fv
(lOdog lOagr-small)
'hunter of them' (them = small dogs)
The second constraint on the complements of Kilega synthetic compounds is that,
when countable, they must be plural, otherwise as Kinyalolo puts it. examples like (59a)
will imply the eating of the same banana several times which is a physical impossibility.
166
TTie third constraint on Kilega nominalized verbs is that they cannot be marked for
tense and aspect, which implies that no IP and ASP" projections are involved in synthetic
compound formation.
Given the foregoing facts, Kinyalolo then proposes a structure for synthetic
compounds. In so doing, he finds that Sproat's (1985) analysis and structure of synthetic
compounds given above in section 4.1 could not explain Kilega facts.
Citing Abney (1987). Kinyalolo (1991) argues that the rule for synthetic compound
formation operates on NPs rather than on DPs. This accounts for the ungrammaticality of
(61b) and (62b) provided by Kinyalolo.
He says that (61b) and (62b) (repeated below for
convenience) are bad because following Abney (1987). the category Det comprises of
determiners, pronouns (such as 'that' and 'which' respectively) which project DPs. Sproat's
(1985) argument that only
are input to the rule of synthetic compound formation is
out (says Kinyalolo). not because the articles are missing, but because the rule makes
reference to NPs and not to DPs.
(61)
a. the driver of that green car
b. *that green car driver
(62
a. he drives that
b. *that driver.
Kinyalolo (1991) proposes the structure of synthetic compounds to be (63)
167
Kinyalolo (1991;226)
(63)
W
mitutu
PRO
mundulu
Kinyalolo hypothesizes that the subject NP is base-generated in VP and occupied by an
empty category (EC). The VP is selected by a null head, and the external theta role is
assigned to PRO. In structure (63). PRO subject of the AP will always be bound by its
antecedent. The antecedent of the PRO subject of AP is the PRO adjoined to the VP. This
is possible only if NP is the category of PRO. The compound contains a governor, the
verb //•- 'eat' and a subject NP accessible to PRO which has undergone raising to VP.
Kinyalolo then concludes that predication in Kilega may be treated as a consequence of
binding theory.
Kinyalolo proposes that synthetic compounds in Kilega are syntactic formations
because they may be modified by numerals, adjectives, just like nouns as exemplified by
(64) , and undergo movement as a unit in processes such as passivization as shown in
(65).
(64)
kilia
mitutu
mundulu
umozi / kimozi
a-a-nyam-a
7-eat-fv 4ripe banana naked
lagr-one/7agr-one 1 agr-A-sleep-fv
'one eater of ripe bananas while naked is asleep'
168
(65)
a.
pro tu-a-libul-il-e
ki-li-a
mitutu
mundulu
u-mozi
Ipl-A-beat-IL-fv 7-eat-fv 4banana naked
lagr-one
'we beat [one eater of ripe bananas while naked]
b.
ki-li-a
mitutu mundulu u-mozi
a-a-libul-il-u-e
7-eat-fv 4banana naked
lagr-one
Ipl-A-beat-IL-pass-fv
'[one eater of ripe bananas while naked] was beaten'
Thus the facts in Kilega seem to pattern with the modificational possibilities of [mu-...-i]
phrases in Grkijyu.
On the question about the status of the noun class prefix. Kinyalolo (1991)
following Carstens (1991) assumes that 'the morphological process of affixation are
nothing more than adjunction of an
markers as
5.3.1
to a Y®. The demerits of an analysis of noun class
are discussed in B&M (1995) and the results there apply to Gikuyu.
Application of Kinyalolo (1991) to Gfkuyu
There are several important observations made by Kinyalolo that accord with what
we observed for G\0(i,")k\0(ur)y\0(u.'). The data presented in (59) are easily seen to be
similar to Grkiiyu as provided in section 3.2 of this chapter. With respect to the constraints
on synthetic compounds that Kinyalolo observes, the first, which is demonstrated by (60).
where Kilega does not permit pronominalization of the complement by object prefixation.
is possible with Grkuyu as is illustrated in sections 3.1.2 for [mu -...-i] types. The other two
constraints, that the complement must be plural if countable and the inadmissibility of
tense, aspect, and modality markers, are also true for Gfkijyu.
With regards to expressions such as (55) and (56). we have already noted in
sections 3.2.1 that demonstratives can be present. Thus the rule of compounding does not
happen in NP or within DPs. but between N and NP for Grkuyu. Clearly. Grkuyu data
demands a proposal different from that availed by Kinyalolo (1991) who proposes that
compounding does happens within NP and not DPs.
169
This brings us to structure (59) which Kinyalolo (1991) proposes for synthetic
compounds. Kinyalolo's structure is inspired by his concern about theta role assignments,
specifically the external argument. However, once it is recognized that the nominalized
form contains the agent role in its lexical specification contributed by the suffixal element,
then we need not capture theta role assignments syntactically.
Finaly. the distributional properties provided by Kinyalolo's data in (60) and (61)
apply to Gfkijyu starightforwardly as extensively presented in section 3.3 with regards to
[mu-...-i] phrases.
Clearly, there are problems that remain unresolved when we apply Kinyalolo's
(1991) analysis to the GHcuyu facts. First, while Kinyalolo data is quite in accord with our
observations of [mu-...-i] phrases, areas in which [mu-...-i] phrases differ from [mu-...-a]
ones remain totally mysterious. This is in addition to explaining why (55b) and (56b).
which are bad in English, are completely grammatical in GiTcQyu. The suggestion that
synthetic compounding happens within NPs and not with DPs is obviously wrong for
Grkijyu. The structure (59) depends on the assumption that there is an empty N that
selects VP. Kinyalolo (1991) does not make (61) compelling partly because there is no
discussion of why (among other possibilities) two VPs cannot be conjoined under the
scope of a common null N--.
5.4
B&M(1995)
B&M present several arguments against deriving synthetic compounds in the
syntax. They provide 5 tests to show that synthetic compounds in Chichewa respect the
lexical integrity of words (Chomsky (1970). Aronoff (1976). Sells (1995 ) etc.). The tests
are: extraction, conjoinability. gapping, inbound anaphoric islands, amd phrasal recursivity.
— T h a n k s ui Pri)l'. Joan Bresnan for pointing this out.
170
We will illustrate their arguments here by considering two of these; extraction and
anaphoric islandhood.
B&M say that constituents of words cannot be extracted by such syntactic
operations as relativization. clefting or topicalization. Thus in (66) from Chichewa (from
B&M (1995:203)), because the verb gut has undergone manner nominalization, its locative
can no longer be extracted.
(66)
a.
Mw-ana w-anu
a-ma-zond-a
1-child
1-your
ISB-PRS HAB-hate-IND
[ka-ik-idwe
ka
ma-kasu m' mi-nda
12-put-NOM 12ASC 6-hoe
18 4-rield
'Your child hates the way the hoes are put in the fields'
b.
*M' mi-nda m-mene mw-ana
w-anu
a-ma-zond-a
18 4-field
18-REL 1-child
1-your
ISB-PRS HAB-hate-IND
[ka-ik-idwe
ka
ma-kasu
] mu-li mi-kango
12-put-NOM 12ASC
6-hoe
18-be 4-lion
'In the fields where this child hates the way hoes are put. there are lions'
B&M argue that the manner nominalization is headed by a noun which is morphologically
derived from a verb-^.
^ ^ T h c cases in (66) are comparable to constructions involving the intlnitive/gerund where such extraction is
possible a s s h o w n by the examples below. Here B&M say that the infinitive/gerund phrase is headed by a verb.
B&M ( 1995:202)
(I)
a.
b
.Vlw-ana
w-anu
a-ma-zond-a
1-child
I-your
ISP-PRS HAB-hate-IND
[ku-ika ma-kasu m' mi-nda]
12-pui
6-hoe
18 4-field
'Your child hates putting the hoes in the fields'
M ' mi-nda
m-mene mw-ana
w-anu
a-ma-zond-a
IS 4-field
18-REL l-child
1-your
ISB-PRS HAB-hate-IND
|ka-ika
ma-kasu
] mu-li mi-kango
12-put
6-hoe
18-bc 4-lion
'In the fields where this child hates to put hoes, there are lions'
B&M ( 1 9 9 5 ) say in addition that this verbal argument restriction is further motivated by applied verb
constructions.
171
With respect to anaphoric islandhood. B&M argue that anaphoric and deictic uses
of pronouns are expected to occur within the phrases and not words. They observe that
English phrases such as eater of them and climber of this contrast with *them-eater (cf.
man-eater), *thiS'climber (cf. rock-climber). The former cases are phrasal and the latter
cases are bad because anaphoric and deictic uses of pronouns can not occur within words.
This being the case B&M argue that Chichewa synthetic compounds are clearly
morphologically derived words because of evidence very similar to that of English. They
note such examples as koncla maungu
'love pumpkins' with which the synthetic
compound mkonciamaungu 'pumpkin lover' is derived, also contrast with kondci in-o 'love
them' and its correlative *mkondaiwo 'them-lover'.
In answer to the apparently phrasal properties introduced by Carstens (1991). with
examples such as mwandika vitabu vya mapenzi 'a writer of romantic books' and mchimba
kisima na kaburi 'a well and grave digger', B&M point out that such cases are prevalent in
English compounds as well as exemplified by raw oyster eater. American history teacher.
They argue that the presence of phrasal strings within words is not enough evidence for
word generation using syntactic principles.
B&M also discuss the status of the noun class prefix, and their findings do explain
some of the behavior and distribution of Grkuyu class markers.
With all the tests that they employ, B&M show that Chichewa synthetic
compounds and noun class marker (excluding the locatives) respect the lexical integrity of
words.
5.4,1
Application of B&M (1995) to Grkuyu
Applying the B&M tests to Grkuyu we find properties that are quite distinct from
the Chichewa data. With respect to extractability. GfkQyu [mu-...-i] phrases permit such
extraction as seen in 3.3.4 above and in fact such data (the [mu-...-i] types) is absent
172
altogether in Chichewa. Their proposal does explain Grkuyu [mu-...-a] with respect to
their inextractability as well as the failure to implement the deictic and anaphoric usage of
the complement. The lexical analysis however, predicts that [mu-...-a] compounds will
disallow phrasal recursivity within their compounds which is not the case as seen in
sections 3.3.1.
The presence of phrases within words is clearly evident in GrkuyiJ. and the
question remains whether GrkuyO pre.sents sufficient evidence for the formation of words
by phrase structure principles. By examining the behavior and distribution of [mu-...-i]
phrases, we have come out with conflicting results. On the one hand we see that Grkijyu
[mu-...-i] phrases are products of phrasal formation, while GrkuyiJ [mii-...-a] compounds
fit the description of the expressions B&M call 'sobriquets'. These cases do not di.splay
genuine phra.sal recursivity as indicated. B&M (1995:228) point out that the earmarks of
lexicalization are lexical gaps and the existence of lexical integrity properties. For Grkiiyu
we have exposed some of these properties with [mQ-...-a] compounds in their
inextractibility and anaphoric islandhood just like in Chichewa synthetic compounds. We
think that the existence of words like those in (67) renders support to a lexicalization
process proposed by B&M. In these examples (67a) shows that in GHcuyu we have
lexicalized [mu-...-a] compounds, also evidenced in Swahili by (67b). both examples are
just words and not phrases.
(67)
a.
munyuamaf
lit. 'water drinker'
'a kind of tree'
b.
kivunja mbavu
lit. 'rib-breaker'
'something extremely humorous'
173
5.5
Summary
In all the studies above, we have seen that none avails an adequate explanation of
the phenomena displayed by Grkuyu. This is not at all suprising, given that Kilega
synthetic compound are phrasal (Kinyalolo (1991)) but Chichewa (B&M (1995)) ones are
lexical and the Shona (Myers (1987)) cases though presented as phrasal (within a
phonological domain) are not studied for lexical/phrasal properties. Given that Orkuyu has
the whole spectrum—lexical, quasi-lexical, and phrasal—it is not suprising that each of the
previous studies provides only a partial explanation for Grkuyu. The conflict is within the
lexical syntax divide itself.
We now attempt to answer the questions that prompted the writing of this chapter.
6.0
Proposal
What is [he identity of the nominalizer?
The final vowel -a and -/ are the nominalizing morphemes for Grkuyu agentive
deverbal nouns. The first evidence of this comes from verb stems that end with a vowel
different from the stative verb-final vowel -a. The forms in (68) provide such stems and
upon nominalization. the -a is used in place of the lexical ending.
(68)
a. he u-horo
'give news'
—>
mu-ha u-horo 'news-giver'
b. te mbiya
'lose money'
—>
mu-ta mbiya
'money-loser'
The second type of evidence comes from a look at the morphology of other types of
nominalizations. Every time we change the type of nominalization we get different suffixal
morphology as seen in (69). but not the noun class prefix necessarily .
(69)
Stem
Verb
a. thunj- thiTnja
b. tliiTnj- thiTnja
'gloss'
'slaughter'
'slaughter'
Nominal
GrthiTnjrro
muthnnje
'gloss'
'slaughter location'
'slaughtered'
Nom Type
Location
Deadjectival
174
c. thjTnj- thiTnja
d. cokcokia
e. frTra
'slaughter"
'return'
'tell'
muthiTnjrre
macokio
mwilro
'slaughter manner'
'reply'
'tell one-self
Manner
Result
Reflexive
What is the status of the Noun class?
In Bantu, noun class markers adjoin to noun stems to form root nouns and derived
nouns. The noun class prefix must be present for a noun to exist. If it is maintained that
the class marker has phrasal status (as done by Myers (1987)). it is not clear how that
phrasal status is to be explained in the formation of nouns from root nominal stems
exemplified by (70). The stems -rigu and ndu- mean nothing without prefixes. The class
marker cannot have phrasal status in these examples, without invoking some form of
syntactic formation of words, which in turn requires a set of supplementary theoretical
assumptions.
It is also not clear what an analysis which takes the class prefix as a
nominalizer would say in cases such as (70b), which would require an explanation for why
nominalization (class prefixation) happens more than once, in effect nominalizing a
nominalization. something that is clearly incoherent.
(70)
a
i-rigu
5-banana
b.
ka-mu-ndu
12-1-person
"little person"
If it is admitted that class markers are morphological elements which participate in word
formation processes and that they are not available to syntactic manipulation, then this
explains why derived nouns can have the class prefix change according to the semantics of
the noun class without changing the suffixal material as illustrated in (71).
Case (71a)
represents the [mu-...-i] phrases we have been studying. (71b) shows that we pluralize with
the class prefix a-. In (71c). we diminutize the deverbal noun by the prefix ka- . We form
plural diminutive with tu- as seen in (7Id)
175
(71)
a.
mu-ror-i
mburi
1 -look after- Nzer
1 Ogoat
'one who looks after goats'
b.
a-ror-i
mburi
2-look after-Nzer
1 Ogoat
'ones who look after goats'
c.
ka-ror-i
mburi
12-look after-Nzer
1 Ogoat
'a little one who looks after goats'
d.
tij-ror-i
mburi
13-1ook after-Nzer
1 Ogoat
'little ones who look after goats'
In answer to the question of the identity of the nominalizer and the status of the noun class,
this study proposes the structure provided in (72). [mu-...-a] types disallow any kind of
affixation and thus behave like root nouns. Since they are deverbal by composition, we
have proposed that affixation is disallowed because nominalization (affixing -a ) precedes
all affixation. The word structure (72) indicates that -a is affixed to the verb root.
(72)
mu
Nstem
In the [mu-...-i] phrases, where all verbal behavior is permitted, nominalization (affixing -i
) follows all affixation. As with the gerund [-NPobj]' the verb stem of [mu-...-i] combines
with -i to create an Nstem and the Nstem combines with the cla.ss marker mii - to form a
noun, hence the structure (72').
176
(72')
N
^CXJl
6.2
Exto
Solution to Problem 1
How can both the lexical!phrasal ambiguity of agentive nouns be represented?
The solution to the issue of how the status of [mu-...-a] compounds in Grkuyu can
be captured given that they display lexical/phrasal ambiguity, is quite evident from the
structures that were provided in sections 3. Grkuyu
[mu-...-a] types are compounded
under N as in (46) repeated here as (73) and Grkuyii [mu-...-i] are phrases which attach to
the phrasal projection NP (X"i^) as in (47) repeated here as (73').
(73)
[mu-..V..-a]
DP
A
[mij-..V..-i]
6.3
DP
A
Solution to Problem 2
Why different morphological shapes?
Consider four pieces of evidence from Gfkuyu. Clearly, we have four flavors of agentive
nominalization in Grkuyu. The one in (74) is a lexicalized phrase which has can only be
177
used as a word. (75) is the one that we have studied in parts of section 3.0. where we have
shown that it is hanging in the middle way between being lexical and phrasal. It is more
phrasal than lexical since it admits very few properties of words. (76) is a phrase in which
all syntactic properties are manifest, while (77) is the open type of [mu-...-i]. which can be
used as a noun and can ignore the stem valency realization.
(74)
mutigairi"
'deceased person who left some offspring'
(75)
m u -end-a andu
1 want-Nzer 2people
'one who loves people'
(76)
mu-end-i andu
1 want-Nzer 2people
'one who loves people'
(77)
mu-end-i
1 want-Nzer
'lover'
6.4
Solution to Problem 3
What happens to the lexicon!syntax divide?
Observe that even though Kinyalolo's (1991) proposal (that synthetic compounds
are products of phrasal formation) is right for Kilega. and also that B&M (1995) lexical
account for Chichewa fits the Chichewa facts, it is clear that neither proposal can account
for the Grkuyu facts. Assuming that our assumptions require us to have either a lexical
account or a syntactic one. observe that we have run out of options. So how do we account
for the facts presented by Grkuyu? Clearly, the answer must lie somewhere in the middle
of the hypothesized lexicon-syntax divide itself.
What all these agentive nouns suggest is that there is some communication within
the lexicon/syntax divide. We present the lexicalization continuum as in (78). The idea in
(78) is that the lexicalization of phrases happens by gradually admitting some properties of
178
wordhood into the phrases, but in a quite constrained manner, so that lexicalizing forms (in
the area labelled LOOSE COMPOUNDING) will permit very few loose properties of
compounding which violate the lexical integrity of words in a very restricted manner and
not in an extensive manner, where they would be all over the map. In the languages
considered in this study we see that Kilega agentive nominalization according to Kinyalolo
(199!) is phrasal while in Grkuyu agentive nouns are identifiable in three different places;
phrasally. in the middle way, and lexically. In Chichewa according to B&M (1995) and
Shona according to Myers (1987) synthetic compounds are a product of word formation.
(78)
LEXICAL
FORM.\TION
Shona
(Myers, 1987)
LOOSE
COMPOUNDING
PHR.ASAL
FORMATION
GTkuyu
nuenda andu
Kilega
(Kinyai(^o, 1991)
Grkuyu
muendi andu
Chichewa
(B&M, 1995)
1
Grkuyu
mutiga in
Kiswahili
kivunja-mbavu
1
1
The conclusion arrived at here is not entirely unprecedented in Bantu. In studying
the Bantu word. B&M (1995) show that in Chichewa. a set of noun class markers (the
locatives) have remained phrasal while the rest lost their syntactic status diachronically. An
incomplete lexicalization process predicts that some elements will traverse the lexicalsyntax divide, and just as the Chichewa locatives have remained phrasal, we expect to find
some synthetic compounds that are totally phrasal, totally lexical ones and some hanging in
the middle-way exhibiting lexical/phrasal ambiguity.
This means that the lexical integrity
179
principle has to be conceived of as a permeable notion (due to processes such as
lexicalization) rather than the hard and fast divide proposed in B&M (1995). It is with a
lexicalization hypothesis that we see why previous solutions, while tenable in the individual
languages studied, fail to apply cross-linguistically or even to the related languages within
the Bantu language family.
7.0
Conclusions
In this chapter, we have shown that there is a lexical/syntactical ambiguity exhibited
by the phenomena of nominalization. Specifically, we have presented ample evidence to
show that despite the fact that agentive deverbal words in Grkiiyu are products of syntactic
formation as proposed by Kinyalolo (1991) for Kilega they do display a number of
characteristics typical of products of lexical formation as argued for Chichewa by B&M
(1995). Since neither a purely lexical account nor a purely syntactic analysis can explain all
the facts in Grkuyu. we have proposed that Grkuyu facts are best captured when we permit
a permeability within the lexicon-syntax divide itself. We have suggested that we can not
discard the notion of lexical integrity altogether because, we see effects of the lexical
integrity clearly manifest in [mij-...-a] compounds.
With regards to the different morphological shapes, we have noted that we have a
split because the forms that are lexicalized are different from those that are phrasal. It is
noteworthy for instance that we have not found any Grkuyu [mu-...-i] phrases that are
lexicalized so as to be names for entities, yet such words abound with [mu-...-a] types
Finally, with regards to the structure, we have also proposed that word structure is
head final while phrase structure is head initial for Gfkuyu . The principle evidence for this
is clearly observed when a range of nominalization types are observed. When it is
admitted that the nominalizing material is suffixal. we are able to derive a descriptive
generalization that, class markers only combine with noun stems to form nouns. That is
180
the reason why we can change the noun class in order to include meanings like pejorative,
diminutive, etc without changing the type of nominalization. It is not possible to change
suffixal material and remain with the same nominalized form.
181
CHAPTER 6
ATTRIBUTIVE PHRASES
1.0
Introduction
As is true with the Bantu noun phrase, there is very little in the literature on Bantu
adjectives and adjectival phrases (Whiteley (1960). Dixon (1977). Moshi (1992). Bresnan
and Mchombo (1995)). In chapter 4, we proposed a lexical domain, where the head N
occurs and a concordial domain where all agreement displaying nominal elements of
category D occur. We noted that despite the fact that adjectives are marked with noun class
morphology, they nevertheless occur within the concordial domain. Since this is apparently
at odds with the proposal provided in chapter 4. in the present one. we will provide
evidence showing that attributive words are deverbal nouns. These deverbal nouns are
reduced relative clauses-'* and relative words (which head relative clauses) are concordial.
Additionally, we will show that deverbal nouns are very similar to agentive nominals in
terms of constituent structure. However, unlike agentive nouns which are f-s heads of the
noun phrase, deverbal nouns are modifiers in terms of grammatical function within the f-s.
As in the case of [mu-...-i] phrases, we employ the Extended Head notion (Bresnan (1996)
to explain the behavior and distribution of attributive words and phrases. It is concluded
that all noun phrase structures having extended heads can be assigned a modifier function.
As will become clear, a study of adjectives leads directly to questions about the tenability of
a cross-linguistic category "adjective".
-•^As Dick Oehrle points out to me. if it is the c a s e that adjectives are reduced relative clauses, there is a
semantic prediction that Bantu languages in general will not have any modifiers like former', which can be
modeled a s predicate modillers. but not a s predicates of individuals, unlike relative clauses. This is the case.
182
According to Barlow (1951:102). there is a 'paucity' of true adjectives in Grkuyu.
Modes of expressing English adjectives do not reveal the actual manner of attribution-^. In
this chapter we look at some of the ways of expressing attibutive notions in Grkuyu. In
Grkuyu, as has been observed to be the case with Chichewa (Bresnan & Mchombo 1995),
root adjectives consist of a closed set of elements. The examples in (1) are examples of
root adjectives in Gfkuyu.
(1)
a. -kan
'stingy'
b. -toi
'ill behaved'
c. -cuwa
'fresh/young produce'
d. -arir
'wide'
e. -Qru
'bad'
In GrkijyQ. the major system of adjectival phrases is couched in derivational processes of
word formation. The three types provided in (2) are some of the principle types of
attributive phrases. (2a) is the participial deverbal adjective which is marked by [Ncl-...-u]
morphology, and (2b) is the patientive deverbal adjective formed by affixation of
[Ncl-...-e] morphology onto the verb stem, and (2c) is an agentive nominalization studied
in the foregoing chapters. The noun class marker occurs as a prefix and the adjectivalizing
vowel occurs as a suffix.
(2)
a.
b.
njogoo
mak-u
9rooster 9worry-Azer
lit. 'a worried rooster'
"a rooster that is worried"
irio
ngarang-e
1 Ofood 1 Ofry-Azer
•fried food"
^ Barlow (1951:102-110) sets out lo explain modes of expressing English adjectives in Grkuyii. instead of Grku>ii
modes of aiiribuiion without regards to English. Consequently. Barlow glossed over or left out altogether the real
Cirkuyu adjectives. TTie [N'cl-...-u| form is bearly mentioned and the (Ncl-...-el form is missing in Barlow's grammar.
183
c.
mwana mwathfk-i
1 child
lobey-Azer
"an obedient child'
Deverbal nouns refer to individuals and can stand alone as in (2').
(2')
a
mak-u
9worry-Azer
"one that is worried'
b.
ngarang-e
lOfry-Azer
'fried ones'
c.
mwathfk-i
lobey-Azer
"an obedient one'
In this chapter we argue that adjectives like those in (1). which are discussed in
section 3.1. are extremely few in CFkilyij and as observed by Dixon (1977; 12), they range
from fewer than ten to about forty or fifty, for Bantu languages in general. TTiese are used
attributively within the noun phrase and they bear the morphology of the head noun. In the
terminology of Whiteley (1960), these would be considered dependent nominals because
they are not restricted to any particular noun class as nouns are. The cases in (2a). (2b). and
(2c) in contrast to the cases in (1) are analyzed as NPs which are used attributively. We will
study them in turn in sections 3.0. 4.0 and 5.0.
2.0
Issues
There are four main issues that arise in the light of evidence from Grkuyu:
Issue I. One of the major observations of chapter 4 is that the distinction between a lexical
domain and a grammatical one captures the generalization that elements formed by the
affixation of the noun class marker are in the lexical domain, while those formed by the
affixation of concordial elements belong to the grammatical one. The issue in question here
then, is why attributive words do not pattern with concordial elements even though they
184
(attributive words) occur within the concordial domain, Table (3) shows that adjectives take
noun class markers.
(3)
Nci
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
tnu
a
mu
mf
i
ma
gf
i
n
n
ru
ka
tu
u
ku
ha
ku
example
muini
aini
muti
mftf
iaego
maaeao
aitr
iu
ngui
ngui
ruuT
kahiT
tijhn
uuru
kuruga
haha
gliku
Gloss
singer
singers
tree
trees
tooth
teeth
chair
chairs
dog
dogs
river
little boy
little boys
evil
to cook
here-Specific
here-General
Adjective 'big'
munene
anene
munene
mfnene
inene
manene
kTnene
nene
nene
nene
rijnene
kanene
tunene
munene
kiinene
hanene
kunene
Issue II. The second issue is with regards to the countervailing mechanisms for the
impoverished root adjective system in Bantu.
Issue III. The third issue concerns the proper representation of mixed category
status items which are used as attributive modifiers.
Issue IV. Then there is also the issue of cross-linguistic typological variance with
respect to attributive phrases and adjectives.
In sections 3.0. 4.0. and 5.0. we present the data to be considered, as well as the
behavioral and distributional properties of the adjectives provided in (1) and (2) above. For
each type of adjective, we will provide a c-structure and an f-structure. As will become
evident,
the typical modifier of nominal heads in Grkuyu. are non-inflected VPs. These
attributive VPs are headed by verbs which are adjectivalized and which carry in their
185
semantic information such as 'staiive' with regards to root adjectives, 'perfect' concerning
the patientive. 'complete' in participial adjective formation, and 'habitual' with regards to
agentive nominals. We begin with a look at data relating to root adjectives and then
proceed to examine the attributive use of participial verb phrases, patientive verb phrases,
and agentive nominals.
3.0
Root adjectives
As pointed with respect to the data in (1) bare adjective stems, are a highly restricted
class of elements. These stems combined with any noun class marker to make adjectives
as provided in (5). These stems (unlike verbs) never take subject agreement morphology.
(5)
a. a-karf
2-stingy
b. mf-arif
4-wide
c. mu-toi
1 -wayward
d. ga-cuwa
12-fresh/young produce
e. rf-uru
5-bad
However, the class 14 prefix converts these adjectival stems into nouns such as u-kari
'stinginess' and Ci-toi 'waywardness'. •
3.1
Properties of Root Adjectives
Root adjectives have properties that distinguish them from other attributive words in
Gfkuyu. Unlike the attributive use of verb phra.ses discussed in sections 3 and 4. root
adjectives cannot have suffixes, but like them (attributive VPs), root adjectives do have a
subcategorization frame and behave like intransitive verbs in that they only require a subject
186
(external) argument. Root adjectives are similar to the derived "adjective" words (whose
discussion follows below) in a number of other interesting properties:
(i) Root adjectives reduplicate as shown in (6), otherwise reduplication in Grkuyu is only
possible with verbs and deverbal words. The process involves a copy of the stem plus the
final vowel (Peng 1989).
(6)
a.
mCi-karakan
3stingy
'a bit stingy'
b.
ga-cuwacuwa
l2-fresh
'a bit more fresh'
(ii) The noun class 14 <7- does not form adjectives but rather converts adjectives to nouns as
shown in (7). This is evidenced by the fact that root adjectives with class 14 pretlxation do
not permit reduplication as seen in (7'). Nouns do not reduplicate in Grkuyu.
(7)
(7')
a.
u-kan
14-stingy
'stinginess'
b.
u-cuwa
14-fresh
'freshness'
a.
*ij-karakarr
14-stingy
'a bit more stinginess'
b.
14-fresh
'a bit more freshness'
iii) Adjectives unlike nouns and like verbs are modified with intensifiers as indicated in (8)
(8)
a.
mu-kari" uuru
Sstingy terrible
'a terribly'stingy one'
187
b.
ga-cuwa muno
12-fresh exceedingly
'exceedingly fresh'
iv) Adjectival words unlike nouns do not allow preprefixation on the stems (9).
(9)
V)
a.
*ka-mu-karf
13-1-stingy
'diminutively stingy'
b.
*tu-mf-arir
I3-4-wide
'diminutively wide'
Inspite of the behavioral contrast between adjectives and nouns, adjectives can be u.sed
referentially. independent of the head noun.
As the gloss to (10) shows, we can
successfully refer with adjectives.
(10)
3.2
a.
mu-karf
Bstingy
'a stingy one'
b.
ga-cuwa
12-fresh
'a small fresh one'
Analysis
In terms of function, the adjective plays a modificational role as shown in (11) by
the equation ( TmOD) =
i. By
the universal principles of structure-function association
Bresnan (1995), constituents adjoined to maximal projections, such as AP in (11). are a
non-argument functions. Root adjectives bear the equation (TH) = (i SUBJ) as part of
their c-s annotations. This annotation says that the value of the head of the mother node is
the same as the value of the subject of the AP. This equation implies Modifier-head
agreement which makes it unnecessary to write the equation (T AGR) = i on the c-s.
188
requiring that the head and its modifier agree in gender marking. Thus in (11). the value of
(i-AGR) for AP would be nothing else but gender class 1.
(T MOD)=X
Th =(iSUBJ)
AP
mundCi
person
T=I
A
mutoi
waywarrd
'a wayward person'
The lexical specification of the terminal elements in (11) is as provided in (12). In these
specifications, we assume that root adjectives select for a subject argument. The noun class
marker introduces an optional
(12)
a.
b.
mundu
mutoi
PRED "pro" feature in the lexical process of word formation
N;
A;
(TPR E D )
(TGEND)
= "person'
= 1
(TPRED)
((TSUBJ PRED)
(TSUBJGEND
= "wayward
= -pro')
<SUBJ>"
= 1
The f-s for (11) is as provided in (13). The lexical specifications of adjectives explain the
optionality of the head exemplified by the examples in (10). Specifically, since root
adjectives have the feature
PRED
'pro', which is optional as part of their lexical
specifications, they can be u.sed to refer in the absence of the head.
189
13)
-
„p.p,
MOD
PRED
GEND
CI ID I
'person'
1
(PRED
GEND
pro'
1
PRED 'wayward <SUBJ>'
Note that the f-s for the modifier provides an interesting distinction between nouns and
adjectives. As pointed out in Dixon (1977). adjectives appear in any noun class (hence are
dependent nominals) unlike nouns which belong to a particular class (independent
nominals). In fact Moshi (1992) argues that adjective stems select their noun classes. One
way to capture the behavior of the root adjectives (and indeed the ones in the ensuing
sections) is provided in the f-s for the modifier. The modifier has a PRED which lexically
subcategorizes for a subject. That subject may be provided by the head noun or may be
provided by the information contributed by the noun class morphology.
4.0
Participial adjectives
There are two kinds of participial adjectives (referred to as such by Barlow 1951:102)
in Grkuyu: a group involving a group of adjective stems and another involving verb
stems. The adjective stems involved in this process are different from those discussed
above in section 3. We begin with participials made with adjective stems and then proceed
to those made with verbs.
4.1
[Ncl-...-h-u] adjectives
In addition to root adjectives, there is a separate class of adjective stems which can
be converted to verbs bv the suffixation of -h- to the base stem as seen in (14). All the
190
resultant words are intransitive with the meaning of 'becoming X'. These must be
distinguished from regular transitives which can also participate in the formation of
participial adjectives.
(14)
a.
b.
c.
d.
e.
Adjective
-nene
-tune
-ceke
-cong'i
-rimu
'big'
'red'
'thin'
'ugly'
'foolish'
Verb
-nene-ha
-tuni-ha
-ceke-ha
-cong'i-ha
-rimu-ha
'become
'redden'
'become
'become
'become
big'
thin'
ugly'
foolish'
Stems in (15) can best be classified with participial verbs because, unlike root and
patientive adjectives, these stems permit the derivation of participial adjectives as seen in
(15). I assume that to derive participial adjectives from these select stems, first the stems
must be converted to verbs by -h affixation and then to participial adjectives by -u
affixation.
(15)
Adjective
u.
b.
c.
d.
e.
-nene
-tune
-ceke
-cong'i
—rimu
'big'
'red'
'thin'
'ugly'
'foolish'
Participial
Adjective
-nene-h-u
-tune-h-u
-ceke-h-u
-cong'i-h-u
-rimu-h-u
'that which has become big'
'that which has become red'
'that which has become thin'
'that which has become ugly'
'that which has become foolish'
The participial forms in (15) have behavior and distribution that is identical to participial
adjectives derived from base verbs. We proceed to introduce them below.
4.2
[Ncl-...-u] adjectives
Participial adjectives are formed by the nominalization of verb stems of any valency
as exemplified by (16). For this type of deverbal noun, the morphology is [Ncl-...-u] and
the interpretation is 'one who has Xed'. Hence the crucial role here is the participant one.
That participial adjectives take noun class prefixes is not suprising since root adjectives
191
themselves are formed by noun class prefixation. This type of nominalization of verbs is
used to modify entities. All arguments of the input verb may be realized in the expression,
thus, intransitives produce expressions with one external argument, while as expected,
transitive stems produce an adjectival meaning consisting of the external (the participant)
and internal (the patient/theme) arguments. Ditransitive stems in turn produce adjectival
phrases containing three arguments the participant, the object and secondary object. The
expression as a whole designates some completed event or result state.
(16)
a.
njogoo
mak-u
9rooster 9worry -Azer
"a rooster that is worried"
b.
mundu mii-thom-u
thiomi
I person I -study-Azer 1 Olanguage
•person who has studied languages'
c.
muintu
mu-twar-u nguku
ndunyu
Igirl
1-take -Azer lOchicken 9market
"a girl who has taken chicken to the market'
(16b) has the structure shown in (17) of a VP adjoined to an
(17).
£)p
mundu
person
\
muthomu
studied
fhiomi
languages
'person who has studied languages'
which is a DP.
192
The similarity between participial words and the corresponding VPs is captured in
terms of the invariance of the argument structure, rather than in the constituent structure-®.
The difference between participial deverbal nouns and VPs is discussed in section 4.3. In
the ensuing section, we look at the properties of participial adjectives which have propenies
similar to root adjectives and then proceed to examine their predication properties.
4.3
Properties of participial nouns
Participial nouns display the same behavioral properties as the root adjectives as
follows;
i) Participial nouns reduplicate their stems as shown in (18).
(18)
a.
-T»
b.
njogoo
makamak-u
9rooster 9worried-Azer
"a rooster that is worried a little more'
mundu
mu-thomathom-
thiomi
U
1 person i-study-Azer
lOlanguage
•person who has studied languages a little'
ii) Reduplicated participial noun stems disallow class 14 prefixation as seen in (19).
(19)
a.
*u-makamak-u
14-worry-Azer
'more worriedness'
b.
*u-thomathom-u
14-educate-Azer
'more educatedness'
iii) Participial nouns like root adjectives (and unlike root nouns) are modified with
intensifiers as indicated in (20).
(20)
a.
mak-u
uuru
lOworry-Azer
'terribly worried'
terrible
^ R a p p a p o r i (I9K3) makes this proposal for the similarity between sentences and their corresponding NPs. See
also L a c / k d (1993) on Hungarian.
193
b.
ga-thom-u
muno
12-leam-Azer
'very learned'
exceedingly
iv) Participial nouns like root adjectives (and unlike nouns) do not allow preprefixation on
the stems (21).
(21)
a.
*ka-mu-mak-u
13-1 -worried-Azer
'a small worried one'
b.
*tu-a-thom-u
13-2-leam-Azer
'little learned ones'
V )
Like root adjectives, participial nouns can be used referentially. without the head noun as
(22) shows.
(22)
a.
mu-mak-u
1 worr>'-Azer
'worried one'
b.
ga-thom-u
12-leam-Azer
'a small learned one'
There are additional properties that a deverbal attributive word has beyond those
exhibited by root adjectives. Deverbal attributive words have verbal properties as well. In
this section we look at non-finite and finite modificational words
4.3.1
Non-finite participial phrases
Participial nouns can be formed with extended stems. The applicative -/> (23a).
the reciproicai -an in (23b) with unspecified objects in this example, and the combination
194
of the applicative and the reciprocal in (23c) are examples. The causative -iih- is not
permitted in structures involving participial -u.
(23)
a.
mundij
mu-thiTnj -rr-u
andu
mburi
1 person 1 slaughter -A-Azer 2people lOgoat
•person who has slaughtered a goat for people'
b.
bururi
mu-thuk-an-u
3country
3-spoil-R Azer
"a country which is messed up"
c.
andu
a-rakar-an-Fr-u
2person
2-angry-R-A-Azer
•people who are angry at each other'
Participial deverbal adjectives never take inflectional prefixes (these are prefixes marking
tense, aspect, modality, object, and reflexive) as seen in (23'). The future tense marker in
(23'a) makes the expression bad as does the past marker in (23'b). Similarly, the habitual
marker -ag makes (23'c) bad.
(23')
a.
*mundu mu-ga-thunj-rr-u
andu
mburi
I person 1-Fut-slaughter-A-Azer
2peopie lOgoat
•person who will have slaughtered a goat for people'
b.
*bururi
mu-a-thuk-an-u
3country
3-Past-spoil-R Azer
•a country which was messed up'
c.
*andu
2per.son
a-rakar-an-fr-ag-u
2-angry-R-A-Hab-Azer
•people who are angry at each other all the time'
From this section we can infer that one of the ways to distinguish "adjective" phrases from
verb phrases is perhaps by inflectional morphology marking possibilities illustrated by
195
4.3.2
Finite participial phrases
Participial adjectives can also mark tense but the attendant morphology is different.
Thus the notions expressed by (23) can also be done using Vs which are marked for
aspect, as in (24'). These forms bear [Ncl-...-ASPECT-e) morphology. The examples in
(24) illustrate completive (COMPL) in (24a) and (24b).
(24)
a.
andu
a-thiTnj-u
mburi
2 -slaughter -Azer ' Ogoat
'people who have slaughtered goats'
2person
(24')
b.
bQruri
mu-thuuk-u
3person 3-spoil-Azer
"a spoiled country'
a.
andu
ma-thunj-nt-e
2person 2-slaughter-Compl-Azer
•people who have slaughtered goats"
b.
mburi
1 Ogoat
bururi
u-thuuk-iTt-e
3person 3-spoil-Compl-Azer
'a spoiled country'
Unlike the forms bearing [Ncl-...-u] morphology in (23), [Ncl-...-ASPECT-e] ones mark
past as illustrated by (24') which indicates that they make reference to completed activity or
a state that is entered into within the present perfect (24"a) and the pa.st (24"b).
(24")
a.
andu
ma-ra-thnnj-iTt-e
2person 2-Pres-slaughter-Compl-Azer
•people who had slaughtered goats'
b.
bururi
ij-a-thuuk-iTt-e
3person 3-Past-spoil-Perf-Azer
"a country that had spoiled'
mburi
1 Ogoat
The use of concordial markers as prefixes makes it difficult to distinguish attributive forms
in (24') and (24") from declarative expressions. Thus (24') can be read declaratively as
provided in the glosses of (24'").
196
(24"')
a.
andu
ma-thiinj-iit-e
2person
2S-slaughter-Compl-Azer
mburi
1 Ogoat
•people had slaughtered goats'
b.
bururi
Q-thuuk-iTt-e
3person 3-spoil-Compl-Azer
"the country is gone bad'
In the following section, we address the issue of distinguishing attributive from declarative
expressions.
4.3.3
.\Uributive versus declarative phrases
The noun class marker versus subject prefix distinction is not reliable by itself to
distinguish between declarative expressions versus attributive clauses. This is due to the
fact that while classes 1,2. 3. 4. 9. 10. and 11, have special concordial morphology, the
other classes (classes 5. 6. 7, 8, 12, 13, 14. 15. 16. and
17) do not have different
morphology for concordial elements. We can distinguish declarative clauses from
attributive ones by the distribution of the focus particle, by tonal consideration, and also by
relative clause formation.
4.3.3.1
The focus particle
Adjectival phrases cannot take the focus particle as indicated by (25). Under the
declarative expression construal however, these expressions are perfect (25'). The focus
panicle highlights that it is the verb that is the focus of the expression.
(25)
a.
*andu
m~-ma-thiTnj-iTt-e
2person fp-2S-slaughter-Compl-Azer
'people who have slaughtered goats'
b.
(25')
a.
mburi
1 Ogoat
*bijruri nf-u-thuuk-iTt-e
3person 3-spoil-Compl-Azer
"the country that has gone bad'
andu
mbCiri
fp-2S-slaughter-Compl-Azer 1 Ogoat
m'-ma-thnnj-lTt-e
2person
•people have slaughtered goats'
197
b.
4.3.3.2
bururi
nf-Q-thuuk-iTt-e
3person 3-spoiI-Compl-Azer
"the country has gone bad'
Tonal considerations
The prefix of the verb used attributively bears a high tone (26a) while the prefix
of the verb used declaratively (26b) has a low tone . The attributive modifiers of nominal
heads can thus be distinguished from declarative expressions.
(26)
a.
andQ
ma-ra-thiTnj-iTt-e
2person 2-Pres-slaughter-Compl-Azer
•people who had slaughtered goats'
mburi
lOgoat
b.
andu
ma-ra-thiTnj-iTt-e
2person 2-Pres-slaughter-Compl-Azer
•people had slaughtered goats'
mburi
lOgoat
4.3.3.3
Reduced Relative clauses
As mentioned above the other strategy used to distinguish declarative from
attributive clauses is to use the relative word in (27b) for (27a). As expected expressions
such as (27b) where the head noun mundu 'person' is modified with a relative clause can
not be declarative. The other strategy available in the expression of these attributive verb
phrases is the use of the relative word as in (27). Relativization is not part of declarative
expressions.
(27)
a.
b.
andu
ma-thnnj-iTt-e
2person 2S-slaughter-Compl-Azer
•people who have slaughtered goats'
andu
arfa
ma-thiTnj-iTt-e
2person 2Rel 2-slaughter -Azer
•people who have slaughtered goats"
mburi
lOgoat
mburi
lOgoat
The expressions (27a) and (27b) are identical in meaning, indicating that (27a) is it.self a
reduced relative clause . The same is true with the attributive words which we looked at in
198
sections 4.1 and 4.2 above, which can not be inflected for tense or aspect. E.xamples of
[Ncl-.-.-u] pairs of reduced and unreduced relative clauses are as in (16) (repeated below)
and (16') respectively.
(16)
a
njogoo
mak-u
9rooster 9astonish-Azer
"a rooster that is worried"
b.
mundu mu-thom-u
thiomi
I person 1-study -Azer iOlanguage
•people who have studied languages'
c.
(16')
muirftu
mu-twar-u nguku
ndunyu
Igirl
1-tiike-Azer lOchicken 9market
"a girl who has taken chicken to the market'
a.
njogoo
ma
mak-u
9rooster 9Rel
9astonish-Azer
'a rooster that is worried"
b.
mundu un-a
mu-thom-u
thiomi
1 person IRel
1-study-Azer IOlanguage
•people who have studied languages'
c.
muintu iina
mu-twar-u ngukii
Igirl
IRel
l-take-A^er lOchicken
"a girl who has taken chicken to the market"
ndunyii
9market
It is because attributive phrases are reduced relative clauses that we find that even though
root and derived adjectives are marked with noun class morphology, they still occur in the
concordial domain. Relative words bear concordial morphology.
4.3.4
Analysis
As we have pointed out variously in this chapter. VPs can be used attributively in
relation to the head. These VPs are also referential in that they clearly denote individuals
and so the top level predicate has to be nominal. This is a situation that we encountered
with aaentive nouns.
199
How can the mixed properties of [mil-...-uj be captured lexically?
As with agentive nouns (chapter 4 and 5). the category associated with [mu-...-u]
words can not be uniquely determined. They are simultaneously nominal and verbal.
[mu-...-u] word structure is head-final and as with other deverbal nouns, the nominalizing
morpheme is suffixal as in (28) where the verb stem (V'stgrn) combines with a nominalizer
(Nzer) to form a noun stem (Nstem) ^"<1 the noun class (Nd) in turn combines with a
noun stem to form a noun (N). Participial nominalization makes reference to the verb stem
since restricted derivational affixation is possible.
(28)
N
mu
The structure at the phrase level is also similar to that of noun phrases headed by agent
nouns as indicated by (29).
(29)
N(P
r=i
^\
\
T=i
N
VP
[mu-..V..-u]
Like root adjectives, attributive VPs have a modifier function as indicated by the equation
(T m OD) = i
as shown in (30). Also similar to the root adjective's c-s annotations is the
200
equation (TH) = ( i SUBJ) which says that the value of the head of the mother node is the
same as the value of the subject of the VP-"'.
Di
(T
MOD) =X
{TH)=(iSUBJ)
(TH)= ^
miindu
person
muthomu
studied
(TOBJ) = I
DP
thiomi
languages
'a person who has studied languages'
lit. 'a language studied person'
In this lexical specification, the suffix -u introduces the aspect attribute whose value is
COMPLETE action/state into the lexical features of the base stem.
(31)
<SUBJ. OBJ>"
a.
thom-
V:
(TPRED)
= "read
b.
thom-u
N:
(TPRED)
(TASPECT)
= "read
c.
mu-thom-u
N:
(TPRED )
= -read <SUBJ. OBJ>((TSUBJ PRED ) = 'pro')
(TSUBJGEND = 1
(TASPECT)
= COMPL
<SUBJ. OBJ>"
= COMPL
The f-s for (30) is provided in (32). In (32). the SUBJ of the modifier
(MOD) PRED is
identified as being the same as the head. This being the case, the gender marking can never
^ T h e equation ( i AGR) =c (J- A G R ) c a n be omitted because the equation ( T H ) = ( -I SUBJ) has the same
et'fcct. since it' the head of the mother n o d e is identical to the subject of Ihe AP. there is n o way of getting gender
marking that is distinct between the head and the subject of the AP.
201
be non-identical between the head and the SUBJ of the modifier PRED. In (32) the curved
line indicates the controll relation, meaning that the controlling material belongs in both
places at once (Sells, 1985).
(32)
HEAD
PRED
GEND
SUBJ
MOD
'person'
1
(PRED
GEND
'pro' ) ^
1
Qgj PRED 'languages'
GEND 10^
ASPECT
COMPL
PRED 'study < SUBJ. OBJ >'
The lexical rule that is relevant in the formation of participial verbs which are used
attributively in Gfkuyu takes the argument associated with the subject of a verb and pairs it
with the agent-like argument as indicated in (33).
(33)
(T PRED) ='slaughter < (TsUBJ) (TOBJ) >'
Agent
Theme
The interaction of the passive form with other lexical rules is taken to be evidence that
passive is a lexical rule (Bresnan. 1982). We find that all verbs do perform the work that is
done by adjectives in English-type languages. This is because suffixal morphology -u
converts verbs to participial forms which can be used attributively.
202
5.0
Patientive phrases
Patientive adjectives formed with [Ncl-...-e] morphology provide a considerable
amount of Grkuyu adjectives. Only stems which select and subcategorize for theme-like
arguments participate in patientive adjective formation. Consider the sentences in (34) and
contrast them with their respective nominalized counterparts in (35).
From the
constructions in (34) we get the ones in (35) by two processes; one involves the
suppression (Grimshaw (1991) of the agent argument, and the second involves the
promotion of the thematic argument to subject. The process is clearly that involved in
passivization. hence the optional 'by phrase' indicated in (35"). For patientive adjectives to
be possible, the internal argument of the predication has to be the subject as seen in (35).
(34)
(35)
(35')
a.
Njoki
nf-a-karang-a
irio
Njoki
FP-lS-fry-fv
lOfood
"Njoki has fried the food'
b.
Kamau nf-a-tiim-a
mania mucif
Kamau FP-lS-send-fv 61etter 3home
"Kamau has sent letters home'
a.
irio
ngarang-e
lOfood FP-lS-fry-Azer
"fried food'
b.
mania
matum-e
mucif
61etter
6send-Azer 3home
"sent letters'
a.
irio
ngarang-e
ni
lOfood FP-IS -fry -Azer by
"food fried by Njoki'
b
5.1
Njoki
Njoki
marua
matum-e
mucif
nf
Kamau
61etter
6send-Azer
3home
by
Kamau
'letters which have been sent home by Kamau'
Adjectival properties
i) Patientive adjectives reduplicate their stems in the same way that participial ones (14) do.
203
ii) Patientive adjectives cannot be converted to nouns by class 14 prefixation as is shown to
be the case with participial ones in (15).
iii) Patientive adjectives like root (8) and participial (20) ones are
modified with
intensifiers.
iv) Patientive adjectives like participial (21) and root (9) ones (and unlike nouns) do not
allow preprefixation on the stems.
v) Like root (10) and participial adjectives (22). patientive adjectives can be used
referentially. without the head noun.
vi) Like participial attributive clauses (23). patientive adjectives are attributive VPs and do
permit affixation of derivational morphology.
vii) As expected with passivized verbs, object prefixation and reflexivization are not
permitted with patientive nominals. This is because the object has been promoted to the
external argument position, consequently there is no object (hence no object prefixation
would be possible). Similarly, for reflexivization to be possible, the active verb form (and
not the patientive (passive) verb form) must be used.
viii) Like the participial (23'), inflectional affixation relating to tense/aspect is not permitted
in [Ncl-...-e] VPs. However there are parallel expressions which utilize concordial
morphology and the verbal passive -w have only the declarative reading and not the
attributive one. This is in contrast to participial forms as seen in section 4.3.3 where the
distinction between attributive and declarative clauses is subject to other considerations (25)
and (26).
ix) Patientive VPs which are used attributively are reduced relative phrases as the case in
with the participial ones.
204
5.2
Analysis
How can the mixed properties of [mu-.-.-el be captured?
As with [mu-...-u] words, the category associated with [mQ-...-e] words can not be
uniquely determined. They are simultaneously nominal and verbal and bear a
modificationai function. As with other deverbal nouns. [mu-...-e] word structure is head
final and its formation makes reference to the verb stem which explains why derivational
affixation is possible within them.
(36)
N
^tem
The structure at the phrase level is also similar to that of noun phrases headed by agent
nouns as indicated by (37).
(37)
T
N
VP
[mu-..V..-e]
Like root adjectives and participial nouns, patient nouns have a modifier function as
indicated by the equation ( TMOD)
=i
and the equation ( TH ) = (i
SUBJ) in
the c-
structure (38) indicates that the value of the head of the mother node is the same as the
value of the subject of the patient nouns used attributively.
205
(38)
DJ
(T MOD)=I
(TH )=(ISUBJ)
( T h ) = -imburi
^oat
T=i
T=i
N
VP
thiinje
slaughtered
(T(iOBl^j^j ) = i
pp J
(TOBJ) = ^
Kamau
In the lexical specification for thiTnj-e
is provided in (39). we see that the suffix -e
contributes the attribute ASPECT specified with the value PERFECT into the lexical
features of the base stem.
(39)
a.
thilnj-
V;
b.
thiTnj-e
N:
(TPRED)
= "slaughter <SUBJ. 0BJ>"
(TPRED)
= "slaughter <SUBJ. 0BJ>"
= PERF
(T ASPECT)
c.
-thiTnj-e
N;
(TPRED)
((T s UBJPRED )
(T SUBJGEND
(TASPECT)
= "slaughter <(OBLa). SUBJ>"
="pro-)
= 9
=PERF
Patientive attributive words differ from participial ones by the lexical rules that
derive them. The lexical rule for patientive words is such that they resemble the passive
rule in that the rule involves the taking of the argument associated with the object of the
active verb and making it the subject, and optionally assigning the displaced subject to an
206
oblique argument function (or omitting it altogether). This lexical rule is the reason why
intransitive verb forms can not be used to make patientive words; namely intransitives by
definition have no internal argument. The input entry is as in (40a) and the out put as in
(40b).
(40)
a.
(T PRED) ='slaughter<
(TSUBJ)(T 0 BJ)>'
Agent
b.
(T PRED) = 'slaughter < (T
Theme
) (TSUBJ) >'
Agent
Theme
The functional structure for (38) is provided in (41). where the SUBJ of the modifier
(MOD) PRED is the same as the head. Just as in the participial cases, the gender marking
must be identical between the head and the SUBJ of the modifier PRED. The control
relation, between the head and the SUBJ of the modifier is indicated by the curved line in
(41) indicating that the controlling material belongs in both places at once (Sells, 1985).
207
(41)
_
"
PRED
GEND
'goat'
9
CI to I {.PRED
_GEND
'pro' );
1
MOD
PRED 'by<—>'
OBJ "P^ED 'Kamau'
GEND 1
ASPECT
PERF
PRED 'slaughter < SUBJ,
6.0
)>'
Agentive nominalization
Virtually all types of stems participate in this form of nominalization as was shown
in chapter 4. In the syntactic structure we assumed in chapter 3. provided in (4) above, we
showed that agentive deverbal nominals function as heads. In this section we present data
to show that agentive noun phrases can be used attributively. This is possible because
attributive phrases are reduced relative phrases and relative words take noun phrase
complements. In (42). the agentive noun phrase are part of reduced relative clauses.
(42)
a.
mundu mu-thiTnj-i
mburi
1 person 1 slaughter-Nzer 1 Ogoat
'a person who loves goats'
b.
mwana mwend-i
andu
1 child
llove-Azer
2people
"a child who loves people"
208
Evidence that these phrases function like participant nouns and patient nouns is evidenced
by the modificational possibilities below. In this linear position, [mu-...-i] can be modified
by intensifier words as discussed extensively in chapter 4.
[mu-...-i] phrases add information to the head noun in (43).
(T
MOD )=i
(TH)=(iSUBJ)
mundu
person
(TOBJ) = i
DP
mwendi
lover
A
'a person who loves people'
7.0
andu
people
Summary
Throughout sections 3, 4. 5, and 6, we have studied the system of attribution within
the Grkuyu noun phrase. We have provided a uniform treatment of modifier-head
relationship with regards to the c-structure and f-structure. We have shown that root
adjectives, participial attributive nouns, and patient nouns have a modifier function as
indicated by the equation ( TMOD) =
X
and modifier-head agreement which is guaranteed
by the equation (TH) = (-1 SUBJ) annotated in the c-structure indicating that the value of
the head of the mother node is the same as the value of the subject of the attributive word.
Within the functional structure there is a control relation between the head and the SUBJ of
209
the modifier indicating that the controlling material belongs in the Head position and the
subject of modifier position at once (Sells. 1985).
The four types of attributive words differ in their lexical specifications as has been
characterized in their lexical entries in (44a) for root adjectives. (44b) for participial
attributive nouns (44c) for patientive attributive use of nouns, and (44d) for agent nouns.
Root adjectives do not have the attribute feature ASPECT, while participial, patientive. and
agentive nouns used attributively do.
(44)
a.
mutoi
A;
(TPRED)
((TSUBJ FRED)
(tSUBJGEND
b.
mu-thiTnj-u
N:
(TPRED)
= •slaughterer <SUBJ.
((TSUBJ PRED) = 'pro')
(TSUBJ GEND
(TASPECT)
c.
mu-thilnj-e
N:
= "wayward <SUBJ>"
= 'pro")
= 1
(TPRED)
OBJ>'
= 1
=COMPL
= 'slaughterer<SUBJ. OBJ>'
((TSUBJ PRED) = 'PRO')
(TSUBJ GEND = I
(TASPECT)
= PERF
d.
mu-thiTnj-i
N:
( TPRED )
= 'slaughterer <SUBJ, OBJ>'
((TSUBJ PRED) ='PRO')
(TSUBJGEND = 1
(T ASPECT)
=HAB
We have also shown that root adjectives differ from participial and patientive
adjectives in that root adjectives are a very small set of elements which only select their
noun class in word formation processes. Participial and patientive "adjectives" are mixed
category elements which are derived by different lexical rules. TTie participial involves the
association of the subject of the active verb to a participant role, while the patientive is
derived by a lexical rule identical to the passive rule.
210
Proof that participial and patientive words really resemble agent nouns is
discernible from the table in (45). In this table we see that participial and patientive words
have a lot of behavioral properties that are identical to those of verbs, and some that
correspond to those of nouns. The root adjective behaves like a verb in that its stem can be
reduplicated, but for most other properties (it can not be modifiable by another adjective) it
(root adjective) behaves like a root noun in that it is not modified by adverbs, it bears noun
class (not subject prefix) morphology, it does not subcategorize and does not permit
derivational and inflectional affixation. Participial words are simihir in behavior to verbs in
that their stems can be reduplicated, be modified with intensifiers. and they subcategorize.
They are also like root nouns in that they are not modified with adverbs and they bear noun
cla.ss morphology, not subject prefix verbal morphology. Patientive forms also exhibit a
mixture of behavioral properties in that they like verbs can be modified by adverbs, be
reduplicated, subcategorize. and admit derivational morphology. They are also like nouns
in that they bear noun class morphology. Despite their similarity to both verbs and nouns,
root adjectives, participial, and patientive words can not be uniquely identified as belonging
to either N or V categories. There is quite compelling evidence to place them in either
category each of which models only part of the observed facts. Specifically, the fact that
they take noun class marking is strong indication that they may be nouns, yet the presence
of a.spectual marking in the lexical entries of the participial and the patientive is reason
enough to consider participial and patientive forms, verbs. Additionally, the fact that all
attributive words can be utilized without heads is a property that can not be associated with
either category N or V and their (conventional) projections in standard X' theory. Figure
(45) indicates the differences and similarities between the various forms described in the
sections above.
211
(45)
Property
Reduplication
modified with adverbs
modified with intensifiers
modified with adjectives
Noun class morphology
Subject prefix
morphology
subcategorizes
optional head
Derivational suffixes
applicative
reciprocal
causative
Prefixal morphology
object marker
tense marker
aspect marker
Root
Adjective
Participial
[Ncl-...u]
V
V
X
X
Patientive
[Ncl-...-e]
Agentive
[Ncl-...-i]
Verb Noun
V
V
V
V
V
X
X
X
V
V
V
V
V
V
V
V
X
V
V
X
X
X
X
V
X
V
V
V
V
V
V
V
X
X
-
-
V
V
V
V
-
X
V
V
V
-
X
X
V
X
V
-
-
X
X
X
-
X
X
X
V
V
V
X
V
V
V
V
V
v
X
X
X
-
-
-
-
-
-
The category "adjective" consists of a small closed class in Grkuyu. Deverbal words have
many properties in common with verbs (Dixon 1977, Givon, 1984, and Thompson 1988)
Suffixal vowels are the deverbal adjectivalizers, the vowel -u and -e attach to the verb
stems in the lexicon and turns the verbs into attributive elements. In addition -u and -e
suffixation inu-oduce aspectual features to the base stem.
8.0
Previous Studies
Previous studies have covered different areas of the facts presented in this Chapter.
Carstens (1991). studies the structural placement of adjectives in the noun phrase structure.
Sproat (1985) and Kinyalolo (1991) look at the discharging of theta-roles in Chichewa and
Kilega synthetic compounds respectively. Moshi (1992) studies the categorial status of
Kichaga adjectives and Bresnan and Mchombo (1995) describe the properties of adjectives
212
in Chichewa. The paucity of adjectives noted earlier for Grkiiyu by Barlow (1951) has been
widely observed accross Bantu in subsequent studies.
Dixon (1977) notes this scarcity and observes that Bantu languages have a
membership ranging from less than 10 to forty or fifty. Venda (Doke 1954; 166-167) has
about twenty. Bemba has less than twenty. Luganda about thirty six, Swahili about fifty.
Dixon says that for languages such as English and Dyirbal, which have adjectives as a
major class, the semantic content of the class is fairly constant from language to language.
He also says that languages with a limited class of adjectives also show similarities in the
concepts expressed by adjectives. He says that Igbo (Kwa sub-group of Niger-Congo)
with about 8 adjectives (which make four antonym pairs (Welmers and Welmers. 1969;
Welmers, 1973)) is very similar in concept to the genetically unrelated Hausa (a Chadic)
language. Dixon notes that adjectives are 'property concepts' identifiable as dimension, age.
value, human propeny, physical property, and speed.
Whiteley (1960) distinguishes nouns from adjectives by (i) defining nouns as
"independent nominals" and adjectives (including demonstratives, possessives, numerals,
&
emphatics) as "dependent nominals". (ii) Nouns are short series nominals and
adjectives are long series nominals. Whiteley (1960) distinguished nouns from adjectives
by calling nouns "short series nominals" and adjectives "long series nominals". He did
this to distinguish adjectives from other dependent nominals. He argued that dependent
nominals do not belong to any specific grammatical class but depend on the noun which is
the controller of morphological marking.
With respect to Zulu. Cope (1963) defined adjectives by their morphological
marking. To him. dependent adjectives are those bearing the same prefix as the head noun,
and an independent adjective are ones that bears verbal prefix morphology (for other
languages, see Sharman 1963. Guthrie 1967/71). Doke (1967), Dixon (1977) distinguish
213
nouns from verbs by person/number agreement, a lack of tense-aspect marking. Moshi
(1992) finds such characteristics limiting when confronted with Kivunjo-Chaga facts.
Moshi's (1992) study of Kichaga adjectives observes that such notions as
"attributive" and "predicative" are not very useful in Bantu, as they are in Indo-European
languages. This is due to the fact that these terms are defined syntactically. An adjective is
attributive if it is precedes the noun it modifies (as in a poor sultan ) and in the predicate
position if it occurs in the post-verbal position (as in the sultan was poor ). Further, she
observes that while it is the case that adjectives in Indo-European languages can head
phra.ses (Bolinger 1967; Siegel. 1980; Croft. 1986) and even (in a broader sense) include
determiners, quantifiers, some verbs, prepositional phrases, relative clauses (Quirk et al.
1985. Lyons 1968. 1977), the situation in Bantu is rather different. Moshi finds that the
criteria (largely based on English) are problematic because Kivunjo-Chaga adjectives fit
both the broad and narrow sense. Moshi, following Whiteley (1960), considers a noun to
be an independent nominal depending on whether it belongs to a specific noun class; it
cannot appear without the relevant class prefix. Adjectives on the other hand may appear in
a variety of noun classes. With regards to the morphological marking (Cope (1963)
Kivunjo-Chaga does have a distinction between independent adjectives (46) and dependent
ones (47). In both these cases, the adjectives seem to select their class prefix (whereas
nouns are restricted to belong to one class).
Moshi (1992:116)
(46)
m-ndu m-leshi
m-ana m-ca
m-fi -mtutu
ma-imba-ma-ili
n-ginda ngi-tutu
(47)
m-ndu a-shimbi
m-fi u-ganyi
ma-imba gha-ganyi
n-ginda i-shimbi
tall person
good child
small arrow
white com
small banana tree
a fat person
a big arrow
big ears of com
thick banana tree
(class
(class
(class
(class
(class
1)
1)
3)
6)
9)
(class 1)
(class 3)
(class 6)
(class 9)
214
Moshi also says that Kivunjo-Chaga adjectives are in conformity to the semantic classes
suggested by Dixon (1977) which are mentioned above.
8.1
Application to Griiuyu
Adjectives in GrkOyii do not fall into the categories mentioned by Dixon (1977),
rather, the bulk of them (participial and patientive types) are based on aspectual notions
STATIVE (root adjectives). COMPLETE (participial nouns), PERFECT (patientive
nouns), and HABITUAL (agentive nouns). Dixon's (1977) proposal that u set of'propeny
concepts' identifiable as dimension, age, value, human property, physical property, and
speed characterize the semantics of adjectives can only describe root adjectives in Grkuyu.
Whiteley's (1960) hypothesis of dependent and independent nominals does describe the
facts studied here as well as the morphological variation that Moshi (1992) provides from
Kivunjo-Chaga.
GrkCiyu makes up for the paucity of adjectives by utilizing verb phrases
attributively. Also unlike these previous studies we have proposed that adjectives and
attributive VPs are predicates whose subcategorization depends on the transitivity of the
stem. Additionally, adjectives and attributive VPs are reduced relative phrases and occur in
the concordial domain of the noun phrase.
9.0
Solutions
The solutions to the problems posed at the beginning of this chapter have been
encountered throughout the sections above. The first issue concerns the correlation between
morphological shape of the elements and their constituent structure placement. The second
issue pertains to how attributive phrases fit in the structure proposed in chapter 2. The third
issue is on category type and the fourth one looks at the typology of adjectives and
attribution across Bantu.
215
9.1
Issue I
The morphological distinction between noun class bearing elements and concord
marked ones is best represented by drawing a distinction between the concordial and the
lexical domain, with elements marked with noun class morphology occurring in the lexical
domain and those marked with subject prefix morphology occurring within the concordial
domain. Adjectives and verb phrases used attributively appear to contradict the domain
split because they bear noun class morphology yet they occur within the concordial
domain. We have suggested that this is due to the fact that all attributive phrases are
reduced relative clauses, and relative words which head relative phrases bear concordial
morphology. This leads directly to the second issue.
9.2
Issue 11
With respect to structure-function association, we have seen that attributive words
attach to the
DP and that they are annotated with the equation (TMOD) = i . We have also
proposed that for all adjectives and verb phrases used attributively, the value of the head of
the mother node is the same as the value of the subject of the attributive word as indicated
by the equation (TH) = (i SUBJ) annotated in the c-structure. Within the functional
structure there is a control relation between the head and the SUBJ of the modifier
indicating that the controlling material belongs in the Head position and the subject of
modifier position at once. With the equation (TH) = (i SUBJ) it follows the AGR value
for the head is identical to the AGR value of the subject of the adjective or attributively used
verb.
9.3
Issue III
We have noted that even though attributive words (root adjectives, participial, and
patientive words) have some semblance with both verbs and nouns, they are not easily
identified as belonging to either N or V categories. We have provided evidence which
216
permits their placement in both category N and V) each of which models only part of the
distributional and behavioral properties. Like nouns, they take noun class marking yet the
presence of aspectual marking in the lexical entries of the participial and the patientive is
equally compelling and does suggest that participial and patientive forms are verbs. We
have therefore concluded that attributive words with verbal bases are like agent nouns (in
chapter 4). Their mixed category status of participant and patient words can also be
explained by the notion of extended heads that is introduced in chapter 4 (in the case of
agent nouns).
9.4
Issue IV
As we have observed in section 8.0. other Bantu languages resemble Grkuyu in that
adjectives are marked by noun class morphology and that like nouns, adjectives can appear
in any noun class (Dixon, (1977), Whiteley (1960), Moshi (1992). Just as noted by Cope
(1963) with respect to Zulu, we do have a class of adjectives which bears subject prefix
morphology and another class that bears noun class marking. In Grkuyu. non-fmite
attributive verb phrases are marked with noun class morphology, while non-fmite verb
phrases take concordial morphology.
The semantic criteria proposed by Dixon (1977) may work to classify root
adjectives but attributive verb phrases are subject to semantic criteria that are based on the
aspectual features introduced to the base semantics of the verb by lexical rules. One lexical
rule maps the subject of the verb to the participant role and the other one maps the object
argument of the base verb to the patient role.
10.0
Conclusions
In this chapter, we have proposed that root adjectives and deverbal nouns used
attributively are part of the system of relative phrases. They have a MODIFIER function
within the f-s of noun phrase. We have seen that attributive words have similarities to both
217
nouns and verbs. We have seen that the 'paucity' of adjectives referred to by Barlow (1951)
is compensated for in Grkuyu by the attributive use of nominalized verbs. This innovation
in the grammar of the language is marked by the fact that verbs used attributively take noun
class prefix morphology. Despite the fact that attributive words are marked with noun
class morphology (which was proposed to occur in the lexical domain), they occur in the
concordial domain where concordial morphology is expected. We have argued, in response
to this apparent contradiction, that all the attributive clauses are reduced relative clauses.
Relative words which head relative clauses bear concordial marking.
218
CHAPTER 7
CONCLUSION
1.0
Introduction
This dissertation has examined Bantu nominalizations drawing evidence primarily
from GrkQyu. a Bantu language which has a very rich system of deverbal nouns. I have
shown that nominalized verbs in Grkuyu bear both noun morphology (noun class
marking), and verbal inflections and extensions. Deverbal nouns display a set of behavioral
and distributional properties unique to them, and by virtue of the properties some general
statements can be made about them. Deverbal nouns are not idiosyncratic elements in
Grkuyu and Bantu grammar. Word and phrase composition of these deverbal elements
brings to the fore the saliency of the lexical integrity of words contradistinguished from
syntactic processes of phrase formation. In terms of their sub-lexical make-up, root nouns
take affixes which encode gender and number information within the noun phrase. Root
verbs also bear concordial morphology and pronominal marking in addition to being hosts
of inflectional and derivational affixation. I have shown that deverbal nouns may be
characterized as the intersection of a set of root N properties with a set of root V properties.
This composition of properties is maintained from the sub-lexical level to the phrasal level.
2.0
The issues
Throughout this study. I have sought to answer the following general questions:
(1)
a.
How do we explain the category split exhibited by
infmitive/gerunds?
b.
How are nouns distinguished from verbs in Bantu?
c.
How can the mixed properties of [mu-...-i]. [mu-...-u] and
[mtJ-.-.-e] be characterized so as to explain their N and V
behavior?
219
d.
What are the implications of the spectrum of agentive
nominals to the lexical integrity of words?
e.
What is the difference between a mixed category and a
synthetic compounds
f.
What are the countervailing mechanisms for the
impoverished root adjective systems in Bantu?
g.
How do we map information from affixes to functions
without depending on category and structure?
In seeking answers to these questions. I have introduced tests as premises for the
proposals made. I have looked at morphological issues, infinitive/gerunds, category type
and word structure, phrasal representation, the lexicon-syntax divide and make-up
strategies for the scarce adjective items. In what follows, I provide a summary of these
issues.
2.1
Tests
In this study. I have introduced various tests for every claim. Each claim and
proposal that I make is buttressed by distributional and behavioral evidence of category
type, category level, wordhood and phrasehood. The results are that pure categories have
the properties listed in (2) and (3). and mixed categories have an intersection of properties
in (2) and (3) as provided in (4). From these empirical tests, it is apparent that
infinitive/gerunds (gerund [-t-NPobj]
gerund [-NPobj]) are ambiguous between
category N and V (a virtual disjunction of N and V properties) and that mixed categories
have an intersection of N/V properties.
(2)
Pure verbs and gerund [+NPobj]
a.
permit extraction as thematic objects
b.
allow object prefixation
220
(3)
(3)
c.
permii free word order without object prefix
d.
disallow relativization
e.
cannot be nf-clefted
f
disallow use of associative phrases
g.
can be formed with auxiliaries
h.
permit negation
i.
very restricted use of tense and aspect
j.
take verbal modification (adverbs, intensifiers. emphatic interjections)
Pure nouns and gerund [-NPobj]
a.
do not have thematic objects
b.
disallow object prefixation
c.
permit free word order only with prefix
d.
permit relativization
e.
can undergo nr-clefting
f
permit use of associative phrases
g.
restricted from semantically empty verbs
h.
disallow negation
i.
tense and aspect disallowed
j.
take nominal modification (adjectives, determiners, possesive pronouns etc.)
Mixed categories
a.
permit extraction as thematic objects
b.
allow object prefixation
c.
do not permit free word order without object prefix
d.
allow relativization
e.
can be nr-clefted
f.
allow use of associative phrases
g.
can not be formed with auxiliaries
h.
do not permit negation
i.
no tense/aspect/aspect marking is permitted
j.
take verbal modification (adverbs, intensifiers, emphatic interjections)
j". take nominal modification (adjectives, determiners, possesive pronouns etc.)
2.2
The morphology
Noun class morphology which is found on nouns and adjectives has been the
subject of a number of studies. The literature (Mufwene. 1980; Myers. 1987; Kinyalolo.
1991; Bresnan and Mchombo, 1995) offers conflicting proposals on the status noun class
marker. In this study I have maintained the Bresnan and Mchombo (1995) proposal that
noun class markers are sub-lexical elements. I have also proposed that the noun class
marker combines only with the noun stem to make nouns. Adjectives bear noun class
marking because they are nominalization. Where the noun class marker appears to be
attached to the verb root. I have pointed out that the suffixal material is the nominalized
element, thereby upholding conventional wisdom that the category determining
morphology for words is derivational and suffixal.
In contrast to the noun and adjective, the morphology of modifiers (demonstratives.
a.s.sociatives. and quantifiers) is concordial and is identical to that found on verbs (familiarly
called the subject prefix within verbs). Concord morphology may contribute agreement or
pronominal information in the functional-structure (Bresnan and Mchombo. 1987). Verb
stems permit infiectional and derivational affixation and have much richer morphology
222
than root nouns. It is because deverbal nouns literally combine the features of both nouns
and verbs that they display N/V behavior throughout the grammar.
2.3
Inflnitive/gerunds
With regards to infinitive/gerunds studied in Chapter 2. I have assembled their
properties and observed two types: gerund [+NPobj]
between gerund [+NPobj] and gerund [-NPobj]
gerund [-NPobj]- The distinction
ihe same as that conventionally known
to exist between V and N. The best way to capture the distinction in behavior and
distribution of deverbal words is to recognize that there is a root level and a stem level. It is
at the sub-lexical level that the two types of infinitive/gerunds may be distinguished. The
Bantu verb stem is the locus of derivational and inflectional affixation (Mchombo. 1993;
1995). The gerund [+NPobj] type is verbal hence it permits inflectional and derivational
affixation among other properties. Gerund [-NPobj] words are formed by the
nominalization of the verb root. The verb root can not be inflected or extended by affixation
which explains why gerund [-NPobj] types exhibit no verbal properties. The point at
which nominalization occurs in the sub-lexicai level (either at the verb root or the verb
stem) explains the categorial split exhibited by inflnitve/gerunds.
The existence of infinitive/gerunds which are nouns finds independent support
from class 14 words denoting abstract concepts. Class 14 words such as (5) are primarily
verb based. In chaper 2, it has been shown that class 14 words do not include the thematic
objects as illustrated by (6).
(5)
u-tungat-i
14-serve-Nzer
'service'
(6)
*u-tungat-i
kanitha
14-serve9church
'church service'
223
The prediction that can be made based on evidence from infinitive/gerunds is that class 14
words involve the nominalization of the verb root. Thus permitting (5) and disallowing (6)
predicts that u-words will exhibit the patterns observed with gerund [-NPobj] types and not
that of gerund [+NPobj] ones. This expectation pans out because all u-words behave like
root. They can. as a consequence, be modified with the typical nominal modifiers, such as
adjectives (7) and associative phrases (8).
(7)
Q-tungat-i
mwega
14-serve3good
'good service'
(8)
u-tungat-i
wa
i 4-serve3 Assoc
'church service'
kanitha
9church
Unlike verbs u-words are neither modified with adverbials (9) nor can their verb bases be
extended by derivational affixation as illustrated by causativization attempted in (10).
(9)
*u-tungati
wega
14-serve
well
'servicing well'
(10)
*u-tungat-ith-i-a
14-serve-C-C-fv
'cause to cause to serve'
The analysis in chapter 2 indicates that the noun class marker combines with a noun stem
to make nouns. Noun class markers are not the category determining elements of the
words they are part of. but rather the elements which encode information about gender
class and number (Bresnan and Mchombo, 1995). The category determining material in
Grkuyu word structure (and Bantu generally) is suffixal.
2.4
Category and word structure types
There is order in the way deverbal nouns behave. Root nouns and nominalized verb
roots have identical properties. Root verbs and gerund [-nNPobj]
have identical
224
distribution, while nominalized verb stems combine the behavior of root nouns and root
verbs.
2.4.1
Root nouns
Root nouns differ from the verbal root nominalizations only in terms of the sub-
lexical structure. Root nouns such as -ndu 'thing' are made of the combination of root
nouns with noun class markers as indicated by the examples in (11).
(11)
a.
kij-ndu
17-thing
•place' (general)
b.
ij-ndu
14-thing
"a matter'
c.
mu-ndu
17-thing
•person'
The structure of words such as those in (11) is as provided in (12). where the NRoot
combines with noun class markers (Ncl) to form root nouns.
(12)
N
Clearly, the element that determines the category of root nouns is the invariant NRoot ^irid
not the varying noun class marker.
2.4.2
Nominalized verb roots
Nominalized verb roots include gerund [-NPobj] types (13). class 14 words
(abstract concepts with u- noun class marker) such as (14) and the totally lexicalized [mG-
225
...-a] items such as (15). These examples are not manipulable by affixation beyond the Ncl
and the nominalizer (Nzer).
(13)
kuna
15eat
"to eat/eating"
(14)
u-tungat-i
14-serve-Nzer
'service'
(15)
mO-kungambura
'rainbow'
Words such as (13). (14). and (15) have the word structure given in (16). where the VRoot
is nominalized by a suffix and the Ncl combines with the result to form a noun.
(16)
N
stem
As pointed out in chapter 2. gerund [+NPobj] such as (17) have the behavior and
distribution of verb phrases such as (18).
(17)
kQna
mTanga
15eat
4cassava
"to eat cassave ' eating cassava
(18)
mianga
3sgS-Pres-eat-fv 4ca.ssava
"she/he is eatins cassava"
a-ra-rf-a
226
Both the head of gerund [+NPobj] and the pure verb have a sub-lexical structure as shown
in (19) where the verb stem combines with the final vowel (typically -a ) to form an
extended verb stem. Verb stems combine with subject prefixes to form root verbs. The
verb stem is the locus of all inflections and extensions.
(19 )
V
Subject Prefix
gu-
stem
stem
final vowel
:cx)t
2.4.3
Nominalized verb stems
Mixed categories studied here are all verb based and include agentive nominals
[mu-...-i] (20). participial nouns [mQ-...-u] (21). and patientive nouns [mu-...-e] (22).
(20)
mu-thiTnj-i
mburi
1 slaughter-Nzer 1 Ogoat
"slaughterer of goats"
(21)
njogoo
mak-u
Qrooster 9worry-Nzer
'a rooster that is worried'
(22)
mburi
thnnj-e
1 Ogoat
lOslaughter-Nzer
o
c
"a goat which has been slaughtered"
Mixed category items simultaneously combine nominal and verbal attributes. They exhibit
the intersection of N and V behavioral and distributional properties. Their sub-lexical
structure (23) explains why this is so. It is because agentive. patientive. and participial
227
deverbal nouns involve the nominalization of the verb stem, that the nixed N/V properties
are manifest from the sub-lexical level to the phrase level. The structure (23) is the
nominalization of that given in (19)
(23)
N
-Stem
2.5
Phrasal structure
It is apparent that nominalization has a sub-lexical source, and the depth of
nominalization in the sub-lexical level explains the split in category of infinitive/gerunds
and the mixed category status of agentive, patientive, and participial nouns. At the phrase
level, infinitive/gerunds with complements have structures identical to those of verb
phrases, while those without complements have structures resembling those of root nouns.
More interestingly however, are the mixed category items. Mixed N/V items are best
represented using of the idea of Extended Heads proposed in Grimshaw (1991) and
Bresnan (1996) illustrated by (24).
(24)
DP
228
With (24). the mixed properties of [mu-...-i], [mu-...-u]. and [mu-...-e] words listed in
(4) can be explained as elaborated in chapter 4.
The analysis of mixed category items suggests that there is concordial domain and a
lexical one. The concordial domain consists of nominal modifiers which are close-class items
such as demonstratives, pronouns and quantifiers. The lexical domain is the locus of
projections of open-class items such as nouns, verbs, and adverbs. The two domains are
illustrated by (25) as elaborated in chapter 4.
(25)
DP
lexical elements
2,6
concordial elements
The Lexicon-Syntax Divide
Grkijyu ha.s two types of agentive nominalizations exemplified by (26) and (27).
(26)
mu-end -a
andu
1 ACT love-Nzer
2people
'a person who loves people'
(27)
mu-end -i
andu
1 ACT love-Nzer 2people
'a person who loves people'
Agentive nouns such as (26) and (27) have invited different analyses from different
linguistic frameworks exemplified by Myers (1987). Sproat (1985). Kinyalolo (1991),
Bresnan and Mchombo (1995). and Mugane (1995). In part, this is because these derived
nouns display a number opposing properties such as their verbal and nominal behavior, as
229
well as word and phrasal status. Basically, there are three main issues: The first issue is
with regards to the identity of the nominalizing morpheme vis a viz the status of the noun
class marker. Some studies such as Myers (1987). Sproat (1985),
propose that
nominalization is prefixal and phrasal. Mugane (1995) proposes that nominalization is
suffixal and lexical. Bresnan and Mchombo (1995) argue that with the exception of the
locative classes, the noun class marker is a lexical element manipulable only by word
formation properties. Clearly, there is no agreement on the status of the class marker and
the identity of the nominalizer. The second issue relates to the lexical/phrasal ambiguity
displayed by agent nouns in Bantu.
In Kilega (Kinyalolo. 1991) agent nouns with
complements are phrasal, while in Chichewa (Bresnan and Mchombo. 1995) they are
products of lexical formation. In Gilcuyu (Mugane. 1995 and chapter 6) they display lexical
as well as syntactical properties. There is need to reconcile the different proposals. Given
the two foregoing issues, there arises a third one relating to the lexical integrity of words.
Given the behavior of agent nouns, it appears that the lexical integrity of words is a violable
principle in the theory of grammar. This is so because, by stipulating that there are word
formation processes that are subject to rules which are distinct from those of phrase
formation (Bresnan and Mchombo 1995). agentive nouns fail to fit on either side of the
divide.
In chapter 5. 1 have argued that word structure is affix-final while phrase structure
is head initial for Grkuyu. It is also shown that there is a descriptive generalization to be
gained by considering the nominalizing material as suffixal and the class marker as a
lexical element which is manipulable only by word formation processes. It is evident that
there is lexical/phrasal ambiguity exhibited by the fact that while Grkijyu agentive deverbal
nouns may be products of phrasal formation (Kinyalolo (1991) for Kilega) they also
exhibit properties which are typical of products of lexical formation (Bresnan &Mchombo
230
(1995) for Chichewa). Faced by the dilemma that neither a purely lexical account nor a
purely syntactic analysis can explain all the facts in Grkuyu agent nouns. I have proposed
that the lexicon-syntax divide must be construed as being permeable and that that
permeability has a diachronic explanation at least for the Gfkuyu facts. Given that there are
lexical integrity effects which are clearly manifest, I have concluded that the principle of
lexical integrity is needed in grammatical theory. With regards to the forms in (26) and
(27). it is concluded that [mu-...-a] types are lexicalized as well as lexicalizing forms while
[mu-...-i] ones are robustly phrasal.
2.7
Adjectives
According to Barlow (1951). there is a paucity of true adjectives in Grkuyu. This
shortage of adjectives has been well documented across Bantu (Doke. 1954; Dixon. 1977;
Moshi, 1992; Bresnan and Mchombo. 1995; etc.). Dixon (1977) finds that Bantu
languages have a membership ranging between less than ten to fifty adjectives. Dixon
observes that these adjectives express property concepts which include dimension, age,
value, human property, physical property, and speed. By these criteria, it is likely that
Grkiiyii has fewer than ten true adjectives. Previous studies have not addressed the issue of
how Bantu languages make up for the scarcity of true adjectives. Does the paucity of
adjectives conversely curtail attribution in Bantu languages? In answer to this question,
chapter 6 shows that mixed N/V items (Mugane, 1996) are used as attributive modifiers of
nouns. I have shown that modes of attribution in Bantu languages have to be studied
without parochial regards to languages which have adjectives as a major class. As Moshi
(1992) says, the Indo-European based terms 'attributive' and 'predicative' are syntactically
defined. Adjectives are called 'attributive' when they occur pre-nominally as in 'a rich man'
and 'predicative' when they occur post-verbally as in 'the man is rich'. Bantu adjectives are
post-nominal, and. as chaper 6 shows, a large number of words functioning as adjectives
231
are nominalized verbs. In analyzing the facts, 1 have propose that the Lexical-Functional
Grammar (LFG) framework of Bresnan (1982; 1995) which allows for the mapping of
linguistic information on parallel levels, provides an elegant mechanism for explaining
Bantu attributive elements. With LFG. it is possible to associate information borne by
morphological components to functions independent of category and constituent-structure.
This being the case, it is possible to associate mixed N/V elements to the modifier function
in the functional-structure without requiring that there exist a category "adjective" and
syntactically defined constituent-structure positions. Thus the impoverished adjective
system in Grkuyu (and in Bantu) has the countervailing mechanism of using verbal
nominalizations attributively.
232
REFERENCES
Abney. S. (1987) The English Noun Phrase in its Sentential Aspectt. Ph.D. diss.
MIT. Cambridge.
Alsina, A. (1994) "A Theory of Complex Predicates: Evidence from Causatives
in Bantu and Romance.' To appear in Complex Predicates, ed. by
Alsina, Bresnan. and Sells. CSLI Publications.
Alsina. A. (1993) Predicate Composition: A Theory of Syntactic Function Alternations.
Ph.D dissertation. Stanford. Stanford University.
Anderson. M. (1979) Noun Phrase Structure. PhD dissertation. University
of Connecticut.
Andrews. A. and Christopher Manning (1993) Information Spreading and Levels of
Representation in LFG. Report No. CSLI-93-176.
Armstrong, L. (1967) The Phonetic and Tonal Structure of Kikuyu. London:
Dawsons of Pall Mall.
Archangeli, D. (1988) 'Aspects of Underspecification Theory' Phonology 5. pp. 183-207
Aronoff. M. (1976) Word Formation in Generative Grammar.MIT Press. Cambridge.
Mass.
Ashton. E.O (1947) Sn ahili Grammar (including Intonation). London; Longmans.
Bach. E. (1981) "On Time Tense and Aspect: An Essay on English Metaphysics". In:
Peter Cole (ed.). Radical Pragmatics, New York: Academic Press, pp. 62-81.
Baker. M. (1988) Incorporation. A Theory of Grammatical Function Changing.
Chicago; Univ. of Chicago Pre.s.s.
Barlow. A.H. (1951) Studies in Kikuyu Grammar and Idiom. Edinburgh:
William Blackwood and Sons Ltd.
233
Barss. A. and Howard Lasnik (1986) 'A note on Anaphora and Double Objects'
LI 17:347-354
Benson. T.G. (1964) Kikuyu English Dictionary. London: Oxford University
Press.
Bergvall. V.L. (1986) "A Typology of Empty Categories for Kikuyu and Swahili.'
In Gerrit J. Gimmendaal (ed.) Current Approaches to African Linguistics.
Vol. 3 No.6 Foris Publications. Dordrecht. Holland.
Bergvall. V.L. (1987) "The Position and Properties of In Situ and Right-Moved Questions
in Kikuyu.' In: David Odden (ed) Current Approaches to African Linguistics.
Foris Publications. Dordrecht. Holland.
Bleek. W.H.I (1862) A Comparative Grammar of South African Languages.
Trubner. and Co.. London.
Bolinger. D. L (1967) "Adjectives in English: Attribution and Predication." in:
Lingua 18:1-34.
Bresnan. J. (1982) The Mental Representation of Grammatical Relations.
Cambridge. Massachusetts: The MIT Press.
Bresnan. J. (1994a) "Category Mismatches', paper presented to the plenary Session
of the 25th Annual Conference on African Linguistics. Rutgers University.
March 26. 1994.
Bresnan. J. (1994b) "Locative Inversion and the Architecture of Universal Grammar'.
Language 70.
Bresnan. J. (1995) Lecture Notes on Lexical-Functional Grammar. Department
of Linguistics. Stanford University.
Bresnan. J. (1996) Optimal Syntax: Notes on Projection, Heads, and Optimality.
ms.Stanford. CA: Stanford University Department of Linguistics.
234
Bresnan. J. and Kanerva. J.M. (1989) "Locative Inversion in Chichewa; A Case Study of
Factorization in Grammar". Linguistic Inquiry 20. pp. 1 -50.
Bresnan, J. and S.A. Mciiombo (1987) Topic Pronoun, and Agreement in
Chichewa.' Language 63.741-782.
Bresnan. J., and S. A. Mchombo (1995) 'The Lexical Integrity Principle: Evidence
From Bantu.' Natural Language and Linguistic Theory.
Bresnan, J. and Moshi. L. (1990) 'Object Asymmetries in Comparative Bantu Syntax'
Linguistic Inquiry 21:147-85
Carstens, V.M.( 1991 )The Morphology and Syntax of Determiner Phrases in
Kisn ahili. Ph.D. diss. University of California, Los Angeles.
Carstens, V. and Kinyalolo, K.K.W. (1991) On /P Structure: Tense, Aspect, and
Agreement., ms., UCLA/Comeil University.
Chao, W. (1987) On Ellipsis. Ph.D. diss. University of Massachusetts, Amherst.
Chomsky, N. (1970) "Remarks on Nominalization.' In R. Jacobs and P. Rosenbaum,
eds.. Readings in English Transformation Grammar. Waltham. MA: Ginn.
Cinque. G. (1980) "On Extraction from NP in Italian', Journal of Italian Linguistics. 5,
47-99.
Clark. E.V. and Herbert H. Clark (1979) "When Nouns Surface as Verbs' Language 55.
pp.767-811
Clements, G.N (1986) "Compensatory Lengthening and Consonant Gemination in
Luganda.' in Wetzels & Sezer (eds.)Smi//e5 in Compensatory Lengthening.
Dordrecht: Foris Publications.
Clements. G.N (1985) "The Geometry of Phonological Features.' In: Phonology yearbook
2. C Ewen and J. Anderson. London: Cambridge University Pre.ss.
235
Contini-Morava. E. (1991) 'Deictic Explicitness and Event Continuity in Swahili
Discourse."
Ma 83.277-318
Cope. A.T (1963) •Nominal-Verbal Constructions in Zulu." In: African Language Studies
4: 69-97.
Corbett. G. G. and Alfred D. Mtenje (1987) "Gender Agreement in Chichewa.'
Studies in African Linguistics 18. 1-38.
Croft. W.A (1986) Categories and Relations in Syntax: The Clause Level Organization
of Information. Ph.D dissertation. Stanford, Stanford University.
Culy. C.( 1990) The Syntax and Semantics of Internally Headed Relative
Clauses. Ph.D dissertation. Stanford, Stanford University.
Dalrymple. M. (1990) Syntactic Constraints on Anaphoric Binding. Ph.D dissertation.
Stanford, Stanford University.
Demuth. K. and J. Gruber. (1994) "Constraining XP sequeences." Paper presented at
the 25th ACAL, Rutgers University, NewBrunswick, NJ.
Dixon, R.M.W (1977) "Where have all the adjectives gone?" In: Studies in Language
1.1; 1-80.
Doke, C.M. (1967) The Southern Bantu Languages. London: International
African Institute.
Doke, C.M. (1954) The Southern Bantu Languages. Oxford University Press, London.
Dowty. D. (1991) "Thematic Proto-Roles and Argument Selection." Language
67; 547-619.
Foley, W. and Robert van Valin (1984) Functional Synatx and Universal Grammar.
Cambridge: Cambridge University Press.
Fortune. G. (1955) An Analytical Grammar of Shona. Longmans, London.
236
Fortune. G. (1985) Shona Grammatical Constructions: Parts 1 and 2. 3rd Edition.
Mercury Press. Harare.
Given, T. (1984) Syntax: A Functional-Typological Introduction. Benjamins. Amsterdam.
Gregersen. E. (1967) 'Prefix and Pronoun in Bantu." UAL 33.3 Part II Supplement
Grimshaw. J. (1990) Argument Structure. MIT Press. Cambridge Massachusetts.
Grimshaw. J. (1991) Extended Projections, ms. Department of Linguistics and Center for
Cognitive Science. Rutgers University.
Guthrie. M. (1948) The Classification of Bantu Languages. Oxford Univ. Press.
London.
Guthrie, M. (1967) The Classification of Bantu Languages, London; Dawsons of Pall
Mall.
Guthrie, M. (1967-1971) Comparative Bantu. Vol. 1-4. London. Gregg International
Publishers Ltd.
Halvorsen, P.K (1983) 'Semantics for Lexical-Functional Grammar." Linguistic Inquiry
14:567-615.
Halvorsen. P.K. and R.M. Kaplan (1988) "Projection and Semantics Description in
Lexical-Functional Grammar." In: Proceedings of the International Conference
on Fifth Generation Computer Systems. FGCS 88; 11 16-1122. Tokyo; Institute
of New Generation Systems.
Heine. B. and Mechthild Reh (1984) Granimaticalization and Reanalysis in African
Languages. Hamburg: Helmut Buske Verlag.
Hellan. L. (1986) 'The Headedness of NPs in Norwegian." In Features and
Projections. P. Muysken & H. Van Riemsijk. eds.. Foris Publications.
Dordrecht. Holland.
237
Herbert. R. (1977) "Phonetic Analysis in Phonological Description Prenasalized
consonants and Meinhofs Rule.' Lingua 43:339-373.
Herbert. R. (1978) "Morphological Re-anaiysis in the Bantu Nasal Class.' African
Studies. 51:123-157.
Hyman, L. and Alessandro Duranti (1982) "On the Object Relation in Bantu."
in P. Hopper and S. Thompson, eds.. Syntax and Semantics 15:
Studies in Transitivity. New York: Academic Pres.s. 217-239.
Jackendoff. R. (1977) X' Syntax: A Study of Phrase Structure. Cambridge.
Massachusetts: The MIT Press.
Jackendoff. R. (1977) "Constraints on Phrase Structure Rules." In Formal Syntax.
Peter W. Culicover. Thomas Wasow. and Adrian Akmajian.
Johnson. M. (1979) "The Natural History of Meinhofs Law in Bantu."
Studies in African Linguistics 10, 261-272.
Kaplan. R.M and Joan Bresnan (1982) "Lexical-Functional Grammar: A Formal System
for Grammatical Representation ." In: Joan Bresnan {&d.)The Mental Representation
of Grammatical Relations. Cambridge, Massachusetts: TTie MIT Press.
Keyser. S.J. and Roeper. T. (1984) "On the Middle and Ergative Constructions in
English'. Linguistic Inquiry, 15, pp.499-505.
Kimenyi. A. (1980) ARelational Grammar of Kinyarwanda. Berkeley: University of
California Press.
Kinyalolo. K. (1991) Syntactic Dependencies and the SPEC-head Agreement
Hypothesis in Kilega. Ph.D. diss. University of California. Los Angeles.
King. T. (1992) Configuring Topic and Focus in Russian. Doctoral dissertation
Stanford university.
238
Kiparsky. P. (1987) Morphology and Grammatical Relations. Manuscript. Department
of Linguistics, Stanford University.
Koster. J. (1978) "Why Subject Sentences Don't Exist, in S.J. Keyser (ed.). Recent
Transformational Studies in European Languages, MIT Press. Cambridge MA,
53-64.
Kratzer, A. (1988) Stage-Level and Individual-Level Predicates, ms. University of
Massachusetts. Amherst, MA.
Kroeger, P. (1991) Phrase Structure and Grammatical Relations in Tagalog.
Doctoral dissertation. Stanford University.
Laczko. T. (1993) Predicates in the Hungarian Noun Phrase. Debrecen.
Lapointe. S. (1990) "Dual Projecting Categories and the Syntax of Mixed Category
Phrases.' ESCOL Proceedings. Cornell University.
Larson. R. (1988) "On the Double Object Construction.' LI 19.3:335-392.
Leakey. L.S.B. (1959) First Lessons in Kikuyu. Nairobi: The Eagle Press.
Lieber. R. (1992) Deconstructing Morphology; Word Formation in Syntactic Theory.
The University of Chicago Press. Chicago.
Lobeck A. (1985) Syntactic Constraints on VP Ellipsis. Ph.D. diss.
University of Washington.
Lyons, (1968) Introduction to Theoretical Linguistics. London and New York:
Cambridge University Press.
Lyons, (1977) Semantics. Volume 2. Cambridge: Cambridge University Press.
Marantz. A. (1984) On the Nature of Grammatical Relations. Cambridge, MIT Press.
239
Marantz. A. (1990) Implications of Asvnimctries in Double Object Constructions.
ms.. UNH-Chapel Hill.
Masunaga. K. (1983) 'The Applied Suffix in Kikuyu." In: J. Kaye and Koopman
(eds.) Current Approaches to African Linguistics. 2 .5, pp. 283-291.
Mbaabu, I. (1985) Sarufi ya Kiswahili. Nairobi: Kanya Publishing and Book
Marketing Co. Ltd.
Mchombo. S.A. (1978) A Critical Appraisal of the Place of Derivational Morphology
within Transformational Grammar, Considered with Primary Reference to
Chichewa and Swahili. Ph.D. diss. School of Oriental and African Studies.
University of London.
Mchombo. S. (1993) "Reflexive and Reciprocal in Chichewa". In Sam A. Mchombo. (ed.)
Theoretical Aspects of Bantu Grammar, pp. 181-207. CSLI/The University of
Chicago Press. Stanford University.
Mchombo. S. (1995) "Reciprocals in Bantu: Quantification in Morphology?" Stanford
University Linguistics Colloquium. Nov. 17. 1995.
McWhorter. J. (1992) Ni and the copula system in Swahili, a diachronic approach.
Diachronica, International Journal of Historical Linguistics. John Benjamins.
Amsterdam. Holland.
Meinhof. C. (1906) Grundzuge einer vergleichenden Crammatik der Bantusprachen.
Berlin.
Meinhof. C (1962) Introduction to the Phonology of Bantu Languages, (Translated by
Warmelo. N.J.) Berlin: Augustin in Gluckstadt.
Meussen, A.E (1962)"Meinhofs Rule in Bantu." African Language Studies III. 25-29.
Moshi. L. (1988) "A Functional Typology of ni in Kivunjo (Chaga).' Studies in Linguistic
Sciences 18.1:105-134.
240
Moshi. L. (1992) The Category Adjective in Kivunjo-Chaga.' In Africa und Ubersee.
Band 75. Dietrich Reimer Verlag, 111-131.
Mtenje. A. (1990) 'Feature-Spreading in Phonology; Evidence from Chiyao.'
Linguistic Analysis. 20:1-2. 87-103.
Mufwene. S. 1980. "Bantu Class Prefixes: Inflectional or Derivational?" In Jody
Kreiman and Almerindo E. Ojeda, eds.. Papers from the 16th Regional
Meeting, Chicago Linguistic Society, April 17-18, CLS: Chicago, 246-258.
Mugane. J. (forthcoming) A Paradigmatic Grammar of Grkuyii. Stanford Monographs
on African Languages.
Mugane, J.M (1996) "The Reciprocal in Grkuyu." Paper Presented at the 27th
Annual Conference on African Linguistics. Univesity of Florida. Gainesville.
Mugane. J..M (1995) "Gilcuyu NP Morphosyntax." Paper Presented at the 26th
Annual Conference on African Linguistics. Santa Monica, California.
Mugane, J.M (1993) "On Nasal Assimilation in GDcQyu." Paper Presented at the 24th
Annual Conference on African Linguistics. Ohio State University.
Columbus.
Myers, S. (1987) Tone and the Structure of Words in Shona. Ph.D. diss. University of
Massachusetts. Amherst.
Napoli D.J. (1985) "Verb Phrase Deletion in English: a Base in Generative Analysis.'
Journal of Linguistics 21:281-319.
Nino, M. (1996) Split Inflection and the Architecture of Grammar: the case of Finnish
Verbal Morphology, ms. Stanford, CA: Stanford University Depanment
of Linguistics.
Nordlinger, R. (1995) 'Split Tense and Mood Inflection in Wambaya", BLS 21.
pp.226-236.
241
Oehrle. R. (1976) The Grammatical Status of the English Dative Alternation..
PhD diss. MIT.
Peng. L. (1989) "Constraining Redundancy Rules: The Southern Kikuyu Case."
University of Arizona manuscript.
Postal. P. (1969) 'Anaphoric Islands', in Robert I. Binnick. Alice Davidson. Georgia M.
Green and Jerry L. Morgan (eds.). CLS 5. Chicago.
Pullum. G. (1993) 'English Nominal Gerund Phrases with Verb Phrase Heads'.
Linguistics 29, pp. 763-799.
Quirk. R. et al (1985) A Comprehensive Grammar of the English Language.
London: Longman.
Rappaport. M. (1983) "On the Nature of Derived Nominals." in: Papers in
Le.xical-Functional Grammar.. Lori Levin. Malka Rappaport. and Annie Zaenen
(eds.). Bloomington; Indiana University Lingui.stics Club.
Rizzi. L. (1990) Relativized Minimality. Cambridge. MIT Press.
Rugemalira. J.M (1993) Runyambo Verb Structure and Constraints on Predicate
Structure.
Ph.D diss. University of California. Berkeley.
Saiki. M. (1986) "A New Look at Japanese Relative Clauses: A Lexical Functional
Grammar Approach." In: Descriptive and Applied Linguistics . 19:219-230.
Selkirk. E. (1977) 'Some Remarks on the Noun Phrase Structure." In Formal
Synta.x. Peter W. Culicover. Thomas Wasow. and Adrian Akmajian.
Sells. P. (1985) Lectures on Contemporary Syntactic Theories. Stanford CA: Center for
the Study of Language and Information.
Sharman. J.C (1963) 'Nominal and Pronominal Prefixes in Bemba." in African Language
Studies 4:98-127.
242
Siegel, M. A (1980) Capturing Adjective. New York: Garland.
Sproat. R. (1985) On Deriving the Lexicon Ph.D diss.. MIT. Cambridge.
Stowell. Timothy (1981) Origins of Phrase Structure. Ph.D. diss.. MIT. Cambridge.
Tabor, W. (1994) Syntactic Innovation: A Connectionist Model.. Doctoral dissertation
Stanford university.
Thompson. S.A (1988) "A Discourse Approach toCross-linguistic Category Adjective."
In: John Hawkins ed.. Explanations for Language Universals.. London: Basil
Black well.
Valois. D. (1991) "The internal Syntax of DP and Adjective Placement in French
and English." NELS 21. ed. Tim Sherer. 367-381.
Wald. B. (1973) Variation in the system of tense markers of Mombasa Swahili.
Ph. D. dissenation. Columbia University. University Microfilms International.
Ward. G. Richard Sproat. and Gail Mckoon (1991) 'A Pragmatic Analysis of So-called
Anaphoric Islands'. Language 67. 439-474.
Wasow. T. (1977) "Transformations and the Lexicon." In: Formal Syntax. Peter W.
Culicover. Thomas Wasow, and Adrian Akmajian.
Wasow. T. and Roeper.T (1972) "On the Subject of Gerunds'. Foundations of Language
8. pp.44-6I.
Welmers W. (1973) African Language Structures. Los Angeles: Univ. of California Press.
Whiteley. W.H (1960) "Some problems of Syntax of Sentences in a Bantu Language
of East Africa.' In: A.W. De Droot. A.J.B.N Reichling and E.M. Uhlenbeck
(eds.): Lingua 9.4: 148-174. Armsterdam: North-Holland Publishing Co.
Williams. E. (1981) "Argument Structure and Morphology." Linguistic Review
1:81-114.
243
Zaenen. A. (1984) "Double Objects in Kikuyu?" In: C. Rosen and L. Zaring, eds..
Cornell Working Papers in Linguistics 5. Ithaca: Cornell University 199-206.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement