1

THE PHRASE STRUCTURES OF ETHIOPIAN OROMO

by

Baye Yimain

Thesis submitted for the Degree of PhD in Linguistics, Department

of Phonetics and Linguistics, School of Oriental and African
Studies, University of London.

August 1986
 2

ABSTRACT

This study is about the phrase structures of Oromo. It examines

various constituent structures in the light of the X-bar theory of

Jackendoff (1977) and recent developments. The theory assumes that

constituent structures are hierarchically organized syntactic expressions

of lexical categories.

The study starts with the identification of the lexical categories

of the language. Four such categories have been recognized. Each

category subcategorizes other maximal categories as complements or

specifiers to form its minimal/maximal projections.

The theory predicts a uniform three-level projection for all categories.

The study shows that only nominals and verbals are characterized by this

potential. The other two categories fall short of a bar. The complements

across the categories are functional arguments in the minimal projections

and restrictive and appositive modifiers in the intermediate and maximal

projections respectively. All of them occur in argument positions.

At the intermediate and maximal levels are also found specifiers of

two types: those which are of quantifying and/or intensifying function,

and those which are of deictic or referential use. They are generated

at the intermediate and maximal levels respectively, as adjuncts.

Specifiers and complements assume non-head positions. The category

by which they are subcategorized assumes a head position. The study shows

that this position is by and large final. This fixes the parameter of

Oromo as a head-final language.

 3

TABLE OF CONTENTS
Page

ABSTRACT 2

ACKNOWLEDGEMENTS 5

LIST OF ABBREVIATIONS 7

CHAPTER ONE GENERAL INTRODUCTION 8

1.0 About the Language 8
1.1 Previous Study 9
1.2 The Present Study 12
1.3 The Theoretical Framework 14
1.4 The Transcription 41
Notes to Chapter One 42

CHAPTER TWO LEXICAL CATEGORIES 43

2.0 Introduction 43
2.1 Nominals 44
2.2 Verbs 48
2.3 Adjectives 49
2.4 Adpositionals 59
2.5 Adverbials 63
2.5.1 Time Adverbials 63
2.5.2 Place Adverbials 68
2.5.3 Manner Adverbials 71
2.6 Specifiers 74
2.6.1 Articles 74
2.6.2 Quantifiers 75
2.7 Particles 77
2.8 Syntactic Distinctive Features 78
2.9 Summary 81
Notes to Chapter Two 83

CHAPTER THREE : COMPLEMENT ONE 84

3.0 Introduction 84
3.1 X' Complements 85
3.1.1 Nominals 88
3.1.2 Verbals 93
3.1.3 Adjectivals 115
3 .1.4 Adpositionals 116
3.2 Lexical Redundancy Rules 120
3.2.1 Derivations 120
3.2.1.1 Nominals 120
3.2.1.2 Verbals 125
3.2.1.2.1 Causatives 126
3.2.1.2.2 Passives 133
3.2.1.2.3 Middles 135
3.2.1.3 Adjectivals 142
3.3 Summary 145
Notes to Chapter 3 149
CHAPTER FOUR COMPLEMENT TWO 150
4.0 Introduction 150
4.1 X'' Complements 153
4.1.1 N'' Complements 153
4.1.1.1 Simple Nominals 154
4.1.1.2 Derived Nominals 162
4.1.2 V'1 Complements 167
4.1.3 A'' Complements 175
4.1.4 P'' Complements 177
4.2 X''' Complements 186
4.2.1 N''' Complements 186
4.2.2 V''' Complements 190
4.3 Summary 193
Notes to Chapter Four 195

CHAPTER FIVE : CLAUSAL COMPLEMENTS 198

5.0 Introduction 198
5.1 Clausal Arguments 199
5.1.1 The Akka-clause 200
5.1.2 The f-clause 213
5.2 Adnominal Clauses 228
5.3 The Position of the Complementizer 249
5.4 Summary 264
Notes to Chapter Five 268

CHAPTER SIX : SPECIFIERS 269

6.0 Introduction 269
6.1 Nominal Specifiers 271
6.1.1 Simple Nominal Specifiers 272
6.1.1.1 Articles 272
6.1.1.1.1 Definite Articles 273
6.1.1.1.2 Indefinite Articles 281
6.1.1.2 Quantifiers 283
6.1.1.2.1 Definite Quantifiers 283
6.1.1.2.1.1 Numerals 284
6.1.1.2.1.2 Measure Phrases 288
6.1.1.2.1.3 Classier Phrases 290
6.1.1.2.2 Indefinite Quantifiers 291
6.1.2 Derived Nominal Specifiers 301
6.2 Non-Nominal Specifiers 307
6.2.1 Verbals 307
6.2.2 Adjectivals 310
6.2.3 Adpositionals 314
6.3 Summary 317
Notes to Chapter 6 320

CHAPTER SEVEN CONCLUS ION 321

APPENDIX 328
REFERENCES 341
 5

ACKNOWLEDGEMENTS

I am indebted to a number of people without whose co-operation,

support and encouragement the completion of this study would not have

been possible.

First in my list is Richard Hayward, my supervisor. He has taught

me how to do Linguistics. His constant encouragement, patient guidance

and insightful comments and suggestions at every stage in the preparation

of the study have helped me to develop a great degree of awareness and

interest in the field.

My heartfelt thanks also go to W. Chao for reading nearly all the

chapters and for commenting on every one of them. I have made a number

of changes with respect to an earlier version by incorporating most of

the points she has suggested. Responsibility for making such decisions

is, however, ailmine.

I would also like to thank my friends and colleagues in Addis Ababa

University for giving me all the encouragement I needed when I made the

decision to work on Oromo, a language which I neither speak nor understand.

Mekonnen Disassa and Eshetu Kebede, both linguists in their own right and

native speakers of the language have been excellent informants. It would

be ungrateful of me if I passed by without mentioning how cheerfully they

agreed to give up a great deal of their office hours for this particular

purpose. I should say the same about my informants in London: Mohammed

Hasan, Getachew Hawaz, Fayyisaa Lemi. and Dereje Deressa, all of whom have

shown a great deal of interest in what I have been doing. I thank them

all for their time and patience.

I should also thank Sara M. Messana for helping me in translating

Italian texts.
 6

Finally, I should mention my friend, Alemayehu Haile, who joined

the School while I was in the middle of my studies, and with whom I have

shared all my 'crazy' ideas about the syntax of Oromo. I am grateful

for his encouragement and for the time we spent together in the 'wilds'

of London.

Last, but by no means least, is my wife, who has given me all the

moral and material support I needed while she herself had all alone to

face the hard realities of married life, a challenge we had just started

together especially well when I had to leave for London. This study

may be a reward for both of us, but more rewarding are Tsega-Zeab and

Kaleab, our twins who were born while it was in progress.



ERRATA

Page Line Corrections

63 16 Mathews Matthews

65 2 mean - mead

94 9 v [- .^,t ] - V + [- #1

105 4
-

325 21 relatin -* relation

329 1 head of list = V

341 19 Agjectives ± Adjectives

341 29 Af far ± Qafar

342 30 Langua - Language

343 30 Affar -4- Qafar

344 3 Affar - Qafar

344 19 Beinding ± Binding
 8

CHAPTER ONE

1.0 About the Language

Oromo is one of the languages of the Lowland East Cushitic group

within the Cushitic family of the Afro-Asiatic phylum. It is predominantly

spoken in Ethiopia though several dialects are also spoken in Northern

Kenya. Within Ethiopia alone it has a population of between 8 and 10

million speakers according to some estimates (Bender and Mulugeta 1976:195;

Graggl976:166). Within Ethiopia it is spoken over a vast area of land

stretching from the mid-north to the far south, and from the far east to

the western border of the country.

According to Heine (1980:55), the dialects spoken outside Ethiopia

include the following:

(i) Gabra
(ii) Boorana
(iii) Sakuye
(iv) Garreh
(v) Ajuran
(vi) Orma
(vii) Munyo
(viii) Waata.

Though adequate information is lacking concerning the types of variations

and their isoglosses within Ethiopia, the following regions seem to have

been recognized as distinct dialect areas:

(i) Meèaa (Western Ethiopia)

(ii) Tulamaa (Central Ethiopia)
(iii) Kottu (Eastern Ethiopia)
(iv) Rayya (Northern Ethiopia)
(v9 Boorana (Southern Ethiopia).

The feeling among scholars in the field is that, though there are

such regional variations, on the grounds of mutual intelligibility and

sharing of core vocabulary items, the various regions constitute a single

language community (cf. Bender and Mulugeta 1976; Andrzejewski 1960).

 9
According to Gragg (1976:174), Bender and Mulugeta (1976), Wako

Tola (1981) and Haimanot Moges (1983), the phonemic inventory of the

language has the following consonants and vowels:

Consonants: Labials Alveolars Palatals Velars Glottals

Stops: vis (p ) t c k

vd b d 5 g

ejective F
implosive : d

Continuants: vis f s s h

vd (z)

Nasals m n

Liquids: 1, r

Glides: w y

Vowels:

1. U

e 0

Each consonant and vowel has a long counterpart.

Since the purpose here is only to introduce the basic sounds as they will

be represented in the orthography in this study, we will not go into the

details of the phonology proper.

1.1 Previous Study

The literature relating to Oromo dates back to the mid-nineteenth

century. On the basis of its content and purpose. it may be broadly

divided into three types:

(i) Word lists and/or dictionaries

(ii) Pedagogical grammars
(iii) Descriptive sketches.
 10

The works of Viterbo (1892), Pretorius (1893), Foot (1913),

Ducati (1937), Da Thiene (1939) and Gragg (1982) belong to type (i).

onesimus (1894), Hodson and Walker (1922), Moreno (1939), Nordfeldt (1947),

and Launhardt (1973) comprise type (ii). The rest of the list in the

bibliography belongs to type (iii).

Leaving aside Gragg's (1982) dictionary. the attempts made by those in

types (i) and (ii) are utilitarian in purpose and traditional in approach.

The main interest of these scholars was to present the language in such a

way that anyone who did not have any knowledge of it could learn it in a

short period of time and use it in day-to-day communication. In other

words, the main objective of such works was to describe the language in one

way or another as a means to a practical end. The end might, of course,

vary slightly from author to author. For Hodson and Walker it was

commercial. This is clearly stated in their preface to the Grammar of the

Galla or Oromo Language where they say:

The grammars on Galla are not of great practical

utility, and it was thus necessary to plan the
lesson afresh keeping always in mind the needs
of the trader and his desire for practical
sentences used in ordinary life (1922:8).

For Moreno, it was translation. His intention was to write a grammar

which would enable the Italian to communicate in the language and the Oromo

to help as an interpreter. Like Hodson and Walker, he too says, "Le frasi

degli esercizi sono congegnate in modo che l'italiano aiuti a comprendere ii

galla, e il galla faciliti la traduzione dall'italiano" (1939:10).

Such works, being constrained by non-linguistic factors, try to present

the language as a whole. Phonology, morphology, and syntax along with

texts and a host of exercises are presented all in one book though hardly

any of them is dealt with at length. Descriptions are limited to surface

phenomena, and explanations are more often notional than formal. In fact,

in some cases, theoretical assumptions about language-learning and teaching

 11

current then seem to have been deliberately avoided. Such is the case,

for example, with Hodson and Walker who say:

It would have been possible to write a grammar

based on the latest direct and oral methods,
but such are not suited to the wilds where
intelligent teachers may not be found (1922:8).

whether they have been successful in accomplishing their aims or not

is not something to be definite about, and it is not important for the

purpose we are after. What is definite and important from our point of

view is the fact of their laying the groundwork for whatever has been done

since then in Cushitic studies in general and concerning Oromo in particular.

The word-lists and dictionaries referred to above have provided good sources

in terms of data, and the various pedagogical grammars have also acted as

spring-boards in terms of providing directions for the descriptive works of

the third category, and, needless to say, for the present study.

The works falling within the third category, like those of Hodson and

Walker, and Moreno, are surface descriptions of one or other aspect of the

language. Excepting Gragg (1976), which is an attempt made to give an

overall picture of Oromo as spoken in Wallagga (Meaa), the rest are

either morphological or phonological in emphasis. There is very little on

phonetics, particularly on suprasegmentals and still less on syntax.

These works, as opposed to those in groups one and two, are characterized

by their description, which one may say is formal. Furthermore, they are

limited to one particular dialect area. Most of them deal with the Meaa

dialect as indeed do the works of some of the earlier scholars such as

Onesimus (1894), Nordfeldt (1947) and Launhardt (1973). To a certain

extent, similar efforts have been made to describe the features of the

Boorana dialect. The accounts of Webster (1959) and Andrzejewski (1960)

focus mainly on this dialect.

 12

Quite recently, some research works based on the Kottuu (Eastern)

variety, have appeared. These are Owens' 1985a and 1985b analyses of

'causatives' and 'nominal relations' respectively. Unlike most other

previous works, these are based on the theoretical assumptions of a

particular framework.

Tilahun Gamta's (forthcoming) Dictionary of Oromo is another promising

work to look forward to. To date nearly all that has been done on this

language has been done by people who have had very little or no working

knowledge of the language. This dictionary is an exception in the sense

that it is the first of its kind to be prepared by a native speaker.

From the present state of affairs, one may gather that there is a

growing interest in Oromo, and that there is also a lot to be done in

the areas of phonetics, dialectology and needless to say, syntax, and as

oragg (1976:173) has said, 'Any work done along these lines.. .is bound to

break new ground and open new perspectives'.

1.2 The Present Study

Looking at the literature on Oromo, it is not impossible to guess the

direction linguistic researchers will have to take. The gap in each of

the areas mentioned is enormous. In some cases, there is a total absence

of materials. Such is the situation, for example, concerning the variety

spoken in Rayya (Wello). In others, the available material is scanty.

To date there has not been anything done on the phrase structures of

the language. This is basic to the study of the typoloqy of languages

and to the investigation of the principles governing the distribution and

relations of categories within a particular language. The various attempts

made hitherto, though commendable for the various purposes they were

designed for, have not been geared towards this. The present study is,

therefore, an attempt made with the intent of narrowing down a gap in this
 13

area. The stud y tries to examine various constituent structures and

formulate the rules that govern the relation of elements within such

struCtuxeS.

Like some of the earlier works, this study is limited to the form of

the language used in Western Ethiopia (Meaa). The reason is partly

accidental and partly pragmatic. My Oromo-speaking colleagues are from Wallagaa

and Ilubabor, and the available literature is mostly based on this variety of

the language. The study is, therefore, primarily based on data elicited from

my colleagues. Subsequent references for similar purposes have been limited

to Gragg's (1982) dictionary which came out just as my study was to get

underway; I am indeed qreatly indebted to Gragg for saving me a lot of

trouble.

There have been instances, however, when speakers of the language from

other regions have been made use of as informants as the study was progressing,

but again, the materials have been checked against their equivalents in the

dictionary, and whenever a form was found to be entirely different, it was

always the one in the dictionary which was used. One such example is the

word for 'money' which is /mahaallek/ in Harar, /ganzaba/ in Shewa (Tulamaa),

and /horii/ in Meaa (Wallagga). Another is the form for 'yesterday'

which is /kalee". halee/ in Harar but /kaleessa/ in Meaa. The form for

'she' also varies between /iee/ in Meâcaa and /isii/ in Tulaxnaa. Except

in such minor cases, references have been limited to the one particular

dialect. However, it...is believed that such apparent differences, being

so much a property of the lexicon, may not have any significant effect on

the principles that underlie the organization of the constituent structures.

'o this extent, the syntactic descriptions that we will be dealing with

throughout the study may be said to be Oromo in form though one may say

that they are Mein substance.

 14

1.3 The Theoretical Framework

As stated earlier on, the attempts made at different times in the past

to describe the language, have been geared towards some utilitarian purposes.

This is to say that the scholar's main interest was the practical use of the

language more than anything else. To achieve this, it might not be

necessary to start the work with a well-defined set of assum ptions about

language, and with a set of strict procedures as to how to go about analyzing

a body of data in the light of the assumptions made. In other words, some

kind of theoretical framework could have been used unless this was thought

unnecessary for the purpose the scholars were after. As one may infer

from the quotation cited earlier, the approach of Hodson and Walker suggests

that this was indeed the case. Such works suffer from the drawback that

this type of approach would lead into, particularly when judged from a

purely linguistic point of view which requires some degree of adequacy.

In this regard the present study may be said to have offset the

drawback of the previous works, since it is based on a strictly defined

theoretical framework. This emerges from the purpose of the study itself,

which is one of making a modest attempt to discover the network of relations

underlying the language and to account for the native speaker's ability to

judge structures as grammatical or ungrammatical. There is no one

particular way of 'achieving' this goal. There are a number of assumptions

and approaches behind the frameworks linguists have proposed over the years.

The choice of one framework over another seems to be partly determined by

the individual's backqround. The framework adopted here is what is known

as the Extended Standard Theory (EST) developed in Chomsky (1970, 1972,

1973, 1976, 1977), Ennds (1976) , and outlined in Radford (1981)

This framework is a development of the ideas embodied in the Standard

Theory of Chomsky (1965). For reasons of brevity, I shall only state the

reasons that led to the revision and list the characteristics of the new

version.
 15

In the Standard Theory, it was felt that:

(i) Deep Structure was too abstract

(ii) The transformational component was too strong
(iii) The categorial component was too restrictive in
some sense and not restrictive enough in another.

In order to do away with the problems that had led to this situation,

it was necessary to make some revisions with respect to each of these

components by way of decreasing the degree of abstractness and the power of

transformation and by constrainin g the rewrite rules of the categorial

component. The revised form is an attempt to achieve this end.

As in Aspects, it is also assumed in EST that the organization of

grammar presupposes the existence of the following five distinct but

interacting components:

(i) The lexicon

(ii) The categorial component
(iii) The transformational component
(iv) The phonological component
(v) The semantic component

The first two constitute what has often been called the base component

(Chomsky 1977:71). The second and the third jointly constitute the

syntactic component. The last two are interpretive in the sense that they

relate to the ways the structures generated by the syntactic component are

pronounced and associated with objects and events in a given context of

situation.

(i) The Lexicon

This is the part of the base which is the repository of idiosyncratic

information about the lexical and grammatical formatives of a language.

In Aspects its role was limited to the entries of such formatives with

the specification of their phonological, syntactic and semantic features,

together with a set of lexical insertion rules which specified which item

in the lexicon substitutes for which terminal symbol in the structures

generated by the categorial rules.

 16

In the revised version (EST), lexical redundancy rules, morpheme

structure rules and rules of allomorphy have also been assigned to the

lexicon. This increasing role is a concomitant of a position which

Chomsky (1970) took to reduce the power of the transformational component,

which until then had been concerned with the derivation of not only syntactic

categories but also lexical categories (Cf. Lees 1960). Chomsky argued

that because of the irregular nature of word-formation processes, it was

wrong to derive words by means of syntactic transformations. The latter

should be limited to the derivation of syntactic structures, and that all

processes of word formation should be relegated to derivational morphology,

which operates in the lexicon. Hence one characteristic feature of EST

is a new division of labour between the transformational component and

the lexicon, which in the words of Mark Aronoff (1976:6) also lead to

'the birth of morphology or at least the declaration of its domain which

is simultaneous with, arid contained in Chomsky's "Remarks on Nominalization"

(1970)

word-formation processes, now qone to the lexicon and being handled

lexically, the transformational component is left with a few obligatory

syntactic rules whose application is both specific and predictable.

(ii) The Categorial Component

This is a set of context-free phrase structure rules which defines

the hierarchical structure of syntactic categories and the linear ordering

of the elements forming them. These rules generate a set of phrase markers

whose initial symbol is a maximal category such as an s or s. The process

continues until a point is reached where no further syntactic derivation

is possible. At this stage lexical items are drawn from the lexicon and

are inserted under each category label according to rules which relate to

lexical insertions and selectional restriction features. Some positions

 17

may be left vacant or filled by empty elements whose nature is specified

by some principles operating at some level in the derivation of the structures.

The categorial rules in the pre-EST version were too restricted with

regard to the phrase types they could generate. These were limited to

two levels of hierarchies in each syntactic category. These are the

phrasal categories such as NP, VP, etc., which are maximal, arid the lexical

categories N, V, etc., which are their minimal primitives. In other words,

there was no room for intermediate categories, that is, for categories

smaller than the maximal and larger than the minimal categories. The

elements constituting such categories were accounted for in terms of

syntactic transformations; such was the case, for example, with prenominal

adjectives, comparatives, etc. (cf. Selkirk 1977). In the new version,

such structures are generated by categorial rules in the position in which

they appear in surface structures as intermediate categories.

As stated earlier, this component was also unrestricted in another

sense. The rules which derive one phrase marker from another were not of

a type which would guarantee that in the derivation of a category, the

categorial membership of the head of the derived category was the same as

that of the category from which it was derived. In other words, there

was no condition that could maintain the categorial relationship of any two

derivationally related categories.

In order to avoid such anomalous situations and the problems arising

therefrom, the component had to undergo some changes. One of these

changes was the attempt made to relax the rules so that they could

accommodate a number of intermediate categories. The other was a constraint

qoverning the relationship between a maximal category and the head of its

derived constituent. The heads of the two categories should be in the

same lexical category.

 18

Other restrictions relate to the dominance and precedence relationship

of constituents branching from the same maximal node. Put formally this

runs as follows:

If a node X dominates two nodes Y and Z, if Y

precedes Z, then any node dominated by Y must
precede both Z and any node dominated by Z
(Radford 1981:83).

This is a well-formedriess condition on the derivation of a phrase marker

from a higher phrase marker.

With such changes implemented, the new version, popularly known as

the 'X-bar' Convention, came out first in Chomsky (1970), though the ideas

in it had been apparent in the works of Zellig Harris as early as the 1940s

(cf. Chomsky 1970:211) and later in Lyons (1968) who says:

Phrase structure grammars fail to formalise the fact

that NP and VP are not merely mnemonically convenient
symbols, but stand for sentence constituents which
are necessarily nominal and verbal respectively,
because they have N and V as an obligatory major
constituent. What is required, and what was assumed
in traditional grammar, is some way of relating
sentence constituents of the form XP to X (where X is
any major category: N, V, etc.)...(p.33).

An elaborated version of this convention is found in Jackendoff (1977).

Since the present study is primarily concerned with the identification

of the possible constituent structures of Oromo in the light of the ideas

embodied in this new version, it might be necessary briefly to point out

its major claims, the methods it employs to achieve them, and the problems

it has encountered.

Jackendoff (1977:29) makes the following three claims for the X-bar

convention as a theory of syntactic categories in universal grammar (UG):

Ci) Universal grammar provides a set of syntactic distinctive

features in terms of which the possible lexical categories
of a particular language are defined.

(ii) Each lexical category X defines the set of syntactic

categories Xn that can be developed from it. The syntactic
category x and the lexical category X are related by a phrase
structure schema of the type in (1):
 19

(1) X > (c j ) ... (ci).. .X 1 ... (c 1 ) ... (ck)

where 1 = < n < 3, and either c 1 = y ''' for some
lexical category y or c. = a grarunatical formative (p.36).

(iii) Rules of grammar are stated in terms of syntactic feature

complexes and bar/prime notation. In other words, the
domains where syntactic rules operate are specified in
terms of syntactic distinctive features and bar notation,
the latter indicating the level at which the rules are
expected to operate.

As stated above, the syntactic distinctive features which specify the

possible lexical categories of a language are provided by UG. A lanquage

has to choose a set of its features on the basis of the degree of naturalness

its syntactic rules may gain when they operate on or across syntactic

categories at a particular bar level. In other words, it is the anticipated

degree of generalization to be made that determines which set of features a

particular language should take from UG.

The features for English as used in Chomsky (1970) are [±N, ±VJ.

This set is based on the recognition of nouns, verbs and adjectives as the

major lexical categories of the language. Jackendoff (1977) adds

prepositions and adverbs to the list of lexical cate gories, and also

recognizes particles, determiners, modals and quantifiers as minort

categories. To define each of these categories, he had to reformulate

chomsky's features and also introduce new ones. These are shown below:

(a) (± su.b.1
(b) (± obj.]
(c) [± comp.]
(d) [± det.]

These features are based on what a category can or cannot have. For

example, nouns are defined by the feature [+ sub, - obj.] because they can

have subjects but not objects (cf. Chapter Two for a discussion of this point).

Regarding claim (ii) and the schema in (1), Jackendoff believes that

this schema is also provided by UG. The underlying idea here is that each

constituent in a language is endocentric and that the head of the initial

 20

n. n-i
constituent X is X . The symbols before and after it represent a major

syntactic category functioning as a specifier or complement of this head,

or as a grammatical formative of one type or another; such as those for

aspect ornumbers for example.

That the head of the category n is shows the extent to which

phrase structure rules have been constrained. This is in response to the

feeling that such rules had been unrestricted in terms of the potential

structures they could generate. The fact that they are now limited to

generating structures whose heads are lexically related means that there

cannot be derivations of the type:

n____ n
x>...y...

An important point implicit in claim (ii) is that constituents are

hierarchically structured units. Their minimal expression is a lexical

category X, whereas their maximal expression is X 11 . Between these two

categories, there could be a number of categories falling at different levels.

The role of the X-bar convention, as a theory of phrase structure

rules, is to examine the syntactic properties of such categories and to

determine their positions. This ultimately gives us the value of n in

the maximal category of X". But this is not as simple as it sounds.

Opinions have always been at variance about the value of n, though there

IS general agreement on the hierarchically structured nature of constituents

(at least in one type of human language) and on the existence of intermediate

categories. For example, in Chomsky (1970) n is two for nouns and three

for verbs. In Vergnaud (1974) and Seigel (1974) n is four for nouns.

In Dougherty (1968) it is three for nouns and six for verbs. Jackendoff

(1971; 1974a) has two for all categories. In Jackendoff (1977), however,

he allows three for both nouns and verbs, and he argues that this should be

extended in all other categories for reasons of structural parallelism

 21

(pp.35-6). If opinions about a particular language vary to such an

extent, it might not be impossible to forecast how much more conflict

might arise when taking many languages into account. It appears that it

is for this reason that Jackendoff says that the best theory is one which

provides 'just enough structure to make the relevant structural differences

and no more' (p.35).

Another controversial point about this convention relates to the

projection line of verbs. Jackendoff claims that the line includes s's

and s's which means that any constituent in a clause is a part of the

projection of V. This makes subject arguments in both Ss and NPs as

projections of V and N, functioning as specifiers of some sort. But as

specifiers belonging to the same functional category they should be subject

to similar syntactic rules; but the y are not, since there are rules which

operate only on VP5 leaving the subject, that is, the specifier, intact,

as we shall observe in Chapter Five. Suggestions have,therefore, been

forwarded to limit the maximal projection of v to v'' (') and thereby make

distinctions between it (V''' ) ) and S (cf. Hornstein 1978).

As stated earlier on, the existence of intermediate categories,

irrespective of how many they might be, is not a matter to be doubted.

The desirability of describing the way such categories behave also does not

seem to be in question. It is hence the purpose of this study to identify

and examine the properties of the constituent structures of Oromo in the

light of the claims of the X-bar convention as expounded in Jackendoff

(1977) and briefly presented here.

(iii) The Transformational Component

As stated earlier, this is a part of the syntactic component. Its

function is to convert the deep structures which are generated by the

categorial rules and lexicalized from the lexicon, into their corresponding

surface structures (S-structures). As already mentioned, in Apects this

 22

component was too unrestricted since its rules (T-rules) could opera.te

at both syntactic and lexical levels. In order to derive certain lexical

items by means of transformational rules, it was at times necessary to

postulate underlying structures which were much more abstract than was

otherwise needed for deriving ordinary syntactic surface structures.

In EST, the rules of the component have been limited just to the

derivation of syntactic structures; and all derivations of lexical items

both regular and idiosyncratic have been assigned to the lexicon where they

can be taken care of by derivational morphology.

It is not only the domain of the transformational component which has been limitec

the rules themselves have also been drastically reduced and their applications

have also been highly constrained by a number of conditions (cf. Ross 1967;

Chomsky 1973, 1977). There is now only one rule, viz., 'move a'; a being

an abstraction of a syntactic category, NP, VP, etc., at the deep structure

level of representation. This rule moves a from its base position and

substitutes it for, or adjoins it to another category. Such an operation

has the following characteristics (Chomsky 1981:56):

(1) Movement is always from a thematic position to a

non-thematic position.
(ii) Movement observes subjacency.
(iii) Movement leaves a trace of a moved category which
must satisfy certain well-formedness conditions.

In short, what these conditions say is that a should move to a position

which is unfilled by a referring expression in deep structure, and that when

it moves, it should not cross more than one bounding node, that is, one NP

or S; Furthermore, its trace should form a structural configuration with

a lexical head and should also obey certain principles which govern the

distribution of categories.

Since our concern here is with the categorial component more so than

with the rule 'move a' we need not elaborate this point any further. We
 23

shall instead raise a point which relates to the typology of Oromo, with

regard to the parameter configurationality versus non-configurationality.

The X-bar convention presented in Jackendoff (1977) as a theory of

phrase structures is based on the assumption that English is a configurational

language, as opposed to other languages such as Walpiri (Hale 1978)

(cited in Chomsky 1981). This suggests that we need to make the same kind

of assumption about Oromo and then go into the details of showing how this

is reflected in the language in the light of the claims made by the theory.

This also seems essential from the practical point of view, since the

entire study is geared towards the examining of the possible constituent

structures of the language. However, since a proposal has already been

made by Owens (1984) to the effect that this language is non-configurational,

we cannot start with such an assumption. Either we have to agree with

the proposal and proceed with the description accordingly, or argue against

it with some concrete evidence.

The proposal was made in relation to Owens t lexical-functional analysis

of Oromo causatives. Accordingly the following has been presented as the

base rule for Oromo sentences:

I S > NP (NP) (NP) V1

Owens has not forwarded any argument in favour of this rule nor has he

given us any features which would characterize Oromo as a non-configurational

language.

Although the discussion in the next chapters is believed to show that

this language is configurational, it might still be necessary to mention

a few points about Owens' proposal in the light of the properties of non-

configurational languages. According to Hale (1980) (as stated in

Bouchard 1984:159), the following are believed to be properties of such

languages:


24

(i) Free word order

(ii) Discontinuous expressions
(iii) No nvement transformations
(iv) Rich case systems
(v) No pleonastic NPs
(vi) Free pronoun drop.,

Now, from the base rule in (I), it is clear that Oron has NPs.

The elements forming these NPs follow a strict order. Hence (la) but

not (ib) or (ic) is possthle.

1(a) [nama guddaa kanal

NP
man big this
'this big man'

(b)* [kana na.rna guddaa]

his man big

(c)* [guddaa kana nama]

NP
big this man

The ungrammaticality of such structures shows that Oromo does not have

the characteristic features indicated in (1). As the base rule in (I) shows,

the language is also verb-final. The verb cannot occur in initial or

medial position without the structure being ill-formed. This again shows

that the same type of order is maintained between constituents forming

clauses. Consider (2) below.

2(a) [nam-i-i kun.i hoolaa guddaa [bit-el]

Sman_sgl_nom this-nom sheep big buy-pf.
'this man bought a big sheep'

(b)* ([bit-el nath-i-i kun-i hoolaa goodaa]

S buy-pf man-sqi-nom this-nom sheep big

The second characteristic feature is that non-configurational languages

allow discontinuous expressions. What this would mean is that in a sentence

like (3) below, the two nouns, N 1 and N 2 could be related to any of the

adjectives or the demonstrative elements in the string.

 25

3. (nam-i-1 guddaa-n dubarti eer-tuu kana rukkut-e]

Sman_sgi_nom. big-nom. woman tall-f this hit-pf
'the big man hit this tall woman'.

If Oromo were a non-configurational language, the noun /nam-i-i/'man-nom'

could be linked to the adjective /guddaa-n/ 'big-nom' or /deertuu/ 'tall-f',

or to the demonstrative /kana/ 'this'. And th same could also be true

of /dubartii/ 'woman'. But as can be observed from the sentence, the

nouns form structural relations with the adjective or demonstrative

immediately following them. This structural relationship is also

manifested morphologically; the adjective /guddaa/ 'big' shows up the

case affix I-ni in agreement with /nam-i-i/ just as in the same way

/deertuu/ 'tall' displays the feminine affix /-tuu/ following the feminine

noun /dubbartii/ 'woman'.

According to Bouchard (1984:157ff), in languages like Walpiri or

2
Japanese which are cited as being non-configurational, the verb could

also be related to either one of the nouns in (3). The effect of this on

the sentence would be that any of the nouns could be the subject or the

object. In other words, subject and object relationship is not something

which is structurally determined in such languages. In Oromo, the subject

is always the left most NP, i.e., [NP, S], and the object is the NP

immediately preceding the verb. The ungrammaticality of (4b) below shows

this.

4(a) [nam-i-i [&altuu rukkut-e]J

S
man-sql-nom ç hit-pf
'The man hit ci.

(b)* [&altuu [nam-i-i rukkut-e]]

S.
C man-sgl-nom hit-pf

The ungrammaticality of such structures shows that there is structural

configuration between, say, the verb and its object argument and that it is
 26

this configuration which determines the role and the case assignment of the

argument. In other words, the assignment of thematic roles and case in

configurational languages presupposes structural relations between the

assigning and the receiving elements. In the structures above, the

subject and the object nouns fall into such relations in (a) but not in (b).

If Oromo were a non-configurational language both structures would be

perfectly acceptable since in such languages case relations are not determined

structurally but are assumed by the arguments irrespective of their

positions in structures of sentences (cf. Chomsky 1981:133ff.). In other,

words, /aaltuu/ and /naInii/ 'the man' could have assumed their respective

cases in both (4a) and (4b) and since the type of case each would have

assumed could have been identified from their forms, both structures would

have been grammatical. But as we can observe, only (4a) is well formed.

The third characteristic feature of non-configurational languages is

the absence of movement rules. As stated briefly earlier on, such rules

move elements from one position in deep structure to another position in

s-structure. The position to which the elements move must be empty.

This suggests that the process can take place only if there is an empty

position in deep structures. In other words, movement presupposes the

existence of such empty positions.

As in other configurational languages, such as English, Oromo has such

positions. There are verbs which do not select external arguments (in the

sense of Williams (1981)). The position where such an argument would be

expected is empty at the level of deep structures. In other words, it

is unfilled by a lexical element. However, at the level of the corresponding

S-structure, the empty position may be filled by a lexical argument. This

involves movement. The moved element leaves its trace behind to satisfy

the subcategorization property of the head. Such verbs include the



27

copulative verb /fakkaat-/ 'seem' and passive verbs in general. With

regard to /fakkaat-/ 'seem', we will enter into a detailed discussion

in Chapter Five. Here, we will consider only structures involving passives.

5(a) (i) [fard-i gurgur-am-el

Shorse_nom sell-ps-pf
'A horse was sold'.

(jj)* [farda qurqur-am-e]

Shorse sell-ps-pf

(b) (i) [man-ni gub-am-e

Shouse_nom burn-ps-pf
'A house was burnt'.

(jj)* [mana gub-am-e]

Shouse burn_ps-pf

Within the framework adopted here, passivization is an instance of the

general rule of 'move ci'. The subcategorization properties of verbs with

or without passive morphology is the same. The effect of passive morphology

is on the case and thematic () role assigning property of verbs. Such

verbs are believed to have no potential for assigning case to their internal

arguments and a thematic role to their external argument (cf. Chomsky 1981).

The position where an external argument is expected to appear is hence empty

at the level of deep structure.

The representation of the structures in (5) is accordingly as follows:

6(a) NPe (farda gurgur-am-e]]

S Vorse sell-ps-pf

(b) (NPe (nama gub-ain-e)]

Viiouse burn-ps-pf

The arguments /farda/ 'horse' and /mana/ 'house' must be case marked

for the structures to be grammatical (cf. Choinsky and Lasnik 1977;

 28

Chomsky 1981, 1982). But in the position they appear in (6) these

arguments cannot receive case for the reason already stated. In order

to receive case, they must move to the position of NPe where they can receive

nominative case from Inflection. The ungrammatical structures in both

(a ii) and (b ii) show exactly this situation.

We may say at this point that this may be true if the language is

configurational, if it is not, the movement may not be necessary since the

nouns could assume their cases in situ. In other words, how do we know

that movement has taken place in (5)? What if the nouns have assumed their

cases in situ; Oromo being a non-confiqurational language as proposed?

The fact that movement has taken place in (5) is deducible from the

situations found in other structures of sentences with both subject and

object positions filled in deep structure. As mentioned earlier on in

this section, the subject argument in such sentences is the prominent NP,

that is, the NP which is the left-most. This NP has the nominative case

marker I-ni. The object argument is always the one before the verb. If

it were the case that Oromo was non-configurational, such distributional

restrictions would not have been necessary. The subject could have occurred

anywhere in the sentence. But this is not the case. As the examples in

(4) clearly show, the subject argument cannot occur in the position preceding

a transitive verb. From this it follows that in the structures in (6),

the arguments have to be generated in object position since the verbs are

transitive. In order for these arguments to occur as subjects, they must

move to the position they occupy in the corresponding structure in (5)

since in the position they appear in in (6) no NP with nominative case

marking would be allowed without the resulting structure being ill-formed.

The other characteristic feature of a non-configurational language is

that such languages have rich case systems. Oronio, like English, has


29

nominative, accusative and genitive cases. The accusative is the unmarked

one. The nominative is indicated xnorphologically by the affix I-ni and

the genitive by the configuration [N NP]. If the language had been

NP
non-configurational, it would perhaps have had a lot more cases, each

indicated by a distinct morphological element, and in such cases the word

order would have been a little freer. But this is again not the case.

For example, a discontinuous genitive construction is not possible, as the

example in (7b) shows.

7(a) (mana Tulluu]

NP
house of T.
'T's house'.

(b) * [(mana] kana (Tulluu] I

NP
house this of T.

(C) ((mana Tulluu] kana]

9ouse of T this
'This house of Tulluu'.

The same may be said about accusative NPs. They are recognized as

having this case only if they occur in positions strictly subcategorized

by the category which assigns this case. If they occurred away from the

category (as in structures like (4b) for example) the resulting structure

would be ill-formed. 3 If on the other hand, each case is assumed and is

also morphologically realized, then syntactic configurations might turn

out to be minimally significant. And if a language is characterized by

such features, that language might be argued to be a non-configurational

one. The data presented so far do not lead us to draw such a conclusion

about Oromo.

The last characteristic feature of a non-configurational language is

free pro-drop. This means that NP arguments may be phonetically null.

In other words, sentences without a surface subject are possible if a



30

language is non-configurational. This is not, however, a feature which is

exclusive to such languages. There are well-known configurational

languages like Italian and Spanish, for example, which allow null subjects.

But in such languages it is only the subject which may be dropped, whereas

in a non-configurational language it is both the subject and the object

which can be dropped. The difference in this case may be said to be one

of degree. Hence, whereas in a language like Italian the verb and its

object may appear in surface structure, in a non-configurational language

the verb alone may appear. In this respect Amharic could be a good example,

though it cannot be conclusively said that it is a non-configurational

language. In Axnharic, structures like the following are possible:

8(a) [NPe NPe matta-hu-at]

S
hit -I -her
'I hit her'.

(b) (NPe NPe matta-ai-]

S
hit -she -me
'She hit me'.

In Oromo such is not the case since the corresponding structures

in (9) are ungrammatical.

9(a)* (NPe NPe rukkut-e I

S
hit-3ms-g.
'He hit'.

(b)* (NPe NPe rukkut-t-e]

S
hit-f-pE.
'She hit'.

Since Amharic has both object- and subject-referring affixes (Clitics)

in its verb, it is possible for both NPs to be missing. On the other hand,

Oroino shows only subject-referring elements in its verbal inflection, hence

only the subject is permitted to be missing.

 31

From this it folloWS that the base rule (I) which Owens has proposed

for Oromo leads to wrong predictions, because according to this rule,

object arguments may be missing. Moreover, on the other hand, subject

arguments would appear unable to be missing. If taken seriously, this

may mean that missing objects but not subjects are recoverable from the

form of a verb. As can be observed from (9), the verb in this language

does not have object-referring elements. In other words, (AGR)eement

is not rich enough in Oromo (as it is in Ainharic) to license object dropping.

On the other hand, the subject, whose feature is recoverable from the verb,

is indicated as if it were obligatory. This means that structures like

(10) would be ungrammatical, which is simply not the case.

10(a) [NPe hoolaa bit-el

S
sheep buy-pf. -
'(He) bought (a) sheep'.

(b) [NPe hoolaa bit-an-il

S
sheep buy-3pl-pf.
'They bought (a) sheep'.

From the situation we have observed thus far, with regard to the

feature pro-drop, it is not possible to maintain that Oromo is a non-

configurational language. It can only be said that it is a null subject

language like Italian or Spanish.

One property of a non-configurational language which Oromo does have

is the absence of pleonastic elements. Such elements do not exist. The

position where they would be expected to appear is vacant. But, the

absence of this particular feature alone cannot lead to the conclusion that

the language is non-configurational. It may only suggest that a language

does not have to satisfy all the requirements listed above in order to be

identified as belonging to languages of a certain parameter.

 32

There is one other point which we need to consider with relation to

the base rule (I). The rule shows that there is no VP in this language.

This is what one would expect given the proposal that the language is non-

configurational. As already discussed, in such languages there is no

structural configuration of the type we find in configurational languages

like English, between head and non-head elements. The subcategorization

properties of verbs, for example, would be satisfied only by the presence of

nouns. In other words, it is not a requirement that the nouns be tightly

bound to the verbs and that the two form a structural unit, a VP. In

fact, because of the 'flat' nature of such languages, there is a possibility

for every structure of the type in (II) to be ambiguous if case were to be

assigned (Cf. Bouchard 198k).

II.

NP1 NP2 V Infi.

In (II), both NP 1 and NP2 are governed by V and also by Inf 1. Hence

there is a possibility for either one of these NPs to be the subject or the

object. The effect of this is that the structure may be ambiguous between

the interpretations: (1) where NP 1 is the subject, and (2) when it is the

object. If Oromo were such a language, structures such as (4 b) would have

been grammatical, but as we saw they were not; which means that only NP1

can be the subject. From this Thay follow the question: Why is it that

only NP 1 can be the subject? What prevents NP 2 from being the subject, and

NP 1 from being the object? Such questions may be answered satisfactorily

if we assume a strict structural configuration between, say, NP 2 and V1

and that this configurational relationship is responsible for determining

the function of the NP in a sentence. In other words, we have to assume



33

that NP 2 can only be an object because its relationship is with the verb

and in the same way we have to believe that NP 1 cannot be the object

because its relationship is not with the verb, but with something else.

As regards the assumption that NP2 forms a structural unit with the verb,

there are a number of constituency tests which we can use to prove such

an assumption to be right. The rest of the discussion is intended to

show this.

The assumption that the verb and the NP preceding it form a structural

unit implies that there is cohesion between the two, and hence, any thing

that affects one also affects the other. The following examples show this

to be true.

12(a) [Thlluu-n hoolaa gurgur-e I

ST_nom sheep buy-pf.
'T bought a sheep'.

(b)? [Tulluu-n gurgur-e]

ST_nom buy-pf
'T bought'.

13(a) [Thlluu-n Dabalaa-daf aannan kenn-e- (ef)]

T-nom D to milk give-pf-to
'T gave milk to D'.

(b)? [Thlluu-n Daballa-daf kenn-e-(ef)]

ST D-f give-pf-to
'T gave to D'.

14(a) [Tulluu-n barsiisaa hin-ta?-a]

S
teacher cm-become-inipf.
'T will become a teacher'.

(b)? [Tulluu-n hin-ta?-a]

ST_nom cm-become-impf.
'T will become'.
 34

15(a) [Thlluu-n wattadarii fakkaat-a]
ST
soldier look-impf.
'Tulluu looks like a soldier'.
(b)? [Thlluu-n fakkaat-a]
S
T-nom look-impf/
'Tulluu looks like'.

The (b) structures of (12-15) cannot initiate discourse. They may

be used only in strictly defined contexts set by sentences with structures

such as those in (a). This suggests that the verbs in such structures

cannot occur without a preceding NP under normal circumstances, that is,

in the unmarked case. This implies that such NPs are indeed strictly

subcategorized arguments and hence cannot be omitted. Notice, however,

that according to Owens' base rule such structures as those in (b) above

should have been possible, since the rule allows such NPs to be optional.

The fact that the (b) structures in (12-15) are marginally acceptable

may also indicate that the NPs in question are not only obligatory but also

form a structural unit (a VP) with the verbs. In other words, it might be

because of the absence of a fully-fledged VP that such structures are

possible only in restricted contexts. thether this is sheer intuitive

guesswork or something empirical will have to be proved. And there are

syntactic facts which can justify the claim that this is more than intuitive

appeal.

In (16) below, the verb and the preceding NP delete on identity with a

parallel constituent in a running discourse. Neither the verb nor the NP

alone can undergo this process. The ungrammatical structures show this.

16(a) [Tulluu-n (aannan hin - 'aalat-a]]. [Fayyiisaa-n

S.
T-nom milk cm - like-impe. F-nom
[aannan hin - aalat-a]
milk cm. like-impf.
'Tulluu likes milk. Fayyiisaa likes milk'.


35

(b) [Thlluu-n (aannan hin - 'aalat-aJ]. [Fayyiisaa-n-iss ( 1]

ST S
milk cm - like-impf. too
'Tullu likes milk, so does Fayyiisaa'.

(c)* [Thlluu-n [aannan hin - aalat-a]]. [Fayyiisaa-n (aannan [-1]

ST_nom milk cm like-impf. SF_nom milk

(d) * [Thlluu-n [aannan hin - 'aalat-a1]/ [Fayyiisaa-n [[ —1

ST N
milk cm like-impf. 5F-nom
V
hin - jaalat-aJ}
cm - like-impf.

This situation suggests that gapping can operate only on strings of

words which form a single constituent. If Oron were a non-configurational

language, we would not have such restrictions on the application of the rule,

and structures like (16 c-d) would also have been well-formed.

Further support for this argument comes from pro-forms. The verb

and the NP as a unit but not as independent entities may serve as an

antecedent to a pro-form, which is again a possibility only if the two

(the verb and the NP) are in a configurational relationship. Consider the

following:

17. Speaker A: [Tulluu-n (abbaa - saa rukkut-e]1

ST father - his hit - pf.
'Tulluu his hi rather'.

Speaker B: [Tulluu-n! inn-i [akka-na] hin-god-u]

ST_nom he-nom as - this neg-do-impf.
'Tulluu! He w6n't do like this'.

In B /akka-na/<akka-kana/ 'like this' refers only to the material in

the inner bracket in A. The reference cannot include Tulluu because the

latter is still present in B as subject of the clause. In other words,

/akka-na/ can stand only in place of the material which is present in A but

missing in B. This material, its antecedent, consists of the NP/abbaa-ssa/

'his father' and /rukkut-/ 'hit'. Since antecedent-'anaphoric' relationship

 36

5
is one to one, we have to conclude that the antecedent /abbaa-saa rukkut-/

'hit his father' forms a single constituent.

As a single constituent, /abbaa-saa rukkut-/ 'hit his father' can be

conjoined with other similar structures. Hence (iSa) but not (18b) or (C).

18(a) (haada - saa arrabs-e] -ett tabbaa - saa rukkut-el

Dther - his insult-pf and ather - his hit-pf
'(he) insulted his mother and hit his father'.

(b)* [haada - saa arrabs-e] -ett (abbaa-saa]

VP
mother - his insult-pf and father-his

(c)? [haada - saa arrabs-e] -ett [rukkut-e]

VP V
mother - his. insult-pf and hit

The fact that only /abbaa-saa rukkut-/ 'hit his father' falls into a

structure of co-ordination with /haada-saa arrabs-/ 'insult his mother'

shows that it is a single constituent and that only it as a unit, but not

any of its internal elements, can occur as a conjunct on a par with /haada-

saa arrabs-/ 'insult his mother'. The ungrammaticality of (b) and the

marginality of (c) are indicative of this restriction. In fact, the

situation in (a) clearly shows that /rukkut'-/ 'hit' requires a preceding NP

as a lexical property, just as /arrabs-/ 'insult' demands one. If Oromo

were a non-configurational language, the kind of conjoined structure we

would have had might have been one which involved lexical (X°) elements,

but not phrasal (X') structures. In other words, we would have had

structures like (19) but not (18a).

19. [arrabs-el -ett (rukkut-el

Vinitf and Vhjtpf
'(he) insulted and hit'.

That there is a constituent VP is also evident from rules involving

movement. The verb and its NP move as a unit whenever the rule applies.

Hence (20a), but not (20b), is grammatical.



37

20. (Tulluu-n farda gurgur-e]

ST_nom horse sell-pf
'Tulluu sold a horse'.

(a) (farda urur (Tulluu-n t]]

Shorse sell-pf ST_nom
Literally, 'Horse sold, Ttilluu'.

(b)* [gurgur-e [Tulluu-n farda ti]

S S
sell-pf T-nom horse

Since in non-configurational languages, word order is free, structures like

(20 , would be perfectly acceptable. But in Oromo, as stated earlier on

and as the example (20 b) shows, the verb is always final.

Finally, elements which form a single constituent can hardly be

interrupted. This would have been possible if it had been the case that

the NP /abbaa-saa/ 'his father' and the verb were not a single VP constituent

in (21)

21(a) (i) [Tulluu-n [duguinaa-n] [abbaa saa rukkut-e]]

S,r. truth-in VPfather_his hit-pf
'Tulluu certainly hit his father'.

(ii)? Thlluu-n (abbaa - saa [duguixa-n] rukkut-e]]

T-nom ather - his truth-in hit-pf
Literally, 'Tulluu his father certainly hit'.

(b) (i) [Tulluu-n [akka na-itti-fakkaat-u] [abbaa - saa rukkut-e]]

ST_nom as me to - seem-impf father - his hit-p
'Tulluu, as it seems to me, hit his father'.

(ii)? [Thlluu-n (abbaa - saa [akka na-itti-fakkaat-u] rukkut-e]]

ST_nom ather - his as me to seem-impf hit-pf.
Literally, 'Tu.11uu his father, as it seems to me,hit'.

Cc) (i) [Tulluu-n (dugumaa-n] [barsiisaa hin-ta ? -a] I

ST truth in eacher am become-ixripf
'Tulluu will certainly becoth [a] teacher'.


38

(ii)? [Tulluu-n [barsiisaa (duguniaa-n] hinta?-a]]

ST_nom eacher truth-in cm-become-impf
'Literally, 'T teacher, certainly will become'.

The structures in (ii) are marked though they may riot be excluded as

being ungrammatical. In each case, the NP is followed by a pause, as

the comma shows. In copular structures, the adverbial /duguniaa-n/

'certainly' cannot occur between the verb and the predicate NP. The

structures below are illustrative of this.

22(a) [Tulluu-n (duguznaa-n]

gooftaa-da]]
S VP
T-nom truth-in lord is
"fliIluu is certainly a lord'.

(b)* [Thlluu-n (gooftaa [dugumaa-n]

. - dal]
ST ord truth-in is

(c)* [Tulluu-n [gooftaa (dugumaa-nJ tur-e]]

S VP
T-nom lord truth-in be-pf.

Such structures illustrate the extreme case where an NP can be bound

to the verb which subcategorizes it in forming a syntactic unit. This

would not have been the case if Oromo were a non- configurational language as

scrambling is a feature of such languages.

From the data observed thus far, the most plausible conclusion to

draw would be that this languaqe has a VP, for in all the cases observed

the verb and its preceding NP form a string which meets Radford's (1981:69)

definition of a constituent which runs as follows:

A given string' of elements j a constituent just in case

it has one or nre of the following properties:
(i) It behaves distributionally as a single structural
Unit - i.e. it recurs as a single unit in a variety
of other sentence positions.
(ii) It can be co-ordinated with another similar string.
(iii)It does not readily permit intrusion of parenthetical
elements internally...
(iv) It can be replaced by, or serve as the antecedent of,
a proform.
Cv) It can be omitted, under appropriate discourse conditions.
 39

As has been shown throughout, strings like /abbaasaa rukkut-/

'hit his father' do meet these conditions, which means that they are VPs.

If this is the case, then it has to be concluded that Oromo is a

configurational language.

As stated earlier, Owens' proposal is based on the analysis of

causatives. Causativization as a morphological process has the effect of

increasing the number of the internal arguments of verbs. A transitive

verb such as /mii -/ 'wash' which strictly subcategorizes only one

argument, can get an additional argument whenever it has the causative

affix /-siis-/ This will be dealt with in due course in some detail in

Chapter Three, but for the purpose of the argument here, let us compare the

following:

23(a) [Tulluu-n wayaa mii-e1

ST_nom. clothes wash-pf.
'Tulluu washed clothes'.

(b) [Thlluu-n Dabalaa wayaa mii-isiis-e]

ST_nom. D-acc clothes wash-cs-pf.
'Tulluu caused Dabala to wash clothes''.

In structures such as (23b), the internal argument which is the

farthest from the verb may be optional since structures such as Cc) below

are possible:

(c) [Thlluu-n wayaa mii-isiis-e]

ST_nom. clothes wash-cs-pf.
'Tulluu caused (cofneone) wash clothes', or
'Tulluu got clothes washed'.

But the NP /wayaa/ 'clothes' which is subcategorized in accord with the

inherent property of the verb /mii-/ cannot be so omitted since such a

structure as (d) would be only marginally acceptable

 40

(d)? [Thlluu-n Dabalaa mii-isiis-e]

ST_nom. D-acc. wash-is-pf.
'Tulluu caused Dabalaa to wash'.

The same would be true with structures of causatives built on

intransitive verb stems if the complement had been missing.

24(a) [biaan-i damf-e]

St boil-pf.
'('rhe) water boiled'.

(b) [Tulluu-n biaan damf -is-el

5T-nom. water-acc. boil-cs-pf.
'Tiilluu boiled [th'} water'.

(c)? [Tulluu-n damf-is-e]

S
T-nom. boil- is-pf.
'Tulluu boiled'.

As the situation in (23-24) shows, it may be possible to say in

general terms that no argument of a verb is subject to omission. Strictly

speaking, however, one may say that the left-most argument of a transitive-

based causative verb may be omitted as (23c) indicates. But this induces

us only to make distinctions between types of arguments but not to say that

such arguments do not fall into configurations with the verbs. Both

types which we may call 'inherent' and 'acquired' arguments, form together

with the verb a single structural unit which satisfies all the conditions

of constituency we have referred to earlier.

If the arguments presented thus far are valid, then, the base rule

for Oron should be as follows (subject to changes to be made subsequently,

see 5.3):

II S > (NP) VP

Assuming this to be correct, then the next thing we need to consider

is the possible hierarchies within constituents and to determine the value

 41

of n in X. The rest of the study is in the main about this. For

purposes of ease of exposition the study is divided into the following

chapters.

The chapter which follows immediately deals with the identification

of X, which is an abstraction of the possible major lexical categories

of the language. Chapter Three examines the minimal projections of X,

whereas Chapter Four discusses its intermediate and maximal projections.

These two chapters will determine the value of n. The fifth chapter will

concentrate on the internal structures of two types of clausal complements

in the light of the discussions in Chapters Three and Four, together with

a discussion on the rule 'move a' and the principles governing it. Finally,

we will have a chapter on the specifier of X and a summary of the major

points discussed throughout the study. This is followed by an appendix

of the vocabulary items used in the illustrative structures throughout

the study.

1.4 Transcription

The transcription used throughout is largely systematic phonemic

(i.e., it is relatively morphophonemic). Vowel and consonant length is

indicated by doubling the letter symbols. In nouns lengthening of a stem

final vowel shows the genitive relationship.

Although Oromo is a tonal accent language, this has not been included

in the transcription for it has no direct relevance for the present purpose.
 42

NOTES TO CHAPTER ONE

1. He also calls this rule as function mapping rule.

2. Actually recent work on Japanese syntax has shown that there is a

VP in this language (cf. Shishido, N., 1985), as quoted in Nagai, N.
(1985).

3. This does not mean that adjacency is strictly observed in all cases.
Accusative case assignment does not seem to obey this condition
(cf. Chapter 5).

4. This element /hin-/ appears as a prefix to main verbs in declarative

clauses. It seems to be similar to the Arbore /?an"?in/, which
Hayward (1985) calls preverbal selector. For the present purpose
I will call it simply a clause marker (cm) though it might be argued
that it is a kind of complementizer (cf. Chapter Five).

5. It may be possible for a proform to have splitting antecedents in

which case it gets its reference from two sources (cf. Chomsky 1981).
 43
CHAPTER 'IS1O

LEXICAL CATEGORIES

2.0 Introduction

In this chapter the lexical categories of Oromo will be identified

and subsequently classified into types and sub-types. Both the

identification and the classification will mainly be based on syntactic

facts. Morphological evidence will be cited as supplementary support for

arguments for which the syntactic evidence appears insufficient.

The identification is necessary since the categorial rules which we

will be formulating throughout the chapters are the syntactic expression of

a set of lexical units. It is the development of such basic units into

syntactic categories that the categorial component deals with; which

means that for an adequate description of the grammar of a language, the

identification of such basic units and the classification of them into

sets of categories is absolutely essential. In other words, in order to

formulate a categorial rule that expands a syntactic category such as, for

example, an NP, we have to make sure first that there is a lexical category N

distinct from any other category or.sets of categories.

From the descriptions made by early grammarians, such as Hodson and

Walker (1922), Moreno (1939), Nordfeldt (1947), etc., one may get the

impression that Oromo has all the categories known to exist. The description

of such grammarians with respect to categories could be said to be based

largely on semantic notions, although some have tried to take formal

properties into account in their classifications of forms into sub-classes.

Moreno (1939:98), for example, recognizes only one category which he calls

particelle (=Particle) for what the others would consider as independent

categories of adverbs, prepositions, conjunctions and interjections.

 44

Formally the individual items in such "categories" are characterized by

their having no inflectional or derivational ixrphology. In other words,

they belong to a class which, in contrast to other classes, is closed or

unproductive.

In our attempt to establish the type of lexical categories that

characterize Oromo, we will argue in favour of Moreno's attempt to use

formal criteria for the set of categories he has established.

The identification of lexical items and their classification into

categories involves recognizing forms which recur throughout a text or a

running discourse as independent lexical units having distinct phonetic

shapes and meaning. Those items "that have essentially the same

distribution and that recur as a structural unit in a variety of different

sentence positions and sentence types" may then be recognised as belonging

to a category of their own as distinct from "other items which are mutually

substitutable in other positions in the same or different structures in

which the former are excluded" (Radford 1981:48). In other words, the

classification of forms into form classes or categories takes into account

the positions where the forms in guestion either co-occur or substitute for

one another in structures of sentences.

Each such class of items may have a number of subdivisions which is

based again on facts pertaining to distributional or formal irregularities,

or on deviations from what is understood to be characteristic of the class

as a whole. It is in the light of such facts that we will attempt to

establish (or re-examine) the lexical categories of Oromo.

2.1 Nominals

Forms which can occur in the position of /hoolaa/ 'sheep' in structures

of the type in (1) will be considered as belonging to the category of nouns.



45

1(a) ([hoolaa] bit-uu-n] gaarii-da

sheep buy-to-nom. good-is
S•
'To buy/buying sheep is good

(b) [[hoolaa] bay?ee

sheep many
'Many sheep.

(C) [[hoolaaJ kanaJ

sheep this
This sheep'.

(d) [akka hoolaaj

'tike sheep'.

In the structures in (1) /hoolaa/'sheep' may be replaced by such forms

as /farda/ 'horse' or /mana/ 'house' but not by others like /guddaa/ 'big'

or /deema/ 'go'. Hence (2), but not (3), is grammatical.

2(a) ((farda] bit-uu-n] gaarii-da

horse buy-to-nom. good-is
'Buying horse is good'.

(b) [(mana] bay?ee]

house many
'Many houses'.

(c) [akka [fardaa]]

'Like horse'.

1
3(a)* ((guddaa] bit-uu-n] gaarii-da
big buy-to-nom. good-is

(b)* [[deeua] bit-uu-n] gaarii-dal

go buy-to-nom. good-is

(c)* (akka [deema]]

like go.

Personal and interrogative pronouns may also be included in the

category since structures like those in (4) are grammatical.

 46

4(a) (akka [isaa]]

like him

(b) [akka [iii]]

like her

(c) (akka [eei?iu]]

like who

(d) [[maal] —in]

what with.

But they do not occur in all of the positions indicated for /hoolaa/

'sheep' in (1). In this respect, they are similar to proper nouns which

are also restricted to certain positions.

5(a) [(fard-i] kun-i] guddaa-da

horse-nom. this-nom. big-is
'This horse is big'.

(b)* [[at-i] kun-i] guddaa-da

you-nom. this-nom. big-is

(c)* [[Tulluu-n} kun-i] guddaa.-da

T-nom. this-nom. big-is.

6(a) ([fard-i] guddaa-n] du?-e

horse-nom. big-nom. die-pf.
(b)* [[at-i] guddaa-nl du?-e
you-nom. big-nom. die-pf.
(c)* [[rnaal-i] guddaa-n] du?-e
what-nom. big-nom. die-pf.

(d) * [[Tulluu-n] guddaa-nl du ?-e

T-nom. big-nom. die-pf.

The ungrammatical structures suggest that pronouns and proper nouns

cannot replace nouns like /farda/ 'horse' in positions preceding ndifiers

or specifiers. The same type of substitutional restrictions exist between

forms like /farda/ 'horse' and others such as /deemuu/ 'going/to go' in some


47

positions. Hence (7b) but not (8b) is possible.

7(a) (fard-i] gaarii-da

horse-nom. good-is
'Horse is good'.

(b) (deem-uu-n] gaarii-da

go-to-nom. good-is
'Going/to go is good'.

8 (a) [[fard-i] guddaa-n] gaarii-da

horse-nom. big-nom. goodis
'(A) big horse is good'.

(b) * [[deem-uu-n] fagoo-n] gaarii-da

go-to-nom. far-nom. good- is

But in the position of /farda/ 'horse' in (7a) above, all the forms

considered so far freely substitute for one another. This is indicated

in (9) below.

9(a) [Tullu-nI gaarii-da

T-nom. good-is
'Tullu is good'.

(b) (at-i]
gaarii-da
you-nom. good-is
'You are good'.

Cc) [ira-a-uu-n) gaarii-da

exist-mid-to-nom. good-is
'Living is good'.

Taking their syntactic similarities into account, we will consider

all such forms as sub-classes of one major category. We will use the

term nominal to refer to the category, and the terms nouns, pronouns,

infinitives, etc., to refer to the sub-classes when such distinctions

are necessary.


48

2.2 Verbs

Forms which occur in clause-final positions belong to a category which

is distinct from that of nominals. Consider the following examples:

10(a) [Tulluu-n farda bit-el

ST_nom. horse buy-pf.
'Tulluu bought a horse'.

(b) (Tulluu-n kaleessa duf-el

ST_nom. yesterday come-pf.
'Tulluu came yesterday'.

(c) [Tulluu-n deeraa-dal

STno tall -is
'Tulluu is tall'.

No forms other than those belonging to this category can occur in the

position indicated for /bit-/ 'buy', for example, since structures such as

those in (11) are ungrammatical.

ll(a)* Tulluu-n farda [namal

T-nom. horse man

(b)* Tulluu-n farda [guddaal

T-nom horse big

On the basis of the type of constituents they are associated with,

verbs may be divided into a number of sub-classes. Anticipating the

discussions about such sub-classes in the next chaflter, we will use the

term verbal to refer to the category as a whole.

The two categories, nominals and verbals, are central to the syntax

of any language. They constitute what Lyons (1977:430) calls the nucleus

of any proposition. They are hence believed to belong to the strong

universals of human language. This is in contrast to the position of other

categories which a particular language may or may not have as independent

classes. Such categories include adjectives and adverbs.

 49

Using the same type of syntactic devices we have employed in the

identification of the two major categories, we will try to answer the

question whether or not Oromo has such peripheral categories.

2.3 Adjectives

In order to recognize a category adjective on a par with that of

nominals or verbals, we need to have elements whose distribution is different

from those of verbals and nominals. In other words, there must be a position

in a constituent structure which is exclusively for adjectives.

One such position in a clause is the slot following the noun mama!

'man' in (12) below:

12. [Tulluu-n [nama]guddaa]J hin-beek-a

ST _nom NPman big cm-know-impf.
'Tulluu knows a big man'.

In the position of /guddaa/ 'big' above, verbals cannot occur without

the structure being ill-formed, because as stated earlier, their position

is clause-final. In this position only forms like /gaarii/ 'good', or

/sooressa/ 'rich' can occur as shown in (13) below:

13(a) [Tulluu-n [nama [gaarii]] hin-beek-a]

ST_nom. man good cm.- know-impf.
'Tu!luu knows a good man'.

(b) [Tulluu-n [nama [sooressaJJ hin-beek-a]

ST_nom man rich cm.-know-impf.
'Tulluu knows a rich man'.

(c)* [Tulluu-n mama [bit-el] hin-beek-a]

ST_nom. NPman buy-pf. cm.-know-impf.

In some types of structures, nouns may occur in the position of the

adjectives in (13). Consider the following, for example:

14(a) 2
[Tulluu-n [nama [Wallaggaa]] arg-eJ
ST_nom. man of-Wallaggaa see-pf.
(literally, 'Tulluu saw a Wallaggaa man' (A man from Wallaggaa).
 50

(b) [Tulluu-n [afaan [Oromo]] hin-beek-a]

ST_nOm. NPmouth of-Oromo cm.-know-impf.
(literally, 'Tulluu knows mouth of Oromo',
'Tulluu knows the Oromo language'.

Such structures may suggest that nouns and adjectives have similar

distributions and hence they may be treated as belonging to one major

category. Before we adopt this as a plausible conclusion, we need to

consider a number of other positions and check if they substitute for each

other in such positions as well.

15(a) [Tulluu-n [bay?ee [deeraa]] -cia]

ST_nom very tall -is
'Tulluu is very tall'.

(b)* [Tulluu-n [bay?ee [nama]] -da

ST_nom. very man is.

In structures such as (15) only adjectives can occur following the

degree word /bay?ee/ 'very'. The ungrainmaticality of (l5b) illustrates

this. In the same manner, nouns but not adjectives can occur preceding

specifiers such as demonstratives or numerals. Consider the following:

16. i(a) [Tulluu-n [naina [kana]] arrabs-e]

ST_nom NPman this insult-pf.
'Tulluu insulted this man'.

(b)* (Tulluu-n [guddaa [kanal] arrabs-e

5T-nom. big this insult-pf.

ii(a) [Tulluu-n (nama [lama]] arg-e]

S NP
T-nom man two see-pf.
'Tulluu saw two men'.

(b)* [Tulluu-n [guddaa [lama]] arg-e]

ST_nom. big two see-pf.

Furthermore, nouns but not adjectives can occur in subject or object

positions as we can gather from the following structures:

 51

17. i(a) [[nam-ni] [du?-e]

S man-nom. die-pf.
'(A) man died'.

(b)? [[guddaa-n] du?-e3]

S big-nom. die-pf.
?'big died'.

ii(a) [Tullu-ri [hoolaa} bit-el

ST_nom. sheep buy-pf.
'Tulluu bought a sheep'.

(b)? [Tulluu-n [guddaal bit-e41

ST_nom big buy-pf.
'Tulluu bought big'.

In addition to the syntactic differences noted above, there are also

morphological features which distinguish adjectives from nouns. Such

features include gender markers. Only adjectives are characterized by

the presence of gender markers. Observe the paradigms below:

18. mas. fern, gloss

gurraa-a gurraa-ttii 'black'
daala-a daala-ttii 'grey'
soore-ssa soore-ttii 'rich'
ham-aa ham-tuu 'evil'
deer-aa deer-tuu 'tall'
gudd-aa gudd-oo 'big'
furd-aa furd-oo 'fat'

In addition to this, adjectives but not nouns reduplicate their

initial syllables to show nuniber. This is again demonstrated by the

examples below:

19. . ol. doss

gurraaCca gu-gguraaa 'black'
deeraa de-ddeeraa 'tall'
diimaa di-ddiimaa 'red'
•1 ,
abaa j a-j .abaa 'strong'.
 52

In nouns the plural is indicated by the suffix /-oota/.

In Baye (1981) I argued that adjectives and nouns could be treated

as sub-classes within the general category of substantives. This was

based on some morphological similarities that the two categories seem to

share. For example, both exhibit the feature case by the same affix I-ni.
Furthermore, adjectives may use the suffix /-oota/ in addition to or instead

of the type of reduplication described above in connection with the feature

number, as in the following:

20. nom. gloss

hoolaa hoolaa-n 'sheep'
guddaa guddaa-n 'big'

nama nam-oota 'men'

gurraaa guggurraaa/
gurraa-oota/
'black'
gu-ggurraa-oota
'blacks'

Such apparent similarities are, however, attributable to agreement

which holds between heads and non-heads in structures of modification. As

shown in earlier examples, adjectives and specifiers agree in number, gender

and case with their heads in NP structure. In other words, such features

are basically characteristic of nominals and that other elements acquire

them only if they form syntactic units with nouns. The argument becomes

weak particularly in the face of the syntactic facts we have observed. A

more reasonable approach would have been one which treated them as distinct

categories.

Another question which we may need to pay attention to is whether or

not adjectives belong to the category of verbals. This is particularly

important given the fact that languages which in the terminology of Dixon

(1977:20) are adjective-deficient tend to use intransitive verbs to express

all sorts of adjectival concepts.

 53

Oromo does not seem to belong with such languages. It has independent

adjectival forms for all the attributes 5 Dixon mentions. There is, however,

one problem that emerges from the syntactic criterion we have set up for

distinguishing verbals from non-verbals. It has been stated that verbals

occur in clause-final position, which implies that no other categories can

occur in this position. But in structures such as the following, it appears

that adjectives can also occur in this position without the structures being

ill-formed.

21(a) [Tulluu-n gurraaa)

S
T-nom tall-is
'Tulluu is tall'.

(b) [Tulluu-n sooressa]

ST rich-is

The question that arises from consideration of structures such as these

relates to the syntactic status of the final elements. We have to prove

that they are predicates having Tulluu as an argument or that they are

complements of a phonetically empty copulative verb. From other copular

structures it appears that such elements can meaningfully be analysed as

complements rather than as predicates, i.e., verbs. Consider the following

examples:

22(a) [Tulluu-n deeraa-da]

STnom tall-is
'Tulluu is tall'.

(b) [Tulluu-n guddaa-da]

ST_nom. big-is
'Tulluu is big'.

(c) [Tulluu-n hamaa-da]

STno bad-is
'Tulluu is bad/evil'.
 54

In these and other similar structures, there is an overt copula /da/

following the adjectival complements. Such structures would be ill-formed

if they occurred without this verb. The structures in (21) may be said to

have similar representations underlyingly. There is some evidence in

support of this claim. The negative counterparts of both (21) and (22)

are identical.

23(a) [Tulluu-ri gurraaaa miti I

S
T-nom. black not-is
'Tulluu is not black'.

(b) [Tulluu-n deeraa miti]

5T-nom tall not-is.

If the forms in question in (21) were verbals, then their negative forms

would not have contained /miti/ but /hin-n-/, as is always the case with other

verbs, and that would have caused ungrammaticality as in (24b-c).

24(a) [Tulluu-n hin-duf-n-eJ

ST_nom. neg-come-neg-pf.
'Tulluu did not come'.

(b)* [Tulluu-ri hin-gurraaa-n-e I

ST neg-biack-neg-pf.

(c)* [Tulluu-n hin-sooressa-n-e

T-nom. neg-rich-neg-pf.

Such structures as (21) are possible only in the imperfective aspect.

In the corresponding perfective aspect, there is always a form in the position

where we would have expected /da/ 'be'.

25(a) [Tul].uu-n deeraa tur-e]

ST_nOm. tall be-pf.
'Tulluu was tall'.

(b) [Tulluu-n gurraaa tur-e]

S
T-nom. black be-pf.
'Tulluu was black'.
 55

25(c)
[Tul].uu-n sooressa tur-e]
ST
rich be-pf.
'Tulluu was rich'.

Notice that if forms like /deeraa/ 'tall', or /gurraa/ 'black', were

verbals, they would have had the various spectual features marked on them.

In other words, the structures in (25) would have been ill-formed. But

they are not. In fact the opposite would have been true if they had occurred

with the adjectives showing the aspectual features. Observe (26), for

example, corresponding to (25 a)

26.* [Tulluu-n deer-el

ST_nom tall-pf.

Structures such as this would be perfectly grammatical if they occurred

in the form in (27):

27(a) [Tulluu-n deer-at-el

ST_nom. tall-mid-pf.
'Tall got tall'.

(b) [Tulluu-n gurraa-at-e]

S
T-nom. black-mid-pf.
'Tulluu got black'.

(c) [Tulluu-n aba-at-e]

ST_nom. strong-mid-pf.
'Tulluu got strong'.

Such forms are not adjectives, however. They are verbs derived from their

corresponding adjectival sources. The manner of the derivation and the

categories involved will be discussed in the next chapter. For the moment,

we only need to note that the aspectual element I-el is attached to the verbal

stems /deer-at-/, /gurraa-at-/, etc., as we would expect, but not to the

adjectival roots /deer-/ 'tall', or /gurraa-/ 'black'.

If we are on the right track so far, then the next thing we need to do

is explain the contexts in which copular structures are possible without a

 56

surface copulative verb. First of all, it is not only adjectival

complements of the type in (21) which occur without a following copula;

certain nouns also behave like this (characterized by the same context).

This.is evidenced by the following structures.

28(a) [Tulluu-n nama]

ST man-is
'Tulluu is (a) man'.

(b) (kun-i mana]

Sthin house-is
'This is (a) house'.

But notice also the following:

29(a) [Tulluu-n barsiisaa-da}

ST teacher—is
'Tulluu is a teacher'.

(b) [aaltuu-n dubart-ii-da

Sy
(-nom. woman-is
'aaltuu is (a) woman'.

These structures would be ungrammatical if they occurred without the

copula Ida!, just as in the same way those in (28) would have been ill-formed

if they had occurred with it. If we argue that some adjectives are verbals,

we have also to argue the same for some nominals, since they too appear in

positions which appear to be clause-final. This would lead to a lot of

problems for which neither syntactic nor semantic solutions could have been

provided.

The simplest thing to do is to assume the presence of /da/ in the

underlying representations of all such structures and account for its absence

from surface structures in terms of some low-level rules operating in the

derivations of such clauses. In surface structures this element occurs

as a kind of suffix attached to its complement. And it is only in some

instances that it is absent. This suggests that we might need to examine

the phonological shape of those complements in order to tell where the

copula does and where it does not appear. In all the structures considered,

it is when the copula occurs following a complement ending in a short vowel

that it disappears. In other words, the conditioning factor is phonological.

In surface structures, the element appears as an affix. It does not

seem plausible to assume that it is an affix at the underlying level of

representation particularly in view of the fact that other copulative forms

such as /tur-/ or the negative /miti/ occur as independent elements

separated from their preceding complements. We may argue that there is a

copula encliticization rule which attaches /da/ to the complement. This

is followed by one of two processes which results in the disappearance of

the copula from surface structures of the type (21) or (28). It may be

argued that there is a rule which deletes the copula whenever the complement

onto which it has been encliticized is one which ends in a short vowel.

There is some phonological evidence which may be cited in support of this

claim. This comes from the form of the complement. Following the deletion

of the copula, the complements seem to lengthen their final vowel. Let us

observe (21 a) repeated here as (30):

30. [Tulluu-n gurraacca da] >

S
T-nom. black is

[Tulluu-n gurraaa-da] =>

T-nom black is

[Tul luu-n gurraaaa] >

T-nom. black

[Tulluu-n gurraaa1
T-nom. black-is
'Tulluu is black'.
 58

The same may be said to be true of the structures in (28) where the

complements occur without /da/ 'is'.

There is, however, a problem with this kind of argument. According

to Chomsky and Lasnik (1977) no element with semantic content can be

deleted, which means that we cannot explain the absence of /da/ in terms

of such a rule. We may argue that the copula has no semantic content and

to this effect we may even cite some languages which have no copulative

verbs. But even this kind of argument does not seem to help us either

because Chomsky (1981) further constrains the rule by stating that no

element with phonetic content can be deleted without recoverability. The

analysis in (30) is a clear violation of this constraint which Chomsky

assumes to be universal. We have, therefore, to look for an alternative

analysis.

We have said that there is a lengthening of the final vowel of the

adjectival or nominal complement following the deletion of Ida!. We may

argue that this may make the deletion recoverable, or on the other hand,

we may argue that the change in the quality of the vowel is not a result

of deletion, but that of reduction. We may say that the copula Ida! is

not deleted but reduced to I-al. It is as a result of this vowel that

the complements appear to have lengthened their final vowel. In other

words, what we have is a rule which deletes part of an item and attaches

the remaining part to a preceding element.

The process is not idiosyncratic. It takes place in cirefted

adpositional phrases as well. Consider the following for example:

31(a) [farda deeraa-da-n]

horse tall-is-by
'It is by (a) tall horse that...'
 59

(b) [farda gurraaaa-n]] /<gurraaa-da-n/

horse black-is-by
'It is by a black horse that...'

In (a) /deeraa/ 'tall' ends in a long vowel. The copula Ida!

appears in its full form. In (b) /gurraaa/ 'black' ends in a short

vowel. The copula is thus reduced to I-al and appears as such. (Cf.

Saeed (1982) for a similar kind of reduction in Somali.)

From the data presented thus far, and from the explanation given for

structures which lack a copulative verb in surface structures, a most

plausthle conclusion would be one which recognizes a category adjective

on a par with the other two major categories we have already established.

2.4 Adpositionals

These are elements whose syntactic relation is with NPs or clauses.

They differ from all the three categories we have recognized so far, both

in their forms and distributions. They do not occur in any of the

positions associated with each of the other categories, nor do they show

any of the morphological features the others exhibit. As has been stated

earlier on, the other categories have inflectional affixes which designate

such grammatical features as number gender, aspect, etc. Adpositionals

do not have any such features. They belong to what are called closed

(as opposed to open) classes. In Oromo only the latter have such

morphological features.

In constituent structures, their positions may precede or follow

the category with which they form a syntactic unit. On the basis of this,

they may be referred to by the less general terms preposition or

postpositions. The following are examples of the former:



60

32(a) [Tulluu-n [waaye [ofi-saal] dubb-a-uu hinaalat-al

S NP
T-nom. about head-his talk-mid-to cm.--like-impf.
'Tulluu likes to talk about himself'.

(b) [Tulluu-n [gara [konna-saall deem-uu hin-barbaad-a]

ST_nom. to NPfarm_his go-to cm.-like-impf.
'Tulluu wants to go to his farm'.

In (32) /waaye/ 'about' and /gara/ 'to' occur preceding the nominals.

In the structures below we have instances of adpositions occurring after

nominals.

33(a) (Tulluu-n [[abbaa-saal wain] duf-el

S NP
T-nom. father-his with come-pf.
'Tulluu came with his father'.

(b) [Tulluu-n [(siigguutii] - dan] leenca aj3ees-e]

S NP kill-pf.
T-nom. pistol-with lion
'Tulluu killed a lion with a pistol'.

In both (32) and (33), the adpositional elements cannot occur without

the preceding or following nominals respectively, nor can they be replaced

by elements belonging to any of the other categories.

In the examples above, we have seen pre-/post-positions occurring

with nominals. In what follows we have examples of them occurring with

clauses:

34(a) [[erga [Tulluu-n deem-eel] Fayyiisaa-n duf-el

S S
after T-nom. go-pf. F-nom. come-pf.
'After Tulluu went, Fayyiisaa came'.

(b) [[akka [Tulluu-n deem-eel] Fayyiissa-n duf-eJ

S S
as T-noin. go-pf. F-nom. come-pf.
'As (soon as) Thlluu went Fayyiisaa came'.

(c) [Tulluu-n [[mana-saa gub-a-uu-saa-ti] -f] bay?ee

S S .
T-nom house-his burn-mid-to-his-is-for very
hin-gadde }
cm.- be-sad.
(literally), 'Tulluu is sad because of the burning of his house'.
 61

The structures in inner brackets are subordinate clauses of adverbial functions.

The elements which occur with such clauses have often been called

subordinate conjunctions (Hodson and Walker 1922:116). But these same

elements can also occur with nominals as the following examples demonstrate:

35(a) [boodee [sa?aati tokkao]] deem- ii]

S NP
after hour one go-imp.
'Go after one hour'.

(b) [Tulluu-n (akka (abbaa-saa]] deeraa-da]

S NP
T-nom. as father-his tall-is
'Thlluu is tall like his father'.

(c) [aannan-ni [[n amaa] -f I gaarii-da]

Smilk _nom . NPman_ for good is
'Milk is good for man'.

Such situations seem to suggest that there are no syntactic distinctions

between pre-/post-positions and the so-called subordinative conjunctions.

The same element is called pre-/post-position when it occurs with an NP

and a subordinate conjunctive when it occurs with a clause. In other

words, there is nothing inherent in it which makes it one but not the other,

other than the clausal or non-clausal status of the category which it takes

as its complement. If this is the deciding factor, then assuming with

chomsky (1970) and Jackendoff (1977) that NPs and clauses are similar

both in their internal structures and in the type of syntactic rules they

are subject to, we may argue that the elements which form structural units

with them may as well be similar. In other words, if the distinction

between ?Ps and clauses is so fine that they may be considered as one,

it is possible to say the same about pre-/post-positions and conjunctions

for the reason that they do not exist independent of their nominal or

clausal complements. Hence adopting this position and borrowing a term



62

from Comrie (1981:85), we will continue to call pre-/post-positions and

conjunctions adpositions, and we will also recognize only one category -

the category of adpositionals - on a par with those of verbals, nominals

and adjectivals.

Subordinate conjunctions thus absorbed into and treated as adpositions,

we are now left with only a few co-ordinative conjunctions. Syntactically

such elements do not form a structural unit with a single NP or a clause.

This distinguishes them from adpositionals in general. They occur between

two categories and as such any two categories whose categorial membership

and structural level is the same. Consider the following examples:

36. i(a) [[Tulluu-fi Fayyiisaa-n] duf-an-i]

T. and F-nom. come-pl-pf.
'Tulluu and Fayyiisaa came'.

(b)* [Tulluu-f duf-e]

ST_and come-pf.

ii(a) (buddena xaat-e] --ett (biaari dug-e]

VP

bread eat-pf -and water drink-pf.
'(He) ate bread and drank water'.

(b)? [buddena aat-e] -ett [dug-e]

Vbread eat-pf -and drink-pf.
Ate bread and drank.

In (ib) I-fl 'and' occurs with one NP and as a result the structure

has turnedout to be ungrammatical. In (jib) it occurs between a VP arid

a V rather than between two VPs. The result is again the same as that

of (ib).

In all structures of NPs or VPs, with a co-ordinative conjunction,

there is no head and non-head relationship between the conjunction and any

of the conjuncts. This causes a problem for the theory of X-bar syntax

which has been developed on the assumption that constituent structures are
 63

headed and that the relationship between any two elements within a

constituent structure is that of head and non-head. In the structures

concerned, neither the conjunction nor the NP or the VP constitutes the

head. Jackendoff (1977:50) considers such constructions as exceptions to

the general phrase structure schema he has proposed. It has also been

argued that co-ordinative conjunctions may be left aside as connectives,

logical or pragmatic, essential in the analysis of discourse, since the type

of relationship they show is one which exists across constituents but not

within a constituent. Hence in the analysis of any conjoined structures

one may disregard the (conjoining) elements by taking into account only the

lexical elements which form the constituent within each conjunct. This

position is adopted for the type of analysis envisaged here.

2.5 Adverbials

Adverbs are to verbs as adjectives are to nouns. Their function is

to restrict the action(s) denoted by verbs to certain contexts. The

contexts may be temporal, spatial, causal, etc. Mathews (1981:122

borrowing a term from Tesniere, calls them the 'circumstants' as opposed

to the 'actants', the latter beina restricted to designating the

participants and their actions. In other words, adverbials may be treated

as elements having the function of relating actors and actions to certain

circumstantial contexts.

In what follows, we will examine a number of such contexts and the

manner in which they get expressed.

2.5.1 Time Adverbials

What have traditionally been called adverbs of time can be grouped

into the following two sets:

 134

37 (a) (h)amma 'now'

kaleessa 'yesterday'
har?a 'today'
bor(u) 'tomorrow'
dafinoo 'Monday'

(b) (i.) bara kana

year this 'this year'

duur duur
before before 'long ago'

bara baraa-n
year year-by 'forever'

ya.roo kana-tti
time this-at 'at this time'

'a1kaba-tti
beginning-at 'at the beginning'

(ii) hamma hamma

now now 'frequently'

taka takka
one one 'sometimes'
guyya guyya-tti
day day-by 'daily'

har?a bor(u)
today tomorrow 'always'

Those in set (a) and others like them are lexical in the sense that

they are morphologically indivisible units. They denote a particular

point in time. Their structural relationship is with verbs since their

function is to relate the actions which the latter express to some point

in time. Consider the following examples:

38(a) [Tulluu-n [([bor(u)]) hin-duf-al]

ST_nom. tomorrow cm.come-impf.
'Tulluu will come tomorrow).


65

38(b) [Thlluu-n [([har?a]) [daadii hin-dug-u]]

ST_nom. today mean neg.drink-impf.
'Tulluu will not drink mead today'.

But such forms do also occur in positions where nominals would be

expected. The following are some examples:

39(a) [(har?-i] guyyaa gaarii miti]

S
today-nom day good not-is
'Today is not a good day'.

(b) [[dafinoo-n] guyyaa hoii-ti]

S
Monday-nom. day of-work-is
'Monday is a working day'.

In (39), the forms in the inner brackets occur as subjects of their

respective clauses. Morphologically this is indicated by their having the

nominative case marker /-i-n/. Since case is a feature of nominals, a

possible conclusion about the forms in question would be to say that they

are also nominals. This may be further substantiated by structures of

the type in (40):

40(a) [aannan-ni kun-(i) [[boru-f] [gaarii hin-ta?-a]]]

Smilk_flom this-nom. tomorrow-for good cm. become-impf.
'This milk will be good for tomorrow'.

(b) [Tulluu-n ([har?aa-f] [birri lama hin-barbaad-a]]]

ST_nom. today-for birr two .m-want-impf.
'Tulluu wants two birr for today'.

In the above structures, both /boru/ 'tomorrow' and /har?aa/ 'today'

occur with a following adpositional element. As stated earlier, adpositionals

occur only with nominal or clausal complements, which means that the forms

in question must be nominal for the structures to be well-formed.

Forms like /dafinoo/ 'Monday' can also occur as heads of relative

clauses. This is possible only with forms belonging to the category of

 66

nominals, but it is not the case for adverbials. Observe the following

examples:

41(a) [Tullu-n [dafinoo [darb-el] gara Gimbii deem-e]

S NP S
T-nom. Monday past-pf. to Gixnbii. go-pf.
(literally), 'Tulluu went to Gimbii on Monday which has
passed'.
'Tulluu went to Gimbii last Monday'.

(b) [Tulluu-n [dafinoo [duf-u]] hoolaa hin-bit-a]

5Tnom. Monday Scome_impf. sheep cm.buy-impf.
'Tulluu will buy sheep next Monday'.

Going back to those in set (b),. we may treat them in the same manner

we have treated those in set (a). The only difference between the two sets

is the fact that those in set(b)(i))are phrasal whereas those in (i) are

lexical NPs.

42(a) [Tulluu-n [[barana] [hoolaa hin-bit-u]]]

ST_nom. year-this sheep neg.buy-impf.
'Tulluu won't buy sheep this year'.

(b) [[barr-i kun-iJ bara gadee-a]

year-nom. this-nom. year bad-is
'This year is a bad year'.

These structures are parallel to those in (39). /bara-na/ 'this year'

occursin an adverbial position in (a) and in a nominal position in (b).

Its function is also that of an adverb in the former and that of a noun in

the latter.

Those in (b Ii) are what are called frequency adverbs. Their function

is to show the number of times an action takes place within a period of time.

In this respect, they may be treated along the same lines as other forms

which show degree or intensity. Compare the following examples:

43 (a)(i) (Thlluu-n [guyya-guyya [hin-kaat-a]]]

ST_nom. day day cm.run-impf.
'Thlluu runs always'.
 67

43 (a) (ji) [Thlluu-n [taka takka [hin-duf-a]]]

ST_nom one one cm.come-impf.
'Tulluu comes sometimes'.

[Tulluu-n [bay?ee furdaa] -da]

b S AP
T-nom. very fat-is
'Tulluu is very fat'.

[Tulluu-n [hoolaa bay?ee] 8 kab-a]

ST_nom. NPsheep many has-inipf.
'Tulluu has many sheep'.

The frequency adverbs in (43 a) are structurally related to the verbal

heads just as in the same way /bay?ee/ 'very/many' is related to the

adjectival and nominal heads in (b). In each case, what is indicated is

the frequency, degree or quantity of an entity or of an attribute or an

action. From this, we may infer that frequency adverbs are similar to

degree or amount forms and hence they may be treated as belonging to the

same category we shall propose for the latter.

Now going back to those forms in sets (a) and (b), we have said

that they belong to the category of nozninals. If this is so, then we have

to answer a question concerning their functional status as adverbs. Nouns

are characteristically understood as playing the roles of agents and

patients. Syntactically this is reflected by their occurrences in subject

and object positions. The position they occupy in those structures where

they function as adverbs are positions where adpositional phrases or clauses

of various adverbial functions are expected.

One possible way of responding to this situation is to assume that in

the positions where they appear as adverbs, such nouns occur as complements

of abstract adpositional elements. In other words, what appear in surface

structures as NPs are headless adpositional phrases.9



68

If this assumption is plausible, then what we have as adverbs of

time are adpositional phrases.

2.5.2 Place Adverbials

These include forms like the following:

44 i. a(i) 'there'

asi 'here'

keessa 'inside'

aalaa 'outside'

ol(i) up'

'ala ,'

Their position is as shown below:

45(a) [Tulluu-n [a(i)] 3ira]
S
here exist-impf.
'Tulluu is here'.

(b) [Tulluu-n [as(i)] jir-a]

S
here exist-impf.
'Tulluu is here'.

Such forms do also occur in the form they appear in in (ii) below:

(ii) a-itti 'towards there'

as-itti 'towards here'
keessa-tti 'towards inside'
ool-itti 'towards above'
ala-tti 'towards under'

Both from their forms and glosses it seems quite obvious that these too

are'adpositional phrases, which means that forms like /as(i)/ and Ia(i)/

'there' are nominals and not adverbials since as we will observe in the

next chapter, only the former can occur as complements of adpositional heads.

Structures such as those below where such forms occur in subject

position may give further support to this claim.

 69

46(a) [keess-i bal?aa-da]

Sinside_nom good-is
'(The) inside is wide'.

(b) (all-i aggaa-da]

5outernom. good-is
?'(The) outer is good'.

However, forms like /as(i)/ 'here' have very restricted distributions.

They cannot occur as heads of relative clauses or allow adjectives or

specifiers to occur with them. This raises some doubts about their status

as nominals. In fact, there are structures in which they occur as

specifiers rather than as heads. Consider the following:

47(a) [Tulluu-n [biyya as-iiti] duf-e]

ST_nom Pregion here-from come-pf.
(literally), 'Tulluu came from here region';
'Tulluu came from this region'.

(b) [Tulluu-n [ganda a-iiti] duf-el

S Adp.
T-nom. village there-from come-pf.
'Tulluu came from that village'.

The bracketed structures are adpositional phrases. The head is

/-iiti/ 'from'. The rest constitutes the NP complements in which /biyya/

'region' and /ganda/ 'village' occur as heads and /as! 'here' and /a&/ 'there'

as specifiers. In other words, the internal structures of such

constituents is as shown in the tree below:

48.

N SP -iiti 'from'

biyya/ganda as/au
'region/village' 'here/there'

In such structures, /as/ 'here' or /a/ 'there' can be replaced by

the forms /kana/ 'this' and /sana/ 'that'. These are forms which as we

shall argue in the next chapter, constitute a sub-class within the general

class of specifiers.
 70

49(a) [Tulluu-n [biyya kana-ti] duf-e]

ST_nom. region this-from come-pf.
'Tulluu came from this region'.

(b) [Tulluu-n [biyya san-iiti] duf-e]

T-nom. region that-from come-pf.
'Tulluu came from that region'.

From these and other similar structures a more plausible conclusion to

draw would be that forms like /as/ or /a/ are specifiers and that

structures like /a-iti/ and /as-iiti/ are reduced adpositional phrases

derived from a full underlying representation of the type in (50):

50. [ [ [e] [as/au]] (-(i)itil].

AdpNP SP p

The same may be said about forms like /aalaa/ 'outside', /klessa/

'inside', etc. since structures such as the following are possible:

51(a) biyya aalaa-ti

region outside-from
'from a foreign land'.

(b) mana kiessa-ti

house inside-from
'from inside (the) house'.

However, in such structures /aalaa/ 'outside' or /klessa/ 'inside'

may not be treated as specifiers if we restrict the term to deictics such

as /kana/ 'this' and /sana/ 'that', or /as/ 'here' and /a/ 'there'.

It might be argued that they are complements. But this is not very

important. What is important here is that they are part of the NPs which

occur as complements of the adpositional element /-iiti/ and that it is the

adpositional phrase which functions as an adverb of place in structures

such as (52)
 71

52. [Tulluu-n Ibiyya aalaa-ti] duf-el

ST_nom. land outside-from come-pf.
'Tulluu came from a foreign land'.

If what has been said so far is sound, then we may conclude that there

are no lexical adverbs of place.

2.5.3 Manner Adverbials

These are forms connected with questions asking 'how?' rather than

with questions asking 'when?' or 'where?'. They refer to the way an action

has been carried out. In some cases, this may include reference to the

means by which an action has been effected. The following are some examples:

53(a) (Tulluu-n (doksaa-rt dubb-at-e I

S Adp
T-nom. secret-in speak-mid-pf.
'Tulluu spoke secretly'.

(b) [Tullu-n [afaa l-umaa-ii] dubb-at-e]

ST_nom 'cleverness-with speak-mid-pf.
(literally), 'Tulluu spoke with cleverness'.

(c) [Tulluu-n [dug-umaa-n 1 dubb-at-e

S Adp
T-nom. true-ness-in speak-mid-pf.
'Tulluu spoke truly'.

The expressions of manner are adpositional phrases. The adpositional

element is the postpositional enclitic /-n/.

It might perhaps be argued that this /-n/ is a derivational affix

and that the forms in question are adverbs derived from nominal stems.

The suggestion that they are derived adverbs is not implausible, but the

idea that they are derived from nominal sources is difficult to believe.

Semantically adverbs are related more to adjectives than to nouns since

both of them are qualifying expressions of nouns and verbs.

Hence, the kind of derivation we would expect is one which assumes

adjectival rather than nominal sources. This can be verified by facts

from other languages, such as English.

 72

Other expressions of manner include what Hodson and Walker (1922:99)

call balancing of verbs. This term refers to concatenating verbs. The

following are some examples:

54(a) [Tulluu-n [daddaf-eel duf-el

ST_nom. hurry-pf. come-pf.
(Literally), 'Tulluu having hurried came';
'Tulluu caine quickly'.

(b) [Tulluu-n [ab-eess-ee] ho-et-e]

ST_nom. strong-c s-pf. work-mid-pf.
(Literally), 'Tulluu having made strong worked';
'Tulluu worked hard'.

The elements within the inner brackets are verbs. This is easily

deducible from the aspectual features which they exhibit, or from their

clause-final positions in structures such as those below:

55(a) [aaltuu-n inana-ii deem-uu-daaf [daddaf-t-e]I

s,'
ç-nom. house-her go-to-for hurry-f-pf.
'aa1tuu hurried to go to her house'.

(b) [Tulluu-n muaa-saa [5ab-eess-e]]

ST_nom. child-his strong-cs-pf.
'Tulluu made his child strong'.

In either case, what is noticeable is that such forms are not adverbs

though their function is adverbial. They may not be simple verbs either.

They may be analyzed as subordinate clauses with a (pro)nominal subject,

but without a subordinative conjunction. In other words, it might be

argued that structures like (54a) are derived p rom underlying structures

of the type of (56):

56. [Tulluu-n [ [e] ! [Pro daddafe]]] [duf-e]]

S Adp SS VP
T-nom. hurry-pf. come-pf.
'Tulluu came hurried'.

Alternatively such structures might be considered to be conjuncts with

the subject of /duf-/ 'come' deleted on identity with Tulluu. But this
 73

does not seem to be the case because (1) there is no co-ordinative

conjunction linking the conjuncts; (2) the relation between them is not

chronological in the sense that the coming did not take place after the

hurrying, but simultaneously, and (3) the interpretation of /daddaf-/

'hurrying' is not independent of that of /duf-/ 'come'. In other words,

it is not intended to give the sense that Tulluu hurried but concerns the

manner in which he did the coming.

Other adverbial expressions follow the same pattern. For example,

reason and instrumental adverbials employ the same type of adpositional

phrases or clauses as the following examples demonstrate:

57(a) [Tulluu-n [eeboo-(da)an] leena a5ees-eJ

S Adp
T-nom spear-with lion kill-pf.
'Tulluu killed a lion with a spear'.

(b) [Tulluu-n [waan dukkubsat-ee-f] gara mana koia deem-e

S T_nom. MPbecause sick-be-pf-for to-house of-medicine go-pf.
'Tulluu went to hospital because he fell ill'.

The general situation as we have observed so far seems to suggest

clearly that adverbial functions are not lexicalized. They are expressed

in adpositional phrases or clauses. From this may follow a further

conclusion that a lexical category adverb does not exist in this language.

There is only a functional category which gets expressed in the form

which has been described throughout this section.

This is not, however, as straightforward as it may have appeared.

There are some elements which cannot be accounted for in the light of

the discussions presented so far. These include forms like /irriigii/

'perhaps'(?). Though they are insignificantly small in number compared to

the host of adpositional phrases, their existence may be indicative of

the existence of a category adverb, and to that extent we may even need to
 74

recognize such a category. But such a category is not lexically

productive since as we have observed throughout, the various adverbial

functions receive their expression in terms of phrases and/or clauses.

A better alternative would be to recognize a minor category for such forms

and for others which do not have the potential for maximal projections.

We will adopt this alternative in this study.

2.6 Specifiers

The categories we have established so far are major categories.

They can occur as heads of constituents. Among the elements which they

optionally select to form a maximal category are specifiers. Semantically

these may be understood as being entity- or quantity-denoting elements.

The following is simply a list of them, due to be discussed in greater

detail in Chapter Six.

2.6.1 Articles

These are entity-denoting elements within the class of specifiers.

Their syntactic relation is with nominals. They include the following:

i. deictics: kana 'this'

sana 'that'

ii. pronominals:
a. possessives: koo 'my'
kee 'your'
etc.

b. interrogatives: kam(i) 'which'

ee'h ju 'whose'

c. indefinites: tokko 'alone'

wan 'any'
homaa 'none'

All of these are substitutable. As regards their position they follow the

head, as the example below demonstrates:



75

58(a) [farda kana]

NP This horse
horse this

(b) [farda kain(i)?]

NP Which horset.
horse which

(c) [farda tokko]

NP 'A horse'.
horse one

2.6.2 Quantifiers

As stated above, these include all elements which refer to quantity

or amount. Positionally they precede articles when both occur in the

same noun phrase.

59(a) [farda lama]

NPh two 'Two horses'.

(b) [Tulluu-n [fardoota lama-n sana] bit-eJ

ST_nom. horse two ? those buy-pf.
'Tulluu bought those two horses'.

The following are some more examples of such phrases:

60(a) [hoolaa bay?ee]

NPsheep many 'Many sheep'.

(b) [hoolaa homaa]

NPsheep no 'No sheep'.

(c) [hoolaa tinnoo]

NP '(A) few sheep'.
sheep few

The class may be extended to include expressions of degree or

intensity, which also involve the same forms /bay?ee/ and /tinnoo/.

61(a) [Tulluu-n [[bay?ee] deeraa] -da]

ST_nom very tall -is
'Tulluu is very tall'.

(b) [Tulluu-rt [[tinnoo] furdaa] -da]

AP.
T-nom. little far -is
'Tulluu is a little fat'.
 76

Again, these same forms may occur with the corresponding verbal heads

to express intensity or degree. Their position is the same as in (61).

62(a) [Tulluu-n [[bayee] deer-at-el]

ST_nOm. very tail-mid-pf.
(Literally) ?'Tulluu very tailed';
'Tuliuu became very tall'.

(b) [Tulluu-n [[tinnoo] furd-at-e]J

S vp•.
T-nom little fat-mid-pf.
'Tulluu became a little bit fat'.

In (63) below they occur with adpositional phrases with a similar sort

of function that they have with the other categories.

63(a) [Tulluu-n [tbay'?eel suut-uma-tti] mana-saa-rra ba?-e]

ST_nom. A
dPvery slow-ness-with house-his-from leave-pf.
'Tulluu came out of his house very slowly'.

(b)*
[Tulluu-n bay7ee mana-saa-rra ba?-e]
ST
very house-his-from leave-pf.

The grammaticality of (63a) and the ungraimnaticality of (b) suggests

that /bay?ee/ 'very' is structurally related to the manner adverbial

/sutt-uma-tti/ 'slowly', and not to the locative adverbial /mana-saa-rra/

though the latter, like the former is also an adpositional phrase. The

difference may have to do with the inherent properties of the complement

nouns of /suttuma/ 'slowness' and /mana/ 'house'. But this is not very

important. What is important to note is the fact that it does occur with

adpositional phrases of some sort and that the position it assumes is the

same position it occupied in the structures (61-62).

Such structural similarities may give support to the view that the

class of quantifiers includes such forms. Anticipating the discussion

in Chapter Six, we shall take this position here and, whenever necessary,

we shall make distinctions between nominal and non-nominal quantifiers.

 77

2.7 Particles

Not all items, grammatical or lexical, fall neatly into the categories

established thus far. There are a number of elements which cannot be

assigned to any of these categories. The following are but a few examples:

64. -kaa 'then' (?)

yaa vocative(?)
-ilee 'even' (?)
ee 'yes'
ii no'
boo
mee 'please' (?)
]SSOO

Some of these may be significant in discourse analysis. Some may be

expressions of suddenly felt emotions (exclamations). All of them are

distinguishable from forms in all other categories by their inability to

occur as heads or as non-heads in constituent structures. Their relation

may be to an entire utterance or set of utterances, which means that the

type of syntactic rules which govern the relation of elements within a single

constituent may not give a proper account of them.

It is therefore necessary to have a separate category for all such

elements. A category 'particle' may serve this purpose.

The major categories and the category we have proposed for specifiers

are categories which occur as heads or as non-heads in forming maximal

categories such as NPs, VPs, etc. In addition to these, two other categories

may be recognized. These categories are: Infi. for all verbal (infl)ections,

and comp. for complementizers. The former is claimed to be the head of a

sentence (S) and the latter that of S (cf. Chomsky 1981; Stowell 1981).

We will argue along the same lines in Chapter Five. For the present purpose,

we will assume only the categories already established, and we shall deal

with the rules that govern their hierarchical structures.

 78

2.8 Syntactic Distinctive Features

The categories we have recognized may be decomposed into features

which are assumed to be provided by rIG.

The definition of categories in terms of such features is necessary

for making cross-category generalizations which X-bar syntax aims at

achieving.

As stated in Chapter One, Chomsky (1970, 1971) used the features

[±V ±N] to define the four major lexical categories: verbs, adjectives,

nouns and prepositions. Jackendoff (1977) recognizes adverbs, particles,

and modals as minor categories and also makes distinctions among the various

elements within the class of specifiers. In order to define both the major

and the minor categories as distinct from each other and from the various

specifying elements, he introduces the following features:

[± sub.]
[± obj.]
[± comp.]
[± det.J (p.33).

Jackendoff believes that the names of these features have rio theoretical

significance. They are just heuristic devices designed to keep the various

categories separate arid also to enable us to collapse them into some

natural classes whenever this is necessary. The difference between these

features and those introduced by Chomsky lies in the fact that the former

are based on whether a category does or does not have the feature(s),

whereas those of Chomsky are based on what a category is or is not. In

other words, Jackendoff's features are based on co-occurrence restrictions.

In order for a category to be [+ obj.] for example, it must co-occur

with an object NP. The same may be said about the other features. And

to this extent the features may indeed be called syntactic distinctive

features.
 79

But they are not without problems. For example, verbs and prepositions

are characterized by the feature [+ obj.] in contrast to nouns and

adjectives, the latter being 1- obj.]. And in order to distinguish them

from modals and particles, the feature [+ comp.] has been employed. Again

only prepositions and verbs are characterized by this new feature. The

question which arises from this state of affairs relates to the difference

between the two features [+ obj.J and [+ comp.]. How is an object

different from a complement? Or is the feature 1+ comp.] a mere label

introduced for convenience? Jackendoff does not tell us the difference.

But from what he says about the names of the features in general, it may be

inferred that these also are convenient labels introduced to distinguish

the categories in question from those others. This makes the features

rather ad hoc, for as we shall observe in the next chapter, both categories

are characterized by NPs which function as direct or indirect objects. In

other words, there is no distinction between objects and complements at the

minimal level of projection of the categories concerned.

Further problems arise with the feature [+ obj.] itself. This feature

is supposed to distinguish verbs and prepositions from the other major

categories. But it distinguishes prepositions and the class of transitive

verbs only, since only these are characterized by object complements.

This may also suggest that other types of verbs do not belong to the category.

And we may perhaps be forced to include them with the categories which are

[- obj.], that is, with nominals and adjectivals. This would create a more

serious problem at the level where they operate as heads of constituents.

Furthermore it may be argued that the feature [+ obj.1 includes

nominals which are derived from transitive verbs, for they too are

characterized by objects. Compare the following, for example:

 80

65(a) [Tulluu-n [hoolaa bit-eJ

ST_nom. sheep buy-pf.
'Tulluu bought (a) sheep'.

(b) ([hoola bit-uu-nI gaarii-dal

S sheep buy-toling-nom. good-is
(Literally), 'Sheep buying is good'.

/bit-uu/ 'buying' like its verbal source /bit-/ 'buy' is characterized

by an object NP which means that the feature [+ obj.] does not take derived

nominals into account.

The features introduced by Chomsky do not seem to have such problems

since they are based on what categories are. For this reason we may adopt

them for purposes of generalizations that we may need to make in cross-

categorial terms.

As Jackendoff (1977:31) has pointed out, the association of features

with categories should be based on the notion of which categories can go

together to make a natural class. This presupposes the existence of certain

rules which govern some categories but not others. Anticipating the

discussions in subsequent chapters, we shall adopt the following features

for the various lexical categories we have established:

nominals [+N -VI

verbals [+v -N]

adjectivals [-4-N +V]

adpositionals [-N -VI

Regarding the two minor categories of specifiers and particles, we may

introduce the feature [SPEC.] and distinguish them as [+SPEC.] and [-SPEC.]

respectively. And in order to make distinctions between nominal and

non-nominal specifiers, the use of the feature [±N] may be extended. This

goes in line with the argument that specifiers are not lexical categories
 81

in the sense that they do not subcategorize other categories or occur as

heads of constituents, but rather appear as adjuncts to other major

categories (cf. Jackendoff 1977). Hence their association with the

feature of the categories with which they form syntactic units is valid.

Other minor distinctions such as those between articles and quantifiers or

intensifiers may follow the same pattern.

2.9 Summary

In this chapter we have established four major and two minor categories.

The identification has been based to a large extent on syntactic facts.

These categories have been defined in terms of syntactic features. The

definition is believed to be essential for making strong generalizations

about rules which operate across categories.

These major categories have the potential to develop into maximal

categories. The categorial component, which we will be dealing with in

subsequent chapters, is a specification of this potential.

We have also recognized two miner categories: specifiers and particles.

The former comprises all elements which form syntactic units with one or the

other of the major categories. They denote the entity, quantity or

intensity of such categories. The same may not be said about the latter.

They do not seem to co-occur with a particular lexical category to form a

syntactic unit in which they constitute either the head or the non-head.

Their relation may be said to be with an entire utterance or set of

utterances, in which case their behaviour may not be captured by syntactic

rules. We may perhaps need to consider them as discourse formatives

rather than as lexical or grammatical formatives. For the present purpose

they have been separated from specifiers by means of the syntactic feature

[-SPEc.1
 82

In traditional accounts a category adverb has been recognized (cf.

Hodson and Walker 1922). We have argued against this. The various

adverbial expressions are phrasal categories, involving mainly adpositional

phrases or subordinate clauses. Some are what are called time nouns.

But it can be argued that they are reduced adpositional phrases.

In his account of the categories of Oromo, Moreno (1939) has included

adverbial in his category of particles which also includes pre-/post-positions,

conjunctions as well as what we have called particles. Though we follow

him in not recognizing a category adverb, we differ from him in making

distinctions between pre-/post-positionals and subordinative conjunctions

on the one hand, and particles on the other. We have not only made such

distinction but also collapsed pre-/post-positions and subordinative

conjunctions into one broad category - the category of adpositionals.

We have also suggested that a category coxnp. for complementizers and

Infl. for inflections of verbs may be necessary. These are recognized

in the literature as heads of S and S respectively; we shall consider

them in Chapter Five.

 o
0'

NOTES TO CHAPTER TWO

1. May be possible only as a reduced constituent.

2. These are genitive NPs and are different from adjectives.

3. Possible as a reduced Np.

4. Possible as a reduced Np.

5. Such attributes include dimensions, size, colour, etc.

6. The copula da has various alternants which are grammatically

conditioned. One of these is /-ti/ (cf. Eshetu 1981).

7. The adpositional element may be only /-n/. /da/ may be the same
copula we have discussed. Such structures seem to be clefts.

8. We will recognize two /bay?ees/ here.

9. Alternatively, we may argue that these are nominals functioning

as adverbs. But this may cause a problem. If they are NPs,
they must be case marked and in order to receive case they must
be adjacent to a transitive verb. But such NPs can also occur
in structures in which the verb is intransitive, and when they
occur in such structures where the verb is a transitive one, they
do not occur adjacent to it. The alternative argument that they
are headed by an adpositional element solves this problem if
following Bresnan and Grimshaw (1979) we assume that the NPs get
case from the feature of the empty adpositional element. (Cf.
Larson 1985) for an alternative analysis of such NPs.
 84
CHAPTER THREE

COMPLEMENT ONE

3,0 Introduction

In the preceding chapter, we have identified four major and some minor

lexical categories. In the present chapter, we shall address ourselves to

the inherent lexical properties of individual lexical or grammatical

formatives in each of these categories.

Such properties are expressed in the lexicon where each item is entered

together with the specification of its categorial membership, subcategorization

potential, phonological description and semantic interpretation(s). The

categorial description specifies the form class of the item in question,

whereas its subcategorization frames show the possible sister constituents

which it selects at its minimal level of projection. The phonological and

semantic information relate t the way it is pronounced and interpreted

respectively.

Other than providing such information, the lexicon is also the place

where word formation processes and the rules accounting for such processes

are indicated. Of the major categories we shall be considering, all but

adpositionals are characterized by some such rules. This will be taken up

towards the end of the chapter.

Since our main concern here is the syntactic specifications of the

categories, we shall begin with the description of the constituents which

the major categories select at the minimal level of their projection. The

minor categories will be dealt with only in relation to the major ones by

which they are optionally selected at an intermediate or maximal level of

projection.

According to the claim of X-bar theory, each major category has a

complement at every level in its projection line. At the X° level, the

 85

complements are believed to be obligatory 1 and hence may not be omitted

without some particular context. In the light of this claim, we shall now

examine the types of complements each lexical category X selects to form

its syntactic category X'.

3.1 X' Complements

As stated above, every lexical item X enters the lexicon with a

specification of some syntactic frames. Such frames include, inter alia,

information about its form class, and a list of the possible constituents

which it selects as its complements.

Before we go into the details of the analysis of such complements, it

would seem quite necessary to have a working definition of some sort for

the term complement itself, as much of the discussion in this and in the

next two chapters depends on what we mean by this term.

According to Chomsky (1970:210) and Jackendoff (1977:37), a complement

is an abbreviatory term for "some concatenations of ordinary syntactic

categories". Such concatenations include all the materials following the

head of a constituent in a VO language like English. In OV languages like

Oromo, such a definition would lead to the inclusion of all the materials

preceding the head, that is the verb, in a VP, as complements. This is

problematic.

Firstly, in Oromo, both specifiers and complements are prehead in VP,

PP, and AP as we shall see in due course, which means that specifiers are

going to be treated as complements if we adopt Jackendoff's definition as it

stands. Secondly, the defining phrase "some concatenations of ordinary

syntactic categories", appears to be unduly vague. It does not tell us

what types of syntactic categories are "ordinary", nor why. Are specifiers

which are NPs, such as measure phrases or genitive N?s, "ordinary", compared
 86

say, to deictics? If we adopt the definition as it stands now, it does

not seem possible for anyone to tell whether a syntactic category is part

of the specifier or the complement in a language like Oromo, in terms of

its position in relation to the head, or in terms of the defining phrase

"concatenations of ordinary syntactic categories". Thirdly, as we shall

observe in this and in the chapter following, simple nominals, unlike

all other categories, are left-headed, which means again that everything

else, including specifiers, which also follow their heads, is going to be

included in the complement. It does, therefore, appear that this

definition would lead to wrong predictions if applied across categories.

Jackendoff seems to have realized the problems that such a definition would

lead into, for he says, "the definition is a convenient device without any

theoretical import" (ibid.).

His definition seems to be based mainly on VPs and NPs of the type

"destruction of the city" (p.40) in which "of the city" is the complement

of the nominal head "destruction". Such categories are derivationally

related, and a definition given for one may apply for the other. But such

a definition fails to take into account structures of NPsof the type "big

boys" in which the complement "big" precedes the head "boys", or still more

structures of NPs in which the head is both preceded and followed by

complements as in the "big boys who came yesterday". According to the

definition given above, only the relative clause "who came yesterday"

would be the complement of the head "boys", since only it follows the head,

though we are told (p.74) that both adjectives and relative clauses are

complements of N''.

If the definition is a convenient device and has no theoretical significance,

then it may be adopted with all the qualifications necessary, so that it

 87

accounts for all the differences anng the various categories in terms of

their positions as heads and in terms of the type of complement structures

they permit at every level in their projection lines.

In the light of this, we shall define complements in terms of their

category labels, positions and functions. In terms of categories, only

maximal projections of major lexical categories can be complements. This

means NP, AP, etc. may be complements since they are projections of such

major lexical categories as Ns and As. This distinguishes them from the

elements we have mentioned in the preceding chapter as specifiers and

particles, since the latter are not projections of any of the major lexical

categories we have identified. In terms of positions, complements are

found in argument positions (= A.—positions). These are positions where

object NPs, and adpositional phrases or clauses of various adverbial

functions are found. This is in VPs and PPs. The same may be said about

constituents which occur in parallel positions in the projection lines of

the other major lexical categories.

In terms of distribution, we may define complements as maximal

categories occurring between any of the elements we have listed as specifiers

and the head of a constituent structure at any one particular bar level.

This makes specifiers the most peripheral elements in constituent structures.

Functionally, complements are either objects or modifiers. Those which

are objects have direct thematic relations with the lexical head by which

they are strictly subcategorized. These are specified in the lexicon in

the manner to be described throughout this chapter. Those which function

as ndifiers will be taken up in the next chapter as complements of X'' or

I uI
 88

3.1.1 Nominals (N')

In this subsection we shall consider only simple nominals.

Infinitivals or gerundives are assumed to be characterized by the same

frames that their verbal sources are associated with. For example, the
2
infinitival/bit-uu/'to buy' and the verb/bit-/'buy' from which it is

derived subcategorize the same complement types. The difference between

them is thus categorial rather than subcatgorial, and may be handled by a

general redundancy rule in a manner to be explained. With regard to simple

nominals, it may be said that their complements at this level include some

genitive NPs. The following are illustrative examples:

1(a) [Tulluu-n [ [mana] [agaa]] i5aar-e]

S N'N N'''
T-nom. house of-stone build-pf.
(literally), 'Tulluu build [a] house of stone'.

(b) [Tulluu-n [ (siree] [sibiilaa]] kab-a]

S N'N N'''
T-nom. bed of-steel has-impf.
(literally), 'Thlluu has a bed of iron'.

(c) (Tulluu-n [ [foon] [hoolaa]] hin-barbaad-a]

S N'N N'''
T-nom. meat of-sheep cm-want-impf.
(literally), 'Tulluu wants mutton'.

The complements in (1) define the head, N, in terms of the material

from which it is made or has originated. In this respect they may be

called genitives of 'source' in contrast to other genitives such as those

of 'possession', 'location', etc,. The latter are analysable as specifiers,

and complements of N'' respectively. We shall consider these in Chapters

Four and Six.

Syntactically, the genitives in (1) are tightly bound to their heads

and may not be easily separated from them. This is apparent from the

ungrainmaticality of (2) below:



89

2(a)* (Tulluu-n ( [mana] [guddaa] [dagaa]] iaar-e1

S N'N A'' N'''
T-nom. house big of-stone build-pf.

(b) * [Tulluu-n [ (siree] [bareed-duuj [sibiilaaj I kab-al

N'N ''
S
T-nom. bed beautiful-f. of-steel has-impf.

As will be discussed in depth in the next chapter, adjectives are

complements of N'', which means that they cannot occur within N' without

this causing ungrammaticality such as that seen in the above examples.

Such structures would be grammatical only if the adjective followed the

entire N' as in (3) below.

3(a) [Thlluu-n [ [ (mana] [dagaa]] [guddaa]] iaar-e]

S N''N' N N''' A''
T-nom. house of-stone big build-pf.
'Tulluu built a big stone house'.

(b) [Tulluu-n [ ( [siree] [sibiilaa]] [bareed-duu]] kab-a]

S N''N' N A'' beautiful-f. has-ixnpf.
T-nom bed of-steel
'Tulluu has a beautiful metal bed'.

Nominals like /mana/ 'house' may, hence, be characterized by the

following frame:

/mana/ : N ± (-N'''] 'house'

The frame shows that /mana/ belongs to the category of nominals, and

that it selects an N''' (NP) as its modifying complement to form a syntactic

category N'. It also tells us that its position is one preceding the

complement.

Other genitive NPs which may also be considered as complements of N' are

those which show 'purpose'. The fllowing are some examples of N' with

such complements.

4(a) ( [sa?a] (aannanii]]

N'
N cow of-milk
'A cow raised for milk' (i.e., milch cow).
 90

4 (b) I [garbuul (farsool I

N' N N'''
barley of-beer
'barley for beer'.

(c) [ [hoolaa] [foonii]]

N' N N'''
sheep of-meat
'sheep for meat'.

The head-complement relation in such structures seems to be the

converse of the relation we observed with respect to the structures of N's

in (1). The difference in meaning between the purposive genitive of

/hoolaa foon! 'sheep for meat' in (3c), and the source genitive of /foon

hoolaa/ 'meat of sheep' in (ic) seems to have resulted from the

configurational differences the two structures exhibit. The head of an N'

with a source genitive complement can become a purposive complement of

another N'. This type of relationship may be taken as characteristic of

N' complements. Other genitive NPs such as those which show 'location',

for example, do not seem to follow the same pattern as the examples in (5)

show.

5 (a) (1) (workii Wallaggaa]

gold of-W
Gold from Wallaggaa

(jj)* [Wallagga workii]

Wallagga of-gold.

(b) (1) [damma Gimbii]

honey of-C.
Honey from Gimbii.

(jj) * (Gilflbli dammaa]

Gimbii of-honey.

Notice that in purposive constructions, both /damma/ 'honey' and

/workii/ 'gold' can occur in the position where they appear in the (b)

structures above, as (6) below shows.



91

6 (a) (1) (amartii workii]

ring of-gold
'ring of gold'.

(ii) [workii amartii}

gold of-ring
'gold for ring'.

(b) (i) (daadii dainmaa]

mead of-honey
'mead from honey'.

(ii) [damma daadii]

honey of-mead
'honey for mead'.

Further distinctions seem to arise in copular constructions of the

type shown in (7-8) where again purposive and source genitives behave in

a manner that distinguishes them from other types of genitives. For

example, corresponding to /daadii dammaa/ 'mead of honey' and /amartii

workii/ 'ring of gold', both (7 a) (i-u). , and (7 b) (i-u) are possible.

But corresponding to /workii Wallaggaa/ 'gold of Wallagga' and /dainma

kaleessaa/ 'honey of yesterday', only (8 a) (i) and (8 b) (i) are

permissible.

7 (a) (i) daadii-n kan dammaa-ti

mead-nom. of honey is
'(The] mead is from honey'.

(ii) damm-i kan daadii-tti

honey-nom. of mead is
'(The] honey is for mead'.

(b) (i) ainartii-n kan workii-tti

ring-nom. of gold-is
(literally), '[The] ring is of gold'.

(ii) workii-n kan amartii-tti

gold-nom. of ring is
'[The] gold is for ring'.


92

8(a)(i) daimn-i kan kaleessaa-ti

honey-nom. of yesterday is
(literally), '[The] honey is of yesterday'.

(jj) * kaleess-i kan dammaa-ti

yesterday-noin. of honey is

(b) (1) warkii-n kan Wallaggaa-ti

gold-nom. of W. is
'[The] gold is from Wallagga'.

(ii)*Wallagga_n kan workii-tti

W-nom. of gold is

Source genitives occur following their heads but preceding other

genitives such as those which show place or time. This may suggest that

the latter may be N'' complements parallel to the place and time adverbials

of V'', which will be discussed in the next chapter. For the moment let

us observe the following:

9 (a) ( [amartii workii] Wallaggaa]

SN'
ring of-gold of-W.
'A ring of gold of Wallaggaa'.

(b)* [ Jainartii Wallaggaal workii]

SN'ring of Wallaggaa of gold.

Such distributional differences between source and purposive on the

one hand, and genitives of other functions on the other, may support the

claim that the two types of genitives belong to different bar levels.

The latter may be analysed as complements or specifiers generated at a

higher level in the projection of N.

To make distinctions between source and purposive genitives in the

entries of nominals the subscript (pur)posive may be employed in the manner

shown below for /daadii/ 'mead':

/daadii/ : N±[ —N'''] 'mead'

pur.
 93

subcategorization frames such as this show the marked cases, that is,

those cases in which the heads have complements. Those which do not have

complements need not be specified as such, for this is obvious from the

presence of the marked ones. Hence the entries of heads which do not have

complements would include only categorial, phonetic and semantic information.

Thus, a nominal Like /waaka/ 'God', for example, would be characterized by

an entry of the following type:

/waaka/ : N 'God'.3

3.1.2 Verbals (V')

Verbals have a much wider choice of complements than nominals. On

the basis of the type of complements they select at this (minimal) level,

they may be divided and subdivided into types and subtypes. The following

is a discussion of each type..

rpe One

The verbs in this group do not select any complements. Their grammatical

functional relation is solely with their external arguments, as in the sense

of Williams (1981)

The following are examples of structures with such verbs:

10 (a) (Thlluu-n deer-at-el

ST_nom. tall-mid-pf.
'Thlluu got tall'.

(b) [Thlluu-n furd-at-el

S
T-nom. fat-mid-pf.
'Tulluu got fat'.

(c) [Tulluu-n hin-gammadJej

ST_nom. cm-please-pf.
'Tulluu rejoiced'.
 94

Such verbs describe a state of being their subject (external)

argument enters into. In the case of the above examples, Tulluu is the

experiencer of the state of being 'tall' or 'fat'. For this reason such

verbs have often been called stative verbs (cf. Lyons 1977).

As in the case of the nominals already discussed, these too may be

said to constitute the unmarked cases; the marked ones being those which

occur with complements. The entry for such verbs is along the following

lines, as shown for /deer-at-/ 'become tall':

/dee-at-/ : V(-#] 4 'become tall'.

Other verbs which may or may not be stative but which are characterized

by the same frame include those in (11) below:

ii. (a) [Kuasii-n doo?-e]

S
ball-nom. burst-pf.
'[The] ball burst'.

(b) [Thlluu-n kufa?-e]

ST_flom. cough-pf.

(c) [keerrens-i hark-el

Sleopard_nom. roar-pf.
'[A] leopard roared'.

(d) [nw-ni du?-e]

Sman_nom. die-pf.
'(A] man died'.

Though such verbs fall into the same frame as those in (10), the two

are not the same in every respect since only the latter show events, for

which reason the term eventive may be used in describing them. This is,

however, a nrphological (and probably semantic) difference and has no

bearing on the classification which is purely syntactic.

Type Two

In contrast to the verbs in Type One, those in Two are characterized

 95

by the presence of some form of complement. The complements are mainly

adpositional phrases. The following are representative examples.

12 (a) (Tulluu-n [ [as-iiti] [ool-e]]]

S V' Adp''here-from V
T-nom. sta-pf.
(literally), 'Tulluu stayed (spent) the day from here'.

(b) [Tulluu-n ( (gara mana-saa] [gal-el]]

S V'Adp'' to house-his V
T-nom. enter-pf.
'Tulluu entered into his house'.

(c) [Tulluu-n [ mana-saa-ti] [ba?-e]]]

S V'Adp'' V
T-nom. house-his-from leave-pf.
(literally), 'Tulluu went out from his house'.

The complements in (12) are locative adverbial phrases. The

subcategorization frames of the verbs are as shown below for /001-I 'stay':

/ool-/ : V + [Adp'' I 'stay/spend the day'.

Dir.

The subscript (Dir)ectional is necessary in order to avoid the

over-generation of structures with other types of adpositional phrases.

Notice that whereas the complements of the type of verbs in (12) are

obligatory, the complements of those in (13) below are not.

13(a) (i) (Tulluu-n [ (farda-rra] [kuf-e]]]

S V' Adp''
T-nom. horse-from fell-pf.
'Tulluu fell from a horse'.

(ii) [Tulluu-n [ (kuf-e]]]

S V' Vf11f
T-nom.
'Tulluu fell'.

(b) (i) [Thlluu-n ( [siree-rra] [raf-e]]]

S V1 Adp'' V.
T-nom. bed-on lie-pf.
'Thlluu lay on [a] bed'.

(ii).[Thlluu-n ( [raf-e]]]
S V' Vljepf
T-nom.
'Thlluu lay'.
 96

The frame suggested for the verbs in (12) will have to be ndified in

the manner shown below in order for it to accomndate such verbs as well.

/kuf-/ : V ±(Adp''] 'fall'.

Abl.
As stated above, the specifications of the type of adpositional phrase

is important in order to avoid anomalous structures such as (14)

5
14. ?[Tulluu-n [ ramana-saa] [raf-e]]]
S V'Adp'' V.
T-nom. to house-his lie-pf.

The verbs in this group are generally known as intransitives, a term

which may also be extended to include stative verbs as well. The difference

between the two is the absence or the presence of some form of complement.

Action verbs, of which the verbs in (11) or (12) are examples, generally opt

for one in order to give the action some spatial context.

Type Three

The verbs in this group have a wider range of subcategorization potential

than those we have observed so far. Their complements include nominals and

adjectivals. Consider the following examples.

15 (a) [Tulluu-n [ [deeraa] Fda]]]

ST V' A''11 V.
'S
'Tulluu is tall'.

(b) [Tulluu-n ( [barsiisaa] [ta?-e]]]

ST0 V
V' N' ''teacher become-pf.
'Tulluu became a teacher'.

The verbs in (15) are what are called copulatives. Their syntactic

function is to 'link' the adjectival or nominal complement with the external

argument, Tulluu. Their subcategorization frame may be indicated as

follows:

/da! : V + [LN''', A'____ ] ' is'.



97

In (15 b) /ta?-e/ 'become' has a nominal complement which is

co-referential with the external argument Tulluu. In (16) below, this

same verb occurs with a clausal C?) complement.

16 i(a) [Thlluu-n ( [akka [pro deem-u]] [ta?-e]]]

S V', S V
T-nom. as' go-impf. become-pf.
(literally), 'Tulluu becomes like one who is to go'
'Tulluu tends to go'.

(b) [Caaltuu-n [ [pro deem-t-u]] [taa-t-e]]]

S.. V's1 t S V
ç-nom. as go-f-impf. become-f-pf.
'Caaltuu tends to go'.

ii (a) [Thlluu-n [ [akka [pro kolf-u]] (ta?-e]1}

S V'S, , S V
T-nom as laugh-impf. become-pf.
'Tulluu tends to smile'.

(b) [Caaltuu-n [ [akka [pro kofal-t-u]] [taa-t-e]]]

s '. V1 S Vbecome-f-pf.
C-nom. 'as' laugh-f-impf.
'Caaltuu tends to smile'.

As the labelling indicates, it is possible to argue that the complements

here are clauses. The subject is pro in each case, and there is co-reference

between this subject (pro) and the subject of the matrix clause. In other

words, there is no co-reference between the clause as a unit and the external

argument of the verb /ta?- / 'become', as such. This is not, however,

peculiar to this particular verb. In (17) below, for example, we have an

instance of the verb 'be' occurring with an infinitival complement with a

pro subject which is co-referential with the matrix subject.

17 (a) lThlluu-n [ [ [pro deem-uu-saa]] -tii]]

ST_nom V S Sgo_to/ing_his is
'Tulluu is about to leave'.

(b) [çaaltuu-n [ [ [pro deem-uu-sii]] -til]]

S.. v'S .
ç-nom. go-to/ing-her is
(literally), 'aaltuu her going is'
'Caaltuu is about to leave'.
 98

In Chapter Five we shall argue that the position of pro is a governed

position, and hence a lexical subject which may be null can occur in it.

For the present purpose, let us assume that the complements are clauses,

and as in (16), there is coreference between the subject of the matrix

clause and pro, but not between this subject and the entire clause.

Following this, we may ndify the entry we have proposed for Ida! 'is' or

/ta?-/ 'become' in the manner shown below so that it accounts for such

facts as well:

/ta?-/ : V + [[AP NP J-] 'become'.

The interpretation of structures such as (16) and(17) shows what the

subject of the matrix clause tends to 'be' or 'become/do' rather than what

he 'is' or 'has become'. The latter interpretation seems to be possible

only when the complement is either an NP or AP as in (15) above.

However, in structures like those in (16), the complements seem to be

in accord with the analysis of adpositional phrases. This is not without

some intuitive appeal. There are copular structures which have

adpositional phrases as complements. Consider the following.

18 (a) [Tulluu-n I [akka [namaa] I ta?-eI I

S V'Adp NP
T-nom. as man become-pf.
'Tulluu became like a man'.

(b) [aaltuu-n [ [akka [haadaa]] taa-t-e]]

S.p V' Adp NP
ç-nom. AS nther become-f-pf.
'&altuu became like a nether'.

In (18) /akka/ 'as' is a prepositional head, the NP following it is

its complement. The entire adpositional phrase has the adverbial function

of showing degree. The complement structures in (16) may be argued to be

adpositional phrases as well, partly because of their being headed by the

same element /akka/ 'as', and partly because of the possibility of

 99

interpreting them as adverbial phrases of degree. In both (16) and (18),

the subject of the matrix verb is compared with the complement of /akka/

'as' for a certain value, which is not stated, but is implicit from the

entire structure. It may, hence, be said that the complement of the

matrix verb is not the clause or the adpositional phrase as has been assumed

so far, but an NP/A? which is empty. This goes in line with the analysis

of degree phrases or clauses as complements of V I, for which we shall argue

in the next chapter. (18a) may, for example, be understood as being

similar to (19) below.

19. [Tulluu-n [ [akka (namaa]] [ sooressa ta?-e]]]

S V'' Adp'' V'
T-nom. as man rich become-pf.
'Tulluu became rich like a man'.

In (19) /akka namaa/ 'like a man' is a comparative phrase. Thlluu is

compared with /namaa/ 'man', the complement of /akka/, for the value of

being rich. Such phrases or clauses of degree are outside V', which, as

shown by the labelled brackets, comprises the verb /ta?-/ 'become' and its

adjectival complement /sooressa/ 'rich'.

If this is the case, then the complement structure in (16) may have to be

treated along the same line, since, as we have hinted at above, they are

introduced by the 'same' element /akka/. Hence, we may say that underlying

(16(i) is the structure in (20)

20. (Thlluu-n ( [akka [e [ (pro deem-u]]]] [ta?-e]]]

S VI' Adp'' NP S S V'
T-nom. as go-impf. become-pf.
(literally), 'Tulluu became like one who is to go'.
'Tulluu tends to go'.

What we have in (20) is an adpositional phrase which has /akka/ as its

head and an empty NP as its complement. This empty NP has itself a

relative clause complement. The entire adpositional phrase functions as a

 100

degree adverbial parallel to the adpositional phrase in (19) and is as such

outside V'. But notice that there is no complement, adjectival or

nominal, within V', nor does it seem possible to have one since such

structures as (21) are ill-formed.

21* [Thlluu-n I [akka [e !deemul]] [sooressa ta?-e]]]

S V'' Adp'' NP S V1
T-nom. as go-impf. rich became-pf

This goes contrary to the subcategorization potential of /ta?-/

'became', which we proposed earlier. Furthermore, if the adpositional

phrase has the function of a degree adverbial, then the comparison must

be between Tulluu and the empty NP, again parallel to what we saw in (19).

This does not seem to be intuitively appealing, because in (19), the

comparison is between two overt entities, Thlluu and mama! 'man', but

here we have only Tulluu and an entity which has rio phonetic realization at

all. Even if it were possible to compare them, we would still lack the

'value' for which they would be compared since the verb /ta?-/ 'become',

has no adjectival complement corresponding to /sooressa/ of (19).

This contradictory situation forces us to stick to our earlier analysis

of structures like /akka deemuu/ as clauses. The element /akka/ 'as(?)'

may be treated as a complementizer rather than as a preposition. The

clause is strictly subcategorized by the verb /ta?-/ 'become' and its function

may be that of an argument complement rather than that of a degree phrase.

This may explain the reason why structures like (21) are excluded as being

ill-formed. The verbal head would have occurred with two complements and

this would have gone against the subcategorization properties of the verb,

for it can only have an NP or AP or a clause but not a combination of these.

Assuming this to be on the right track, we will now go back to (15).

The copula Ida,! 'is' occurs in the form it appears in (15a) in so far as


.101

the complement has no genitive element. Whenever there is such an

element, the form is /-ti(i)/ as in (17) above, or as in the grammatical

structures in (22) below.

22 (a) (1) [Thlluu-n barsijsaa-da]

ST teacher is
'Tulluu is [a] teacher'.

(ii) [Thlluu-n barsiisaa-koo-tii (*da)]

ST_nom. teacher my is
'Thlluu is my teacher'.

(b)(i) (jjj-n barsiiftuu-da]

she-nom. teacher-f. is
'She is a teacher'.

[iii-n barsiiftuu-da Tulluu-tii (*da)1

S
she-nom teacher-f of-T. -is
'She is Tulluu's teacher'.

(c)(i) [man-ni guddaa-da]

Shouse_nom. big is
'(The] house is big'.

(ii) (man-ni kan Tulluu-tii (*da)]

Sh of Tulluu is
(literally) '(The] house is of Tulluu'
'[The] house is Tulluu's'.

The situation seems to suggest that in order to bar the ungrammatical

structures in (22), we need to enrich the subcategorization frame we have

formulated for /-da/ by adding information about allonrphs. This is

necessary given the fact that the distribution of the allomorphs is one

which is syntactically governed. In other words, it is not a case where

we could take one allomorph as the base from which the other is derived by

a general phonological rule. In the light of this, then, we may associate

/.tii/ with the frame shown below.

 102

/-tii/ : V(N''' ] 'is'

Gen-

Apart from such cases involving allonrphy, there are also other

copulative verbs which do show some degree of regularity, and these need

mentioning. They include verbs like /fakkaat-/ 'resemble' or 'look' as

in (23)

23 (a) [Thlluu-ri abbaa-saa fakkaat-a]

S 1 1
T-nom. father-his resemble-impf.
'Thlluu resembles his father'.

(b) [Thlluu-n deeraa fakkaat-a]

S
T-nom. tall look-impf.
'Tulluu looks tall'.

(c) [Tulluu-n nama gaarii fakkaat-a]

ST_nom. man good seem-impf.
'Tulluu seems (to be] a good man'.

In (23) /fakkaat-/ 'resemble' or 'look', selects either an NP or AP

as its complement. As in the other copular constructions we have already

examined, the complement in each case is coreferential or near coreferential

with the external argument. The subcategorization frame of this verb may

hence be said to be the same as that we have formulated for any of the

other copulative verbs. But notice the structure in (24) vis--vis the

one in (23(a)):

24 (a) [Tulluu-n [ [mana kan iaar-e] [fakkaat-a]]1

ST_nom. V Shouse comp.build-pf. Vseem_impf.
'It seems Tulluu built a house'.

(b) (aaltuu-n [ ! hoolaa kan bit-t-e] [fakkaat-a]]]

S_nom. V Ssheep comp. buy-f-pf. Vseem_impf.
'It seems aaltuu has bought [a] sheep'.

The complement of /fakkaat-/ here is not a simple NP; it is a clause.

This is obvious from the presence of the (coxnp)lementizer /kan/. But this

/kan/ is what has traditionally been called a relative pronoun (Hodson and
 103

Walker 1922:77). If this were the case, then the complement in (24)

would have to be an NP with a relative clause complement. The absence

of this NP in surface structures might be explained in terms of an empty

category. The NP may also be said to have no phonetic entity though its

context is identifiable from the features of the external argument. But

such analysis does not seem to be plausible for two reasons. Firstly,

there is some discrepancy of agreement between the matrix verb and the

external argument in (24). This is particularly true of the second sentence.

The external argument, i.e., &altuu, is a feminine noun, but the verb,

which normally agrees in gender with 'its' subject, fails to do so here.

In other words, the verb is in the same form we would expect it to appear

in if the external argument were 3ms Furthermore, this same structure

may also occur in the form in (25) below, i.e., with the verb showing the

relevant agreement phenomena we would expect.

25. (&altuu-n [ [hoolaa kan bit-t-e] [takkaat-t-iJ]]

Sw V'S V
ç-nom. sheep comp. buy-f-pf. seem-f-pf.
'It seemed (that] aaltuu bought [a] sheep'.

The question which arises from this state of affairs is: Why does the

verb show the feminine morpheme /-t-/ in (25) but not in (24 b) .? Could it

be that the subject in (24 b) is not &altuu?

Secondly, if the complement in both structures of (24) is a relative

clause with its head optionally missing in surface structure, then,

structures without a missing head should be possible. But this again

does not seem to be the case as the corresponding structure in (26) below

shows:

26* t&altuu-n [ [ (dubartii] !hoolaa kan bit-t-e]]

S Vt N''' N'' S
c-nom. woman sheep comp. buy-f-pf.
[takkaat-a] I I
V
seem-impf.
?'aaltuu seemed a woman who bought a sheep'.
 104

The dubious ungrammaticality of (26) may induce us to believe that

the complement here may not be a relative clause at all, for in that case,

the structure would not have been ungrammatical. But it is. The

ungrmm.ticality might be said to have resulted from the lack of agreement

between aaltuu and the matrix verb, which is an indisputable fact. But

the question which also follows from this is why does this happen in this

particular structure? (24 b) above is grammatical without there being

any such agreement between &altuu and this same matrix verb? If the

verb had shown the agreement feature in (26), the structure would have been

grammatical, but its meaning would have been entirely different from the

meaning of the structure in (25).

27. [çaaltuu-n [ [ [dubbartii] [hoolaa kan bit-t-e]]

II
ç-.
S
nom.
v1 N'
woman
S
sheep comp. buy-f-pf.
[takkaat-t-il] I
root-f-pf.
'&altuu looks like a woman who has bought [a] sheep'.

The complement here is an NP with a relative clause complement,

there is also agreement between the matrix verb and the feminine subject

aaltuu. But as stated above, the structure has a different interpretation.

Here /fakkaat-/ means 'look like' and is similar to /fakkaat-/ 'resemble'

in the structures in (23), in terms of its complement types. In both, the

complement is an NP, with or without a relative clause complement. In

(27) /fakkaat-/ 'seem' is different from either of the two /fakkaat/s in

that (1) its complement is a clause rather than an NP, and (2) its external

argument is an empty NP. Tbis may explain the reason why the verb shows

agreement features at times and fails to do so at other times. To make

the point a little clearer, let us consider the following structures.

 105

28(a) [nam-oon-ni [aaltuu kan aalat-an-i] fakkaat-a]

S
man-pi-nom. 9. comp. love-pl-pf. seem-impf.
[It} -seemed the men loved &altuu'.

(b) nam-oon-ni (&altuu kan aalat-an-i] fakkaat-an--i

S...
man-pi-nom. ç. comp. love-pl-pf. seem-pf-pf.
'The men seemed that [they] loved aaltuu'.

The situation in (28) is parallel to that we have observed in relation

to the structure in (24 b) and in the corresponding (25). As there, the

verb here shows the agreement feature in (b) in relation to the plural

subject /nam-oon-nh/ 'men-nom', although superficially, this same noun

appears to be the subject in (28 a) too. In other words, in those cases

where the verb does 'not' seem to show any agreement features, its subject

is the empty NP. (cf. ote 6).

In languages like English such empty positions are filled by pleonastic

elements. Oromo does not have such elements. The position is, hence,

left vacant or covered by a raised subject of a complement clause. In

short, then, the structure in (28 b) is one which is derived from a

corresponding structure of the type in (29).

29. [NPe [ [ [nam-oon-ni kan aalt-uu aalat-an-j]] [fakkaat-a] TI

S V'SS V
man-pl-nom comp. c. love-pl-pf. seem-irnpf.
(It] seems that [the] men loved c5aaltuu'.

Notice that the embedded verb is in agreement with its subject /nam-oorx-

nil 'men-nom' .ith respect to the feature of number as expected, just as in the

same way the matrix verb agrees with its empty subject NP.

If this line of argument is sound, then the subcategorization frame of

this verb must be along the following lines:

/fakkaat-/ : V + [S—] 'seem'.

The frame tells us what the complement of the verb is. It does not tell

us that its subject is an empty NP. For such particular cases, it is

 4.

necessary to include in the frame information idiosyncratic to the

particular element. In this case, the frame above may be modified in

the manner shown below:

V + [NPe[S]—]

Now, if /fakkaat-/ subcategorizes a clause, and if we maintain the

claim made earlier that it is a copulative verb, then we need to answer a

question pertaining to the relation between the clausal complement and the

empty NP subject. Earlier on we said that the relation between the external

and internal arguments in such copular structures is one of coreference,

which if maintained, would also mean that in (29) the empty NP and the

clausal complement are in similar relations. And it seems to be so. The

empty NP subject may be treated as being the same as the state of affairs

which the clause expresses. In other words, what 'seems' is what is

expressed by the clause.

The internal structure of the complement clause, the nature of /kan/

and the manner in which such structures as (28 b) are to be derived will be

the subject of Chapter Five. For the moment, assuming that the argument

is on the right track, we shall move on to the next type of verb.

Type Four

The verbs in this group select simple nominals or clauses as their

complements. They are also characterized by their potential for passive

morphology. For purposes of presentation, they may be divided into three

subtypes. Using traditional terms, we may call them semi-transitives,

transitives and ditransitives. 7 In what follows, we shall be concerned

with each subtype.

Ci) Semi-trans itives

The verbs in this subgroup have complements consisting of simple NPs

which do not necessarily have to appear in surface structure. Most of them

are cognate objects. The following are some examples:

 -a
a

30(a) [Tulluu-n [ ('aata] [hin-aat-a]]]

ST V' N ''' fd Vtif
'Tulluu will eat food'.

(b) [Tulluu-n ( [sirbal [sirb-e]]]

S V'N''' V.
T-nom. song sing-pf.
'Thlluu sang a song'.

(c) [Tulluu-n [ [uffaata-saaj [uffat-eJ 1]

S V'N''' V
T-nom. clothes-his put on-pf.
'Thlluu put on his clothes'.

Other similar verbs but with non-cognate object complements include

those seen in the structures below:

31(a) [Tulluu-n [ [nama] [hin-sodaat-a]]}

S V' N'''
T-nom. man Vcm_ fear-impf.
tThlluu fears men'.

(b) [aaltuu-n [ (inuaayyo] [dee--e1]] /<dee-t-e]

S- V'N''' V
ç-nom. baby deliver-f-pf.
'aaltuu gave birth to a baby'.

The complements of such verbs may be null in surface structure without

any loss of meaning or without the sentence becoming ungrammatical. The

reason for this seems to be related to our knowledge of the world. For

example, once we know that aaltuu is a woman, we do not need to be explicit

about what she would give birth to. This is pretty obvious from our

knowledge that a woman gives birth to a baby. The same may be said about

the verb in (31 a . Unless Tulluu is fearful of animals (in which case,

the complement would need to appear in surface structure), it is not

necessary to mention mama! 'man' for this is understandable from our

knowledge of people who find it difficult to mix with others easily. In

other words, syntactically the complements are always there, but for

pragmatic reasons they do not obligatorily appear in surface structures.

 108

Since our concern here is the representation of such objects in

D-structure, we need to specify the entry of such verbs in the manner

shown here for /xaat-/ 'eat':

/Aaat-/ v + (N''' —] 'eat'.

(ii) Transitives

Unlike the semi-transitives, these are characterized by complements

which cannot be missing without the resulting structure being incomplete

or totally ill-formed. Consider the following structures:

32 i(a) [Tulluu-n ( (biaan] (fid-e]]]

S V'N''' V
T-nom. water bring-pf.
'Tulluu brought water'.

(b) ?[Tulluu-n fid-el

ST_nom bring-pf.
'Tulluu brought'.

ii(a) [Tulluu-n [ [farsoo] (aalat-a]]]

S V' N''' V.
T-nom. beer like-impf.
'Tulluu likes beer'.
(b) ?[Tulluu-n aalat-a]
ST_nom. like-impf.
'Tulluu likes'.

iii(a) [Tulluu-n hoolaa bit-el

ST_nom. sheep buy-pf.
'Thlluu bought [a] sheep'.

(b) ? (Tulluu-n bit-e]

ST_nom. buy-pf.
'Tulluu bought'.

The structure in (b) may be possible only in restricted contexts, such

as when the participants of a discourse are in the middle of their

conversation and when each knows what the discourse is all about. Only in

such cases can such structures be used. Their subcategorization frames are

essentially the same as that proposed for /icaat-/ 'eat', above.

 109

(iii) Di-transitives

The verbs in the previous two subgroups, subcategorize only one

complement. In contrast to these are verbs which are characterized by

one NP and one adpositional phrase or a clause, as the case may be.

We shall first consider those with phrasal complements.

8
33(a) [Tulluu-n [ [Fayyiisaa-daf] [aannan] [kenn-e(-ef)]]]
S V' Adp'' N''' V
T-nom. F. tO milk give-pf-to
'Tulluu gave milk to Fayyiisaa'.

(b) ? [Tulluu-n [ [Fayyiisaa-daf] [kenn-e (-ef)]]]

S V' Adp'''
T-nom. F to give -p f-to

(c) ?[Thlluu-n [ [aannan] [kenn-e]]}

S V'N''' V.
T-nom. milk give-pf.
'Tulluu gave milk'.

34(a) [Thlluu-n [ [aaltuu-daf] [doksa] [hini-e(-ef)]]]

S V' Adp'' N'''
T-nom. c. to secret Vt lift
'Tulluu told [as] secret to çalltuu'.

(b) ?[Tulluu-n [ [aaltuu-daf] [him-e(-ef)]]]

S V1 Adp'' . V
T7nom. c. to tell-pf-to

(c) ?[Thlluu-n ( [doksaj [him-e]]]

S V1 N''' V
T.nom. secret tell-pf.
'Tulluu told [a] secret'

The structures in (b) and Cc) are restricted to the same type of

context we described in connection with the structures shown in (32 b)

above.

Such verbs may be associated with the entry shown for /him-/ 'tell':

/him-/ : V + [Adp'' N''' —] 'tell'

dat.

The subscript (dat)ive is necessary in order to avoid other phrases

of the same category from occurring in this position. For example, in

(35) below, the verb /kaa?-/ 'put' like /him-/ 'tell' above, subcategorizes


110

an adpositional phrase as well as an NP as its complements. But the

adpositional phrase here is locative in function, which means that unless

it is labelled as (loc)ative in the entry, it may not be distinguished

from a dative one, or from other types of adpositional phrases for that

matter.

35(a) [Tulluu-n [ [siree-rra] [wayaa] [kaa?-e]1]

S V' Adp'' N''' V
T-nom. bed on clothes put-pf.
'Tulluu put clothes on [the] bed'.

(b) ,?Tulluu-n wayaa kaa?-e]

T-nom. clothes put-pf.
'Tulluu put clothes'.

The complements we have considered thus far axe all phrases. They

have been either NPs or adpositional phrases, or both as in the examples

in (35). In what follows, we shall consider some verbs within the same

class of transitives which require clausal complements. First, we shall

deal with those which subcategorize finite clauses. These include epistemic

verbs like /beek-/ 'know', /daga?-/ 'hear', /arg-/ 'see', etc. The

following are illustrative examples of these.

36(a) [Thlluu-n ( (akka [Fayyiisaa-n deem-e]] [beek-a] 1]

S V's t S V
T-nom. 'as F-nom. go-pf. know-impf.
(literally), 'Thlluu knows as Fayyiisaa went'.
'Tulluu knows that Fayyiisaa has left'.

(b) [Tulluu-n [ [akka [iii-n duf-t-e]] [daga?-e]]]

S V'S S • V
T-nom. 'as' she-nom. come-f-pf. hear-pf.
'Tulluuheardthat she came'.

(c) [Tulluu-n [ !' [nam-i-i dukkubsat-e] I [arg-e] I]

S V'S, , S V
T-nom. as man-sgl-nom.be-sick-pf. see-pf.
'Tulluu saw that the man was sick'. -

Such verbs like other transitive verbs, do also occur with simple NP

complements as the examples in (37) show.



ill

37(a) [Tulluu-n ( [nama] [beek-a]]]

S V'N''' V
T-nom. man know-impf.
(literally), 'Tulluu knows man',
'Tulluu knows people'.

(b) (Tulluu-n ( [sagalee] [daga?-e]]]

S V'N'''
T-nom. noise hear-pf.
'Tulluu heard [a] noise'.

Accordingly, their subcategorization frame may have to be as shown

here for /daga?-/ 'hear':

/daga?-/ : V + [[S N'''}—] 'hear'.

The clausal complements above are finite in the sense that their verbs

show (ASP)ectual and (AGR)eement features. It is also possible for such

types of verbs to select clauses whose lexical heads are characterized by

some pronominal elements. Such clauses may be called 'finite infinitivals'.

38(a) (Tulluu-n [ [ [Fayyiisaa-n deem-uu-saa]] [beek-a]]]

ST_nom V1 S SF_nom . go-to-his Vknow_impf.
'Tulluu knows Fayyiisaa's going'.

(b) [Tulluu-n [ ( (&altuu-n deem-uu4ii]] [daga?-e]]]

ST_nOm. -nom. go-to-her Vhear_pf.
'Tulluu heard Calltuu's going',

Whereas verbs like /daga?-/ 'hear' are characterized by finite clauses

of one sort or another, others like /barbaad-/ 'want', which we may call

'desiderative verbs', are associated with both finite and infinitival

clauses. The following examples are illustrative of this.

39(a) [Thlluu-n [ [ [deem-uu]] [barbaad-a]]]

S V'SS V
T-nom. go-to want-impf.
'Tulluu wants to go'.

(b) [Thlluu-n [ ! [aaltuu-n akka duf-t-u]] [barbaad-a]] I

ST_nom. -nom. as come-f-impf. Vwant_impf.
(literally), 'Tulluu wants that Caaltuu come',
'Tulluu wants &altuu to come'.
 112

Whenever the complement is one of the type seen in (b), i.e., a tensed

clause, the verb is always in its subjunctive (which is imperfective as far

as aspect is concerned) form, since structures like (40) with the verb in

the perfective aspect are ill-formed.

4.0* . [Tulluu-n [ ! [&altuu-n akka duf-t-e]] [barbaad-a}]}

S V
T-nom. ç-nom. as come-f-pf. want-impf.

Like the epistemic verbs, /barbaad-/ 'want' does also occur with a

simple NP complement, as in (41)

41. [Tulluu-n [ [horii] [hin-barbaad-a]]]

S V' N'''
T-nom. money Vcm_ want-iinpf.
'Tulluu wants Ixrney'.

The subcategorization frame of such verbs may hence be said to be the

same as that we suggested above for verbs like /daga?-/ 'hear'. In fact,

this same frame may be extended to account for the situation we have with

verbs like /aman-/ 'believe' since they too are associated with clausal or

simple nominal complements. Observe the following.

42(a) [Tulluu-n ( [nama] [hin-aman-a]]]

S Vt N'''
T-nom. man Vcm_ believe-iinpf.
'Tulluu trusts people'.

(b) [Tulluu-n [ [akka [Caaltuu-n duf-t-e] I (hin-aman..)jJ

S V'S s. V
T-nom. as C-nom. come-f-pf. cm- believe-impf.
(Tulluu believes that &altuu came (has come)-'.

Cc) [Tulluu-n [ [ [aaltuu-n duf-uu-ii]] [hin-aman-a]]J

S V'SS. V
T-nom. C-nom. come-to-her cm-believe-impf.
?'Tulluu believes alltuu's going'.

In contrast to the verbs we have considered which may take either

tensed or non-tensed clauses as well as simple NP complements, there are a

few verbs which are strictly limited to one of the two clause types. The

verbs /nkkar-/ 'try' and /danda?-/ 'be able' select only non-tensed clauses,

whereas the quotative verb /ed-/ 'say' and the question verb /gaaf-/ 'ask',


.1.

occur only with tensed clauses. Consider the following examples as

illustrative of the former two:

43(a) [Tulluu-n ( ! [wayaa miiuu(-f)]] [mokkar-el)J

S V'SS V -
T-nom. clothes wash-to--for try'-pf.
(literally), 'Thlluu tried for to wash clothes',
'Tulluu tried to wash clothes'.

(b) (Tulluu-n [ [ [wayaa mii-uu1] [hin-anda?-a]]]

S V'SS
T-nom. clothes wash-to cm—be-able-impf.
'Tulluu is able to wash clothes'.

(c)* [Thlluu-n [ [akka [wayaa mii-uu] I [hin-danda?-a] 1]

S V' S I..
T-nom. as clothes wash-impf. cm- be-able-impf.

Just as /danda?-/ 'be able' is sensitive to non-tensed clauses, so is

/'e-/ 'say' to tensed clauses. This is obvious from the grammaticality

of the structures in (a) and the ungramxnaticality of the corresponding

structures in (b) of (44).

44 (a) (i) (Tulluu-n ( [ [aannan na-n-barbaad-a]] [ed-e] 11

ST_nom. S Smiu V
I-cm want-impf. say-pf.
'Tulluu said "I want milk"'.

(jj)*[Thlluu-n [ ! [aannan na-n-barbaad-uuj] [ed-e]]]

V S V
5 milk I-cm-want-to say-pf.

(b)(i) [Tulluu-n I I [aaltuu-n duf-t-el] (ed-e] 11

S Vt S S.. V
T-nom. c-nom. come-f-pf. say-pf.
'Tulluu said, "aaltuu came"'.
(ii)*[Tulluu_n I I [aa1tuu-n duf-uu]] [je-e]]J
S V'SS-. V
T-nom. ç-nom. come-to say-pf.

The complement clauses in (44) are not only finite but also simple in

the sense that they are not introduced by overt complementizers or other

subordinating elements. In other words, the structures are in the form of

direct statements. And this seems to be predominantly the case not only

in Oron but also in Amharic, and probably in other Ethiopian languages.

Indirect statements are hardly used if they are used at all.

 4

- In the structures above /'ed-/ selects a declarative clause complement.

In (45) below it occurs with an interrogative clause.

45(a) (Thlluu-n I I [makaa-n-kee eeiu]] (ed-e}]]

ST_nom. V S S narle_.nom_your who-is V'say_pf.
'Tulluu said, "What is your name?"'

(b) [Tulluu-n [ [ [biyy-i-kee eessa]] (ed-e]]]

S V'SS V1
T-nom. country-nom-your where-is say-pf.
'Tulluu said, "Where is your country?"'

Though questions are predominantly direct, it is not absolutely impossthle

to make them indirect whenever this is desirable. When such is the case,

the verb used is /gaafat-/ 'ask' as the examples in (46) demonstrate.

46(a) [Tulluu-n [ [ (makaa-n-koo akka eezAu ta?-eJ] [na

S V' S S N'''
T-nom. name-nom-my as who become-pf. me
[gaaf-at-e]]]
V
ask -mid-pf.
(literally), 'Thlluu asked me (as to] what my name became',
'Thlluu asked me what my name was'.

(b) [Tulluu-n I [ [ biyy-i-koo akka eessa ta?-eJ] [na]

S V' S S N'''
T-nom. country-nom-my as become-pf. me
[gaaf-at-e]]]
V
ask-mid-pf.
(literally), 'Tulluu asked me [as to] where my country became',
'Thlluu asked me where my country was'.

These same questions may be asked in a manner that also involves the

verb /e-/ 'say'. But such structures appear to be co-ordinative rather

than subordinate clauses. Consider (47)

47(a) [Thlluu-n [ [ (makaa-n-kee eeAuTI (5ed-e]I

S V'SS V
T-nom. name-nom-your who-is say-pf.
-ett I (na] (gaaf-at-e]]1
V'N''' V
having(?) me ask-mid-pf.
(literally), 'Thlluu asked me having said "What is your name?"
 '4

(b) [Tulluu-n [ [ [biyy-i-kee eessa]] [ed-e]]

S V'SS V
T-nom. country-nom-your where say-pf.
-ett [ [nal [gaff-at-el]]]
V' N''' V
having me ask-mid-pf.
(literally), 'Tulluu asked me having said "Where is your
country?'

In some cases /3'ed-/ 'say' may be used with a clause and a simple NP

as in:

48. [Tulluu-n [ [ [biyy-i-kee eessal J [naa-n] [5e-e]]]

S V'SS N''' V
T-nom. country-nom-your where me-cm say-pf.
(literally), 'Tulluu said me "Where is your country?"'

From the examples above, verbs like /5ed-/ 'say' or /gaafat-/ 'ask'

may be associated with the following entries:

/jed-/ : V + [S(N''')-1 'say'

/gaafat-/ : V + [S, N'''—] 'ask'.

3.1.3 Adjectivals (A')

Like nominals, adjectivals may be characterized by optional complements.

This is particularly true of what Dixon (1977:63) calls 'value adjectives'.

They select adpositional phrases as complements. The following are some

such examples:

49(a) [arakee ug-uu-n [ [koraa-f] [gaarii]] -da]

S . A'Adp'' A
arakee drink-to-nom. cold for good -is
'Drinking arakee is good for the cold [weather]'.

(b) [daaii dug-uu-n [ [nanomsaa-f] [hamaa]1 -da]

Smead to'nom A' Ad P'' body...for Abad -is
'Drinking mead is bad for the body'.

Cc) [Tulluu [ [sirba sirb-uu-n] [imaa]] -daj

S A'Adp''
T-nom song sing-to-at Astrong -is
'Tulluu is good at singing songs'.


4
.L

The complements here may be said to be similar to those of N' since

they seem to have the same function of showing the purpose in terms of

which the external argument is evaluated as 'good' or 'bad'.

Such adjectivals may be identified by the sample entry shown for

/gaarii/ 'good'. (See 3.2.1.3 for an alternative analysis.)

/gaarii/ A ± [Adp'' —1 'good'

Pur.

3.1.4 Adpositionals

Like verbals, adpositionals may be divided into types and subtypes.

The first division is between those which select simple nominal complements

and those which select clausal complements. The following are examples

of each.

50 (a) ( j ) [Thlluu-n ( [gara] [nama koriaa]] deem-el

S Adp' Adp N'''
T-nom. to house of-medicine go-pf.
'Thlluu went to hospital'.

(ii) [Tulluu-n [ [waaye] [waraanaa]] dubb-at-e]

S Ad'Ad
T-nom. about war speak-mid-pf.
'Thlluu spoke about war'.

(iii) [Tulluu-n [ [akka] [abbaa-saa]] riama gaarii-da]

S Adp' Adp N'''
T-nom. as father-his man- good is
'Thlluu is (as] good like [as] his father'.

(b) (i) [Tulluu-n [ [erga] [iiifl duf-t-e]] deem-el

S Adp' Adp S
T-nom. after she-nom. come-f-pf. go-pf.
'Tulluu went after she came'.

(ii) [ [ (yoomuu] ! (&altuu-n duf-t-e]]1 Thlluu-n hin-deem-a]

SAdp'Adp SS.
when ç-nom. come-f-pf. T-nom. -go--impf.
'Tullun will go when aaltuu has come'.

(iii)[ [ [yoo] ( [kana ed-e]]] soba e-e]

SAdp'Adp. SS .
if this say-pf. lie say-pf.
'If he said this, he said a lie' (Gragg 1982:407).
 -L

Both types of adpositionals may be represented by the following

entries:

/gara/ : Adp + 'to/towards'

/yoomuu/ : Adp + [S] 'when'.

Some adpositionals may freely select either one of the two complement

types. Consider the following, for example:

51(a)(i) [ ( [akka] ( [Thlluu-n duf-e]]] 9 &altuu-n deem-t-e]

SAdp'Adp SS
as T-nom. come-pf. ç-nom. go-f-pf.
'As [soon as] Tulluu came aaltuu went'.

(b)(i) [Tulluu-n I [erga] ( [iii-n duf-t-elJ} as-i tur-e]

S Adp'Adp SS
T-nom. after she-nom come-f-pf. here-from be-pf.
'Tulluu was here after she came'.

(ii) [Tulluu-n [ [erga] [sa?aati torba]] as-i tur-el

S Adp' Adp N'''
T-nom. after hour seven here-from be-pf.
'Tulluu was here after seven o'clock'.

In (51 a). as opposed to (50 a) (iii) /akka/ occurs with a clause. In much

the same way /erga/ 'after' occurs with a clause in (51 b) (i) and with a

simple NP in (51 b) (ii), This calls for a subcategorization frame of the

following type:

/erga/ : Adp + [—L N '' ' SJ] 'after'.

The division we have made above is based on the types of complements

that adpositionals are associated with. Another way of dividing them

would be by taking their positions as heads into account. The structures

we have observed so far are left-headed. But there are others which are

right-headed. The examples below have such structures.

52(a) [Tulluu-n I [abbaa-saa] [woin]] duf-e I

S Adp' N''' . Adp
T-nom. father-his with come-pf.
'Thlluu came with his father'.

(b) [Tulluu-n I [hoii1 Ibooda]] farsoo hin-4ug-al

S Adp' N''' Adp
T-nom. work after beer cm-drink
'Tulluu drinks beer after work'.


118

52 (c) [Tulluu-n ( [kawwee] [-dan leena ajjees-ej

STnom Adp' N' gun AdPwjth kill-pf.
lion
'Tu].luu killed a lion with a gun'.

(d) [Tulluu-n (Fayyii.saa-1 [-daf]] aannan kenn-e (-ef)]

S Ad'N'''
T-nom. F. to milk give-pf-to
'Tulluu gave milk to Fayyissaa'.

Such adpositional elements may also occur with infinitival complements

as the following examples show.

53(a) [Tulluu-n t [deem-uu-] [-fi]] barbaad-a]

S Adp"S Adp
T-nom. go-to- for want-impf.
(literally), 'Tulluu wants for going', 'Tulluu wants to go'.

(b) [Caaltuu-n [ ! [hoolaa bit-uu]] (-f ii] deem-t-e]

S Adp'SS Adp
L-nom. sheep buy-to- for go-f-pf.
(literally), 'aaltuu went for buying (a] sheep',
'aaltuu Went to buy sheep'.

The subcategorization frame for such adpositionals is substantially

the same as that required for those which are left-headed. What is different

is the position of the head in relation to its complement. This must be

indicated in entries along with other idiosyncratic properties for which

the lexicon is believed to be the reservoir, (cf. Selkirk 1982). Hence,

the frames for /wo5jin/ 'with' and I-f i/ 'for' may be shown as follows:

fwojjin/ : Adp + [N'' '-1 'with'

/-f j/ : Adp + [S ] 'for'.

pur.

A question which we need to raise at this point is whether adpositions

assume a single position in deep structure. Or whether we should formulate

two phrase structure rules for them: one for all those which are left-headed

and another for all those which are not.

 119

From the point of view of economy, it seems advantageous to assume

one underlying position for both types and account for surface differences

by means of movement rules. But in order to do this, we need to have son

means of choosing one or other of the two contending positions. Since

both involve movement, the choice of one over the other could be made

arbitrarily. However, in quantitative terms, the number of postpositions

is greater than the number of prepositions, which means that if we assume

adpositiorial phrases to be right-headed, our work wouldbe less than if we

assumed otherwise. For purposes of comparison observe the following:

Adpositions

Prepositions Postpositioris
akka - 'like' bira - 'beside'
gara - 'to' booddee - 'after'
hamma - 'until' duuba - 'behind'
waayee - 'about' gidduu - 'between'
Conjunctions gubbaa - 'over'
akka - 'as' -irra - 'on'
otuu - 'while' -itti-- 'to'
yoomuu - 'when' gala - 'under'
erga - 'after' keessa - Inside'
yoo - 'if' malee - 'without'
sababi - 'because' - 'with'
- 'by', 'at'
-fi - 'for'
-iti - 'from'
dura - 'before'
-(da}f - 'to'

From the data above and from Greenberg's (1963:79) generalization

(which is also based on such observations of data) about SO y languages

being postpositional, we may adopt right-headedness for adpositional

phrases/clauses in Oromo. Following this assumption then, a very general

entry for any adpositional element would be like the one shown below:
 120

X : Adp + [CN''' s}—]

What we have attempted to do so far is show the possible constituents

that the basic lexical items (X°) in each of the major categories are

associated with in forming their syntactic categories (X'). The next thing

we need to pay attention to is the specifications of words derived from

other words in terms of similar syntactic frames and in terms of the word

formation rules which are involve in their derivations. The rest of the

chapter will concentrate on these two aspects.

3.2 Lexical Redundancy Rules

As hinted at earlier on, lexical redundancy rules are rules of word

formation. Their domain is the lexicon. It may not be wise to try to

give a complete picture of the processes involved in the derivation of such

lexical items in each of the three major categories, nor might it be

necessary for the purpose we are after, which is primarily the identification

of the complement structures of lexical heads at the minimal level of

projection. It is, therefore, only secondarily and to a very limited extent

that we will be dealing with this complex area of derivatiorial morphology.

3.2.1 Derivations

To the exclusion of adpositionals all the other major categories have

derived forms in addition to their simple forms, which we have already

dealt with in the preceding section. In what follows we shall be concerned

with the properties of the derived forms.

3.2.1.1 Nominals

Nominals may be derived from adjectives, nouns and verbs. The

following are examples derived from the first two sources:

 121

Paradigtrs One

N N
nam- nam-uinmaa

man body/manhood
nag- nag-ummaa
peace peacefulness

moot moot-ummaa

king 'kingdom'
-uxnmaa
A
gaar- gaar-ununaa
good 'goodness'
bareed- bareed-unmiaa
beautiful 'beauty'
furd- furduinmaa
fat 'fatness'

The derivational process involved may be captured by the following

rule. For purposes of generalization features will be used.

Rule One:

x
[^Nj—> [_k} / — -unimaa x-'ness'.

The rule is partly category changing since it converts adjectives into

nouns. A more or less similar type of rule derives nominals from exclusively

adjectival sources. The following are some examples of this.

Paradigms Two

A N
deer- deer-ma
tall 'tallness'
gudd- +ina —> gudd-ina
big 'bigness'
bay?-, bay?-ina
abundant abundance'
 122

Rule Two : /-ina/

[] -> i4i / - - ma x-'ness'.

The derived nominals in both Paradigms One and Two have the selectional

restriction feature [-CONCRETE] in common. Syntactically such nominals are

characterized by their lack of complements.

Notice that the rules do not change only the categorial status of the

forms in question, but also their subcategorization properties as well.

For example, the adjectival /gaarii/ 'good' selects an optional purposive

phrase as its complement, as we have seen in 3.1.3, but the same is not

true of its derivative /gaar-ummaa/ 'goodness'. The latter is characterized

by the absence of any complement at all. Hence its entry specifies only

categorial and other interpretive information:

/gaar-ummaa/ : N 'goodness'.

Regarding nominals with verbal stems consider the following paradigms:

Paradigms Three

V N
i) barsiis- barsiis- aa
'teach' 'teacher'
fayyiis- fayyiis-aa
'save' + -aa -> saviou r'
faars- faars-aa
'sing' singer
ii) haal- haalii
'deny' 'denying'
bit- bitt-u
'p + -ii—> 'buying'
haadd- haadd-ii
shave' 'shaving'
iii) duul- dull-tuu
'war' 'warrior'
gargaar- gargaar-tuu
'help' + -tuu —> 'helper'
bar- bar-tuu
'learn' 'learner'
 123

iv) dug dug-uu

'drink' 'to drink/drinking'
deem- + -uu -> deem-uu
'go' 'to go/going'
duf- duf-uu
'come' come/coming'
v) daan- daan-ia
'punish' 'punishment'
fid- + -ia fid- ia
'bring' 'bringing'
f jig- fiig-ia
'run' 'running'

Each of these rules needs to be indicated in the manner shown below

for the paradigms in (i) above:

Rule Three:

[±] -> [+] / -aa

Of the nominals in (i-v), only those with the ending /-aa/, /-uu/ and

/-tuu/ seem to be characterized by the same subcategorization frames that

their corresponding base forms are associated with. For example, both the

transitive verb /bar-/ 'learn' and its derivative /bar-tuu/ 'learner' have

the same complement structure. The differences are only the function of

these complements and in the position they assume in relation to their heads.

Compare the following:

54 (a)(i) [Tulluu-n [ [afaan faranii] [bar-el]]

S V'N''' V
T-nom. mouth of-foreigner learn-pf.
'Tulluu learnt English'.

(ii) [Tulluu-n [ [bar-tuu] [afaan faranii]] tur-el

ST_nom. N' Nlearn_erN'''mouth of-foreigner be-pf.
(literally), 'Tulluu was learner of English'.

(b) (i) [Tulluu-n [ [hoolaa] [gurgur-e]]]

S V'N''' V
T-nom. sheep buy-pf.
'Tulluu bought a sheep'.
 .4,

54(b) c_li) [ [ [hoolaa] [gurgur-uu-n]] gaarii-da]

SN'N''' N
sheep buy-to-nom. good is
'Buying/to buy sheep is good'.

(c) (1) [Tulluu-n [ [aaltuu-daf I [afaan Sidaamaa]

S Vt Adp''.. N'''
T-nom. ç-nom. to mouth of-Sidama
[barsiis-e (ef) 1]]
Vthft
'Tulluu taught aaltuu Sidama language' (=Amharic).

(ii) [Tulluu-n [ t [barsiisaa] [ [afaan Sidaamaa]

S V. N' N N''' N''
T-nom. teacher of-A.S.
[aaltuuJ I tur-eJ]
be-pf.
'Tulluu was çaaltuu's Amharic teacher'.

As is observable from the structures, in each of the pairs, the verbal

and the corresponding nominal heads seem to have corresponding complements.

In fact, there is no difference even in position between the complements of

the nominals ending in /-uu/ and those of their corresponding verbs. It is

only between the complements of those nouns ending in /-aa/ or /-tuu/ and those

of their verbal sources that such differences appear to exist. From these

situations, it appears that it may not be necessary to have independent

entries for derivatives unless they are substantially different in their

choice of complements. The derivatives may be identified by the same frames

that their sources are associated with. The only information we need to

provide is that which relates to their position as heads. This may be made

explicit by a general statement to the effect that nominals other than

infinitivals are left-headed. Accordingly, the relation between, say, /bar-/

'learn' and /bar-tuu/ 'learner' may be indicated in the manner shown below:

/bar-/ : V + [N' ''-I 'learn'

N[-tuu) 'learner'.

Notice, however, that this type of relationship does not always exist

between any two derivatives. For example, the complements of the verb
 125

/barsiis-/ 'teach' include a clause as well as a simple NP, whereas that

of the corresponding nominal /barsiisaa/ 'teacher' has only simple NPs.

Observe the following examples.

55 (a) (Tulluu-n [ [ [mana iaar-uu]] [ana] [barsiis-e]]]

S V'SS N''' V
T-nom. house build-to me learn-cs-pf.
'Thlluu taught me to build a house'.

(b)* [Tulluu-n [ [barsiisaa] [(a)na] [mana 1'aar-uu1] tur-e]

S N'N N''' S
T-nom. teacher me house build-to be-pf.

(c)* [Thlluu-n [ [barsiisaa] [-koo [mana ijaar-uu]] tur-el

S N'N N''' S
T-nom. teacher house build-to be-pf.
? 'Thlluu was my house-building teacher'.'

But notice also that the structure below is grammatical with an NP

occurring in the position of the infinitival clause in (55)

56 [Tulluu-n ( [barsiisaa] Jafaan Oromo]] tur-e]

S N'N N'''
T-nom. teacher of-mouth of-Oromo be-pf.
'Tulluu was an Oromo teacher'.

The ungrammatical structures in (55) show that /barsiisaa/ cannot

occur with a clausal complement though its verbal source /barsiis-/ 'teach'

can. This suggests that the two cannot be associated with the same frame.

We may hence need to have the following independent entries:

/barsiis-/ : V + [Adp''S N''J—] 'teach'

/barsiisaa/ : N + [- N'''] 'teacher'.

3.2.1.2 Verbals

The mzrphology of Oron verbs is very complex. Only to attempt to

make a complete description of it is to do no justice to it. We shall,

therefore, limit the discussion to those processes which have some bearing

on the subcategorization properties of the forms undergoing them.

Among the forms concerned are: causatives, passives, and middles or

reflexives. Excepting some derivation resulting in reflexives, the rest

 126

are non-category changing processes. The effects they have on the forms

in question are subcategorial rather than categorial.

3.2.1.2.1 Causatives

As hinted at in Chapter One, causativization is a lexical process in

Oromo. To the exclusion of the 'be' verb /-da/, all other verbs may be

causativized. The elements which show this are /-(i)s/ and /-(i)sisvi

(i)siis-/ Their distribution is partly determined by the intransitive-

transitive nature of the verbs undergoing the process. Those which are

intransitive require /_(j) 3/11 whereas those which are transitive take

/-(i)sis-t"-(i)siss-/ Paradigms Four and Five below are illustrative

examples of each type.

Paradigms Four

V N
arreed- arreed-is-
run 'cause to run'
room- + - (i)s—> = çoom-s-
get fat 'fatten'
book- book-is
ferment 'ferment/leaven'.

The morphological process may be summed up as follows:

Rule Four : /..(i)s/

[x-J -> [ [X-J - (i)s -] 'transitive counterpart of X'

Vi VtVi

The subcategorization frames of the resulting forms are different from

those of their intransitive base forms. They are characterized by an NP

complement which they acquire as a result of the morphological process they

have undergone. The entries for /'oom-/ and /ooms-/ show this difference.

/oom-/ : V [-1 'get fat'

/cooms-/ : V + [N''' -1 'fatten'.

 127

Since the subcategorization frame of forms like /ooms-/ 'fatten' is the

same as that of any transitive verb, the affix /-(i)s-/ may be considered

as being a transitivizer, and the verbs as transitivized verbs (Hayward 1975,

1976; Gragg 1976). The following examples give support to this

consideration.

57 (a) [sang-ia-i oom-e]

S
bull-sgl-nom. get-fat-pf.
'The bull got fat'.

(b) [Tulluu-n sang-ia oom-s-eJ

ST_nom. bull-sgl. fat-tr.-pf.
'Tulluu fattened the bull.

Paradigms Five

V V
bit- bit-isiis-
'buy' [biisiis-] 'cause to buy'
arg- -isiis- —> arg-is us-
'see' [agarsiis-} 'cause to see'
abboom- abboom-isiis-
It [abboom-sijs-] 'cause to order'.

The following rule may capture the process involved here.

Rule Five : /-(i)siis/12

[X-] —> ([X-] -isiis-] 'causative'.

V VV
Following structures like:

58. [Tulluu-n [[Fayyiisaa] [hoolaa] bi-isiis-e]]

5T-nom. ' F. sheep buy-cs-pf.
'Tulluu caused Fayyiisaa to buy a sheep',

the subcategorization frame of such verbs as /bi&isiis-/ 'cause to buy',

may be formulated as /biisiis-/ : V+ [N''' N'''—} 'cause to buy'.

As hinted at earlier on, transitivizations and causativizations are

processes which increase subcategorization potentials. Verbs with the

 128

suffix I-is-I or /-isiis- -isis-/ are always a complement ahead of their

base forms. This is noticeable from the examples in (57) and (58) above.

Moreover, since the complement which a verb acquires as a result of its

having either one of these affixes is the external argument of a corresponding

non-transitive or non-causative form, the process may be considered as being

one of internalizing an external argument (cf. Chomsky 1981).

As has been stated in Chapter One, the complement which a verb acquires

as a result of its having a causative affix may be optional, whereas the

complement which it requires in order to satisfy its inherent lexical

property is always obligatory. Hence structures like (58) above may have

the surface realization as in (59a) but not as in (b).

59 (a) [Tulluu-n hoolaa bi-isiis-e]
ST_nom. sheep buy-cs-pf.
'Tulluu had [a] sheep bought'.

(b)? [Tulluu-n Fayyiisaa biisiis-e]

ST_nom. F. buy-cs-pf
'Thlluu cause Fayyiisaa to buy'.

Transitivized verbs, like other transitive verbs, may undergo the same

process, as the following examples dennstrate.

Paradigms Six

V V
damf-is- damf-is-iis-
boil-tr. 'cause to boil'
buu-s- -us -> buu-s-iis-
'get down' 'cause to get down'
daam-i s- daam-is-iis-
extinguish-tr 'to get extinguished'.

Taking the transitivized stems as base forms, the corresponding

causative forms may be derived by the same rule we proposed earlier for

causativizing basic transitive verbs. The only difference between the

 129

paradigms above and those of the basic transitives is in the form of their

causativizing affix. The affix above is I-us-I whereas that of the

basic transitives is /-siis-/. This is, however, a case of allomorphy

and does not seem to have any effect on the subcategorization potential

of the forms in question. What is important to note is that the number

of complements that a verb requires progressively increases as it changes

from intransitive to transitive and then to causative. This is what

the structures in (60) below show.

60 (a) [biaan(i) daxnf-e]

water-nom. boil-pf.
'[The] water boiled'.

(b) [Thlluu-n bi g aan damf-is-e]

ST_nom. water boil-tr-pf.
'Tulluu boiled the water'.

(c) [Tulluu-n Fayyiisaa biaan damf-is-iis-e]

S
T-nom. F. water boil-tr-cs-pf..
'Thlluu caused Fayyiisaa boil [the] water'.

The situation seems to suggest that though such verbs are

derivationally related, they differ from one another with respect to the

number of complements they require, which means that they need to be

identified by independent entries of the type shown below.

/damf-/ : V (----#] 'boil'

/damf-is-/ : V + [N''' -] 'cause to boil'

/damf-s-iis-/ : V + [N''' N''' -] 'cause someone to boil'

There are, however, some problems with such entries. Firstly,

causativization is productive, nearly every verb]. root/stem can undergo

the process. If we therefore show both the root/stem and its causative

form in the lexicon in the manner indicated above, the lexicon will be over-

burdened with both general and idiosyncratic information. Secondly, the

 130

frames of the causative forms show us only the number of complements.

They do not tell us that such complements are arguments whose thematic

roles have been changed from agent to goal. Even if we show the role

of each argument in every entry by a subscript, we will still be faced with

the first problem. In order to avoid such problems, it seems necessary

to have a general redundancy rule of the type in (61) below.

61 (a) V + [A, —] —> V -(i)s- + [A G, —1

(b) V + [A, G —] —> V -siis- + [A G,G —].

where A, = external argument.

What (61) tells us is that any root verb may be transitivized/

causativized by attaching I-is! or /-siis-/, and when this happens, its

external argument, which was an agent, becomes an internal argument, a goal.

Once we introduce such a general rule, it will not be necessary to have an

independent entry for every causative form. The lexicon will hence be

reduced to general statements and to entries of forms which are

derivationally unrelated.

The verbs we have considered thus far may be called morphological

causatives since they are all derived from transitive or intransitive base

forms by affixing /-(i)s--/ or /-siis-/. The verb /go-/ 'make' is a

little different in this respect, since it behaves like causatives 'without'

having a causative affix. Observe the following examples.

62 (a) [Tulluu-n [ [Fayyiisaa] [wattadarii] [god-el]]

S V' N''' N''' V
T-nom. F. soldier xnake-pf.
'Tulluu made Fayyiisaa [a] soldier'.

(b) (Tulluu-n ( [nivaa-saaj [bartuu] [god-el]]

S v II • N''' V
T-nom. child-his student made-pf
'Tulluu made his child a student'.
 131

The relation between Tulluu and Fayyiisaa is direct in the sense that

Tulluu is personally involved in forcing the latter to become a soldier.

And from the gloss it appears that the lower clause, for which Fayyiisaa

is the subject is copular, in which case /go-/ 'make' may be treated as

being the causative counterpart of the copulative verb /-da/ or /ta?-/

'be/become' and Fayyiisaa as its internalized argument.

In structures of indirect causatives, the form used is /go&-isiis-/

as in (63) below.

63 (a) (Thlluu-n [ [Dabalaa] [Fayyiisáa] [wattadarii]

S V' tl N'''
T-nom. D. F. soldier
gocc-isiis-e]]
make-cs-pf.
'Thlluu make Dabalaa force Fayyiisaa to be a soldier'.

(b) [Thlluu-n ( (Dabalaa] [muaa-saa] [bartuu] go-isiis-e}]

S V' N''' N''' N'''
T-nom. D. child-his learner make-cs-pf.
'Tulluu make Dabalaa force his child to be a student'.

The verb in (63) has one nre complement than its counterpart in (62).

The interpretations of the corresponding structures are also different.

In the (a) structures of both (62) and (63), for example, Fayyiisaa is the

causee. However, whereas Tulluu is the direct causer in (62), in (63)

this is not the case. The direct causer in (63) is Dabalaa. Tulluu is an

indirect causer. Except for Tulluu, which is the subject of the clause,

all the other arguments are all in the accusative case.

The question which arises from this situation concerns the manner in

which such arguments get their cases. If we follow the adjacency

condition of Stowell (1981) on case assignment which dictates that the case-

assigning and the case-receiving categories should be next to each other

for case assignment to take place, only that argument which is the closest


132

to the verb could receive accusative case. According to this condition,

the other arguments cannot get case marked without violating this

condition. And since the structures are perfectly granunatical, we cannot

say that the arguments are without case. We need only to account for the

manner in which they receive it.

If, following Chomsky (1981), we assume that case may be assigned

inherently at the level of D-structure and structurally at the level of

S-structure, the argument which is not the closest to the verb may be said

to have received case inherently, and the one which is nearest to it may

be said to have received it structurally. Accordingly, in (63) /wattaddarii/

'soldier' may get its case structurally, whereas /Fayyiisaa/ and /Dabalaa/

may get theirs inherently. However, Chomsky's proposal is based on

non-causative verbs which subcategorize double NPs. The situation we

have here is, thus, slightly different since we are dealing with causative

verbs though the spirit of the explanation is the same.

Alternatively, we may argue that the adjacency condition does not hold

for Oromo. The fact that structures like (64 b) where the NP argument is

separated from its verbal head, are grammatical may give support to this

view.

64 (a) [Thlluu-n Da.balaadaf aannan kenn-e (-ef)]

5 T-nom. D. to m.ilk give-pf-to
'Tulluu gave milk to Dabalaa'.

(b) (Tulluu-n aannan Dabalaadaf kenn-e (-ef) I

5 T-nom. milk D.to give-pf-to
(literally), 'Tulluu gave to Dabalaa milk'.

The difference between (64 a) and (b) is that there is a slightly

prolonged pause following /aannan/ 'milk' in (b). In the light of this,

we may dispense with the condition and assume that causativization as a

 133

nrphological process has the property of increasing not only the number of

internal arguments but also the case-assigning potential of verbs undergoing

this process. In other words, as a verb increases its internal arguments

by progressively changing its form from intransitive to transitive and

then to causative in the manner shown in (60) or in (62-63), its case-

assigning property also increases. This appears plausible given the fact

that in general, there is correlation between the number of arguments and

the number of causative affixes a verb has. This has also been attested

by Hayward (1975:216) and Owens (1985).

3.2.1.2.2 Passives

Like causativization, passivization is lexical. Every verb with the

subcategorization frame [N' '-1 has a corresponding passive form indicated

by the affix I-am-I. But unlike causativization, passivization is also

syntactic since it involves NP nxvement. The following paradigms show

the nrphological process.

Paradigms Seven

• bit- bit-am-
buy t be bought'
gurgur- -am- —> gurgur-am-
sell 'be sold'
aj jees- aj jeef-am
kill be killed

Rule Seven I-am-I

IX-] —> ([X-]-am] tpassive'

The above rule tells us that every transitive verb has a passive

counterpart. It does not tell us whether the converse is also true.

Butfrom the existence of such forms as /deekk-am/ 'be angry' without a



134

corresponding transitive form, we may infer that the converse is not true.

This means then we have to make distinctions between two types of passives:

those with and those without transitive base. This distinction is important

since it is also reflected in the thematic role (8.role) assigning

property of the verbs in question. According to Burzio's (1981)

generalizations (in Chomsky 1981) only passives with transitive base forms

are characterized by their inability to assign theme to their external

arguments and case to their internal arguments. Those passives which

have no transitive base do, however, assign theme to their external

arguments. In this respect they are similar to intransitive verbs. This

is also obvious from the fact that, like the latter, they are associated

with no internal arguments. Compare the following structures.

65 (a) (Thlluu-n deekkam-e]

ST_nom. angry-pf.
'Tulluu got angry'.

(b) [Tulluu-n rukkut-am--e]

- S
T-nom. hit-ps-pf.
'Thlluu was hit'.

Corresponding to (65) are the S-structures in (66):

66 (a) [Tulluu-n deekkam-e]

ST_nom. angry-pf.

(b) (Tulluu-n [ [tJ [rukkut-am-e]]]

S 1 V' N''' 1 V
T-nom. hit-ps-pf.

(c)* (e [Tulluu rukkut-am-el]

SVt
T. hit-ps-pf.

(66) shows that the S-structure in (a) is the same as the surface

structure in (65 a) whereas the S-structure in (b) is different from the

corresponding surface structure in (65). Tulluu in (65 a) is generated

in situ whereas Tulluu in (b) is derived by move c. The movement is

 135

necessary in order for Thlluu to be able to receive case from INFL and to

escape the case filter mentioned earlier on. It cannot receive case in

its base position from the governing verb, because the verb has passive

morphology and according to Chomsky (1981:124) passive morphology has the

effect of absorbing the case-assigning property of transitive verbs. This

makes the movement obligatory as structures like (66 c) are ill-formed.

Passive morphology does not affect the subcategorial status of verbs.

Its effect is only on their 8-marking and case-assigning properties, and

once this is made clear by such general statements as Burzio's, it does not

seem necessary for every passive form of a verb to have an entry independent

of its active counterpart for that would make the lexicon unnecessarily

bulky. Instead, the affix I-am-I may be introduced by a lexical redundancy

rule of the type shown earlier on.

3.2.1.2.3 Middles

Other than the processes which we have observed in the preceding sections,

Oromo has also a process which derives what have been called middle/reflexive

verbs (Gragg 1976) from adjectives, nouns or other verbs. The paradigms

below show some examples:

Paradigms Eight

(1) V

fuud- + add —> fud-add-
a.

take take for oneself
hid- hid-add-
tie tie by oneself
bit- bit-add-
buy buy for oneself
 136

(ii) A V
abaa 5aba-add-
Strong 'get strong'
furdaa furda-add-
'fat' get fat
deeraa deera-add-
tall 'get tall'
(iii) N
1.
hojii ho3-edd
work 'work by/for oneself'
deebuu deeb-odd
thirst 'be thirsty'

The process is partly category-changing as the examples in (ii) and

(iii) show. The forms in (i) and (iii) are similar to those verbs we have

considered as transitives as far as their basic subcategorizatioris are

concerned. The difference lies in the type of external arguments the

middles require. Their subjects are not only agents but may also be

patients or experiencers or beneficiaries. Consider the following examples:

67 (a) (Tulluu-n hoolaa bit-el

ST_nom. sheep buy-pf.
'Tulluu bought [a] sheep'.

(b) [Thlluu-n hoolaa bit-at-el

ST_nom. sheep buy-mid-pf.
'Tulluu bought a sheep for himself'.

(c)* [Tulluu-n Dabalaa-daf hoolaa bit-at-e]

ST_nom. D-for sheep buy-mid-pf.
'Tulluu bought [a] sheep for Dabalaa'.

(d) (Thlluu-n Daballadaf hoolaa bit-e(-ef)]

T-nom. D-for sheep buy-pf-for.

As stated earlier, both middles and transitives are character .ized by a

preceding complement. This is also clear from the examples above.

However, whereas transitives allow any dative object, the middles don't.
 137

The ungrammaticality of (67 c) and the grammaticality of (d) is indicative

of this difference. Structures like (67 a) tell us only the fact that

Tulluu has bought a sheep. They do not tell us for whom he bought the

sheep. In other words, we do not know the beneficiary. Whenever the

beneficiary is the agent himself, the affix I-at-I appears on the verb.

The ungrammaticality of (67 c) emerges from this fact, because there, we

have a situation where the agent and the beneficiary of the action of

buying a sheep differ in reference. The structure would be grammatical

only if it appeared in the form in (67 d) or in (68) below, where the dative

object is coreferential with the agent subject.

68. (Tulluu-n of-ifi hoolaa bit-at-e]

ST_nom. self-for sheep buy-mid-pf.
'Tulluu bought a sheep for himself'.

The syntactic effect of the middle verbs of the type /bit-at-/ 'buy

for oneself' seems to be on the choice of dative objects. Such objects

must be coreferential with the agent subject. For this reason, such

verbs have often been called autobenefactives (Hayward 1975:219).

However, as Hayward has noted, not all verbs with the affix i-at-I

are autobenefactive. In structures like (69 b) below, for example,

the affix gives the meaning that Thlluu personally did the action without

necessarily being the beneficiary.

69 (a) [Thlluu-n balbala uf-at-e]

ST_nom. door close-mid-pf.
'Tulluu closed the door personally'.

Furthermore, in structures involving those verbs with an adjectival

base, I-at-i shows that the subject is the experiencer of the state which

the verb denotes. Consider the following.

 138

70 (a) [Tulluu-n deera-at-e]

ST_nom. tall-mid-pf.
'Tulluu got tall'.
'V
(b) [Thlluu-n j aba-at-el
ST_rlom. strong-m.id-pf.
'Tulluu got strong'.

In both structures, the subject is an experiencer of the state of

becoming 'tall' or 'strong'. Assuming adjectival roots/stems as statives,

Hayward (1975:213) calls the affix an inchoativizing formative.

A proper analysis of the middles may be possible if we approach them

via the type of thematic subjects they require, that is, whether they

select a subject which is a beneficiary, an experiencer, a patient, etc.,

or a subject which is just an agent, as Hayward has suggested. In the

light of this, such verbs may need to have entries that provide information

not only about their complement types but also about the thematic status

of their external arguments. Since the lexicon is the place for both

general and idiosyncratic properties of lexical items, or grammatical

formatives, the entries of such verbs may be enriched by including such

information as well.

Regarding the nrphological process involved in their derivations

a rule of the type shown below may be formulated.

Rule Eight /-at--/'3

(X -1 -> [ IX -] - at-] 'middle'

a V a

A point which we need to stress here is that the meanings of the middles in

the structures observed are not the same as the meanings we get from

structures involving pronominal anaphors, nor are the structures of these

two similar. This is evident from the examples below:



13

71 (a) [Thlluu-n farda bit-at-el

ST_nom horse buy-mid-pf.
'Thlluu bought a horse for himself'.

(b) Thlluu-n of-if farda bit-at-e

T-nom. self-for horse buy-mid-pf.
'Thlluu bought a horse for himself'.

(C) Thlluu-n of-un farda bit-at-e

T-nom. self-by horse buy-mid-pf.
'Tulluu bought a horse by himself for himself'.

72 (a) Thlluu-n of mur-e

T-nom. self cut-pf.
'Thlluu cut himself'.

(b) Tulluu-n of miid-e

T-nom. self hurt-pf
'Tulluu hurt himself'.

(C) Tulluu-n of haad-e

T-nom. self shave-pf.
'Tulluu shaved himself'.

From the structures in (71) it is possible to associate the affix

7-at-I with an adpositional phrase showing the beneficiary of the action

or the agent who personally carries it out. This is evident from the

adpositional elements 7-if-i 'for' in (71 b) and I-un! 'by' in (c).

The pronoun /of-/ 'self' which they govern is both the beneficiary and the

agent respectively. In other words, it is not the goal.

If it were the case that I-at-I showed a reflexive action, then the

presence of the reflexive pronoun /of-I 'self' would have made the

structures sound redundant if not ungrammatical, for, in that case, both

the affix and the pronoun would be doing the same thing. But the structures

are well-formed, which means that the two are not the same either

structurally and/or semantically.

 140

In the structures in (72) the verbs occur with anaphoric pronominal

objects. They also do not show any of the middle affixes which again

means that such affixes are limited to verbs of such structures as (71).

Since their function in such structures is to show the agent,

beneficiary, etc., status of the subject NP, that is, whether the action

is for, or by the subject, and since such notions imply that there is an

adpositional phrase of the type we have seen in (71), the affixes may be

collectively treated as 'pro-forms' of such adpositional phrases. In

this regard they may be said to be similar to the element I-itt-I which

appears in relativized positions of adpositional phrases of the type shown

in the (ii) structures of (73) below.

73(a) (i) [Tulluu-n [ulee-dan] nama rukkut-e]

S Adp''
T-nom. stick-with man hit-pf.
'Tulluu hit a man with a stick'.

(ii) [ [ulee-n [Tulluu-n naina itti-rukkut-e(-n)---]] [VP]]

SN? . S
stick--nom. T-riom. man which-hit-pf-with
'[The] stick with which Tulluu hit a man. . .'

(b)U) [Tulluu-n [gara Gimbii] deem-el

S Ad''
T-nom. to G. go-pf.
'Tulluu went to Gimbii'.

(ii) [ [bakk-j [Tulluu-n itti - deem-e (-n) ---11 (VP]]

SN? S
place-nom. T-nom. which-. go-pf-to(?)
'[The] place to which Tulluu went...'

(c)(i) (Thlluu-n [dafinoo] gara Gimbii deem-e]

S Adp''
T-nom. Monday to G. go-pf.
'Thlluu went to Gimbii on Monday'.

(ii) ( [guyyaa-n (Tulluu-n gara Giinbii itti-deem-e(-en)]] (VPJ]

SN? S
day-nom. T-nom. to G. which-go-pf. on
'[The] day on which Thlluu went to Gimbii'.
 141

As is observable from the (ii) structures, there is always the element

/-itti-/ preceding the verb of the relative clause. It refers to the

head of the relative clause (or the complement of the adpositional head

in the corresponding structures in (i). It is invariably the same for

all such complements of adpositional heads, irrespective of whether the phrases

are locative, temporal, instrumental, etc., just as I-at--I is the same for

the various interpretations of the middles. In this case both I-at-I and

/-itti-/ may be considered as 'pro-forms' referring to the entire

adpositional phrase in the case of the former, i.e., I-at-I, and to the

complement of the adpositional head in the case of the latter. In other

words, I-at-I refers to an 'auto-benefactive' type of phrase whereas /-itti-/

refers to the complement NP in an adpositional phrase of some adverbial

function.

Notice that the presence of I-at-I has some effect on the subcategorization

potential of some types of verbs. For example, transitive verbs like /arg-/

'see' are characterized by one NP complement. But the corresponding

middle form, /arg-at-/ 'see (find) for oneself' indicates the presence of

an adpositional phrase like /of-if/ 'for self'. This is in fact what the

structures in (71 b-c) show. There we have adpositional phrases with

pronominal complements which are coreferential with the subject NPs. In

other words, forms like /arg-at-/ or /bit-at-/ have one more complement

than their corresponding base forms. And in this regard they may be

classified along with such verbs as Ikenn-/ 'give' which selects two

complements: one a direct object and the other an indirect object. The

latter is always an adpositional phrase just like /of-if/ 'for self' is in

the structures in (71). The difference between them is simply that in

one, the complement of the adpositional phrase is an anaphor whereas in

 142

the other it isa (R-)eferring expression. And it appears that the

reason why /of-i/ 'for self' is optional in surface structure is that it

is construable from I-at-I, its 'proform', or from the features of the

external argument. Hence for a middle verb like /arg-at-/ the

subcategorization frame looks like the following.

/arg-at--/ : V + [(Adp'') N'''—] 'find (see) for oneself'.

3.2.1.3 .Adjectivals

In the paradigm of three in the preceding section we have noticed

processes of deverbative agent formations. The same process seems to

derive adjectivals as well. The deverbative affix in either case is the

same /-aa/. Consider the following.

Paradigm Nine

V A
fokkis- fokki-saa
'look ugly' 'ugly'
gonkis- + -aa —> gonkis-aa
be harsh harsh
barbaacc-is barbaais-aa
be.'necessary necessary
kabbanaa.. kabbanaaw-aa
be-cold cold
Rule Nine /-aa/

[X-(is)] —> [[X-(is)] -aa]

V sV

The derived forms above have all the properties of adjectives. They

can occur in both predicative and attributive positions and can also be

preceded by the degree word Ibay?ee/ 'very'. Observe the following

examples.


143

74 (a) [ (nam-i!-i fokkisaa-n kuni-il eessaa duf-el?

S NPman_sgi_nom. ugly-nom this-nom where come-pf.
'Where did this ugly man come from?' (Gragg 1982:180)

(b) (na1n-i-i kuni-i [fokkisaa-da] I

S VP
man-sgl-nom. this-nom. ugly is
'This man is uglyt.

Cc) [bay?ee fokkisaa I

AP
'Very ugly'.

The same is true of the others.

The sulcategorization properties of such adjectivals are basically

the same as those of their corresponding verbs. In both cases the

complement is an adpositional phrase of purpose. Compare the following.

75 (a) (j) [aannan-(i) [ [namaa-fil [hin-barbaais-a]:}]

S Vt Adp'' man-for 'cm-be-necessary-impf.
milk-nom.
'Milk is necessary for man'.

(ii) [aanan.(i) [ [nainaa-fi] [barbaaisaa]I -da]

S ilk A' AdP'' A
for is
'Milk is necessary for man'.

(b) Ci) [ilkaan(i)-ii [ [arg-uu-daf I [hin-fokkis-a]1]

Stooth_nom_herV S see_to_for Vcm_ugly_impf.
'Her teeth appear ugly to look at'.

(ii) [ilkaan(i)-ii [ [arg-uu-daf] [fokkisaa]] -dal

S A'S A
tooth-nom-her see-to-for ugly is
'Her teeth are ugly to look at'.

Such adjectivals, like those simple forms we have observed in

Section 3.1.3, might be associated with optional adpositional phrases of

purpose. And a.s stated earlier in connection with the complements of

simple adjectival heads, the fact that such complements are purposive

adpositional phrases may lead to some degree of generalization about [+N]

categories, since simple nominals can also take the same type of complements,

though of a different category. 14



14

There is, however, some problem with this line of argument. The

problem is that what have been considered as complements of A', that is,

the adpositional phrases in such structures as (75 b) above, might be

complements of the VP Cv''). In other words, the structural relation of

such complements may be with the VP rather than with the adjectival head.

Accordingly, (75(b)) above might be relabelled as shown in (76) below.

7. [ aannan(i)
.
[ [namaa-fi] [ [barbaaccisaa] -dali]
V 1 Adp ' ' V' A
milk-nom. man-for necessary is
!Milk is necessary for man'.

According to this analysis, the purposive phrase /namaa-f 1/ 'for man'

is a part of V I, and not of A'. And as we shall observe in the next

chapter, such relationship is a property of VP adverbials. Such adverbials

can be fronted as in (77 a) , or remain unaffected by gapping as in (b ii)

77(a) [namaa-fi [aannan(i) [ t I [barbaaisaa] -dali]]

S S V't 1 V' A'
man-for milk-nom. necessary is
'For man, milk is necessary'.

(b) (j) (aannan(i) [ [namaa-f ii [ [barbaaisaal -dali]

S V''Ad' V1 A'
milk-nom. man-for necessary is
'Milk is necessary for man'.

( I i) [arakee-ni iimno [ [koraa-f] [ I]]

S V'' Adp'' VI
arakee-nom. also cold-for
'And so is arakee for cold [weather]'.

From such constituency tests it seems that the purposive adpositional

phrases in (75) may indeed be treated as V' 1 complements and that adjectives

may be considered as having no complements. But this again is not without

any problems. In structures like (75 a) (ii), the purposive phrase /namaa-f i/

'for man' and the adjective /barbaaisaa/ 'necessary' can be fronted as a

unit leaving the copula /tur-e/ 'was' intact. Consider the following.


145

78 (a) aannan-(i) [namaa-fi gaarii] tur-e

milk-nom man-for good be-pf.
'Milk was good for man'.

(b) jnamaa-f I gaarii-n 15 (aannan-(i) t tur-e]]

1 S 1
man-for good -? milk-nom. be-pf.
(literally), 'Good for man, milk was'.
/namaa-fi gaarii/ 'good for man' can also be deleted on identity with a

similar constituent as in (79).

aannan-i
79 (a) (namaa-fi gaarii] -da
A
milk-nom. man-for good is
'Milk is good for man'

(b) arakee-n garu [- -] miti

A'
arakee-noin but not-be
'But arakee is not'.

Such facts suggest that /namaa-f i/ 'for man' forms a syntactic unit

with the adjective /gaarii/ 'good' just as it did with V' in forming V'' in

(76). It does, therefore, seem that in copular structures of the type

under consideration, a purposive adpositional phrase may be treated as a

complement of either the adjective as in (79) above, or the VP(V'') as

shown in (76). In short, copular structures with such adpositional phrases

are structurally ambiguous between the two configurations (75 ii) and (76).

Having this in mind, we may associate adjectives with the entry shown

below for /barbaaisaa/ 'necessary' for purposes of maintaining the

generalization referred to earlier on in connection with [+N] categories

and for reasons of completeness.

• / : A[Adp''—]
/barbaacciisaa 'necessary'
pur.

3.3 Summary

We have tried to show two things in this chapter. The first is the
 146

specification of lexical items in each of the major categories in terms of

syntactic semantic and phonological information. Syntactically this

involves formulating subcategorization frames in which are expressed the

possible constituents that an item selects at its niminal level of projection.

These include, inter alia, the argument complements of [-NJ and the

ndifying complements of N-NJ categories.

The discussion we have had with respect to such properties may be

captured by the following rules:

N'—> N (N''')

Adp'
V' —> Adp'' N''' V
N'''
A''
S

A'—> (Adp''') A

Adp'—>f N'''I

l_ J Adp

Two cross-category generalizations seem to emerge from this state of

affairs. The first relates to the types of complements the [+NJ categories

permit. Their complements are optional This is in contrast to the

con1ements of the [-NI categories which are always obligatory.

In terms of headedness, it appears that the four categories fall into

two: those which are left-headed and those which are right-headed. Only

the category [+N-V] belongs exclusively to the former which suggests that

there is some imbalance in the distribution of the categories with respect

to the notion of headedness since only one out of the four major categories

is left-headed.

It might perhaps be necessary to reconsider the position we have

taken with regard to the analysis of adpositional phrases. The category

 147

has both pre- and post-positional elements. The analysis was based on the

fact that the pos1—positions outnumber the prepositions from which also

follows the typology of Oromo as a postpositional language. If we argue

in favour of its being prepositional (contrary to observable data) we may

achieve a nice symmetry between [+V] and (-VI categories since adpositionals

will forni a natural class with nominals, both of which are [-VI and which

are left-headed.

The other thing we have dealt with in this chapter is a brief overview

of the derivational morphology of lexical heads and the effect this has on

subcategorization properties. As Chomsky (1981:126) has noted, morphological

processes have the effect of blocking 8-roles or assigning new 8-roles,

which seems to be true since as we have observed, passivization has the

effect of blocking the assignment of 8-roles to the position where an

external argument is expected, whereas causativization has the effect of

assigning new 8-roles to internalized arguments. Hence a causativized

verb has more complements than its non-causativized counterpart.

These two processes are widely discussed in the literature in relation

to languages like English. The situation in Oromo, and perhaps in others

like it, seems to require us to consider processes which derive a category

of middle verbs along the same lines. The process does not seem to have

a blocking effect on 8-roles. In fact, one may argue that it has the

effect of producing new 8-roles. Whereas the subject of a clause with a

transitive non-middle verb is just an agent, the subject of a clause with a

transitive middle verb may be both an agent and a beneficiary. Following

this we have argued that transitive verbs with middle morphology may be

considered as having an adpositional phrase as a complement in addition

to the Np which they require to satisfy their inherent lexical properties.

 148

The benefactive role may then be associated with this (adpositional)

complement. The effect of such a proposal on the relation of 8-roles and

arguments is that it makes the relation unique as required by Chomsky's

(1981) 8-criterion which dictates that an argument should have one and

only one 8-role. The association of both agent and benefactive 8-roles

to the subject of a clause is a clear violation of this criterion, which

Chomsky assumes to be universal. The association of the (auto)benefactive

role with this complement may avoid this problem.

 149

NOTES TO CHAPTER THREE

1. This is true mainly of the complements of the [-N] categories.

2. There are cases, however, where a derivative differs from its source
in its subcategorization potential (cf. 3.2.1.3).

3. Figuratively it is possible to say /wa4a iyyeessaa/ 'God of the

poor'.

4. The frame is necessary to show that the verb is final in clauses.

5. Such anomalies would not arise if the frame included selectional

restriction features as part of the specification of the item in
question.

6. This does not mean that there is no agreement between the verb and
the subject. what is meant is that the verb does not actually
have overt agreement features when the subject is 3ms. This is
the unmarked form of the verb.

7. These are differences based on surface structures. Underlyingly

all of the verbs belong to the class of transitives.

8. The verb in Oron does not have direct object marking. But it
seems that it shows indirect object by affixing the element I-fl.

9. There are in fact three /akka/s to recognize: the comparative,

the temporal, and the complementizer /akka/.

10. This /-ia/ may not be a deverbative element. It could be the

same singulative or determinative element discussed in Chapter Five.

11. This classification is perhaps not so simple. For example, /deem-/

'go' is intransitive, but it takes /-siis-/ to form the causative.
The reason for such irregularities may be attributed to the semantics
of such verbs. They show volition (cf. Hayward 1975) in addition
to causation. In other words, the process is not of simple
causativization only.

12. As stated above in note 11, this is just one instance of three
alternants. Hayward (1976) argues in terms of Is, us, and Ills,
the numbers indicating the number of arguments a causative verb may
have.

13. Other alternants are I-add-I for first person singular, and f-an-I
for all plurals. (See Hayward 1975).

14. The similarity between the complements of the two categories is,
hence, , only functional.

15. /-n/ may be a focus marker.

 150

CHAPTER FOUR

COMPLEMENT TWO

4.0 Introduction

The major lexical categories we have established in Chapter Two have

the potential for maximal projections. This potential accords with the

theory of X-bar syntax which stipulates that constructions a±e endocentric

and that lexical categories constitute the heads of such constructions.

The projection potential of a lexical category is determined by the

hierarchies of structures in a construction unit in which it constitutes

the head. The hierarchies in such constructions are indicated by bar

notations. For example, for a major lexical category X, the bars which

are associated with each level of hierarchy in the construction unit of

which X is the head, may range from X°to x(0^ where n is a specified

number indicating the maximal projection of X.

In the preceding chapter we have observed various types of constituents

which X selects in forming its minimal projection X'. Following Jackendoff

(1977:57) we have argued that such constituents are functional arguments

strictly subcategorized by X and should be entered in the lexicon as part

of the inherent lexical specifications of X.

In this chapter, we shall consider each of the major lexical categories

again, and examine the type of constituent structures they select to form

the next higher level(s) in their projections. In the course of the

discussion we shall try to see if the Uniform Three-level Hypothesis which

Jackendoff (1977:42) argues for for English applies for Oromo as well.

According to this hypothesis, all major lexical categories have the potential

for a treble bar maximal projection. Furthermore, the structures which

they select at every bar level in their projection lines are believed to be
 151

similar across all categories. We have noticed in the preceding chapter

that this is partially true. The complements are tightly bound to their

heads in the [-N] categories though they do not necessarily belong to the

same category. The complements of intransitive verbs are, for example,

adpositional phrases whereas those of transitive verbs may be clausal or

phrasal (NP) arguments. But both are strictly subcategorized by their

heads, and hence belong to the same bar level, though not to the same

category.

In what follows, we shall consider constituents which occur at the

next higher level. Before we go into that, however, it may be necessary

to make some distinctions among the major categories. Such distinctions

are not far removed from those we have already made in the preceding chapter

with regard to the subcategorization properties of lexical items.

The first distinction to be made is that between nominals on the one

hand and non-nominals on the other. This distinction relates to the

- notion of head and to its position in syntactic categories. As mentioned

earlier, non-derived nominals are left-headed, which means that in a simple

noun phrase, the head occurs preceding its complement, whereas in all other

categories, it occurs following them.

The second distinction is between [-N] and [-'-N] categories. The

complements of the former are tightly bound to their lexical heads at the

minimal level of the projection line. In other words, such complements

are strictly subcategorized so that their omission would lead to

ungrammaticality. It is only those complements which appear at the

intermediate and maximal levels of projections which may be omitted freely.

On the other hand, the complements N-N] categories are not so strictly

subcategorized and hence may be omitted at any level in the projection line.
 152

In other words, they are in effect similar to those complements of f-NI

categories which occur at the intermediate and maximal levels of projection.

At the minimal level of projection, all types of complements occur as

sister of X . These are treated in the lexicon (in the manner shown in the

preceding chapter) as part of the specifications of the inherent properties

of X (i.e. a lexical item, in terms of which they are subcategorized).

They will not be dealt with in this chapter for the main reason that this

would lead to unnecessary redundancy as noted in Chomsky (1981:31) and

Hoekstra (1984:24) where it is argued that such strictly subcategorized

functional arguments 1 should be included in the lexicon as a specification

of their lexical heads.

This means that the categorial component should be restricted to the

specifications of the complements at the level(s) higher than the minimal

projections. In other words, the categorial rules specify the positions

to which those complements which have no direct thematic relations with

their lexical heads and which are hence optionally selected by such heads

are associated in D-structures. These include structures of adverbial

functions which, on the basis of their restrictive or non-restrictive -

nature, are associated with one or another level of projection across all

major categories.

The function of the categorial component is in this sense reduced to

that of determining the levels or nodes with which a complement or

concatenation of complements (or specifiers for that matter), are associated.

It does not seem to show the precedence relations of such complements at any

level, for there do not seem to be uniform co-occurrence restrictions,

and even if there were one, it would be disrupted by the rule move ci. at some

level in the derivation of clauses, as Hoekstra (thid.) has noted in relation

 153

to languages like English which allow movement. The same might be said

about the situation in Oromo since the language is characterized by some

type of movement, and co-occurrence restrictions at the levels higher than

the minimal projections seem to be loose. This will be observed in due course.

Anticipating the arguments that will follow, we shall assume here (n)

in x(0+1) to be 3, and following Jackendoff, we shall call X' complements

'arguments' and X'' and X''' complements 'restrictive' and 'appositive'

complements respectively. The categorial component as a device for

specifying the positions of the latter two is the subject of the rest of

the discussions in this chapter.

4.1 X'' Complements

These are maximal phrasal or clausal constructions which restrict the

reference of the head by supplying information about its attributes, means,

manner, etc., all of which are circumstantial to the event or the

assertion which a strictly subcategorized complement and its head express.

As stated earlier, such complements are optional as opposed to those

complements of X' which are obligatory, and which they immediately precede or

follow, depending on the type of the category which constitutes the head.

In the discussion that follows, we shall concentrate on the syntactic

properties of such complements across all major categories. We shall take up

each category in turn.

4.1.1 N'' Complements

As discussed in Chapter Three, two types of nominals are recognizable:

simple nominals and derived nominals. The latter refer to infinitivals/

gerundives. They have the internal structures of clauses, and hence differ
 154

from other nominals, which may also be derived but may not have such

characteristics. The distinction between the two is also important for

two other reasons: (1) they select different types of complements, and

(2) they differ in their positions as heads of constructions.

4.1.1.1 Simple Nominal (N'') Complements

These include adjectives, restrictive relative clauses, and some

genitive NPs, all of which may have the effect of restricting the head of

the construction in which they occur as modifying complements. Let us

observe the following:

(i) Adjectives

1(a) [ [fard-i [guraa--i]i du? - ci

SN'' A''
horse-nom. black-m-nom. die-pf.
'[A] black horse died'.

(b) [ [dubaartii-n [bareed-duu-nJ] duf-t-e]

S N' A''
woman-nom. beautiful-f-nom. come-f-pf.
'The beautiful woman came'.

In the structures above, and in others like them, the nominal heads

If ard-i/ 'horse-nom' and /dubaartii-n/ 'woman-nom' are followed by the

adjectival complements and are restricted in reference to only those which

are characterized by the attributes which the complements express.

The complements in the above structures and in all other similar

structures are maximal projections. This is in accordance with the

definition of complements as 'concatenations of syntactic categories'.

The adjectives in (1) are hence maximal categories which, in anticipation

of the discussions that will follow soon, are believed to be at the double

bar level.

N'' complements al;ays follow N' complements as the following examples

show.
 155

2(a) [Thlluu-n [ (farsoo [garbuu]J [gaarii]] hin-aalat-a]

S N' N'
T-nom. beer of-barley good cm-like-impf.
'Thlluu likes good beer [made] of barley'.

(b)* [Thlluu-n [ [farsoo [gaariil] [garbuu]] hin-aalat-a]

S N'' N' A'' N'''
T-nom. beer good of-barley cm-like-impf.

3(a) [nam-ni hund-i [ [buddeena [taafii]] [adii]]

S N'' N' N''' A''
man-nom. all-nom. bread of-tef white
hin-barbaad-a]
cm-want-impf]
'Everybody wants white bread [made] of tef.'

(b) * [nam-ni hund-i [ [buddeena [adji]] [taaf ii]

S N''N' A'' N
man-nom. all-nom. bread white tef
hin-barbaad-a]
cm-want-impf.

4(a) [Tulluu-n [ [wayaa [muaa]] [bareed-duu]] bit-el

S N'' N' N''' A''
T-nom. garment of-child beautiful-f. buy-pf.
'Tulluu bought [a] beautiful garment [for] a baby'.

(b)* [Tulluu-n ( (wayya (bareeduu]] [muaa]] bit-e]

S N'' N' A'' N'''
T-nom. garment - beautiful child buy-pf.

The ungraxnmaticality of the (b) structures suggests that in such

structures of NPs, the adjectives can occur only after the genitive NP5.

The latter are complements of N' as has been discussed in the preceding

chapter, which results in their position being immediately after the head

and before other complements.

(ii) Genitive NPs

These may be divided into two sets: those which function as specifiers,

and those which function as complements. As shown in Chapter Three, and

as will also be discussed in depth in Chapter Six, genitive phrases of

possession may belong to the class of specifiers while all other types of

genitive phrases may belong to the class of complements. As complements,

 156

they may fall into two groups: those having some adverbial function and

those without such functions. The former may be treated as complements

of N'' or N''' depending on whether their relation with the head is one of

restriction or of apposition. Those which have a restricting effect on

their heads may be assigned to N'' and those without such an effect to N'''.

Those which do not have adverbial functions may be treated as complements

of N'. These are to be indicated in the lexicon as part of the lexical

specification of the head. Consider the following examples:

5(a) [ [ [daadii-n [danimaa]]2 [Gimbii]] gaarii-da]

S N'' N' N''' N'''
mead-nom. of-honey of-Gimbii good-is
(Literally), 'mead of honey of Giimbii is good'.

(b)* [ [ [daaii-n [Gimbii] [danimaa]] gaarii-da]

S N' ' N' N'' ' N''
mead-nom. of-Gimbii of-honey good-is

6(a) [ [ [daadii-n [daaa]] [kaleessaa]] gaarii tur-e]

S N'' N' N' ' N'
mead-nom of-honey of-yesterday good be-pf.
(literally), 'Mead of honey of yesterday was good',
'Yesterday's honey mead was good'.

(b)* [ [ (daadii-n [kaleessa]] [dammaa] gaarii tur-e]

S N' N' N''
mean-nom. of-yesterday of-honey good be-pf.

The ungrammaticality of the (b) structures may be attributed to the fact

that the locative and temporal genitives /Gimbii/ and /kaleessaa/ 'of-yesterday',

have occurred in the position of an N' complement, or that the 'source'

genitive /danmiaa/ 'of-honey' has been raised from its base position in N' to

another position in the next higher level.

In either case the situation is one which shows that the two types of

genitives cannot freely exchange positions.

The position of the locative and the temporal genitives in the

grammatical structures above is the same position where adjectives were

seen to occur in our earlier examples (2-4).



(iii) Relative Clauses

Unlike genitives and adjectives which are phrasal, relative clauses

are sentential and are, therefore, subject to an independent analysis on

their own. However, in so far as their functions and distributions are

concerned, they are similar to adjectives, and those genitive NPs which

we have considered as N' •' complements. Like them, they always occur

following N' complements. as the following examples illustrate:

7(a) [Tulluu-n ( [daadii (dammaa]] [Fayyiisaa-n . bite]] dug-el

S N''N' N'''
T-nom. mead of-honey F-nom. buy-pf. drink-pf
(literally), 'Tulluu drank the mead of honey [which] Fayyiisaa
bought'.

(b)* (Tulluu-n [ [daadii [Fayyiisaa-n bit-e]] [dammaa]] dug-el

S N''N' S N'''
T-nom. mead F-nom. buy-pf. of-honey drink-pf.

( 8(a)
[ [farsoo-n (garbuu]] !Tulluu-n dug-e]1
SN'' N' N''' S
b eer-nom. of-barley T-nom. drink-pf.
gaarii tur-e]
good be-pf.
'[The] beer of barley (which] Tulluu drank was good'.
(b)* [ ( (farsoo-n [Tulluu-n dug-el] [garbuu]]
S N'' N' S
beer-nom. T-nom. d.rink-pf. of-barley
gaarii tur-e]
good be-pf.

The reason for the ungraxnmaticality of the (b) ctructures is the same

as that stated in connection with the structures in (5) and (6). The

relative clause in such structures cannot occur immediately following the

head if there is an N' complement. This is possible only when the N'

complement is missing, as in (9) below, for example, which is derived from (8 a)

9. [ [ [farsoo-n] [Thlluu-n dug-el] gaarii tur-e]

SN''N' S
beer-nom. T-nom. drink-pf. good be-pf.
'[The] beer [which] Tulluu drank was good'.


158

It is not our purpose here to deal with the internal structure of the

relative clause. Disregarding that for the nment (cf. Chapter 5), we shall

limit the discussion to a description of their position in structures of the

type under consideration. As the examples indicate, the position they

assume in such structures is the same as that where the adjectives and those

genitive NPs of time or location are found in structures such as (5-6).

Like other ndifying complements, relative clauses restrict the reference of

the head, that is, N' in (7-9), to the one which Thlluu 'bought' or 'drank'.

In all the examples we have seen so far, the distinction we have made

between N' and N'' complements is based on distribution. The complements

of N'' occur only after those of N'. This may be a necessary condition,

but it is not a sufficient one. The ungrammaticality of those structures

like (8 b) , for example, could also be explained in terms of precedence

relations within a level of hierarchy. In other words, it may be argued

that what we have labelled as N' and N'' complements could all be treated as

N' complements, but with a strict linear ordering. Accordingly, we could

dispense with N'' as an independent bar level in the projection line of N.

Such arguments would ultimately lead to the conclusion that there may

not be any intermediate categories at all. In order to justify the claim

that there is a distinct level N'', and that the types of complements at

this level are different from those of N', we need to apply some constituency

tests. One such device is gapping. It operates on string of items which

form a single constituent, or on a level within a larger constituent. In

the light of this then, let us consider the structures in (10) below:

10(a) [ [ [farsoo-n (garbuu]] [Thlluu-n (kan) bit-el]

SN'' N' N''' S
beer-nom. of-barley T-nom. comp. buy-pf.
gaarii-da]
good is.
(Literally), '[The] beer of barley which Tulluu bought is good'.
 15

10 (b) [ [ [farsoo-n [garbuu]] Fayyiisaa-n (kan) bit-el]

S'N'' N' N'" S
beer-nom. of-barley F-nom. comp. buy-pf.
gaarii miti]
good neg-is
(Literally), '[The] beer of barley which Fayyiisaa bought is
not good'.

(c) [ [ (farsoo-n [garbuu]l ! Th11uu -n (kan) bit-el] gaarii-dal

S N'' N' N'" S
beer-nom. of-barley T-nom. comp. buy-pf. good is
'[The] beer of barley which Thlluu bought is good'.

[Tulluu-n (kan) bit-el] garuu gaarii miti]]

N'' N'
T-nom. comp. buy-pf. but good neg-is
'But [that] which Tulluu bought is not good'.

In (10 c) , /farsoo-n garbuu/ 'beer of barley', is deleted. The

process does not exclude /garbuu/ 'of-barley', a case which would have been

possible if it were not part of N'. In addition, the relative clause is

not affected by the process which means that it belongs to a different

level. This, plus the strict linear ordering we have observed earlier

on (cf. (8 b)) strengthens the argument that there are hierarchical levels,

and that each level is characterized by a different set of complements.

Now if this is the case, then the next question we have to address

ourselves to is whether or not there is a restriction on the precedence

relation of complements belonging to the same level. Structures of N' 's

with all their complements, that is, genitive locatives, adjective and

relative clauses are very rare but not absolutely impossible. Whenever

they all occur, they maintain a strict linear order. This is noticeable

from the following examples:

11(a) 1 [axnartii [workii]] [Wallaggaal [bareed-duu]

N' ' N' N' ' A'
ring of-gold of-W. beautiful-f.
[Tulluu-n (kan) bit-el ---]
ST_nom. comp. buy-pf.
(Literally), '[A] beautiful ring of gold of Wallaggaa which
Tulluu bought'.


160

11 (b) [ [amartii [workii] [bareed-duu] (Wallaggaa]

N'' N' .
ring of-goid beautiful-f. of-W.
[Tulluu-n (kan) bit-e] ---]
ST_nom. comp. buy-pf.

(c)* [ (amartii [workii] [Tulluu-n (kan) bit-el

N'' N'
ring of-gold -nom. comp. buy-pf.
[bareed-duu] [Wallaggaa] ---I
beautiful-f. of Wallaggaa.

From the grammaticality of (11(a) and the ungrammaticality of (b) and

(c), it may be inferred that the linear ordering of the complements is as

shown in (12) below.

12 N'' —> N' (N''') (A'') (S)

In cases where a genitive NP (N''') is used, a structural ambiguity

may arise if the N' complement which is also a genitive NP is already there.

Consider the following examples.

13(a) [Tulluu-n [ [ainartii [workii]] [Wallaggaall bit-el

S N' N' N' ' N'
T-nom. ring of-gold of W. buy-pf.
'Tulluu bought a gold ring of Wallagga'.

(b) [Tulluu-n [ [daadii [dammaa]] [Gimbii]J dug-e]

S N''' N' N''' N'''
T-nom. mean of-honey of-G. drink-pf.
'Tulluu drank honey mead of Gimbii'.

In (13 a) , the N' complement /wo ii! 'of-gold' and the N'' complement

/Wallaggaa/ 'of-Wallaggaa' occur in contiguity though they belong to

different levels in the hierarchy. In the reading indicated by the

labelled brackets, /amartii wo.r}ii/ 'ring of-gold' constitutes the head of

the larger constituent N'', and /Wallaggaa/ occurs as a complement of this

head. This gives us the interpretation as shown in the gloss. In

another reading, it is also possthle for /workii/ 'of-gold' to be in

configuration not with /amartii/ 'ring', but with /Wallaggaa/, in which it

constitutes the head. The two readings are shown below in (14).
 161

14(a) (=13) (b)

N'' N''

A
amartii worki Wallaggaa
ring of-gold of-Wallaggaa
amartii
ring
'ring of gold of Wallaggaa'
N N'''

A workii
gold
Wallaggaa
of-Wallagga
'ring of Wallaggaa gold'

As is observable from the trees, the ambiguity is between /warkii/

'of-gold' being the complement of N' in (14 a). = (13 a) , or the head of

N'' in (14 b) . Such ambiguities arise not so much from the contiguous

occurrences of the two genitive NPs, as from the lack of any overt genitive

imrphemes. As stated earlier (cf. note 2), such genitive relationships are

indicated by syntactic positions and by phonological means. There is,

therefore, every possibility for the second NP in structures of phrases of

the type in (13) to be read either as a complement of a preceding head or as

a head of a following complement. In short, /wor]ii/ 'of-gold' may be

treated as part of N' or N''.

One way of disainbiguating such structures as (13) is to leave one of

the genitive NP complements phonetically null, which is possible given the

fact that the complements of (+NJ categories are optional. Hence, instead

of (13) above, we may have (15), below.

15(a) [Thlluu-n [amartii [workii]] bit-el

S N' N'''
T-nom. ring of-gold buy-pf.
'Tulluu bought a ring of gold'.

(b) Thlluu-n [workii [Wallaggaal] bit-e

N''' N'''
T-nom. gold of-W. buy-pf.
'Tulluu bought gold of Wallaggaa'.
 162

Another way of disaxnbiguating structures like (13), without missing

a complement, is by paying attention to the location of pause in the clause.

In (14) above, /workii/ 'of-gold' forms a pause group either with /amartii/

'ring', or with /Wallaggaa/, each intended for a different interpretation,

as shown in the glosses in (14).

The situation would remain the same if instead of the locative

genitive in (13), we put a temporal genitive as in (16) below, which has

either one of the readings in (17).

16 [ [daadii [dammaaj] [kaleessaa] ---1

N'' N' N'''
me ad of-honey of-yesterday
(Literally), '[The] mead of honey of yesterday'.

17 (a) -- (b) N''

N' N'''
I I
N N''
/
daadii damxnaa kaleessaa daadii //'\
mead of-honey of-yesterday mead

NI
N N'''

'Mead of honey of yesterday.'

/\
I_______________________________

damma kaleessaa
honey of-yesterday

'mead of yesterday's honey'.

In (17), /dammaa/ 'of-honey', is a complement of /daadii/ 'mead' in (a).

In (b), it forms the head of N'' of which /kaleessaa/ 'of-yesterday' is a

complement. In each case, the interpretation is different.3

4.1.1.2 Derived Nominals

In this subsection, we shall be dealing with one type of deverbative

nominal in the light of the preceding discussions concerning simple nominals.



163

The type we will be concerned with here is one which is sentential in

character. We shall not consider its internal structure directly now.

That will be the subject of Chapter Five. Here we shall concentrate on

its choice of complements at the level we are considering. Let us observe

the following examples.

18 (á) [ [Thlluu-rra [hoolaa bit-uu-n]] hin-danda?-ani-u]

SN'' N'
T-from sheep buy-to-nom. neg-be-able-ps-iinpf.
'Buying/to buy a sheep from Tulluu is impossible'.

(b) [ (eeboo-dan [bineensa a'5ee-s-uu-n]1 gaarii mitii]

SN'' N'
spear-with animal kill-cs-to-nom. good-neg-is
'Killing (an] animal with a spear is not good'.

In (a),the source adpositional, /Tulluu-rra/ 'from Thlluu', and in

(b), the instrumental adpositional /eeboo-dan/ 'with spear', ocur as

complements of their respective heads (N'). Positionally, such complements

precede the strictly subcategorized complements of N' as we can observe

from both structures. In terms of their function, they modify the head by

restricting it to the particular context they designate. Compared to the

complements of N', which are NPs, and are tightly bound to their heads,

these are adpositionals and are not so bound to the head. Hence only they

but not the NP5 of N' may occur optionally. This is observable from the

grammaticality of (19 a) compared with the marginal status of the

corresponding (19 b) below, both of which are derived from (18 a) above.

19( a ) [ [(Adp) (hoolaa bit-uu-n]1 hin-danda?-am-a]

SN'' N'
sheep buy-to--nom. cm-be able-ps-impf.
'Buying/to buy a sheep is possible'.

(b) ?( [Tulluu-rra ((NP) bit-uu-n]] hin-danda?-am-a]

SN'' N'
T-from buy-to-nom. cm-be able-ps-inipf.
'From Tulluu buying/to buy is possible'.

The claim that such adpositional phrases belong to N'' can be further

substantiated if we apply the syntactic device of gapping on either of the

 164

examples in (18). The process deletes the head of N'' on identity with

another phrase in a previous structur in an appropriate discourse

situation. Thus (18 a) may be reduced to (20) below:

20 ( [Tulluu-rra [hoola bit-uu-n] I hin-danda ?-am-a],

SN'' N'
T-from sheep buy-to-nom. cm-be- able-ps-impf.
(Literally), 'From Tulluu to buy a sheep is possible'.

[Fayyiisaa-rra garuu [ hin-danda ?-am-u]

SN'' N'
F-from but neg-be able-ps-impf.
'From Fayyisaa it is not possible'.

The fact that /Tulluu-rra/ 'from Tulluu' has not been deleted along

with its head N' shows that its relationship with the head is at a level

higher than N'. That level is the level if N''. At the level of N',

the relationship we have obtains between /hoolaa/ 'sheep' and the lexical

head /bit-uu/ 'buying/to buy'.

Along with or instead of adpositional phrases, we may also find

subordinate clauses of adverbial functions occurring as complements. The

following are examples with such structures:

21(a) ( [ [yoomuu bokkaa-ri roob-u] [deem-uu-n]] [gaarii mitil]

SN'' S N'
when rain-nom. rain-impf. go-to-nom. good not-is
'Going when it rains is not good'.

(b) [ [ (erga Tulluu-n deem-el [hefl-e&'-uu-n]] [gaarii-da]]

S N'' S N'
after T-nom. go-pf. work-mid-to-nom. good is
'Working after Tuiluu has gone is good'.

In both structures, the subordinate clauses are structurally related

to their infinitival heads /deem-uu-n/ 'going/to go' and /ha53-e-uu-n/

'working/to work', respectively. This is what the labelled brackets show.

However, it might not be impossible to argue here that in such structures

the subordinate clauses may be treated as complements of the predicate

phrase rather than of the noun phrase (N''). But this does not seem to be


165

the case, because in that case, the position of the subordinate clause

has to be in the predicate, preceding the VP (V'), and following the

subject NP, unless we also argue that the position in whiph they appear

in (21) is a derived position. In other words, such structures are derived

from the corresponding structures in (22) below. Again, this does not

seem to be right because the structures in (22) are ungrammatical.

22(a)* [deem-uu-n ( [yoomuu bokka-n roob-u (gaarii miti]]]

S VI'S VI
go-to-nom. when rain-nom. rain-impf. good not-is
(Literally), 'Going when the rain rains is not good'.

(b)? [ho-e-uu-n [ [erga Tulluu-n deem-eJ [gaarii-daJ]J

S VI'S V'
work-mind-to-nom. after T-nom. go-pf. 'ood is
'Working after Thlluu has gone is good'.

If the examples in (22) were grammatical, we could argue that the

structures in (21) had been derived by preposing the subordinate clause.

But even then the argument would still fail to hold because preposing

involves adjunction to S/S, but not to an NP, which means that the clause

would still have no structural relation with the latter. But subject

postposing shows that this is not the case either. In (21a), for example,

/deem-uu-n/ 'going/to go', cannot be postposed without the subordinate

clause. This is obvious from the degree of acceptability of (23) below.

23(a) [ [ti [gaarii miti]] ( [yoomuu bokkaa-n roob-u] [deem-uu-n]]

SN''i V' N''S 1
good not-is when rain-nom rain-impf. go-to-nom.
'Going when it rains is not good'.

(b)* [ ( [yoomuu bokkaa-n roob-u It]] [gaarii miti] [deem-uu-n]]

SN''S 1 V' 1
when rain-nom rain-to good not-is go-to-nom

The situation in both (22) and (23) favours the analysis of the subordinate

clause as labelled in (21).

Assuming this to be correct, we shall now consider the relative position

of subordinate clauses and adpositional phrases when both occur preceding the
 166

same head. Such structures as (24) below may dennstrate this:

24(a) [ [ [yoomuu bokkaa-n roob-u] [namaa wain]

SN'' S when rain-nom. rain-impf. Adp''
man with
[deem-uu-n]] [gaarii miti]]
N' V1
go-to-nom. good neg-is.
'When it rains going with a man is not good'.

(b)? [ [ (nama wa3'in] ! yoou bokkaa-n roob-u] [deem-uu-n]]

S N'' Adp'' S N'
man with when rain rain-impf. go-to-nom.
[gaarii miti]]
V good not-is.
'When it rains going with a man is not good'.

From (24 b) it seems that the clause and the adpositional phrase cannot

easily exchange positions, though they belong to the same level, N''. In

the light of this, a categorial rule of the type below may be postulated to

show their relative order:

25 N'' > (S) (Adp'') N'

In Chapter Two, it was argued that pre-/post-positions and subordinative

conjunctions could be treated as subclasses of the nxre general category -

the category of adpositionals, idiosyncratic properties in their choice of

complements being left aside for the lexicon. If we maintain that analysis,

then the rule in (25) could be reduced to the one in (26) with the diacritic

mark (*) showing the possibility of having concatenations of such complements.

26 N'' > (Adp''*) N'

It does not seem possible to collapse this rule with the one in (12),

which has been proposed for simple nominals, for two reasons. Firstly,

there is a difference between the t nominals in their choice of complements.

The ccmplenionts here are all adpositional phrases/clauses 4 whereas those

in (12) are not. Secondly, even if it were the case that they had some

complements in conon, it s.ould still be difficult to collapse them into

 167

one class for they also differ in the position of their heads. The head

in (26) follows its complements, whereas the head in (12) precedes them.

A possible generalization we can think of is that which puts such deverbative

nominals along with verbals, which is very natural given the fact that

they (deverbatives) are basically derived from verbs. We shall consider

this possibility after we have observed the type of complements verbals

take at the same level at which we have examined the deverbatives.

4.1.2 V'' Complements

The structures which occur as complements of V'' are manner, time,

purpose, instrumental, etc. phrases, and/or clauses of adverbial function.

Traditionally all of these have been known as VP adverbials (Chomsky 1965:

102). They are circumstantial to the action or event which a verbal head

and its strictly subcategorized arguments designate. Their position,

whenever they optionally appear in a constituent, is immediately preceding

V 1 complements in the unmarked case.5

27(a) [Pro [ [kaleessa] [na-barbaada tur-t-ee?]]]

S V' t Adp'' Vt
yesterday me want be-2p-pf.
'Had you been looking for me yesterday?'

(b) [Tulluu-n [ [kawwee-dan] (nama a5ees-e]]]

S V'' Adp'' V'
T-nom. gun-with man kill-pf.
'Tulluu killed [a] man with a gun'.

Both /kaleessa/ 'yesterday' and / lSawwee -dan/ 'with a gun', do normally

occur outside V'. The position they occur at within V 1 is where the

strictly subcategorized arguments (NPs) are found. As discussed in the

preceding chapter, the verbal heads in both structures belong to the subclass

which selects only one NP argument as its complement. In the structures

above, the pronominal / (a) na/ 'me' in (a), and the nominal mama! 'man'


168

in (b) are such argument complements. Since the adverbials are not

arguments directly subcategorized by the verbs their position is always

outside V 1 where they occur as projections of V1.

As shown in the preceding section, the syntactic process of gapping

may also operate here on V 1 only leaving the adverbial intact. This is

what the ungrammatical (28 d) shows.

28(a) [Tulluu-n ( (kaleessa] [hoolaa bit-el]]

S V'' Adp'' V1
T-nom. yesterday sheep buy-pf.
'Tulluu bought (a] sheep yesterday'.

(b) (Fayyiisaa-n [ [har?a] [hoolaa bit-eli]

S V'1 Adp'' V1
F-nom. today sheep buy-pf.
'Fayyi lal sheep today'.

(c) [Thlluu-n [ [kaleessa] [hoolaa bit-el]] [Fayyiisaa-n

S V'1 Adp'' V' S
T-nom. yesterday sheep buy-pf. F-nom.
imixo [ [har?a] [ ]l]
V'' Adp'' today V'
also
'Thlluu bought a sheep yesterday. Fayyiisaa did so today'.

(d)* [Thlluu-n ( (kaleessal [hoolaa bit-elli (Fayyiisaa-n

S V'' Adp'' V' S
T-nom. yesterday sheep buy-pf. F-nom.
imno [ 1]
VII
also.

The ungrammaticality of (d) suggests that /kaleessa/'yesterday' cannot

be deleted along with V' 1 which means that the two do not belong to the

same level of hierarchy. In this connection, it might be possible to

treat /immoo/ 'also' as pro-V 1 since it appears oniy whenever the process

has deleted V', in which case, the ungrammaticality of (28 d) could also

be accounted for in terms of its(/imno/)standing in place of V'' rather

than of V1.

In addition to gapping and the resulting pro-form, there are also

some semantic cues which may enable us to identify V' 1 complements. Such
 169

cues include the scope of negative elements. V I ' complements may be

interpreted as being outside the scope of such elements, whereas V1

complements may not. Consider the following examples.

29ta) (Thlluu-n [nama a'ees-e]]

S V1
T-riom. man kill-pf.
'Tulluu killed [a] man'.

(b) [Tulluu-n [naina hin-aeef-n-e] 1 (1< hin-a'ees-n-e/)

ST_nom. 'i" man neg-kill-neg-pf.
'Tulluu did not kill a man'

In (29 b) the scope of the negative verb extends primarily to the

argument /nama/ 'man' which this verb strictly subcategorizes. If we

want to negate the external argument, it is necessary to leave the internal

argument empty given the context set in (29 b) . Hence (30):

30. [Tulluu-n hin-a3eef-n-e]

ST_nom. neg-kill-neg-pf,
'Tulluu did not killt.

In other words, for a constituent to be unambiguously affected by

negation, it must be adjacent to the form carrying the negative element.

If it is away from the latter there is always a possibility for it to be

interpreted as being unaffected by the negative element.

In the light of this, the adpositional phrase in (31) below may

primarily be interpreted as being outside the scope of negation since there

is the strictly subcategorized argument intervening between it and the

negative verb.

31 [Tulluu-n [ Leeboo-dan] [nama hin-a5'eef-n-e]]]

S V'' Adp'' . V'
T-nom. spear-with man neg-kill-neg-pf.
'Tulluu did not kill a man with a spear'.

For /eeboo-dan/ 'with [a] spear' to be properly affected by the

negative verb, it has to occur close to the verb. This may lead to a


170

situation where we either leave the argument mama! 'man' empty as in

(30), or raise and adjoin it to V'' so that the adpositional phrase could

occur immediately preceding the negative verb as in (32 b) below.

When the latter process takes place, there is always a change in the

intonation contour of the clause. There is a marked pause following

mama! 'man'.

32 (a) [Tulluu-n (eeboo-dan hin-aeef-n-e]]

S V'1
T-nom. spear-with neg-kill-neg-pf.
'Tulluu did not kill with a spear'.

(b) [Tulluu-n [nama - [eeboo-dan t hin-aeef-n-eJ 1]

S V't 1V'' 1
T-nom. man- spear-with neg-kill-neg-pf.
'Tulluu did not kill with a spear'. -

The fact that /eeboo-dan/'with spear' has to be close to the negative

element in order to be strongly affected, suggests that in the unmarked

case, that is, in structures of the type in (31), it is outside the scope

of the negative element whereas the strictly subcategorized argument mama!

'man' is not.

Since the interpretation of negative structures is very complex, it

may not be wise to depend too much on it. We shall therefore cite other

syntactic facts to substantiate further the claim we have made about

adverbials being of V' 1 complements. Earlier on, we have stated that

/immoo/ 'also' might be treated as pro-V' since itoccurs whenever V' is

deleted. In the same manner, we may argue that /-iss/ 'too' could be

treated asapro-V'', since it occurs whenever gapping operates on V''.

This is observable from the following examples:

33 (a) [Tulluu-n[ [eeboo-dan] [leena aees-e]]]

S V"Adp'' V'
T-nom. spear-with lion kill-pf.
'Tulluu killed a lion with a spear'.
 171

33 (b) [Fayyiisaa-n [ [eeboo-dan] [leena a5ees-eJ]]

S V'1 Adp'' V'
F-nom. spear-with lion kill-pf.
'Fayyiisaa killed a lion with a spear'.

(c) (Tulluu-n [ [eeboo-dan] [leeria aees-e]]]

S V'' Ad''
T-nom. spear-withVIlion kill-pf.
[Fayyiisaa-n-iss [ 1]
S V''
F-nom-too
'Thlluu killed a lion with a spear. Fayyiisaa too'.

In (33 c) it is /-iss/ 'too' which substitutes for V'1. The sentence

would be ungrammatical if we put /immoo/ 'also' instead of /-iss/ as in

(d) below.

(d)* Tulluu-n eeboo-dan leena aees-e. Fayyiisaa-n immoo

T-nom. spear-with lion kill-pf. F-nom. also.

From the structures thus far, it seems almost certain that V' and V'1

constitute distinct levels of hierarchies in the projection line of V 1 and

that their pro-forms are also different.

The examples we have considered so far involve temporal and instrumental

adverbials. Other types of adverbials which also occur at the same level

include those in the structures below:

34(a) [Tulluu-n [ [niitii-saa wain] [daadii hin-dug-a]]]

S V'' Adp'' . V'
T-nom. wife-his with mead cm-drink-impf.
'Tulluu drinks mead with his wife'.

(b) (Tulluu-n [ [kaaa-n] [gara Gimbii hin-deem-a] 11

S V''Ad'' V'
T-nom. running-with to G. cm-go-impf.
'Tulluu will go to Gimbii quickly Iwith speed]'.

(C) (Thlluu-n [ (hoolaa bit-uu-fi] (gara Gimbii hin-deem-a]]]

S V''Ad'' V'
T-nom. sheep guy-to-for to G. cm-go-impf.
'Thlluu will go to G.imbii to buy [a] sheep'.

In all such structures, as in those we have already considered, the

position of the adverbials is outside V' for the same reason stated earlier

on in connection with the temporal and instrumental adverbials. In (34 b)



172

and (c) the locative adverbial /gara Gimbii/ 'to Gimbii' is within V1

though it is an adverbial phrase like any one of those we have treated

as V'' complements. The reason for this is inherent to the lexical head

/deem-/ 'go'. As shown in the preceding chapter, this verb

subcategorizes a locative adverbial which means that all other adverbial

phrases can occur only outside V' 1 parallel to the comitative and the

manner adverbials in (34 a) and (b).

Other than simple phrases of adverbial functions, subordinate clauses

with or without overt subordinators may also occur as complements of V't.

Already we have an instance of this in (34 c) where the purposive infinitive

occurs in V''. Others include those in (35) below:

35(a) [Thlluu-n [ [daddaf-ee] [gara Gimbii hin-deem-a]]]

S V''S V'
T-nom. hurry-pf. to G. cm-go-impf.
'Tulluu will go to Gimbii quickly', Literally, 'Tulluu,
hurrying, will go to Gimbii'.

(b) [&altuu-n [ [daddaf-t-ee] [gara Giinbii hin-deem-t-i]]1

S V"S VI
C-nom. hurrieQ-f-pf. to G. cm-go-.f-impf.
'Caaltuu will go to Gimbii quickly'.

In (35) the elements in V' t are basically verbal. They display

agreement phenomena. In (b) for example, the verb has a feminine nxrpheme

/-t-/ in agreement with the feminine subject /aaltuu/.

The presence of to finite verbs in one clause suggests that that

clause is not a simple one. In the structures above /daddaf-t-ee/ may be

considered as being the head of a clause embedded in V'. Its subject is

a (pro)noun which is coreferential with the subject of the matrix clause.

This is deducible from AGR. As a subordinate clause arid as a complement

of V'' its function is that of a manner adverbial. It shows how the

action denoted by the matrix verb is effected. But unlike other subordinate
 173

clauses, it is not introduced by an overt subordinative element. The

element seems to be optional in such clauses, as also in others which show

time or reason. Compare the following, for example:

36(a) [Tulluu-n [ [deena-saa aat-ee] (gara konna-saa deem-el 11

S v•1'•S vu
T-nom. meal-his eat-pf. to farm-his go-pf.
'Thlluu went to his farm after he ate his meal'.

(b) [Tulluu-n ( [erga deena-saa aat-ee] [gara konna-saa deem-el]]

S V''S V1
T-nom. after meal-his eat-pf. to farm-his go-pf.
'Tulluu went to his farm after he ate his meal'.

37(a) [Tulluu-n ( [daadii bay?ee dug-ee] (hin-dukkubs-at-e]]]

S V''S V1
T-nom. mead much drink-pf. cm-sick-mid-pf.
'Tulluu got sick because he drank much mead'.

(b) [Tulluu-n [ (waayee daadii bay?ee dug-eel [hin-dukkubs-at-eJI]

ST_nom. V''
Sbecause mead much arink-pf. 'cmsick-midpf.
'Tulluu got sick because he drank much mead'.

Other parallel structures are also possible, and it seems that in the

structures in (a), the subordinator, which is the head of the clause, is

optional, or alternatively, if we assume that such clauses are base-generated

with a phonetically null head, then we may also have to say that such

clauses are in free variation with those which occur with overt subordinators.

These are, however, formal properties. Functionally, we can say that both

types occur as complements and that their position is as labelled in the

structures in (35).

If this line of argument is sound, then we shall maintain the claim

that subordinate clauses and pre,/post-positional phrases of adverbial

function may be treated as one category. Accordingly the phrase structure

rule that expands V' 1 should be along the following lines:

38 V'1 > ( Adp''*) V1.

 174

This rule does not show the linear ordering of the various

adpositional phrases/clauses. In general, this relationship may be left

free to be determined by the logical or chronological sequence of the

complements themselves in relation to the head V' 1 which they modify.

For example, whereas (39 a) is readily acceptable, (b) and (c) may be

acceptable only marginally.

39(a) (Tulluu-n [ [waayee argat-ee] [wayaa bit-uu-fi]

S V'' Adp'' Adp''
T-nom. because money get-pf. clothes buy-to-for
v.
[niitii-saa wa jj in] [gara Gimbii deem-el]]
Adp'' wife-his with Vt
to G. go-pf.
'Tulluu went to Gimbii with his wife tobuyclothes since
he got money'.

(b)? [Tulluu-n ( [niitii-Saa wa5in] [waayeehorii argat-ee]

S VI' Adp'' Adp''
T-nom. wife-his with because money get-pf.
[wayaa bit-uu-fil [gara Giinbii deem-el]]
Adp''' VI
clothes buy-to-for to G. fo-pf.

(c)? [Tulluu-n [ [wayaa bit-uu-fi] [waayee horii argat-ee]

S V'' Adp'' Adp''
T-nom. clothes buy-to-for because money get-pf.
[niitii-saa wain] [gara Gimbii deem-el]
Ad'' Vt go-pf.
wife-his with to G.

(b) and (c) sound odd, not so much for their structure as for their

logical relations.

In cases of structures in which there are no such problems of logic or

pragmatics, the order seems to be absolutely free as the following

structures show.

40(a) (Tulluu-n [ [kaleessa] (niitii-saa wa55in]

S V'' Adp'' Adp''
T-nom. yesterday wife-his with
[makiina-danl [gara Gimbii deem-el]
Ad'' V'
car-by to G. go-pf.
'Tulluu went to Gimbii with his wife yesterday by car'.
 175

40(b) [Tulluu-n [ [makiina-dan] [niitii-saa wa'inJ

ST V'' AdP''carby Adp''
wife-his with
[kaleessa] [gara Gimbii deem-el]]
Adp'' V1
yesterday to G. go-pf.

(c) (Thlluu-n [ [niitii-saa wa fl in ] [kaleessa]

S V'' Adp'' Adp''
T-nom. wife-his with yesterday
[makiina-dan] [gara Gimbii deem-e]]]
Ad'' V'
car-by to G. go-pf.

Taking the examples in (40) as the unmarked or the core cases as

against those in (39) which are marked, the base rule in (38) may be

taken as holding for all cases of V' 1 complements.

4.1.3 A'' Complements

As stated in Jackendoff (1977:64) adjectivals, like adpositionals,

which we will be considering shortly, 'are consistent with the Three-level

Hypothesis, although they do not push it to its limits as do Ss and NPs'.

Although this statement is made with reference to English, it may also be

taken as being valid for the situation in Oromo. At the level of A',

their complements include some adpositional phrases of purpose, as has been

observed in the preceding chapter.6

At the level of A'', the complements include adpositional phrases of

degree. The phrases are associated with the forms /hamma/ 'as much as',

/akka/ 'as/like', /aala/ 'be better/excel', or /manaa/ 'be worse'. This

is demonstrated by the following examples:

41(a) [Tulluu-n [ [hamma abbaa-saa] (sooressa]] miti]

S A'' Adp''
T-nom. as father-his A rich neg-is.
'Tulluu is not rich as his father'.

(b) (Tulluu-n ( [abbaa-saa-rra (kan) aal-e] [deeraa]] -da]

S A''Ad'' A'
T-nom. father-his-from comp. excell-pf. tall is
 176

(41 b) is difficult to translate. But the idea is as follows: 'Tulluu

is exceedingly taller than his father'.

In both structures the complements are adpositionals and their

position is immediately preceding A' complements as expected. This is

clear from structures like the one below where A' has an overt complement:

42(a) (Tulluu-n [ [akka abbaa-saa] [kaaa-n imaa]] -va]

S A'' Adp'' A'
T-riom. like father-his running-at strong is
'Tulluu is good at running like his father'.

(b) [Tulluu-ri [ [akka abbaa-saa} [wan beekaa]] -da]

S T-nom. A'' Adp'' . . A'
like father-his thing wise is
'Tulluu is wise [at] things like his father', (literally), 'Tulluu
is matter knowing like his father'.

These structures would be accompanied by a prolonged pause if the complements

of A' had occurred preceding those of A'' as in (43) below derived from

(42 a) above.

43. [Tulluu-n [kaaa-n], [ [akka abbaa-saa]

S Adr'' 1 A'' Adr''
T-nom. running-at like father-his
[ imaa]] - da]
A' strong is
(Literally), 'Tulluu is good like his father at running'.

The fact that /kaaa-n/ 'at running' in (43), unlike in (42), constitutes

a pause group on its own suggests that it is a displaced constituent, moved

from its base position in A'.

The complements here are not any different from those discussed in

connection with the previous two categories. The degree phrases may be

treated along the same lines as the various adpositional phrases of V''

and N''. 7 Hence the base rule (44) may be proposed, parallel to the one

in 38:

44. A'' > (Adp''*) A'.

 177

4.1.4 p ' 1 Complements

There does not seem to be any restrictive adpositional phrase or

clause occurring as a complement of P'' in a manner similar to the adverbial

phrases of V'', or the restrictive relative clauses of N''. Nor is it

the case that degree clauses, which we have considered with regard to A''

complements are recurrent in P''. The few instances of such complements

are of the following type:

45(a) [Pro [ [ [erga haad-i-saa duu-t-e]

S V' t Adp'' Adp'
after mother-nom-his die-f-pf.
[hamma bara lama.a-tti] ] I
Ad [gara Gimbii hin-deem-n-e]]
until year two up V1

(Literally), 'After his mother died, up until two years [laterl

he has not gone to Ginibli'.

(b) [ [ [ [erga gannaa-ti] (hainma har ?aa-tti]]]

S V' 1 Adp'' Adp' Adp''
after winter-from until today-up
bokaa-n [hin-roob-n-e]]
Vt
rain-nom. neg-rain-neg-pf.
'After winter, up until today, rain has not rained'.

In such structures as (45 a) /hamma bara lamaa-tti/ 'up until two years'

optionally modifies the head Adp', by showing the amount of time lapsed

since the death of his mother, without him going to Ginibil. In other

words, the phrase, which is itself adpositional, has the effect of restricting

the period of time expressed by Adp'. Without it, the duration between

the death of his mother and the time at which the statement was made would

remain indefinite. This is noticeable from (46) below:

46 [Pro [ [ [erga haad-i-ssa duu-t-ee]]

S V''Ad''Ad'
after mother-nom-his die-f-pf.
[gara Gimbii hin-deem-n-e]]]
V to G. neg-go-neg-pf.
'Since his mother died he has not gone to Ginibii'.
 178

From Adp t alone in (46) it is not possible to tell (in specific terms)

the length of time Thlluu has been away from Gimbii. This is possible

only with the temporal adpositional phrase /hamma bara lamnaa-tti/ 'up until

two years' included in Adp'' in the manner indicated in (45 a) . From the

restrictive effect this phrase has on the interpretation of the clause as a

whole, it may be possible to assume that its structural relationship is as

labelled. However, given the possibility for such complements to occur

in concatenations, it might as well be not impossible to argue that the

structural relationship of /hamma bara lamaa-tti/ 'uo until two years' is

with V I ' rather than with Adp'', in which case the labelling would have to

be as shown in (47) below:

47 [Pro [ [erga haad-i-saa duu-t-ee]

S V'' Adp''
after mother-nom-his-die-f-pf.
thamma bara lamaa-tti] [gara Gimbii hin-deeni-n-e] 1]
Ad'' V'
until year two-up to G. neg-go-neg-pf.
'Since his nther died, up until two years, [he] has not
gone to Gimbii'.

Though this is quite sound at the level of assumption, all the evidence

seems to prove that (47) is not the right analysis. Observe the following,

for example:

48 (a) [Pro [ (erga haad-i-saa duu-t-ee hamma bara lainaa-ttil

S b'' after nther-nom-his-die-f-pf. until year two-up
V" Ad-a

[ugumaa-n [gara Gimbii hin-deem-n-e]]]

Ad'' V1
truth-in to G. neg-go-neg-pf.
'Truly (he] has not gone to Gimbii for two years since/after
his nether died'.

(b)? (Pro [ [erga haad-i-saa duu-t-ee] (dugumaa-n]

Apd''
S V'' Adp after ixther-nom-his-die-f-pf. truth-in
[hainma bara lamaa-ttil [gara Gimbii hin-deem-n-e]]]
Adp'' Vt
until year two-up to G. neg-go-neg-pf.
 17

In (48) /dugumaa-n/ 'truly', which is a sentence adverbial, occurs

between the complements of V' 1 and V' 1 that is, between the two major

nodes or levels, and the structure is grammatical. In (b) , however, this

same phrase occurs between /erga haad-i-saa duu-t-ee/ 'after his mother

died' and /hamma bara lamaa-tti/ 'up until two years', and the result is

that the structure turns out to be dubious. The reason for this can

only be linked with the non-interruptable nature of strings of words

forming syntactic units (cf. Radford 1981:69). In other words, /erga

haai-saa duu-t-e hainma bara lamaa-tti/ in (48 a) is an uninterruptable

syntactic unit.

This argument may also be further substantiated by facts related to

movement. As stated in Chapter One, and as will further be shown in

Chapter Five, structures which form a single constituent move as a unit;

no part of the constituent can be left behind without the resulting

structure being ill-formed. The situation in (49) below is an instance

of this. The structure in (a) is grammatical with the movement of Adp''

as labelled in (45 a) , where (49 b) is only marginally acceptable although

it has the same kind of movement, but operating only on the second Adp'',

that is, on /harnma bara lamaa-tti/ 'up until two years' of (45 a)

49(a) [ [ [erga haad-i-saa duu-t-e]

SAdp'' Adp'
after mother-nom-his-die-f-pf.
[hainina bara lamaa-ittij [Tulluu-n [t
Adp'' 1 S V''1
until year two-up
[gara Gimbii hin-deem-n-e]] JI
Vt0 G. neg-go-neg-pf.
'After his mother died for two years, Tulluu has not gone
to Gimbii'.
 180

49 (b)? j [hamma bara laznaa-tti] [Tulluu-n

S Adp' ST_nom.
' until year two-up
[erga haad-i-saa duu-t-ee] I
V'' Adp'' Adp'
after mother-nom-his-die-f-pf.
[gara Gimbii hin-deem-n-e]]]]
Vito G. neg-go-neg-pf.
(Literally), 'Up until two years, Thlluu after his mother
died, has not gone to Giinbii'.

The marginal status of (49 b) can be explained only if we assume

/hamma bara lamaa-tti/ 'up until two years' to be a part of /erga haad-i-saa

duu-t-ee/ 'after his mother died', for if it were an independent temporal

phrase, its preposing in (b), which is a natural process for any adverbial

phrase to undergo, would not have led to this situation. But this is not

the case here, which suggests that /hainma bara lamaa-tti/ 'up until two

years' and /erga haad-i-saa duu-t-ee/ 'after his mother died' are two chunks

of the same constituent.

There is also another piece of evidence in support of this argument,

though it is a semantic one. Both /erga haa-i-saa duu-t-ee/ and /haimna

bara lamaa.-tti/ are temporal phrases and their function is to modify the

action designated by the head of V'. As there is only one action in the

clause, there can only be one such modifying expression projecting from

the verb, which expresses the action. Therefore, unless there are two

actions as in structures of co-ordination, for example, it is not plausible

to assume that two temporal expressions project from a single verb

designating a single action under normal circumstances.8

Although we have argued that the two temporal phrases are parts of a

larger constituent, we have not yet said explicitly which of the two forms

constitutes the head, though this is something which has been indicated in

the labelling. Again from such constituency devices as omissibility of

 181

complements (Radford 1981:69), though not of heads, and from the

consideration of substitutability of an entire constituent by its head (a

fact of endocentric constructions), though not by its complement, 9 it can

be proved that /erga haad-i-saa duu-t-ee/ 'after his mother died' is the

head and that /hamma bara lamaa-tti/ 'up to two years' is the complement

of Adp'' in (45 a) . Let us consider the following:

50(a) [Tulluu-n [ [erga haad-i-saa duu-t-ee]]

S Adp'' Adp'
T-nom. after mother-nom-his-die-f-pf.
gara Gimbii hin-deem-n-e]
'Tulluu has not gone to Gimbii after his mother died'.

(b)? (Tulluu-n (hamni bara lamaa-tti] gara Gimbii hin-deem-n-e}

S Ad''
T-nom. until year two-up to G. neg-go-neg-pf.
(Literally), 'Tulluu has not gone to Gimbii up until two years'.

It is also evident from the observation of the internal structure of

the phrases and the facts following from the argument about their head!

complement relations, that such structures are left-headed like non-infinitival

nominals, and unlike verbals, which as we have seen are right-headed.

Whether this relation is consistent and natural is a thing one cannot be

definite about now, since there are also structures of the same category

which are right-headed. Structures of purposive and intrsuinental adverbials,

or example, belong to this group, unless we argue that in such structures

what appear to be postpositional clitics are actually case affixes, and may

hence be relegated to noun morphology. 10 Even if we accepted this to be

valid, we would still have problems with adpositional phrases having

independent postpositions as heads, as in /niitii-saa waj'in/ 'with his

wife'.

It seems then that although there is a possibility for a language to

have both case affixes and pre 7'post-positional elements, 1 ' the position of
 182

such elements in relation to the complements which they strictly subcategorize

and govern has to be determined in view of the overall effect this

relationship may have on the grammar. Following Greenberg's (1963:76)

observation and generalization about SOy languages, and the theoretical

framework followed here, it might be more generalizing if not very economical

to assume right-headedness for adpositionals in general.

Whereas the structures cited so far as examples of Adp'' complements

are restrictive, the complements of some adpositional phrases appear to be

non-restrictive. Consider (51) below for example:

51 (Tulluu-n [ [gara Gimbii [gara biyyaa bunaa] I deem-el

S Adp''' Adp' Adp''
T-nom. to G. to land of-coffee go-pf.
'Tulluu went to Gimbii to [the) land of coffee'.

In such structures, the complement may be treated as having an

appositive rather than restrictive role. In fact the complement seems to

be a comment on the complement of Adp', that is, on /Gimbii/, rather than

on the head 12 (Adp') itself. This is deducible from the fact that the

complement /gara biyyaa bunaa/ 'to the land of coffee' is headed by the

same element /gara/ 'to 1 which is also the head of the head /gara Gimbii/

'to Gimbii'. This may lead us to one possible conclusion: that structures

like /gara biyyaa bunaa/ 'to the land of coffee' are Adp''' rather than

Adp'' complements. But, unlike the appositives of X''' complements,

these can be affected by negation, or clefted as we can observe from

(52 a) and (b) respectively.

52(a) [Thlluu-n ( [gara Gimbii] [gara biyyaa bunaa]l

S Adp'' Adp' Adp''
T-nom. to G. to land of-coffee
hin-deem-n-e-]
neg-go-neg-pf.
'Tulluu did not go to Gimbii to [the] land of coffee'.
 183

(b) [Tulluu-n [ [gara biyyaa bunaa] [gara Gimbii-til]13

S Adp'' Adp'' Adp'
T-noin. to land of-coffee to G. is
kan deem-el
comp.go-pf.
'It is to Gimbii to the land of coffee that Tulluu went'.

The ability to be affected by negation or to be clefted is the

characteristic feature of complements. The reason why the complements

in (51) appear to be appositive rather than restrictive seems to be related

to the fact that /Gimbii/ is a proper noun, which means that anything that

follows or precedes such a noun has no defining effect on it as the thing

is inherently defined.

In structures of other adpositional phrases with common noun complements

in Adp t , the complement of Adp'' is always restrLctive, thus supporting the

claim made above. This is observable from the following:

53 (a) [Tul].uu-n [ [eeboo-dan], [abbaa-n-saa (kan) keenn-ee-n] I

S Adp'' Adp' Adp''
T-nom. spear -with father-nom-his give-pf-with
[leenca aees-e] I
V lion kill-pf.
'Tulluu killed a lion with the spear with that which his
father gave him'

(b) [Tulluu-n [ [fard-ia-n] , (kaleessa (kan) bit-ee-n]]

Adp'' AdP' horse _ slg _on AdP''yesterday comp.buy-pf-on.
5Tnom
gara Gijnbii deem-eJ
to G. go-pf.
'Tulluu went to Gimbii on the horse on that which he bought
yesterday'.

From their forms and their glosses, these complements appear to be

relative clauses, and as such non-restrictive ones, because of the prolonged

pause following the heads /eeboo-an/ 'with spear' and /fard-ia-n/ 'on

the horse [back]' respectively. If this is the case, then they have to

be treated as Adp''' rather than Adp'' complements , for as we shall see



184

shortly, prolonged pauses are characteristic of X''' complements. However,

the complements in question here appear as relative clauses only

superficially, because relative clauses occur only as N'' complements as

discussed earlier. What appear here as relative clauses are adpositional

heads, that is Adp', with a strictly subcategorized nominal complement N'''

which itself is a head to a relative clause complement. This relationship

is shown in the tree below:

54 (=53)

Adp''

A
Adp' Adp''

eeboo (da)n N''' Adp

spear with
J
N''

T'
Pro Comp S

kan Vt,,

abbaa-n-sa V'
father-nom-his

N''' V

I
Pro keenn-
1 give
'eeboo-dan abbaa-n-sa kart keenn-en'
'with the spear with [that] which his father gave him'
 185

In (54) it is Adp'' which is the complement of Adp' (eeboo-dan).

The relative clause is the complement of N'' in Adp' which is directly

dominated by the initial Adp''. The two Pros in N'' get their reference

from /eeboo/ 'spear' with which they are in c-command relationships in the

sense of Chomsky (1981:166) and are hence co-indexed with this noun.

If we are on the right track, then the rule that expands Adp'' must

be along the following lines:

55 Adp'' > Adp' (Adp'').

Notice that according to this rule, the complement comes after the

head. This is what we have observed in all the structures considered

earlier on, though in the chapter preceding this what we argued was that

Oromo is postpositional, which means that the head is final. And this

is still the case if we take Adp' /eeboo-dan/ 'with [a] spear' in (54)

above, for example. What this leads us to conclude is that adpositionals

are right-headed at their minimal, and left-headed at their maximal levels

of projections, and accordingly two base rules should be formulated, one

for each level. This is very uneconomical.

The alternative is either to maintain the argument that the language

is postpositional and account for such cases as (54) by a movement rule,

or to reconsider Oromo as a prepositional language, in which case no movement

is needed at the level of Adp''. The rule will apply only on Adp'. In

terms of cost this option appears economical because there will only be one

movement applying at the minimal level whereas if we adopt the other option

there will be two - one at each level, since there are also structures like

/gara Gimbii/ 'to Gimbii' which are prepositional, and which need to be

accounted for as deriving from a postpositional base. Furthermore, the

antecedent-anaphoric relation in (54) seems to favour the prepositional

 186

analysis. The (pro) nominals are co-indexed with a nominal /eeboo/ 'spear'

which is in an antecedent position. This relationship will not be

maintained if the postpositional option is adopted. For these reasons

the rule in (55) may be adopted as a first approximation.

4.2 X''' Complements

In the preceding sections of this chapter, we have considered all

structures which have a restrictive effect on their heads as X'' complements.

In the rest of the chapter, we shall argue that other structures which have

no such effects on their heads are base generable as complements of X'''.

It will also be shown that only two of the major categories, namely, verbals

and nominals, are characterized by such complements.

As previously, we shall take up each category in turn and examine the

characteristics of its complements at this level of projection.

4.2.1 N''' Complements

At the level of N''', the complements associated with simple nominals

include non-restrictive relative clauses and some noun phrases. Consider

the following examples:

56(a) [ ( [nam-i!-i !kaleessa(kan) duf-el],

SN''' N'' S
man-sgl-nom. yesterday comp. come-pf.
(Tulluu-n kan arrabs-e]] [deeraa-da]]
ST_nom. comp. insult-pf. tall is
'The man who came yesterday, whom Tulluu insulted is tall'.

(b) ( [ (mu-i-i !as(i) (kan) tur-ek),

S N''' N'' S
child-sgl-nom. here comp. be-pf
[Gimbii-ti kaleessa kan duf-e]] (bareedaa-da]]
from yesterday comp.come-pf. handsome is
'The child who was here, who- came from Gimbii yesterday, is handsome'.
 187

In (56) the first clause (5) following N'' is a restrictive relative

clause, as already discussed (in 4.1.1). This clause restricts the

reference of the head /nam-i!-i/ 'the man-nom' in (a) to 'the one who

came yesterday'. The clause outside N'' is also a relative clause, but

its role is not that of defining, since the noun has already been defined

by the first clause. Hence, its function is that of providing extra

information about the head.

There is no formal difference between the two relative clauses in

such structures, other than the intonation break. 14 The restrictive

relative clause forms a single pause group with its head /nam-i'i/ 'the

man-nom', whereas the non-restrictive one does not. There is always a

prolonged pause separating it from the rest of the clause.

One other device for distinguishing a non-restrictive relative clause

from a restrictive one is that the pronominal subject of the former always

has a phonetic matrix. The example in (57) is an instance of this.

57 [ ( [nam-j-j], !inn-i Tulluu arrabs-e]], [deeraa-da]]

SN''' N'' S
man-sgl-nom. he-nom. T. insult-pf. tall is
(Literally), 'The man, he insulted Tulluu, is tall',
'The man who insulted Thlluu is tall'.

The pronominal /inn-i/ 'he-nom' is the subject of the clause and is

coreferential with the head. The intonation break is still apparent. This

is indicated by the comma.

The pronominal subject may be dropped in such structures (as in (57))

if the head of the relative clause is a proper noun, as in (58), for example.

58. [ [ (obboo Thlluu nil, kaleessa Gimbii-ti (kan) duf-e]]]

Ul
S N''
Mr. T-nom. yesterday G.from comp.come-pf.
(nama gaarii-da]]
man good is
'Mr. Tulluu, who came yesterday from Gimbii is a good man'.
 188

Appositives, unlike restrictives, can occur preceding their heads.

Whenever this is the case, they form one pause-group with the head (and

are not separated by a connna in the transcription). Let us observe the

following:

59(a) (i) I I [obboo Tulluu-n], [kaleessa Giinbii-ti (kan) duf-an-ij11'

S N''' N'' S
Mr. T-nom. yesterday G. from comp. come-pol-pf.
(du?-an-i]]
die-pl-pf.
'Mr. Tulluu, who came from Gimbii yesterday, died'.

(ii) I I [kaleessa Giinbii-ti kan duf-an-i]

S N''' S
yesterday G-from comp. come-pol-pf.
[obboo Tulluu-n]] [du-an-i]]
N''
Mr. T-nom. die-pl-pf.
'Mr. Tulluu who came from Gimbli yesterday died'.

(b) (i) I [ [ fard-i-i] !Tulluu-n (kan) bit-el] [du?-e]]

SN''N' S
horse-sgl-nom. T-nom. comp. buy-pf. die-pf.
'The horse which Tulluu bought died'.

(ii) ?[ [ [Tulluu-n kan bit-e] [fard-i-i]J [du?-e]]

S N'' S N'
T-nom. comp. buy-pf. horse-sgl-nom. die-pf.
'The horse which Tulluu bought died'.

(59(a ii) derived from (59(a 1) is possible, whereas (59(b ii) derived

from the corresponding (59(b' 1) is only marginally acceptable.

Such differences between restrictives and appositives suggest that

the two clauses need to be treated differently. !n Section 4.1.1.1, we

have treated restrictive relative clauses as complements of N''. What we

are left with here is the non-restrictive (appositive) relative clause

which may be associated with the next higher level, in which case a rule

of the type below is necessary:

60. N'''_- > N'' (S)

 189

Rules like (60) show general cases. There are, however, instances of

structures of NPs with a proper noun head followed by another NP complement.

The structures in (61) are examples of such cases.

61(a) [ [ (óbboo Tulluu-n], [abbaa-n Tolasaa]], [as(i) tur-an-i]]

S N''' N'' N'''
Mr. T-nom. father-nom. of-T. here be-pol-pf.
'Mr. Tulluu, father of Tolasaa, was here'.

(b) ( ( (obboo Tulluu-nl, [abbaa-n farda gurraaaa]],

S N''' N'' N'''
Mr. T-nom. father-nom. horse black
[as(i) tur-an-i]]
here be-pol-pf.
'Mr Tulluu, owner of a black horse, was here'.

In such structures, the complement NP is coreferential with the head /obboo

Tulluu-n/. Functionally, it seems to be doing exactly what an appositive

relative clause could have done. It does not modify the head because the latter

is a proper noun and hence does not need any restrictive modifier.

In some cases, instead of simple NPs, NP5 with a relative clause complement

ay also occur as complements. (62) is such an example.

62 [ [ (obboo Tulluu-n], [ [ (nam-i-iJ

S N''' N'' I
Mr. T-nom. man-sgl-nom.
! ( kaleessa as(i) tur-ari-i]]]]] (du?-an-i]]
SSyesterday here be-pol-pf. die-pol-pf.
'Mr. Tulluu, the man who was here yesterday, died'.

One characteristic feature which distinguishes such complement NPs from

:omplements of the type seen in (61) is their ability to undergo extraposition.

ence corresponding to (62) above 1 (63) below is possible.

63 f[ [ ( obboo Tulluu-n] [t]] [du?_an_i]] I [ [ nam-i-iJ

SS N 1 N N 1 VP . Nt N'N
Mr. T-nom. die-pol-pf. man-sgl-nom.
([kaleessa as(i) tur-an-i]J]]]
SSyesterday here be-pol-pf. 1
'Mr. Tulluu died, the man who was here yesterday'.


190

In the light of the examples in (61-62), the rule in (60) needs to be

modified along the following lines:

64 N''' >N'' (tSN'''J)

4.2.2 v'' 1 Complements

In the section on V' 1 complements, we have argued that their complements

include VP adverbials. In this subsection we shall argue that sentence

adverbials may be considered as V'" complements, parallel to the appositive

clauses in nominals.

The following are structures with such adverbials:

65(a) [Tulluu-n [ [duguma-anl ( [bor(u)]

STnom v''' Adp' truth-in V'' Adp'
'tomorrow

[gara Giinbii hin-deem-a]1]]

v to G. cm-go-impf.
'I will certainly go to Giinbii tomorrow'.

(b) [Thlluu [ [hookan] [ [bor(u)]

S V''1 Adp'' V't Adp''
T-nom. perhaps tomorrow
(gara Gimbii hin-deem-a]]]]
v' -
'Tulluu will perhaps go to Giinbii tomorrow'.

One of the characteristics of sentence adverbials is their ability to move

freely across the sentence. This is observable from (66) below, which is

derived from (65 a) where the adverbial is base-generated in situ.

66(a) [duguma-an] (Tullluu-n [[t] [ [bor(u)]

Ad" 1 S V''t 1 V' 1 Ad"''
truth-in tomorrow
[gara Gimbii hin-deem-a]]]]
V to G. cm-go-impf.
'Certainly Tulluu will go to Giinbii tomorrow'.

(b) [Thlluu-n [ Et] ( [bor(u)]] [duguma-an]

S V''t 1 v'' Adp
T-nom. tomorrow Adp'' truth-in
[gaza Gimbii hin-deem-a]]]]
Vito G.. .cm-go-impf.
'Tullu will certainly go to Gimbii tomorrow'.
 191

66(c)? [Thlluu-n [ [ti [ [bor(u)]] [gara Gimbii hin-deem-a]]]

S V''1 1 VII Adp'' V1
T-nom. tomorrow to G. cm-go-impf.
[duguma-an]
Adp''' 1
truth-in
'Thlluu will go to Giinbii tomorrow certainly'.

When a V'' 1 complement is preposed in the manner suggested by (66 a) , it is

not separated from the rest of the clause by a pause. This distinguishes it from

V I ' complements, because the latter are always followed by a prolonged pause.

Consider (67), for example.

67. [booru], [Thlluu-n [ [t] [gara Gimbii hin-deem-a]]]

Adp'' 1 S V'' lv'
tomorrow T-nom. to G. cm-go-impf.
'Tomorrow, Tulluu will go to Gimbii'.

Going back to (66), it appears that the sentence adverbials may not be

postposed without the resulting structure being dubious. There are, however,

other structures where such is not the case. Consider (68) below in relation

to (66 b) above.

68(a) hookan [Tulluu-n [ [t] [bor (u) [gara Gimbii hin-deem-a] JI]
1 s V''1 1 V'' VI
perhaps T-nom. tomorrow to G. cm-go-impf.
'Perhaps, Tulluu will go to Gimbii tomorrow'.

(b) [Tulluu-n [ [bor(u)hookan [gara Gimbii hin-deem--a]]]]

S V1'1 V'' V'
T-nom. yesterday perhaps to G. cm-go-impf.

Cc) [Tulluu-n [ t (bor(u) [gara Gimbii hin-deem-a]]1 hookan]

S V''1 1 V' 1 VI 1
T-nom. yesterday to G. cni-go-impf. perhaps

A further characteristic of V 1 '' complements is that they are not affected

by negation. Let us observe the following again.

69(a) (Tulluu-n [dugu.ma-an] bar (u) gara Gimbii hin-deem-u]

ST_nom. truth-in tomorrow to G. neg-go-impf.
'Trulluu will truly not go to Gimbii tomorrow'.

(b) (Tulluu-n [duguma-an] nama gaarii miti]

8Tnom. truth-in man good not-is
'Tulluu is truly not a good man'.
 192

In structures such as these, only the V 1 and/or V' 1 complements may be

aid to be within the scope of the negative element. For this reason, questions

ike /egaa eessa hin-deem-a/ 'where will he go then?'or (with the temporal

hrase /bor(u)/ 'tomorrow' as its response), /egaa yoommuu hin-deem-a/ 'when will

,e go then?' are likely to be raised. No questions of these types will arise

.n relation to /duguma-an/ 'truly'.

Finally, V''' complements, again unlike those of V'' or V' cannot be clefted.

:ence (a) but not (b) may be derived from the structure in (70).

70. [Tulluu-n [hookan [bor(u) [hoolaa hin-bit-a]]]

S V''' VI' tomorrow V'
T-nom. perhaps sheep cm-buy-impf.
'Tulluu will perhaps buy a sheep tomorrow'.
(>0)16
(a) [bor(u)-da [Tulluu-n hookan hoolaa kan bit-u]]
Stomorrow is ST_nom. perhaps sheep comp-buy-impf.
'It is tomorrow that Thlluu will perhaps buy [a] sheep'.

(b)* hookan-da Tulluu-n boru hoolaa kan bitru

perhaps-is T-nom. tomorrow sheep comp. buy-impf.

The complements we have considered so far are adpositional phrases. There

re also clauses which may be argued to be V''1 complements. (71) below is an

xample with such a clause.

71. [Tulluu-n [ [akka na-tti fakkaat-u] (nama gaarii miti] 1]

S V''' S V'
T-nom. as me-to seem-impf. man good neg-is.
'As it seems to me,Tulluu is not a good man'.

Syntactically, such modifying clauses behave in much the same way as other

ppositive complements do. They cannot be clefted or affected by negation.

hey can also undergo postposing. The structures in (72) are demonstrative

xamples.

72(a)* ([akka na-tti fakkaatu-da] [Thlluu-n t nama gaarii miti]]

S as me-to seem-impf-is ST_nom. 1 man good neg-is
 193

72 (b) (Thlluu-n (t [nama gaarii , miti] (akka na-tti fakkaat-u]]]

S 1V'' 1
T-nom. man good neg-is as me-to seem-impf.
'Tulluu is not a good man as it seems to me'.

From the differences we have noticed between V' t and V'' 1 complements, it

;eems appropriate to suggest a base rule of the type (73) for V'' 1 complements.

73 V'' 1 > ( Adp'*')V''

L3 Summary

From the discussions we have had so far, some general statements may be

iade about the complement structures of the major categories at the intermediate

md maximal levels of their projections. The complements associated with each

.evel are restrictive and non-restrictive modifiers respectively. For nominals,

these include adjectives, restrictive relative clauses and some genitive NPs of

mdverbial function at the level of N'', and non-restrictive relative clauses

mnd appositive NPs at the level of N'''.

Verbals may have adpositional phrases or clauses of time, instrument,

ianner, etc. at the level of V' 1 . These may be generally called VP adverbials

is opposed to the complements of V'' 1 which are sentence adverbials.

Regarding the complenents of adjectivals and adpositionals, it may be said

:hat their too include adpositional phrases/clauses of degree at the level of

m'' and Adp''. Unlike the other two categories, they also terminate their

)rojection line at the double bar level. This is contrary to Jackendoff's

niform Three-level Hypothesis which we mentioned at the beginning of the chapter.

he hypothesis predicts that all categories have a maximal treble bar expansion.

However, the hypothesis seems to predict the potential capacity for, rather

than the actual occurrence of a maximal projection. Hence the fact that

mominals and verbals expand to the level of three-bars as predicted means that

the hypothesis holds strongly for these two categories, and less strongly for
 194

he other two. It is with this in mind, it appears, that Jackendoff

(1977:64) says, '...there is strong evidence for all three levels in S(=V''')

md NP and for at least two levels in AP and PP'. Although this is said with

reference to English, it may also be extended to apply to the situation we have

Dbserved in Oromo as well.

The base rules we have been formulating throughout the discussion may be

Dresented here again for purposes of exposition and subsequent generalization.

N''' > N'' ([S, N'''J)

N'' > N' (EN''', A'', Si)

V'' 1 > ( Adp'' * ) VI'

V' t > ( Adp'') v'

A'' > (Adp'') A'

Adp'' -> Adp' (Apd'')

t the level X'' the following cross-category generalizations may be made:

[+Vj '' > Adp'' [+V]'

With regard to the notion of headedness, the four categories seem to fall

Lnto two: those which are right-headed and those which are not. [+V] categories

)elong to the former, whereas [-VI categories belong to the latter.

 195

NOTES '10 CHAPTER FOUR

1. This term refers primarily to NP arguments, though some PPs may be

included in the list of strictly subcategorized complements.

2. There is no genitive morpheme. The relation is indicated by the two

NP5 occurring in a possessed-possessor order. Phonologically the final
vowel of the possessor NP becomes long. Hence,
mama! 'man', but [mana namaa] 'a man's house'.
This lengthening could be considered as a genitive morpheme.

3. Such genitive NPs of the type we have considered thus far may be thought
of as being reduced relative clauses. For example, /amartii 'wo4ii/
'ring of gold' in (15) may be assumed to have derived from a structure like
the one below (details have been avoided).
1) [ [amartii [[warkii-rra kan ho5-et-axn-e]]]---
N''' N'' SS
ring gold-from comp. work-inid-ps-pf.
'A ring which is made of gold'.

If this is the case, then /war1ii/ 'of-gold' or /dammaa/ 'of-honey' in

(15) and (16) respectively, which have been treated as N' complements,
will have to be reconsidered as N'' complements since the relative clauses,
from which they have been assumed to have derived arecomplements of N'',
but not of N'. And in accordance with this analysis, the base rule in
(12), which expands N'' as N' (N''') (A'') (S), would have to be modified
in the manner of (ii):
ii) N'' > N' (A'')(S)
Whereas such an analysis works well for some genitives such as /workii/
'of gold', it does not seem to hold for others, because in structures such
as (iii) below, there is no appropriate relative clause to which the
genitive NP could reasonably be traced back.

iii)(a) [ [ (aynatii (watii]]] nan-beek-al
S N'' N' N'
type of-stew I-cm know-impf.
'I know (about] varieties of stew'.

(b) ( [ (karaa (niiilaa]] (fagoo]] Tulluu-n deem-e hin-beek-u]

S N' N' A''
road of-leg far T-nom. go-pf. neg-know-impf.
(Literally), 'Thlluu has never gone a long walk of leg',
'Tulluu has never gone on foot'
Furthermore, if we maintain this type of analysis for all genitive structures,
some of them will have to be accounted for a deiiving from mere than one
relative clause source. Such structures include the following:
iv) [ [mana [margaall
N'' N' N'''
house of-grass
'Grass house'.
 196

If we treat (iv) as a reduced relative clause, its source will have to be

either (a) or (b) below.
a) ( (mana] [[marga-rra (kan) i5'aar-am-e]]]
N''' N' SS
house grass-from comp.build-ps-pf.
'[A] house built from thatched grass'.

b) [ [mana] ([marga tee-siis-uu--f kan ta?-u]]]

N''' N' SS
house grass put-as-to-for comp.become-to
'A house for storing hay'.
The source clause in (a) shows the material from which /mana/ 'house' is
built, whereas the one in (b) indicates the purpose the house is built for.
It is not clear from the structure (iv) which of these two clauses is
underlying /mana margaa/ 'grass house'. Even if we knew from pragmatic
considerations that (a) might be a more likely source for the derivation,
we would still have to face the complex problems that would arise from
the derivation itself. Quite apart from this, we should certainly have
to recognize three types of genitive NPs, viz.:
1. Those which have no relative clause source.
2. Those which can be derived from a single source.
3. Those which may have more than one source.
Instead of deriving them from clausal sources of the type shown about, and
thereby facing the problems arising either from the process itself or from
the identification of the clause type from which a particular genitive NP
derives, we can take the other alternative, which generates them in situ,
that is, as complements of N' or N'' and account for the structural
ambiguities encountered in the manner shown in (14) and (17). This is
guite possible since one of the strong claims of X-bar syntax is its ability
to capture such subtle ambiguities (cf. Hornstein and Lightfoot 1981).

. Time adverbials such as /bor(u)/ 'tomorrow' are considered here as reduced

or headless adpositional phrases. There are cases when they occur with a
head, since we have structures such as:
guyyaa-n har ?aa bokkaa-n hin-roob-a
day-on today rain-nom. cm-rain-ixnpf.
(Literally), 'On day of-today rain will rain'.

5. The marked case is one in which they or X' complements are preposed as in
(1 b) derived from (1 a)
1(a) Tulluu-n eeboo-dan bineensa afl'ees-e
T-nom. spear-with animal kill-pf.
'Tulluu killed an animal with a spear'.

(b) eeboo-dan Tulluu-n tbineensa a5ees-e

1 1
spear-with T-nom. animal kill-pf.
'With a spear, Tulluu killed an animal'.
 197

6. Adjectives which are derived from transitive verbs may subcategorize NPs
as their complements, as in (42 b) . Such structures seem to suggest
that derived adjectives have 8 and case-assigning properties.

7. This refers mainly to complements of infinitival heads.

8. This is to exclude cases of structures like /haimna kaleessa/ 'now

yesterday' which is possible if one wants to emphasise that 'yesterday'
does not refer to some day a long time ago.

9. The complement may substitute the whole constituent with some loss or
change of meaning.

10. If this argument is seriously followed it will have a weakening effect

on the theory of case assignment which is based on government in the
sense of Chomsky (1981; 1982). Every such noun must be assigned case by
the verb irrespective of whether it is adjacent to the case assigner or not.

11. In fact Oron is one such language having both case affixes and
postpositional clitics.

12. The reason for this seems to be related to the fact that /gara/ 'to'
is lexically empty.

13. It is the entire Adp'', and not just the complement, which appears to
be clefted as we can observe from the position of the copula I-ti!.
Under normal circumstances we would expect it to occur following Igara
biyyaa bunaa/, as in,
? gara biyaa bunaa-ti gara Gixnbii kan deem-e
to land of-coffee-is to C. comp. go-pf.
What makes this kind of clefting exceptional is that both the complement,
which undergoes the process, and the head belong to the same category of
adpositionals.

14. This is equivalent to what Jackendoff (1977:63) calls 'comma intonation'.

15. The plural is used as a polite form of address.

16. As discussed in Chapter One, the copula is reducible to 0 following a

stem-final short vowel or consonant.
 198

CHAPTER FIVE

CLAUSAL COMPLEMENTS

5.0 Introduction

In the preceding two chapters, we have discussed two types of

complements. In Chapter Three, we observed the complements of lexical

heads. Such complements are usually obligatory. 1 It has been argued that

these should be included in the lexicon as part of the lexical specifications

of the heads by which they are subcategorized. The second type of complements

is that which we have discussed throughout Chapter Four as projections of X''

or Xw. These are peripheral to the main predicate structure of the

lexical head (X) and its strictly subcategorized functional argument(s).

At the levels of X' and X'', there are complements which are clausal.

This is true of the [-NI categories of verbals and adpositionals. Both

subcategorize sentential complements as their lexical specifications at their

minimal levels of projection. These complements are functional arguments

and occur in thematic positions. At the intermediate and maximal levels of

their projections their complements may include subordinate clauses of various

adverbial functions. These are modifying complements and are outside

the strictly subcategorized argument complements of the head.

With regard to the (+N] categories of nominals and adjectivals, it has

been argued that they too have sentential complements at the level of X''.

At their minimal level they are characterized by genitive NPs of some type,

or by adpositional phrases. What is noticeable here is that all the

categories have clausal complements at one level or another in their

projection line, and that these complements are functionally distinct.

Throughout the preceding chapters we have simply treated all clauses

as complements of either a lexical or a phrasal category, without making any

 199

reference to their internal structures. This chapter is an attempt

towards this end. Here we shall attempt to examine the internal structure

of the argument complements of the [-N], and the modifying complements of

the [+N] categories. It is believed that the analysis of these two types

of clause will help us in determining the position of the complementizer and

the nature of the Infi node, on the basis of which a more comprehensive

generalization about the notion of head in this language will also be made.

It may also enable us to see the type of movement the language allows, the

conditions constraining its application, and the principles governing the

relation between moved categories and their traces.

Tangential as it is, the discussion will be limited just to the two

clause types mentioned. Other types of clauses, such as those with

adverbial functions, will not be dealt with directly for the main reasons

(1) that such clauses are not arguments nor a part of such complements, and

are as such outside main predicate structures; and (2) that their analysis

may not be substantially different from the analysis we shall be proposing

for the argument complements.

The two types of clauses we shall be dealing with may be called nominal

and adnominal clauses. The latter deals with relative clauses, which have

been treated in the preceding chapter as N'' or N''' complements. The main

objective here will be to raise some points about traditional accounts of

such clauses and to propose an alternative analysis.

5.1 Clausal Arguments

As already stated, these are arguments and as such occur in argument

positions as complements of such verbs as /beek-/ 'know', /daga?-/ 'hear',

/arg-/ 'see', which we have collectively called epistemic verbs (cf.

Chapter Three). There are two types of such clauses: finite (tensed) and
 200

non-finite (non-tensed). The former is introduced by the element /akka/

'that', and the latter by an optional adpositional element I-fl 'for'.

For purposes of exposition we shall call them /akka/ and I-fl clauses

respectively.

5.1.1 The Akka-clause

Akka-clauses are finite. Their verbal heads carry (AGR)eement and

(ASP)ectual features. The following are examples of structures with such

clauses.

1(a) [Tulluu-n [ [akka [Fayyiisaa-n duf-el] [hin-beek-a]]]

ST V S that SF
come-pf."cm-know-impf.
'Tulluu knows that Fayyiisaa came'.

(b) [Thlluu-n [ ['aaltuu-n hoolaa bit-t-e]] [aga?-e]]]

S V's S., V'
T-nom. that c-nom. sheep buy-f-pf. hear-pf.
'Tulluu heard that aaltuu bought [a] sheep'.

(c) [Tulluu-n [ [akka [Fayyiisaa-n duf-e]] [arg-e]1]

S V'S S
T-nom. that F-nom. come-pf. see-pf.
'Tulluu saw that Fayyiisaa came'

In such structures the constituents in V' 1 namely the Ss are the akka-

clauses. Like simple nominal arguments, they occur in thematic positions

which the verbal head strictly subcategorizes and 8 (theta) marks in accord

with its inherent lexical property.

The structural relations such clauses have with their verbal heads may

be better observed if presented in the form of a tree structure as in (2):

2.

S V

beek-
 201

Such clauses occur as internal arguments of verbs like /beek-/

'know' or /aga?-/ 'hear'. They cannot occur as external, arguments of

verbs with NP complements, as (3) below is ungrammatical:

3* [ [akka Tulluu-n duf-e] [na aars-e]J

S Sthat T-nom. come-pfYme annoy-pf.
'That Tulluu came annoyed me'.

Such structures would be possible only if the matrix verb were the

copula /da/ 'is', as in (4).

4. ( [akka Tulluu-n duf-e} [ugaa-da]]

SS VP
that T-nom. come-pf. certain is
'That Tulluu came is certain'.

The question which arises from this situation is: why is it that (4)

but not (3) is grammatical? Can this be explained in a principled way?

It might be possible to explain this situation by referring to the

theory of case. According to chomsky (1981:178ff.) case is assigned as a

lexical property of nominals. The subject clause in (3) may hence be

assumed to have been assigned nominative case by ASP/AGR of the matrix Infi.

But case is an abstract relation, it may or may not have morphological

realizations, whether it does or not will depend on the language. In Oromo

nominative case is realized by the affix I-ni. But this element is a feature

of the category [+N-V]. In other words, it is realized only if the head of

the argument which receives the case is a noun. The verbal head of the

clausal argument in (3) is [-N+V], though the clause as a whole is nominal.

Hence it cannot have the nominative marker f-ni. This means that the clause

has been assigned abstract case but that this case has not been (and cannot

be) morphologically realized because of the [-N] nature of the verbal head

of the clause. The reason for the ungrammaticality of (3) might hence be

attrthuted to the lack of the morphological realization of the case rather

than to its abstract relation. In other words, the structure is ill-formed

 202

because the subject clause does not have the nominative marker I-ni. Even

if it had this element, the structure would still have been ill-formed

because the element is a feature of [+N-V] categories only.

If this is the case, then we may ask why (4) is grammatical for it too

is a subject clause lacking the nominative marker I-ni. It seems to me

that in (4) the external argument of the matrix verb is not the clause /akka

Thlluu-n duf-e/ 'that Tulluu came', for in that case the structure would

have been ill-formed for the same reason that (3) was. It could be argued

here that the external argument in (4) is a pleonastic element which in Oromo

is phonetically null. The clause which appears to be the external argument

seems to be an adjunct of SI', similar to the situation in such English

structures as 'It is clear that John came' where 'that John came' is an

adjunct, and 'It' is the subject of the clause (cf. Koster 1978).

Notice that if this argument is correct, it means that the structures

in (1) would have been ill-formed if Oromo had accusative case marking

because the clausal arguments would have failed to show up this morpheme for

the same reason stated above, though they could have received accusative

case from the verb. There is some evidence in support of this assumption.

Infinitival clauses are characterized by the nominative marker I-ni when

they occur in subject positions, and by its absence when they are in object

position, because as we shall observe in the next section, the internal

head of such clauses is nominal and may hence carry the morphological feature

of case.

Regarding movement, in such structures as (1), the entire clause, but

not any of the NPs contained in it, may be optionally raised to topic

position for purposes of emphasis. Thus, (5 a) but not (b) or (c) may

be derived from (1 b) above.



203

5(a) 1 SS1 [aaltuu-n hoolaa bit-t-el]

that q-nom. sheep buy-f-pf.
[Tulluu-n t hin-beek-al]]
S S p_0
1 ctn..know-impf.
'That aaltuu bought a sheep, Tulluu knows'.

(b)* [[hoolaaj ! [Tulluu-n !akka [&altuu-n t bit-t-e]]

1 S S 1
sheep T-nom. that .-nom. buy-f-pf.
hin-beek-al]]
cmrknow-impf.

(c)* [[aa1tuu-n] [ [Tulluu-n [akka [ t hoolaa bit-t-e]]

1 SS S Si
ç-nom. T-nom. that sheep buy-f-pf.
hin-beek-a]]]
cm.- know-impf.

Such structures would be grammatical only if the NPs in question moved

out of the S in which they are base-generated, as in (6) below:

6. [Tulluu-n [hoolaa [akka [aaltuu-n t bit-t-e]]] hin-beek-a]

S S iS SM 1
T-norn. sheep that L-nom. buy-f-pf. cm..know-impf
'Tuliuu knows that 'aaltuu bought [a] sheep'.

The ungranimaticality of (5 b). and (c) and the grammaticality of (6)

suggests that NP movement observes subjacency, a constraint on movement

rules which, according to Chomsky (1973:247), disallows the extraction of

NPs from structural configurations of the type in (7).

7. NP/S

NP
 2fl4

What (7) shows is that movement is possible only within or across

S, but not across the upper S or NP which contains it. This means that

S or NP is a bounding node for Orotno, just as it is for English.

Akka-clauses may also undergo postposing, though this is not as frequent

a process as preposing is. Hence corresponding to (5 a) above, (8) below

is possible.

8. [ [Tulluu-n t hin-beek-a], [akka aaltuu-n hoolaa bit-t-e]]

S ST_nom. cm.know-impf. 1 that -nom. sheep buy-f-pf.
'Tulluu knows that aaltuu bought [a] sheep'.

s stated earlier, akka-clauses are tensed, which means that the Inf 1.

node is characterized by ASP and AGR. The former specifies the completedness-

incompletedness of the action denoted by the verb, whereas the latter

identifies the external argument of the verb in terms of the nominal features

of person, number and gender. Since such features are reflected in the

morphology of verbs, the external argument may easily be dropped, as in (9 5)

derived from the corresponding underlying structure seen in (9 a).

9(a) [Tulluu-n [ !- [isaan(i) hoolaa bit-an-i]1 aga?-eJJ

S V'S S
T-nom. that they-nom sheep buy-pl-pf. hear-pf.
'Tulluu heard that they bought [a] sheep'.

(b) [Tulluu-n [ ! akka [(NP) hoo].aa bit-an-i]] daga?-el]

S V'S S
T-nom. that (they) sheep buy-pl-pf. hear-pf.
'Tulluu heard that they bought [a] sheep'.

However, there are structures of tensed clauses where the subject

cannot be dropped. Such structures include negative clauses, in which

the verbs are in the perfective aspect. In such clauses, the Inf 1. node

is characterized by ASP only. The agreement elements do not appear at

all, with the result that the subject argument cannot be dropped without

the resulting structure being multiply ambiguous. Consider the following,

for example:
 205

10. [Tulluu-n [ [akka [Pro hoolaa hin-bin-n-e]]

S V' S
T-nom. that Pro sheep neg-buy-neg-pf.
hin-beek-al]
cm-know-impf.
'Tulluu knows that Pro did not buy [a] sheep'.

Pro in (10) is not specified in terms of the features person, number

and gender since the verb does not show such features whenever it has the

negative morpheme I-n-I. It seems that in such clauses negative morphology

has the effect of blocking AGR, and whenever this is the case, the subject

argument cannot be Pro or else the interpretation of structures like (10)

would have the reading in (11).

11. [Tulluu-n [ [akka [X hoolaa hin-bin-n-e]] hin-beek-a]]

ST_nom. \ Sthat S sheep rieg-buy-neg-pf. cm;know-impf.
'Tulluu knows that X did not buy sheep'.

This situation suggests that AGR and ASP are independent features in Infi.

The position of X in (11) is a governed positionS which means that a

lexical subject is licensed to occur. This is noticeable from the

structure in (9) where we have /isaani/ 'they' or in (12) where we have

/iii/ 'she'.

12. [Tulluu-n [ [akka [iii-n hoolaa hin-bin-n-e]]

S V'S S
T-nom. that she-riom. sheep neg-buy-neg-pf.
hin-beek-al]
cm. know-impf.
'Tulluu knows that she did not buy sheep'.

In such structures, the case-assigning element must be ASP in Inf 1.

since AGR is not available, for the reason stated above. We shall pursue

this argument in the next section.

Going back to the question of movement, we have noticed that fronting

or postposing of a constituent to a non-argument (A) position 2 is possible,

and that Oromo observes subjacency. In what follows we shall consider

 206

movements of substitution, that is, movement to an argument (AL position.

Such movements are syntactic in the sense that they map D-structure into

S-structure as opposed to some stylistic movements of adjunction which may

operate at the level of the phonological component (Chomsky 1981:18).

Only such movements are optional. This is in relation to languages like

English which are characterized by both obligatory and optional movements.

For Oromo the situation is different. Consider the following structures:

13(a) (i) [nam-oon-rii [kan 'aaltuu aalat-an-i] fakkaat-a}

Sman_pi _nom . Scomp. love-pl-pf. seem-impf.
'It seems that the men loved aaltuu'.

(ii) [nam-oon-ni [kan caaltuu 'aalat-an-iJ fakkaat-uJ

Sman_pi _nom. Scomp. . love-pl-pf. seem-3p1.
(Literally), 'The men seem they loved aaltuu'.

(b) (i) [aaituu-n [kan horii kab-d-uJ fakkaat-aJ

S _nom. Scomp. money hold-f-impf. seem-ixnpf.
'It seems that aa1tuu holds/has money'.

(ii) [aaltuu-n [kan horii kab-d-u] fakkaat-t-i]

S•_nom. Scomp.money hold-f-impf. seem-f-irnpf.
(Literally), '&altuu seems that she holds/has money'.

The structures in both (a) and (b) are paraphrases despite the formal

differences shown by their matrix verbs. In the (i) structures, there

is'no agreement between the matrix verb /fakkaat-/ 'seem' and the subject

arguments. In the (ii) structures, however, there is agreement between

the verb and /nom-oon-ni/ 'men-nom' in (a) and aaltu in (b). For

example, the verb in (a ii) is plural in agreement with the plural subject

/nam-oon-ni/. Similarly /fakkaat-/ 'seem' in (b ii) agrees in gender

with the feminine noun 'aaltuu, though this is not the case in (b i)

The verbs in the embedded clause are consistent in their agreement

relations with /nam-oon-ni/ 'men-nom' in (13 a) and aaltuu in (13 b)

This situation suggests that in those cases where there is no agreement

 207

between /nam-oon--nh/ 'men-nom' or &altuu and the matrix verb, the subject

of the main clause may not be /nam-oon-ni/ or caaltuu. It could be an

empty NP, in which case such structures may be treated as being derived

from an underlying representation of the type in (14):

14(a) [NPe [ [kan [nam-oon-ni [aaltuu 'aalat-ani-i]1J]

S V'S s
comp. man-pi-nom. C. love-pl-pf.
(fakkaat-a]]
V
seem-impf.
'It seems that the men loved &altuu'.

(b) [NPe [ [ [&altuu-n [horii kab-d-u] 111 [fakkaat-a] I

S V'SS V1 V
e-nom. money have-f-impf. seem-impf.
'It seems that 'aaltuu holds/has money'.

In (14) the matrix verb is singular and in agreement with the empty

subject NPe, just as in the same way the embedded verbs show the relevant

features in agreement with their respective subjects.

This approach seems to explain the reason for the apparent discrepancy

of agreement between the structures in (13). There we have noticed that

the matrix verb agrees with the lexical subject in (a ii) and (b ii), but

not in the corresponding structures in (as). In those cases where there is

agreement, we may argue that the subjects of the embedded clauses have

undergone a movement rule which raises them to the empty NPe position in

the matrix clause, and in those cases where there is no such agreement, we

may assume that the movement has not taken place. In other words, the rule

optionally operates on the underlying structure in (14) to derive the

corresponding surface structure in (13).

This approach also goes in line with the lexical property of /fakkaat-/

'seem' described in Chapter Three. There we have indicated that this verb

does not select an external argument. The position where such an argument

is expected i vacant since Oromo does not have non-referring or pleonastic

 208

pronouns to cover it whenever movement has not taken place.

Certain questions arise from this state of affairs:

i) Why does the rule apply only optionally?

ii) Why should such a movement take place at all? (This is a

more serious question from the point of view of syntactic theory.)

The kind of movement seen in corresponding structures in languages

like English is obligatory, because the NP undergoing the movement has to

receive case in order to escape the case filter which bars NPs having

phonetic matrices from structures if they are not case-marked (cf. Chomsky 1981).

Since the position where such NPs are base-generated is not governed -

and hence not case-marked (the clause being non-finite), the NPs cannot

receive case in situ. Hence, they should move to another position where

they can receive case and so escape the filter. One such position is the

subject position of the matrix clause which is empty, but which is governed

and hence case-marked by AGR/ASP in Infl. In other words, the movement in

such languages is independently motivated by the theory of case.

There is no such motivation for the movement of /nam-oon-ni/ 'men-nom'

or aaltuu in the (ii) structurea in (13). As is observable from (14),

the embedded verbs have both aspectual and agreement features, which means

that the clauses are finite, and that the Inf 1. node has AGR/ASP, the

element which, according to Chomsky (1981:170), assigns nominative case

to an external argument. In short, then, both /nam-oon-ni/ 'men-nom'

and &altuu could receive and in fact do receive nominative case within

their own clauses as the structures in (13 i) demonstrate. In such cases,

there is no difference between the base and the corresponding derived

structures in terms of the linear positions of arguments. Such differences

 209

arise only when an argument moves from its base position and lands in another
position in the corresponding S-structure as in the (ii) structures in

(13), the S-structure of which is as shown below:

15(a) [nam-oon-ni [ kan [ t &altuu 'aalat-an-i]] fakkaat-u]

S S S1.
man-pi-nom. love-pl-pf. seezn-3p1-impf.
'[The] men seem [they] loved &altuu'.

(b) [(aaltuu-n (kan [t horii kab-d-u]] fakkaat-t-i]

S_nom. 1 1 money comp. has-f-pf. seem-f-p-impf.
'aaltuu seems [she] holds money'.

Although the movement in (15) is not triggered by the case filter and
3
hence is optional, its application does not seem to violate the 8 criterion
since the moved element lands in a non-thematic ( 8) position, that is, in

a position which has not been 8-marked by the VP of the matrix clause,

and which hence has been empty. In other words, the position has been

vacant until the raising has taken place. The relation between the raised

NP and the trace does not seem to violate the empty category principle

(ECP) of Chomsky (1981:250), which requires that traces should be properly

governed, whereby 'proper government' is meant, in one sense, government

by a c-commanding antecedent, 4 that is, by the moved element, or, in

another sense, government by a lexical head such as (±N ±V]. Infi., the

element which governs the trace in (15), is a governor but not a lexical

governor. Hence the trace is properly governed only by its antecedent.

In either case there is no violation of this principle, in fact, since

Infl. in Oromo is richer than Infi. in languages like English, it may be

considered as a lexical governor, which means that even the trace may be

properly governed by it.

The other principles which govern the relation between a moved NP

and its trace are the binding principles of Chomsky (1981188). One of
 210

these principles states that anaphors must be bound in their governing

categories. 5 NP traces are anaphors, which means that they too must be

bound in their governing categories. The governing category of the

trace in (15) is the matrix clause since only it meets the definition

given by Chomsky (1981:211). And, as predicted by the theory, the trace

may be said to have observed this principle since it is bound by the

antecedent NP for its reference in this clause. Notice, however, that

the trace is also governed by Infi. which has AGR, its(trace-accessible

SUBJECT. This means that it can also get its features from AGR in its

own clause and hence does not need to be bound by the antecedent for its

reference. But this violates the principle which dictates that it should

be bound. It cannot be bound by AGR because AGR is not a binder; only

NPs are binders. In other words, the trace behaves like a pronoun in this

situation, since pronouns are free in their governing categories. This is

a contradictory situation. I leave it open.

Further problems arise when we consider the status of the trace in

relation to case. As stated above /nam-oon-ni/ 'men-nom' or aaltuu-n

have been extracted from a case-marked position and moved to another case-

marked position. If we assume with Choinsky (1981:170) that structural case

assignment takes place at the level of S-structure, then /nam-oon-ni/

'men-nom' and aaltuu-n may receive nominative case from AGR/ASP in Inf 1.

of the matrix clause whenever movement takes place, or from the AGR/ASP

of Infl. in their own clauses when movement has not taken place. The

question which follows from this concerns whethcr or not the trace also

has case.

According to Chomsky (1981:80) the trace of an NP does not have case

since its position is not case-marked. Instead, it is the chain which the

antecedent forms with the trace that gets case. In other words, the
 211

trace cannot have case independent of or other than that which is assigned

to the head of the chain. But the situation we have in (15) does not

seem to fit into this line of argument. The trace may be said to be part

of a chain which is headed by the antecedent, but it does not seem possible

to say that the trace has no case given the fact that it is in a governed

and hence in a case-marked position. If it is case-marked, then it means

the chain is doubly case-marked, once from its position in the lower clause

and a second time from the matrix clause. This is contrary to the

property of chains. They are supposed to have one case only (cf.

Chomsky 1981).

The relation of the trace with AGR in Inf 1. in its own clause and with

its antecedent Np in the matrix clause at the levels of S.S and logical

form (LF) is as follows:

16.

In (16), the relation between the trace (t) and Infl. in the S is

a relation that may involve both binding and case, since as stated above,

Infl. has AGR, the case-assigner. The trace may be coindexed with AGR

or with the NP in the matrix clause for its reference. The NP gets its

case from Inf 1. in S.

 212

The movement we have observed in (15) involves extraction from a

case-marked subject position, and the problems with regard to the

principles of binding and case theory which arise from such extractions

concern the status of the trace left behind by the moved NP. Oromo

appears to violate some of the principles involved when a subject argument

is moved. In what follows we shall consider extraction from an object

position and check whether or not this also leads to the same kind of

problems we have observed with respect to subject extraction. In order

to do this, we have to raise the object altuu in (15) and substitute it

for the empty NPe of the matrix clause in the manner suggested in (17)

below:

17* [&altuu-n [ [kan [nam-oon-ni t 'aalat-an-i]] [fakkaat-t-i]]]

s . . 1v' 1 V
ç-nom. camp. nian-pi-nom. love-pl-pf. seem-f-pf.
(Literally), '&altuu seems that the men loved'.

Unlike in (15), the extraction in (17) leads to ungramznaticality.

The reason for this does not seem to be related to the movement as such,

because: (i) the movement crosses only one bounding node (S), hence there

is no violation of subjacency; (ii) the moved NP lands in 8 pasftion, thus

avoiding violating the 8 criterion, and (iii) the trace of the moved NP

observes the empty category principle, which dictates that it should be

properly governed. The trace is properly governed by the verb /aalat-an-/.

As in (15) the position from which aaltuu has been extracted is a

case-marked position, which means that aaltuu would not have to move for

reasons of case. In other words, the situation in (17) is exactly the

same as that in (15) except for the terms themselves. In (15) it was the

subject which has been raised, here it is the object. But this

difference does not seem to be the real cause for the ungrainxnaticality of

the structure. In fact, the structure may be said to be well-formed from

 213

the point of view of its syntactic derivation. In other words, its

derivation does not violate the principles that govern NP movement.

The problem with (17) may be explained in terms of the principles of

binding, which operate at the level of LF, that is, after movement has

taken place. As stated earlier in connection with the structures in (15),

the trace left behind from a moved NP is an anaphor and according to

principle A of the binding theory of Chomsky (1981:188) it should be bound

in its governing category. The governing category of the trace in (17)

is the lower clause. But there is no NP in this clause with which the

trace can be associated. It cannot be coindexed with /nam-oon-ni/

'men-nom' because /nam-oon-ni/ is in a 8-position and coindexing it with

the trace would violate the 8-criterion. Hence the trace remains without

being bound to any antecedent NP for its reference in the governing

category, which is in violation of the said principle. This means that

&altuu cannot be raised in (17), not so much on account of a principle of

movement such as subjacency, but for principles that operate on the output

of such movement.

From the discussion so far it appears that Oromo is characterized

by NP movement which operates optionally on tensed clauses. The process

observes subjacency. But the trace fails to satisfy the requirement

that it should not bear case, and that it should be bound since it is an

anaphor.

5.1.2 The f-clause

This is what has traditionally been called the infinitive or gerund.

It is morphologically identifiable by the deverbalizing suf fix /-uu-/

The derivation is lexical and is also highly regular. Any verb stem, with

the exception of the copula Ida! 'be' may be nominalized by affixing I-uu-/.


214

The derived form behaves like any other nominal. It occurs in

thematic positions and is also subject to the case filter. It is also

characterized by the same syntactic features that characterize its verbal

base. For example, the verb /i5aar-/ 'build' strictly subcategorizes an

object argument as part of the specification of its inherent lexical

properties. It also selects an external argument for the syntactic subject

position. In the same manner, its nominalized form /iaar-uu/ 'to build!

building' is characterized by an object argument. It may also select an

external argument, which could be Pro. in the marked case. The difference

between the verb and its nominal form may hence be said to be categorial

rather than of subcategorization.

As has been shown in Chapter Three, infinitivals occur as complements

of a set of verbs we have called desiderative verbs. The following are

some examples of such complements.

18(a) [Tulluu-n [ [ [mana iaar-uu] I hin-barbaada]]

S
T-nom.
V'SS house build-to cm.want-impf.
'Thlluu wants to build a house'.

(b) (Tulluu-n ( [ [aannan dug-uu]] hin-aalat-a1]

S v'sS.
T-nom. milk drink-to cm... like-impf.
'Tulluu likes to drink milkt.

The infinitival clause in (18), like the finite akka-clauses we have

already seen, may be preposed as in (19 a) or postposed as in (19 b)

both being derived from (18 a)

19(a) j [mana i'aar-uu] ! (Tulluu-n t hin-barbaad-a]]]

S house build-to S ST_nom 1 cm.- want-impf.
?'To build [al house, Tulluu wants'.

(b)? [Tulluu-n [ t hin-barbaad-a] [mana jar-uuJ]

S V' 1 1
T-nom. cin.-want-impf. house build-to
'Tulluu wants to build a house'.
 215

Whereas preposing of a complement is quite natural and readily

acceptable, judgments vary about postposing, though structures with such

postposed complements are not regarded as being ill-formed.

Again as in akka-clauses, nvement out of infinitival clauses observes

subjacency. This is deducible from the structure in (20) below derived

from (18 b) above.

20? jaannan ( [Tulluu-n I [ [ t dug-uu]] hin-aalat-a]1]]

S. 1 SS v'SSl
milk T-nom. drink-to cm. like-impf.
(Literally), "Milk, Thlluu likes to drink'.

In (20) /aannan/ 'milk' crosses two bounding nodes, the embedding and

the embedded Ss, before it gets to its surface position. Even if we

argued that the uvement was cyclic and that /aannan/ 'milk' first landed

in the position of the complementizer of the clause in which it is base-

generated, and finally got to the topic position by crossing the upper S,

the resulting structure would still be excluded at LF for the reason that

the trace, which, as we have said earlier, is an anaphor, would fail to

satisfy principle (A) of the binding theory, which dictates that it should

be bound to an antecedent in its governing category. The governing

category in (20) is the embedded S since according to the definition we

had, it is this clause which contains the governor of the trace, the verb

/dug-I 'drink' and the accessible subject, Tulluu. But the trace cannot

be bound in this clause because the accessible subject, Thlluu, which is

supposed to be its binder is in an independent thematic position and

coindexing it with this subject would violate the 8-criterion.

Whereas infinitival clauses are similar to akka-clauses with respect

to subjacency, they differ from the latter with respect to their scope of

distribution. As we have observed in the preceding section, akka-clauses

 216

are hinted to the position of objects. In other words, they function

only as internal arguments of verbs. In contrast, infinitivals can occur

not only as objects of verbs as in (19), but also as complements of

adpositions and as subjects of clauses. To this extent infinitivals are

similar to simple nominals. Consider the following examples:

21(a) [ Iaannan dug-uu-n] [gaarii-daJJ

ss. VI
milk drink-to-nom. good is

(b) (Tulluu-n [hoolaa bit-uu-fi] gara Gimbii deem-e]

S Adp''
T-nom. sheep buy-to-for to G. go-pf.
t Thlluu went to Gimbii for buying sheep'.

In (21), the infinitival clause /aannan dug-uu-n/ is in subject

position in (a) and in complement position in (b). As subject of the

clause, it displays the nominative case affix I-ni.

The distributional differences between the two types of clauses seem

to reflect the categorial differences the internal heads of the clauses

show. As stated earlier, the head of the VP in akka-clauses is verbal,

and, as we have assumed in the preceding section, such clauses seem to

resist features of case and hence fail to occur in positions where NPs with

morphological case marking are expected. Contrary to these, however, the

heads of infinitivals are nominal, and as stated above, they are subject

to the case filter. The structure in (21 a) would be ungrammatical if

the head of the infinitival clause, i.e., /duguu/ 'drinking/to drink'

occurred without the nominative marker I-ni. Hence the difference in

distribution between the two clauses is a difference that results from the

verbal vs. nominal nature of the heads.

Going into the internal structure of infinitival clauses, it is

noticeable that in (21) the subject position of the infinitival clause,

which is itself the subject of the matrix clause, is not lexically filled.
 21?

The structure would have been ill-formed if this position had a subject

with a phonetic matrix as in (22) below:

22* [ j [Tulluu-n aannan dug-uu-n]] [gaarii-da]]

sss . VI
T-nom. milk drink-to-nom. good is

The reason for the ungrammaticality of (22) is again the case filter.

The position of Tulluu is an argument position, but it is not a case-marked

position because the clause being non-finite, Inf 1. does not have the

governing and hence the case-assigning element AGR/ASP. But in accordance

with the extended projection principle of Chomsky (1981:25ff) and (1982:10),

which stipulates that all clauses have subjects, the subject of infinitivals

is assumed to be an abstract entity, that is, an entity which does not

have a phonetic matrix. This entity, called PRO, behaves like other

pronominals with respect to certain principles which determine co- or

disjoint references at LF.

In the structures in (21) or (22), the subjects of the infinitival

clauses must be PRO according to the principle mentioned above; which

means that at the level of D-structure (21) is like (23)

23. [ [ [ [PRO aannan dug-uu-nJ] [gaarii-da]]]

ssss V'
milk drink-to-nom good be
'To drink/drinking milk is good'.

As stated earlier, PRO is like a pronoun with respect to its reference,

which means that it may be deictic (disjoint) or anaphoric. When it is

deictic, it has independent reference and when it is anaphoric it gets its

reference from an NP outside its own clause. In (23) above, for example,

PRO is deictic since there is no NP in the matrix clause with which it can

be coindexed for reference.

As opposed to PRO in (23) which is free for its reference, PRO in

(24) below is not free, though the clause in which it occurs is in the

same subject position as the clause in (23)



218

24. [ [ ( [PRO horii kab-uu-nIl [Tulluu [hin-gaminaiis-a]1]]

ssss v' V
nxney have-to-fern. T. cm. please-cs-impf.
'To have/having ney pleases Tulluu'.

Here PRO refers to Tulluu, the object of the matrix verb. Hence PRO

is anaphoric. But unlike other anaphoric pronominals such as reflexives

which are bound in their governing categories in accord with principle A of

the binding theory, PRO is free in its governing category. In this respect

it is like pronouns which may be free in their governing categories. But

in (24), PRO is not free since it gets its reference from the object NP

Tulluu. But Tulluu is not in its governing category. Besides, Tulluu

is not in a c-commanding 6 position, a condition which antecedent-anaphoric

relationships should satisfy. The situation suggests that the relation

between PRO and its controller (binder) is not and cannot be the same as

the relation between art anaphor and its local binder. Instead, it appears

that PRO may have to find an antecedent somewhere in the root clause

irrespective of whether that antecedent is in a c-commanding position or

not. From this follows that the relation between PRO and its controller

may not be determined in terms of syntactic configurations, but, as

Chomsky (1981:76) has said, it may involve 'thematic roles or other semantic

properties of verbs, or perhaps pragmatic conditions of some sort'.

This seems to be true with respect to the structure in (24) and with

others having the same type of matrix verb as (24). The verb /gaininais-/

tplease s is irpho1ogica1ly causative. Its non-causative counterpart is

intransitive, and it is the subject of this intransitive verb which occurs

as the object of the causative form in (24). Consider (25) below.

25. [Tulluu-n gammadd-e]

ST_nom please -pf.
'Tulluu became happy (pleased)'.
 219

The causative counterpart of /gammadd-/ 'please' has the property of

subcategorizing an NP as its complement. Whenever this is the case, it

seems that PRO is controlled by an object NP which may or may not be in

antecedent and hence in c-commanding position. In (24), for example, when

PRO is in a subject sentence, the controlling NP is in object position.

In (26) below, the controller is still an object NP and is also in an

antecedent and c-commanding position:

26. [ [Tulluu-n [[Fayyiisaa] [ [ PRO farda guluf-uu]] barsiis-e]1J

S S_01 S S
F. horse ride-to learn-cs-pf.
'Tulluu taught Fayyiisaa to ride (a] horse'.

The matrix verb /barsiis-/ 'teach' like /gamma-is-/ in (24) is

causative. But unlike the latter, it is a two-place verb, which means that

it requires two complements. The clause with the PRO subject and Fayyiisaa

are expressions of this lexical requirement. In such cases PRO is controlled

by an object NP. So, it seems that whenever PRO is in an object sentence,

the controlling Np is in antecedent position, and whenever it is in subject

sentence, the controlling NP is in a non-antecedent position. The

structures in (24) and (26) above, and the one in (27) below support this

view.

27(a) [Tulluu-n ([Fayyiisaaj [ [ P RO hoolaa bit-uu]] gargaar-e]]

S V SS
T-nom. F. sheep buy-to help-pf.
'Tulluu helped Fayyiisaa to buy [a] sheep'.

(b) [Tulluu-n ( [Fayyiisaa] [ [ PRO ho5ii ho'-isiis-uu]J

S V' SS
T-nom. F. work work-cs-to
hin-barbaad-a] I ]
cm.- want-impf.
'Thlluu wants to work Fayyiisaa'.

In the data provided so far, we have observed two types of P1O: one

which refers freely to anyone, and another which depends on an NP. The NP

in the examples cited is an object NP. Thus PRO in such structures is

 220

object-controlled. In contrast to this, PRO may be controlled by the

subject NP of a matrix clause, as in (28).

28(a) [Thlluu-n [ [ ( PRO wayaa miiuu]1 hin-danda?-a]]

S V'SS
T-nom. clothes wash-to cm. be-able-impf.
'Thlluu is able to wash clothes'.

(b) [Thlluu-n [ [ [ PRO mana i5aar-uuu-f} I hin-ta?-a] I

S V'SS
T-nom. house build-to-for im-think-impf.
(Literally), 'Thlluu thinks for building a house',
'Tulluu plans to build a house'.

Earlier it was suggested that PRO may not be subject to the same

principle of binding theory as are anaphors since it is free in the governing

category, whereas the latter are bound. In other words, PRO, though not

anaphors, may be controlled by or bound to a distant antecedent. In order

to capture PRO and other pronominals in one domain where they may be free

or bound to an antecedent and where such a relationship is expressed in

terms of syntactic configurations, rather than in terms of semantic notions

or pragmatic cues, Manzini (1983) has extended the notion of 'governing

category' as per Chomsky (1981:188) to 'domain governing category'. 7 It is

claimed that anaphors are bound in their governing and domain-governing

categories, whereas pronominals are free in their governing categories.

Whether this theory or those proposed by Bresnan (1982) or Williams

(1980) are fully applicable to Oromo or not is a subject not to be

definitive about at this point. One needs to consider other anaphors,

such as reflexives, reciprocals, NP traces in structures across all maximal

categories. For the moment we may tentatively say that PRO is either

free, or controlled by an antecedent or a non-antecedent N? in the matrix

clause.

In the discussion so far we have observed infinitival structures

with PRO subjects. These constitute only one type of such structures.
 221

The other type includes infinitives which may have lexical subjects.

The following are examples of such structures.

29(a) [((Thlluu-n) deem-uu-n-isaaj [na-aar-s-e]J

S T-nom. go-to-nom-his V me-annoy-cs-pf.
'Tulluu's going annoyed me'.

(b) [Tulluu-n [ ( [(d'aaltuu-n) deem-uu-i'iiJ] daga?-e]]

V S S -nom.
T-nom. go-to-her hear-pf.
'Tulluu heard of &altuu's going'.

In (29) the infinitival structures in both sentences have subjects

which have phonetic matrices. As stated earlier, such subjects can occur

only in positions which are case-marked. In order for such a position to

be case-marked, the Inf 1. node of the clause must have AGR/ASP, or else

the position must be subject to government by a case-assigning element in

the matrix clause. In the structures concerned, the entire infinitival

clause is assigned nominative case in (a) and accusative case in (b).

This is indicated by the presence of the nominative marker I-ni in the head

of the infinitival clause in the former, and by its absence in the latter.

The subject position within the clauses in both cases cannot be governed,

and hence case-assigned, by the matrix Inf 1. or verb because such clauses,

being Ss, are a barrier to government. However, as it is observable from

both structures, the subjects Tulluu in (a) and daaltuu in (b) have the

nominative case-marker i-ni, which shows that nominative case has beeen

assigned to them within their respective clauses. What is significant

to note here is that in (29 b) whereas the entire infinitival clause is

assigned accusative case by the subcategorizing verb /aga7-/ 'hear',

the subject of the clause &altuu is in the nominative case. This

suggests that the clause is assigned accusative case as a unit by this

verb, but the subject of this object clause is assigned nominative case


222

independently. In other words, we have a situation where the part is

assigned a case which is different from that assigned to the whole.

This situation suggests that the subjects of the infinitival clauses

in such structures as (29) are case-marked by an element within their own

clauses. If it were the case that they had the same case features as

their heads, we should perhaps argue in terms of agreement relations,

although this again is difficult given the fact that the structures are

one of subject and predicate and not of head and modifier. In other words,

it is not a case where a non-head category copies the features of its head

for reasons of concord.

What distinguishes the infinitival heads in (29) from other

infinitives is their having such pronorninal affixes as /-saa/ 'his' or

/-ii/ 'her'. It must therefore be the case that it is the presence of

such affixes which licenses the occurrence of lexical subjects in positions

which are otherwise filled by PRO. These ronominal elements may be

considered as being the counterparts of the person marking elements in

verbs in tensed clauses. In other words, they are the person-markers

in infiiit.ivals or gerunds just as the others are in verbs. For purposes

of illustration and comparison let us observe the following paradigms

below, one for the verb /deem-/ 'go', and the other for its infinitival

counterpart /deem-uu-/ 'to go/going'.

30. deem-uu -koo

1. deem-8-
2. deem-t- deem-uu-kee
deem-uu-isaa
3m. deem-8-
f. deem-t- deem-uu-i'ii
deem-an- deem-uu-keenna

deem-t-an- deem-uu-keesan
deem-uu-saani
deem-8-an-
 223

The difference between the paradigms above lies in the possibility

for the verbs to have tense or aspect-marking elements. This is obvious

given the fact that tense or aspect is a feature which distinguishes them

from other categories. In so far as the feature person is concerned,

however, there does not seem to be any difference since the infinitivals,

just like the corresponding verbals, have pronominals which refer to their

possible subjects.

Compare the fellowing structures.

31(a) ( [ati [gara Gimbii deem-t-e]]]

S Vito G. go-2-pf.
'You went to Gimbii'.

(b) [ [&altuu-n [gara Gimbii deem-uu-n-iii...]]]

S S_nom. to G. go-to-nom-her
' 'aaltuu's going to Gimbii...'

Because of the presence of the person-marking elements in both the

verbal and infinitival heads of such structures, the subjects may be dropped

as in (32)

32(a) [Pro [gara Gimbii deem-t-e]]

S V0 G.
go-2-pf.
'You went to Giinbii'.

(b) [Pro [gara Gimbii deem-uu-n-i'ii...J}

S
to G. go-to-nom-her
'Her going to Gimbii...'

This parallel situation induces us to believe that it is the person-

marking elements in both types of structures which determine whether a

lexical subject is possible or not, and also whether that subject may be

dropped or not. The only difference between the person-markers in the

two types of clauses is that in the case of the infinitivals, the markers

are genitives. But this difference does not have any effect on the type
 224

of case the subjects should have since in either case they are in the

nominative case. In other words, the form of the affix in the verb or

in the corresponding infinitive has no bearing on the case form of the

subject.

In parallel structures in languages like English, the subject of the

gerund is in the genitive case and this case is assigned to the noun in

subject position by an inserted (poss)essive element. This is the

situation in (33 b) below, derived from (33 a)

33(a) (John going]

(b) ([John + poss] going]

John's going.

In (33), there is nothing in 'going' that refers to John. Hence,

John is not licensed to occur by AGR in Infi., but by the fact that it

has received the genitive case from the inserted (poss)essive. The

situation we have in Oromo is not the same as what we see here, because

parallel to (33) above, the structure in Oromo would be the ungrammatical

(34) below.
34* [ [Tulluuisaa deem-uu] (na aar-s-e]]
S ST_his go-to Vime annoy-cs-pf.

In Oromo, the genitive element is part of the inifectiori of the gerund,

just like the person-marking elements are part of the inflectional morphology

of verbs.

From this situation the most plausible approach to the analysis of

infinitives or gerurids in Oromo would be that which recognizes the pronominal

elements as AGR in Infl., i.e., as on a par with those in finite clauses.

In other words, the Infl. node in Oromo infinitival clauses has AGR, and

it is this element which assigns nominative case to the subject parallel

to the AGR in finite clauses.

 225

Notice that Infi. only has AGR in infinitival clauses. It does not

have ASP. This may lead to the conclusion that it is only AGR which is

involved in the assignment of nominative case. Chomsky (1981:170) has

taken this position, although in 1980 he had argued-that it was the presence

of tense in Infl. which was responsible for the assignment of this case.

The negative structures we have examined in the preceding section

may suggest that in Oromo ASP alone may also assign nominative case.

As we have observed earlier, negative morphology appears to block AGR..

Verbs in tensed clauses do not show person-marking elements if they

contain a perfect negative marker in them. But the subjects of such

clauses are case-marked. Since in such structures only ASP is available

in Infi., the case-assigner miist be ASP, just as it is AGR in the case of

the infinitivals. From this it follows that both AGR and ASP are case-

assigners, and since Inf 1. may have either one or the other or both or

neither of the two, the node has to be specified in terms of features (cf.

Picallo 1984) rather than in terms of just AGP. or ASP. Accordingly,

clauses may have to be divided into tensed and non-tensed categories.

Those which are tensed have [+ ASP ± AGR] in Inf 1. Those which are non-

tensed have either [-ASP - AGR] or f-ASP + AGR] in Infi. These may be

called non-finite and finite infinitivals respectively.

If this is plausible, then we may also have to distinguish between

clauses in terms of their possible subjects. The subject of a clause is

PRO if Inf 1. is characterized by the features [-ASP - AGRI and pro when it

is characterized by [+ASP + AGR] and a variable when it has the features

[+ASP ± AGRI.

This analysis is based on the assumption that the infinitive or gerund

in Oromo has person-marking inflections. The fact that such elements are
 226

identical in form to the possessive pronouns, which occu as specifiers

of simple NPs (cf. Chapter Six) may induce us to treat them as specifiers

of infinitival heads. In other words, the situation is comparable to

what we see in (35).

35(3) deem-uu-isaa
go-to-his
'his going'.

(b) mana-isaa
house-his
'his house'.

First of all, the two structures differ in their level of projection

/deemuu-saa/is clausal, whereas ,1nana-isaa/ 'his house' is not. Hence the

former but not the latter allows a subject since (36 a) but not (b) is

grammatical.

36 (a) Thlluu-n deem-uu-n-isaa

T-nom. go-to-nom-his.
'Thilu's going'.

(b)* Tulluu-n mana-isaa

T-nom. house-his.

Secondly /-isaa/ 'his' in /mana-isaa/ shows possession, whereas /-aa/

in /deemuu-isaa/ toes not. This is obvious from the structures in (37)

below, corresponding to those in (35) above.

37(a)* Tulluu-n deem-uu kaba

T-nom. go-to has

(b) Tulluu-n mana kaba

T-nom. house has
'Thlluu has a house'.

Thirdly, according to Jackendoff (1977), the specifier of a clause is

the subject itself; in which case the specifier of /deemuu. 'going/to go',

in (35) is the subject PRO or Tulluu as in (36 a) , but not /-saa/.

 227

Furtherur,re, like simple nominals, infinitivals such as /deemuu/ may

be followed by their specifier. Hence corresponding to (36 a) we may

have (38) below.

38. [deemuu Thlluu]

go-to of-T.
(Literally), 'Going of Tulluu'.

This is parallel to the head-specifier relationship in simple N'Ps

of the type:

39. mana Tulluu

house of-T.
(Literally), 'House of Tulluu'.

Notice that in structures like (38) the infinitival does not have /-isaa/.

Its presence would have made the structure ill-formed. The reason why it

is not there seems to be related to the type of structure (38) is in.

The structure is still clausal. The subject Tulluu seems to have undergone

a postposing rule, and received genitive case from the infinitival head.

In other words, in such a structure Tulluu is not in the position where it

was base-generated and where it would have received nominative case from

AGR in Infl. in the manner discussed. The type of postposing and the

problems that arise from it will be discussed in the next chapter where we

deal with specifiers in general.

The analysis of pronominal elements like /-isaa/ in infinitivals as

specifiers does not tell us how nominative case is assigned to subjects of

infinitivals. It cannot be assigned under government because the clause

being non-finite does not have AGR in Inf 1. One possible answer is to

say that the subject is identified and case-marked without government being

involved. But this leads to a situation where we have two types of case

assignment: one under government as in tensed clauses, and another under



228

no government as in non-tensed clauses. This lacks in generalization

unless we also argue that the assignment of case does not depend on

government but on other factors. This is not an attractive solution

because if pursued seriously it would end up with case no more being a

structural notion. For purposes of generalization, and with a view to

having a highly constrained grammar, we shall maintain the claim that the

gerund or infinitive in Oromo is inflected and that case assignment is based

on the notion of government in clauses of every type.

5.2 Adnominal Clauses

What we have been concerned with so far are types of clauses which

occur as complements of V'. The other types of clausal complements we

need to consider are those which we have treated in the preceding chapter

as complements of N'' or N''' - depending on the restrictive or non-restrictive

effect they have on their heads. Such clauses are modifying complements,

and like any other such complements, they are optionally selected by their

NP heads. This is one major point which distinguishes them from the

argument complements of V', which, as we have discussed in the preceding

section, are obligatory.

As stated earlier in this chapter, our main concern here is to

examine the nature of such clauses and also to propose an alternative

analysis for the so-called 'relative pronoun'.

Let us observe the following structures.

40(a) [ [fard-i-i (Thlluu-n (kan) bit-el] [du?-e]]

S N' horse-sgl-nom. 2 T-nom. comp. buy-pf. die-pf.
'The horse which Thlluu bought died'

(b) [ [nain-i-i (kaleessa (kan) duf-el] [bbaa-koo-ti]]

SN'' S
man-sgl-nom. yesterday comp. come-pf. father-my-is
'The man who came yesterday is my father'.
 229

In (40), S is the modifying complement of the head /fardi-i/

'the horse-nom' in (a) and /nami-i/ 'the man-nom' in (b). Both heads

are external arguments of the matrix verbs. In (41) below, we have

structures in which such clauses occur as complements of arguments in two

object positions. The first head occurs as complement of V 1 and the second

as that of P'. In either case, the clause is still a complement of N''.

41(a) [fard-i-i [marga [Tulluu-n kan bit-el] [iaat-e]]

S N''' S
horse-sgl-nom. grass T-nom. comp. buy-pf. eat-pf.
'The horse ate the grass which Tulluu bought'.

(b) [Tulluu-n [waayee [ (nam-ia [kaleessa kan du?-e]]]]

S P' N'''N'' S
T-nom. about man-sgl. yesterday comp. die-pf.
[dubb-at-e]]
speak-mid-pf.
'Tulluu spoke about the man who died yesterday'.

The modifying complements in the examples in (40-41) are headed.

Moreover, they occur following phonetically realized heads. But structures

without such heads are also possible as the examples in (42) show.

42(a) [ [e [kan hoB'-et-u]] [horii hin-argat-a]]

S N''
Scomp. work-mid-impf. money cm. get-impf.
'[One] who works gets money'.

(b) [Tulluu-n [e [kan barbaad-uJ] [hin-beek-u]]

ST_nom. N''
Scomp warit-impf. neg.know-impf.
'Tulluu does not know what he wants'.

(c) [Tulluu-n (gara [ [e [haad-i-saa itti-deem-t-e]]]]

S P' N'''N'' S
T-nom. to mother-nom-his to-go-f-pf.
deem-el I
go-pf.
(Literally), 'Tulluu went to [the place] to which his mother went',
'Tulluu went to where his mother went'.

The headless NPs with the clausal complement are in a subject position

in (a) and in object positions in (b) and (c).

 230

The empty heads in such constructions may be construed from or

identified by the selectional restriction features of the embedded or main

verb. In (42 a) , for example, the verb /hojrett-/ 'work' is one which

basically selects a [+ HUMAN] noun as its external argument. Hence the

head, of which the clause is a modifying complement, must be one which is

characterized by the feature [+ HUMAN]. On the other hand, /barbaad-/

'want' as a transitive verb subcategorizes arguments which may or may not

have the feature [+ HUMAN], in which case the empty head may remain

indeterminate, which again means that one is free to put any noun consonant

with the other selectional restriction features of the verb. The complement

clauses in such examples may hence be understood as being associated with

some heads at some point in their derivations. In other words, structures

like (42) may have derived from a structure of the type in (43).

43(a) [kan ho'3'-et-u]] [horii hin-argat-.a]]

N''
man-nom. comp. work-mid-impf. money cm-get-impf
'A man who works gets money'.

(b) [Tulluu-n [ [wan [kan barbaad-u]] [hin-beek-u]]]

S V'N'' S
T-nom. thing comp. want-impf. neg-know-impf.
tTulluu does not know the thing he wants'.

Following the base rule which generates restrictive relative clauses

as complements of N'', as discussed in the preceding chapter, structures

of NP5 with such complements may be represented in the manner shown below.

44. N'''

I
N''
 231

Following again the discussion in Chapter Four, the internal structure

of S in (44) is as in (45)

45. N'''

N''

N comp. S (=V''')

nam-ni N''' V''

xnan-nom

I I
Pro ho3'ettu
work-mid-impf.

Anticipating the discussion in the next section, we shall assume

here that the complementizer in this language is base-generated in clause

initial position.

In (45), we know from subcategorization properties that verbs like

/hojett-/ 'work' select certain positions as thematic positions. Such

positions must be filled by arguments at the level of D-structure. The

argument may or may not have a phonetic matrix, depending on the type of

clause. In this particular case, the internal argument position of the

verb /ho'ett-/ 'work' may or may not be filled by the argument /ho3Tii/

'work'. The external argument position is filled by a pronominal element

(Pro). This element may surface with a phonetic matrix, as we shall

observe shortly.
 232

As a pronominal element, Pro may be free in its governing category or

be bound by a c-commanding antecedent NP in accordance with the principles

of binding. In (45), Pro is free in S, its governing category, and bound

in N'', which is another governing category according to the definition

given by Chomsky (1981:188), and also according to that of Manzini (1983).

In either case, Pro is coindexed with /nam-ni/ 'man-nom', for its reference.

Notice that in (45) there is an element which has not been accounted

for. This element is /kan/. It is different from Pto. In the

discussions e have had so far, it has been taken for granted that it is a

complementizer. This is contrary to the traditional account, since in Hodson

and Walker (1922:77), Moreno (1939:178), Gragg (1976:191), etc., this

element has been assumed to be the relative pronouns, though arguments are

not advanced in support of this assumption. If this assumption is correct,

then the structure in (45) would have to be represented in the manner of (47).

47. N''

N'

i
nam-ni e N''' V''
man-nom.
/
Pro

kan
'I
ho'3'ettu
work-mid-impf.
 233

The complementizer here is null as there is no other element in the

clause to be associated with the comp node. This means there is an

asymmetry between this type of clause and those which we have considered

as complements of V 1 in the preceding section. There we have the element

/akka/ 'that' playing the role of a complementizer.

If /kan/ is a relative pronoun, then it must be subject to the case

filter like any other argument. In other words, it must be case-marked

and that case-marking should be morphologically realized for the structure

to be grammatical. But as we can gather from the structures in (47)

above and (48) below, this element is invariably the same without this

causing any degree of ungrammaticality.

48. [ (nam-i-i I (Tulluu-n kan rukkut-e]]] ,(deeraa-da]]

SN'' SS
man-sgl-nom. T-nom. comp. hit-pf. tall-is
'The man whom Tulluu hit is tall'.

In (47) /kan/ occurs in 'subject' position. The position is case-

marked since the clause is finite, but the element does not have the

nominative case-marker I-ni. Even if it had this affix, the structure

would still be ill-formed, in (48), /kan/ seems to occur in object position

as per the lexical requirement of the verb /rukkut-/ 'hit'. The question

that arises from this situation is: why is /kan/ without the nominative

case-marker, whereas all other arguments - variables and pronominals alike -

are characterized by it? This is another problem which the analysis of

/kan/ as a relative pronoun fails to explain (other than the asymmetry

problem mentioned above).

From facts in other languages like English in which there are

differences between relative pronouns in the nominative and accusative or

genitive cases, we may wonder why there are no such distinctions in Oromo

(of course there does not necessarily have to be any, as the two languages

are not related in any way).

 234

Furthermore the fact that relative pronouns in English and many other

languages are wh-words may again induce us to question the status of /kan/,

for it is not a wh-element, nor is it like any other pronominals for that

matter, though it appears to occur in thematic positions and so should

receive case like other arguments. Other pronominals vary according to

the case they are assigned. Observe the following:

49(a) [inn-i obbooleessa-koo-ti]

She_nom. brother-my-is
'He is my brother'.

(b) [Thlluu-n isa rukkut-e]

T-nom. him hit-pf.
'Thlluu hit him'.

50(a) [maal-tu duf-e?]

Swhat_nom. come-pf.
'What came?'

(b) [Tulluu-n maal arg-e?]

ST_nom. what see-pf.
'What did Tulluu see?'

As the structures in (49) and (50) show, there is a formal as well as

a functional difference between the pronominals in subject and object

positions. This may further lead us to question why the 'relative pronoun'

is invariably the same for all cases, and also why it is different in form

from all other pronominals including interrogatives.

Pronominals have lexical content in the sense that they refer to an

entity both in context or in isolation, and the interpretation of structures

in which they occur depends partly on their meaning. /Kan/ is not like

them in this respect either, for it has no reference outside itself, and

the interpretation of structures in which it occurs does not depend on it,

but on the meanings of other lexically significant forms. In fact, in



235

simple clauses it does not appear at all, for structures like those below

are both ungrammatical and uninterpretable as simple clauses.

51(a)* [kan hoolaa bit-el

S ? sheep buy-pf.

(b)* (Tulluu-n kan bit-el

S
T-nom. ? buy-pf.

Notice that the positions where /kan/ occurs in (51) are thematic

positions and may be filled by nominal or pronominal arguments, including

relative pronouns, since they too are arguments. Now, if /kan/ is such

a pronoun, why should such structures be excluded? Other pronominals do

occur in such positions without this causing any ungrammaticality. The

structures in (46-47) are examples of this.

The fact that /kan/ does not occur in simple clauses, parallel to

other pronominals, may also mean that its distrthution is limited to

embedded clauses of the type under consideration. Even in such clauses,

its occurrence is often optional except when the clause is headless or when

it (the clause) occurs preceding its head (cf. 4.2). Consider (52).

52(a) [ [hoolaa-n [ (Tulluu-n bit-eli] lguddaa-da]]

SN'' SS
sheep-nom. T-nom. buy-pf big-is
'[The] sheep which Tulluu bought is big'.

(b) [ (nam-i-i [ [kaleessa duf-e]]] [deeraa-da]]

S N' S
man-sgl-nom. come-pf. tall-is
'The man who came yesterday is tall'.

Such facts may push us to doubt whether /kan/ is a relative pronoun

and to look for an alternative analysis of relative clauses. The following

is an attempt at this.

We have said earlier that the compleinentizer is null, and as a result

there is an asymmetry between relative clauses and akka-clauses. We have

also said that /kan/ as 'a relative pronoun' is in no way similar to other
 236

pronouns or arguments in general. In order to avoid the asymmetry between

relative clauses and akka-clauses, and also to account for the differences

between /kan/ and other pronouns, it is necessary and possible to assume

/kan/ to be the complementizer in relative clauses. Its function, like

that of any other complementizer is to introduce the clause as complement

of N'' or N'''. Its base position may also be assumed to be clause

initial though in surface structures it may appear in other positions.

The fact that it is lexically empty, and also why it is optional, may be

attributed to this fact, that is, to its being a complementizer, rather

than an argument of the type which refers freely and which cannot be omitted

so easily.

This approach gains support from copular constructions of the type we

have considered in relation to the verb /fakkaat-/ 'seem' in the preceding

section. There we have argued that this verb subcategorizes a clause

which is introduced by /kan/. For purposes of comparison, we shall

consider some structures again. (53) below is one such example.

53. (NPe I [kan (altuu-n hoolaa bit-t-e]] [fakkaat-a] 1]

S V' S Sv
ç-nom. sheep buy-f-pf. seem-impf.
'It seems that aaltuu bought a sheep'.

As stated earlier, /fakkaat-/ 'seem' does not select an external

argument. Its subject position is thus empty. But it has an internal

argument which is a clause. This is what we have said earlier and also

what we observe in (53). In the complement clause, the lexical

requirements of the verb /bit-/ 'buy' are satisfied. It requires an object

complement, and accordingly we have the nominal /hoolaa/ 'sheep'. It

also needs an external argument, and again we have aaltuu in that position.

Since the clause is a complement, there is /kan/ playing the role of a

complementizer. We cannot argue here that it (kan) is a relative pronoun

 237

because both the external and the internal arguments of the verb are NPs.

Relativization presupposes that one of the two arguments of a verb or that

of a pre- post- position is a pronoun. The pronoun is coreferential with

the antecedent NP which is the head of the clause. In the light of this,

the lower clause in (53) cannot be a relative clause although it contains

the element /kan/. This supports the claim that /kan/ is not a relative

pronoun.

The question which follows from this concerns what the relative pronoun

would be in Oromo if /kan/ is a complementizer. It is possible that this

language, like mharic (Hailu 1972), modern Hebrew (Borer 1984), or Chinese

(Huang 1982), may have no relative pronouns equivalent to the English 'who'

or 'which'. It may also be the case that the reason why such languages

have no wh-movement in their syntax (Huang 1982) could be related to this fact.

In the base position, where a wh-element would be expected, what we

find in Oromo relative clauses is a personal pronoun which may be Pro. In

his 1976 account, Gragg has observed this, although he still believes that

/kan/ is the usual relative pronoun. If both /kan/ and third-person pronouns

are relative pronouns, then one would expect relative clauses with both /kan/

and the third-person pronoun to be ungrammatical. However, this is not

the case, as (54) below demonstrates.

54. [ (nam-i-i [ (inn-i kaleessa hoolaa kan. bit-el]]

SN'' SS
man-sgl-nom. he-nom. yesterday sheep comp. buy-pf.
[fira Tulluu-ti]]
V relative of-T.-is
(Literally), 'The man, he came yesterday is Tulluu's relative',
'The man who came yesterday is Tulluu's relative'.

(54) is perfectly acceptable although the relative clause contains

both /kan/ and /inn-i/ 'he-nom'. The latter is generated in the position

where we would generate the relative pronoun 'who' in a corresponding



238

structure in English. The former is found in the position preceding

the verb. If /kan/ were a relative pronoun it could not have occurred

in this position, since: (1) the relativised argument is a subject rather

than an object NP, and (2) the subcategorization property of the verb /bit-/

'buy' is already satisfied by the presence of the object argument /hoolaa/

'sheep'. This makes /kan/ an adjunct rather than an argument.

As an adjunct (complernentizer) its base position is clause initial.

The position it assumes in (54) is hence a surface position. The reason

for the change in position will be explained later.

According to this analysis, the NP with its relative clause complement

may have the representation as shown in the tree below at the level of

D-structure.

N' S (details have been avoided)

comp. S

N''' V''

A
ii

N'' V

/ /
namii (kan) (inn-i) hoolaa bit-e
1
 239

In the above representation, the pronominal /inn-i/'he-nom' occurs

as an external argument of the verb /bit-/ 'buy' in S. As a pronoun,

it is free in S in accordance with principle B of the theory of binding

of Chomsky (1981). But /inn-i/ is also like an anaphor since it is bound

for its reference in N''. As an anaphor it satisfies principle A of the

same theory of binding which states that an anaphor must be bound in its

governing category. The governing category here is N''.

Such pronominals may be null at the level of surface structure as

the examples in (52) show. This is possible since the identity of such

pronouns is construable from their antecedents with which they share all

the relevant features and/or from AGR in Inf 1. within their own clauses

when they are free.

Notice that such pronominals unlike /kan/ are characterized by case

affixes which suggests that they are case-assigned. In order to get case,

they must occur in a governed and case-marked position. In (55) /inn-i/

'he' is in the nominative case indicated by the affix I-il. The pronoun

can get this case from AGR in Infi. only if it is generated in the position

it appears in in (55). The other thematic position in (55) is the position

occupied by /hoolaa/ 'sheep'. There is no other thematic position with

which /kan/ could be associated. In other words, both the projection and

the extended projection principles are satisfied which means that /kan/

cannot be analysed as an argument, for there is no argument position left

in the clause. This leaves us with the option of treating it as a

complementizer.

If, on the other hand, we argue that it is a relative pronoun, then

we will have to analyse pronom.inals like /inn-i/ in structures like (55)

in terms other than as arguments. But there does not seem to be any other
 240

way we can explain them other than as arguments. We cannot call them

adjuncts because they have features of case, which suggests that they have

been assigned some case, and since only arguments can receive case, we

have to conclude that they are arguments. This excludes /kan/ from the

possibility of being a relative pronoun and hence also from being an argument.

In languages like English relative pronouns are moved to comp. This

constitutes part of the wh-movement in the language. In Oromo, the

pronominal arguments occur in comp. whenever they arc overt, which means

that there is movement. Consider (56 a) derived from (56 b)

56(a)* [Tulluu-n [ [nam-ia [isa [kan [Fayyiisaa-ri t

S V' N'' S S
T-nom. man-sgl. him comp. F-nom.
arrabs-e]]1] [waam-e]]]
V
insult-pf. call-pf.
'Thlluu called the man whom Fayyiisaa insulted'.

(b) [Tulluu-n [ [nam-ia [kan [Fayyiisaa-n isa

S V' N''
T-nom. man-slg. 5comp 5Fnom. him
arrabs-e] 1] [waani-e]] I
V
insult-pf call-pf.

The movement is obligatory since structures of clauses like (56 b)

where the pronoun is overt in situ are hardly acceptable. Such pronouns

can be overt only if they are in comp. The effect of the movement on the

resulting structure is that the clause becomes appositive rather than

restrictive. This is indicated by a prolonged pause following the head

/namicca/ 'the man'.

When the pronoun is in comp., /kan/ has either to delete or move to

the '.rp of the clause for a structure like (56 a) to be well-formed. The

reason for this seems to be related to the principle which governs the

relation between the moved pronoun and its trace. According to this

principle (the empty category principle - ECP) the trace, which is an

 241

empty category, must be properly governed, whereby 'properly governed'

is meant government by a lexical head or by a c-commanding and coindexed

antecedent. In (56 a) the trace fails to be governed by its antecedent

in comp. because of the presence of /kan/ in this same position. However,

since the trace is governed by the lexical head /arrabs-/ 'insult' there

is no violation of this principle. Hence the ungrainmaticality of such

structures as (56 a) cannot be explained in terms of ECP.

This principle accounts for structures like (57) below where the moved

pronoun is the subject of the relative clause.

57(a)* [ [nam-i'i [inni kan [t kaleessa hoolaa bit-e]]]

SN'' S 1 1
man-sgl-nom. comp. yesterday sheep buy-pf.
(deeraa-da]]
tall-is
'The man who bought sheep yesterday is tall'.

(b) [ [nam-ii [inn-i [ t kaleessa hoolaa bit-e]]J

SN'' S Si
man-sgl.-nom. he-nom. yesterday sheep buy-pf.
(deeraa-da]]
tall-is
'The man who bought sheep yesterday is tall'.

In (57 a) the trace is not properly governed by the antecedent because

of the presence of /kan/ in comp. It cannot be governed by Inf 1. in its

own clause because Infi. is not a proper governor. Proper governors are

N, V, A and NP 10 (51 b) is grammatical because the trace is properly

governed by its antecedent in comp. since there Ikan/ has been deleted.

The c-command relation holding between the antecedent and the trace

may appear clearer if the structure is represented in a tree.

 242

58.

NP7

deeraa-da

namii Coinp. S

jnfl- comp. NP VP

HIA kan t
1
kaleessa
hoolaa bite

From the situations in (56-57) it appears that ECP accounts for cases

of extractions from subject positions only. Structures with extractions

from object positions are not accounted for. This may suggest that the

problem with the ungrammatical structures may not have to do with this

principle at all, for in that case (56 a) would have been grammatical

since there is no violation of the principle, but it is not. Such structures

would be grammatical only if either /kan/ or the pronoun is deleted, or if

/kan/ is nved to the VP of the relative clause.

V. . .
59(a) [ [naxru.cci [inn-i ( t kaleessa hoolaa bit-e]]]
SN'' 5 1 Si
man-sgl .-nom. he-nom. yesterday sheep buy-pf
(deeraa-dal]
tall is
'The man who bought sheep yesterday is tall'.
 9
,. A
•2 ..

59(b) ( [namii [inn-i [ t kaleessa hoolaa bit-e]]J

SN'' 5 Si
man-sgl-nom. he-nom. yesterday sheep buy-pf.
(deeraa-da
tall is
'The man who bought sheep yesterday is tall'.

(c) [ [namii [kan (kaleessa hoolaa bit-el]] [deeraa-da]]

SN'' tall-is.
n3an-sgl-nom.comp. yesterday sheep buy-pf.
'The man who bought a sheep yesterday is tall'.

The situation seems to suggest that the pronoun and/kan/ cannot both

appear in the same position, that is, in comp. The presence of one always

excludes the other. This may lead us to a possible conclusion that what

is involved in all the cases of the ungrammatical structures is not something

that can be attributed to any deep principle such as ECP or binding. It

may be said that the structures are correct as far as such principles are

concerned. Their ill-formedness may hence be explained only in terms of

rules (filters) which operate on the output of movement. One such rule

which is relevant to the problem here is the multiply filled comp. filter

of Chomsky and Lasnik (1977:446). According to this rule, no comp. can

contain more than one overt constituent. In the structures in question,

comp. violates this constraint since it contains both the complementizer

/kan/ and the moved pronoun. In order for such structures to be grammatical,

the complementizer has either to be null (delete) or move to some other

place as shown in (59 b)

Assuming this to be correct, we now move to another point that also

relates to this same movement. Before movement takes place a pronom.inal

shares the features of person, gender and number with its antecedent NP,

the head of the relative clause. There is no matching in the feature of

case. However, when movement takes place and the pronominal lands in

comp., it obligatorily takes the case form of its antecedent and becomes


244

identical with it in all its features. Observe the following, for

example.

60(a) (i) [ nam-i-i [(kan) [Thlluu-ri (isa)

SN'' S S
man-sgl-nom. comp. T-nom. he
arrabs-e] 1] [duf-e]]
insult-pf. come-pf.
'The man whom Tulluu insulted came'.

(ii), ( [nam-i-i !inn-i (Tulluu-n t arrabs-e] 1]

SN'' 1 S 1 S
man-sgl-nom. he-nom. T-nom. insult-pf.
[duf-e]]
come-pf.
'The man whom Tulluu insulted came'.

(iii)* [ [ nam-i- [isa [Tulluu-n t arrabs-e]]]

SN''' S 1 1
man-sgl-riom. him T-nom. insult-pf.
(duf-e]]
come-pf.
'The man whom Tulluu insulted came'.

(b) (i) [Thlluu-n ( [nam-ia ((kan) [(inn-i) hoolaa bit-e]]]

S V'N'' S S
T-nom. man-sgl. comp. he-nom. sheep buy-pf.
[arg-e] 1]
see-pf.
'Tulluu saw the man who bought [a] sheep'.

(ii) [Tulluu-n [ [nam-ia [isa [t hoolaa bit-el]]

S V' N'' S 1 1
T-nom. man-sgl. him sheep buy-pf.
[arg-el]]
see-pf.
'Tulluu saw the man who bought (a] sheep'.

According to the projection and the extended projection principles and

the theta (8) criterion of Chomsky (1981, 1982), the relative clauses in

(60) should have all their argument positions filled by NP arguments.

These requirements are satisfied in (60). Both the external and the internal

argument positions are filled by NPs. These NPs satisfy the case filter
 245

since they occur in case-marked positions and have the feature of case

assigned to them in their own clauses. Hence they do not need to move

for reasons of case. However, as stated above, the pronominal subject or

object may move out of its base position and land in comp. for the reason

stated earlier. And whenever such a movement takes place, the pronominal

acquires the case form of its antecedent NP and thereby becomes identical

with it in all features.

The situation seems to suggest that there is what Andrews (1981:lff),

referring to a similar situation in Icelandic, calls case attraction. And

according to McKee (personal communication) relative clauses in ancient

Greek behave in the same way.

As stated earlier, such pronominals have no irdependent reference.

They are hence bound to their antecedents with which they share the features

of person, number and gender in order to satisfy the matching condition

that should exist between any such coreferential elements. This is not,

however, a condition with regard to the feature of case. The pronominal

and its antecedent may be characterized by different case forms since they

occur in different argument posrtions. However, when the pronominal element

leaves its base position and gets closer to its antecedent by landing in

comp., a complete matching of features takes place between the two elements.

However, if we consider Chomsky's (1981:332) chains and their formations

and properties, the analysis we have offered for the structures in (60) in

terms of case attraction does not seem to be valid. According to Chomsky,

a moved element and its trace form a chain. The chain is an argument chain

(A chain) if the head (antecedent) is an argument and is also in an argument

position. If the antecedent is a non-argument element such as a wh-element

in comp., the chain is an A chain. The antecedent of the trace in (60) is

 246

in coinp. and since comp. is an A position, the chain which the pronoun and

its trace form may be said to be an A chain.

The property of chains is that there should be feature agreement between

the head and its members. The chain in (60) does not seem to satisfy this

condition in full, as the members are characterized by different case

features. The trace in (a ii) is in the accusative case since its

position is governed by the verb, whereas the antecedent with which it forms

the A chain is in the nominative case, in agreement with the antecedent NP.

The situation seems to suggest that whenever a pronominal becomes

identical in features with its antecedent NP, the possibility for the chain

formed by this pronomina]. and its trace ceases to exist.

The fact that a imved pronominal takes the case form of its antecedent

has some parallel in structuresof simple noun phrases. Consider the

following for examples.

61(a) (i) [ [nam-i&?'-i eera-n] [duf-e]]

S N'''
man-sgl-nom. tall-nom. come-pf.
'The tall man came'.

(ii) (Tulluu-n deeraa] [arg-e]]]

N'''
T-nom. man-sgl. tall see-pf.
'Tulluu saw a tall man'.

(b) .(i) [ [fard-i-i kun-i] [aadii-da]]

SN'''
horse-sgl-nom. this-nom. white-is
'This horse is white'.

(ii) (Tulluu-n [ [fard-icca kana] [bit-el]]

$ V'N'''
T-nom. horse-sgl. this buy-pf.
'Thlluu bought this horse'.

In (61), the adjective /deeraa/ 'tall' and the demonstrative element

/kana/ 'this' are characterized by the same case affixes that the head of

the noun phrase is associated with. The process here may be argued to be
 247

one of agreement. The non-head elements in the category, that is, in the

NP, agree with the head in the features of person, number, gender and case

by attracting or copying them from the head. The same argument may be

extended to the situation we saw in (60) since the pronoun seems to function

like the demonstrative /kana/ 'this' in (61), when it moves to comp.

A question which is likely to arise here concerns the landing site of

the moved pronominal in (60). It has been argued that the movement is to

the position of comp. and since this position is within S, that is, within

the relative clause, it is difficult to assume that the moved pronominal

element gets the case feature of the antecedent by agreement of the type

we saw in the NPs, in the structures in (61). In order for us to argue

in terms of agreement, the pronoun should move out of the S and land

somewhere in the NP. In other words, the movement should be as shown

below.

62. N''

N"S

nami-i
man-sgl-nom

arrabs -
insult
 248

If we assume that the movement is adjunction to N' or N'', we may

be able to argue in terms of agreement. The pronominal may be said to

have agreed with the head /nam-i-i/ 'the man' which is nominative.

But this would lead to a more serious theoretical question concerning the

motivation of the movement and the extraction and landing sites of the

element. The element must move to an empty position (A position) or it

must be adjoined to a maximal category, that is, to an A position. Since

N' is not a maximal category, /isaa/ 'him' cannot be adjoined to it without

violating the projection principle of Chomsky (1981). Hence the movement

must be adjunction to comp., although this also has its own problem.

The problem with this type of adjunction is connected with the nature

of the moved pronoun. According to the data presented and the analysis

proposed, Inn-u or /isaa/ he' is an argument. If this is the case,

then it cannot move to comp. in (60), for according to Chomsky (1981 :115),

'move a can move a to comp. only if •a contains the feature wi'. In other

words, only wh-eleiaents can move to comp. But /inn-i/ (or /isaa/ in

(60) is not a wh-element, and as stated earlier, Oromo does not have

wh-movement in the syntax. We may, hence, say that Chomsky' s condition

should be limited to languages like English which are characterized by

both NP and wh-movements.

However, if we consider the trace left by the moved pronoun, it looks

like a wh-trace since it is case-marked and its antecedent is in an A

position. If it had been an NP trace, it would not have had case, and

the antecedent would have been in an argument (A) position. From this

situation we may argue that the movement of the pronoun /inn-i/ or /isaa/

'he' within the structures in question is a kind of wh-movement, although

the pronoun is not an actual wh-element. If so, the relation between

 249

the trace and the pronoun could be that of a variable and an A (non-

argument) binder. The problem with this kind of argument consists in the

chain having to be characterized by two cases.

A possible way out of this problem is to assume that in the structures

in question what we have is not a pronoun moved to the position of the

complementizer, but a pronoun which is base-generated in topic position,

that is as sister of S. This pronoun takes the case form of the head

of the relative clause by copying rules. The relativized position may

also be assumed to be filled by Pro. Since there is no movement, there

will not be any chain involving the moved element and the trace left behind.

This alternative may appear neat and economical, and may hence be adopted

temporarily. Its strength, however, will have to be tested in the light

of other structures involving topicalizations.

5.3 The Position of the Complementizer

In the foregoing discussion it has been assumed that elements like

/akka/ 'that' and /kan/ function as complementizers. Their position in

relation to the rest of the clause has also been assumed to be initial.

In this section we shall examine the syntactic and semantic properties of

these and other similar elements, and in order to justify the proposal that

their base position is clause initial, we shall consider two other positions

where such elements also appear in surface structures and see why such

positions cannot be base positions.

Before we go into that, it is necessary to raise a few points about

what complementizers are and how they differ from other elements in clauses.

According to Bresnan (1979:6), complementizers are clause-particles found

in predicate complementation and in comparative and relative clause

 250

constructions. Their function is to distinguish clause types. Such

elements include 'that', 'for', 'than', 'as', arid 'wh' or 'Q' in English.

They distinguish main clauses from subordinate clauses and statements from

questions.

For Oromo we have noted the elements /akka/ 'that' and /kan/ occurring

in predicate complement and relative clauses respectively. These are the

clauses we have discussed throughout this chapter to be complements of verbs

and nominals. Other types of clauses with which complementizers may be

associated include independent clauses of various mood types. We can, for

example, distinguish clauses in indicative mood from clauses in imperative

or optative moods. Compare the following.

63 Ca) [Tulluu-n hoolaa hin-bit-e]

ST_nom. sheep cm... buy-pf.
'Tulluu bought a sheep'.

(b) [Tulluu-n hoolaa hoo-bit-u gaarii-da]

ST_nom. sheep cm... buy-pf.good-is
'It is good if Tulluu buys [a] sheep'.

(c) hoolaa bit-li

sheep buy-imp.
'Buy [a] sheep'.

In (63 a) that the mood is indicative is indicated by the pre-verbal

element /hin-/. As stated in Chapter One, this element occurs only in

declarative independent clauses. It does not occur in subordinate clauses

or in interrogative clauses. The structures in (64) are illustrative of

this.

64(a)* [Tulluu-n !ka Fayyiisaa-n hoolaa hin-bit-el beek-a]

ST_nom. Sthat F-nom. sheep cm. buy-pf. know-impf.
'Tulluu knows that Fayiisaa bought a sheep'.
 251

64(b) (i)?? [Thlluu-n hoolaa hin-bit-ee-ree!?]

ST_nom. S
sheep cm.buy-pf. Q
'Did Tulluu buy a sheep?!'

(ii)?? [eeMuu-tu hin-duf-e?]

Swho_js_jt cm.. come-pf.
'Who came?'

These structures would be perfectly acceptable if they contained no

/hin/. This same element is also excluded from structures of the type

in (63 b-c) above.

In negatives /hin/ occurs but with a different tone, as the examples

in (65) demonstrate.

65(a) [Thlluu-n farda hin-bit-e]

ST_nom. horse cm. buy-pf.
'Tulluu bought a horse'.
/
(b) Thlluu-n farda hin-bin-n-e 1< hin-bit-n-e/
T-nom. horse neg-buy-neg-pf.
'Thlluu did not buy a horse'.
/
The indicative affirmative mood-marker /hin-/, but not the negative

marker /hin-/, may be reduced to 0 or I-ni. When the latter takes place, i-ni

may be encliticized on to the element before it. Hence corresponding

to (65 a-b) (66 a) but not (b) is possible.

66(a) (Tulluu-n boor-n adeem-a]

ST_nom. tomorrow-cm. go-impf.
'Tulluu will go tomorrow' (Gragg:188).

(b)* (Tulluu-n farda-ri bin-n-e]

ST_nom. horse-neg. buy-neg-pf.
'Tulluu did not buy [a] horse'.

Now, if complementizers are elements indicating clause types, then

the class should include not only /akka/ and /kan/, but also all the other

elements we have considered in relation to the various mood types since mood

is also one feature which distinguishes one clause type from another.


252

The next question we need to address ourselves to is whether

complementizers belong to the class of subordinative conjunctions or

form an independent class of their own. In order to answer this question

we need to see the syntax and semantics of complementizers vis^vis

other subordinative elements. As stated earlier, elements like /akka/

'that' and /kan/ occur with clauses which occur as argument complements

of verbals or as modifying complements of nominals. In other words, they

introduce clauses which occur in a strictly subcategorized argument position.

In this regard, they are different from subordinative conjunctions since

the latter are associated with clauses which have adverbial functions. Such

clauses are complements of VPs (V'') and their position is outside the main

predicate structure CV').

Complementizers show that the clause which they introduce is definite

in terms of mood or in terms of the feature tense/aspect. In this sense

they may be considered as definitizing or determining elements (cf. Bresnan

1979). Hence they differ from subordinative conjurictives which as stated

above indicate adverbial clauses relating to the time, manner, reason, etc.

of the action designated by a verbal head. Let us compare the following

structures.

67(a) [ [akka [Tulluu-n deem-e]] [Fayyiisaa-n t duf-e]]

SS S 1
as T-nom. go-pf. F-nom. come-pf.
'As Thlluu went, Fayyiisaa came'.

(b) [Fayyiisaa-n [ [akka [Tulluu-n deem-eu [daga?-e]]]

S V'S S
F-nom. that T-nom. go-pf. hear-pf.
'Thlluu heard that Fayyiisaa went'.

In (67) the same clause /Thlluu-n deem-e/ 'Thlluu went' functions as

an adverb of time in (a) and as a nominal complement of the verb /daga?/

'hear' in (b). The functional distinction is due to the difference in the

 253

meaning of the two homophonous elements. The meaning of /akka/ in (b)

is that the clause which it introduces is a tensed nominal clause occurring

in a position which is strictly subcategorized by the matrix verb. The

same cannot be said about /akka/ in (a) because here the element shows that

the clause is used in a temporal reference to the action denoted by the main

verb. In short, the element changes the clause into an adverbial expression.

As shown in Chapter Four, such clauses are modifiers of V'' and are different

from the argument complements of V'. Such distinctions suggest that

complementizers ought to be treated as either a subclass within the category

of adpositionals or as an independent class of their own. The fact that

there are mood-marking elements which are associated with main clauses may

give support to the view that complementizers should be recognized as

forming an independent class. This is the position taken in Bresnan (1979)

and Chomsky (1981). According to them, all clauses may be treated as

being Ss with (comp)lementizers branching down as a right or left sister of S.

As claimed earlier, the complementizers in Oromo occur basically in

clause initial position. Before we go into the details of substantiating

this claim, let us obseEve the other positions where complementizers and

other mood-indicating elements occur in surface structures. Such structures

include the following.

68(a) [Thlluu-n [akka [Fayyiisaa-n hoolaa gurgur-e]] (daga?-e]]

S S S
T-nom. that F-nom. sheep sell-pf. hear-pf.
'Thlluu heard that Fayyiisaa sold [a] sheep'.

(b) [Thlluu-n ! (Fayyiisaa-n akka hoolaa gurgur-el] [aga?-e]]

S sS
T-nom. F-nom. that sheep sell-pf. hear-pf.

(c) [Tulluu [ [Fayyiisaa-n hoolaa akka gurgur-e]] [aga?-e]]

S SS
T-nom. F-nom. sheep that sell-pf. hear-pf.

(d)* (Thlluu-n [ [Fayyiisaa-n hoolaa gurgur-e akkall [aga?_e]]

S SS
T-nom. F-nom. sheep sell-pf. ithat hear-pf.
 254

As (68) shows /akka/ 'that' can occur in any position preceding the

verb /gurgur-/ 'sell', but not in the position following it. This is also

the case with /kan/ in relative clauses.

69(a) [ [nam-i-i [kan [Tulluu-n arg-e]1] (deeraa-da]]

SN''' S S
man-sgl-nom. comp. T-nom. see-pf. tall-is
'The man Tulluu saw is tall'.

(b) I [nain-i-i ! [Tulluu-ri kan arg-e]]] [deeraa.da}]

S N''' S S_01
man-s gl-nom. comp. see-pf. tall-is
(c)* [ [nam-i3-i [ [Tulluu-n arg-e kari]]] [deeraada]]
SN''' SS
man-sgl-nom. T-nom. see-pf. comp. tall-is

In infinitival (non-finite) clauses the element /-f/ 'for' which might

be argued to be the complementizer occurs in clause final position only, as

the structures in (70) demonstrate.

70(a) [Tulluu-n I Eaanrian dug-uu-(f(i)] [hin-barbaad-a]]]

ST_nom. milk drink-to cm. want-impf.
(Literally), 'Tulluu wants for to drink milk',
'Tulluu wants to drink milk'.

(b) * [Tulluu-n [ [aannan-f dug-uul [hin-barbaad-a]]]

ST_nom. Smilk for drink-to cm. want-impf.

Regarding the elements which show various types of mood, we need to

make distinctions between those which are morphological and those which are
,
syntactic. The element /hin-/ which shows indicative mood may be treated

as being part of the inflectional morphology of the verb. This is evident

from the fact that it occurs as a prefix attached to main verbs in clauses.

The examples in (71) are illustrative of this.

71(a) [Tulluu-n hoolaa hin-bit-al

ST_nom. sheep cm.- buy-impf.
'Tulluu will buy a sheep'.

(b)* [Tulluu-n hin hoolaa bit-a]

ST_nom. cm. sheep buy-impf.

(c)* [ hin Tulluu-n hoolaa bit-a]

S
cm. T-nom. sheep buy-impf.
 255

The same may be said about the imperative mood markers I-u '\ -uu/

since they too occur as suffixes of main verbs.

72(a) deem-u!
go-imp.
Go!

(b) koott-uu!
come-imp.
Come!

Of those which are syntactic, /haa/ behaves in a manner similar to

that of /akka/ or /kan/ as (73) below demonstrates.

73(a) [haa Tulluu-n deem-uu]

Siet T-nom-nom. go-impf.
'Let Tulluu go'.

(b) jTulluu-n haa deem-uuj

ST_nom. let go-impf.

(c)* [Tulluu-n deem-uu haaj

ST_nom. go-impf. let

In interrogative clauses there is always an optional question particle

occurring in final position. The fact that the speaker is asking a question

but not making a statement is also noticeable from the intonation final

contour of the utterance. Consider (74).

74(a) !Tulluu-n hoolaa bit-e-ree?]

ST_nom. sheep buy-pf-Q!
'Did Tulluu buy [a] sheep?!'

(b) [Tulluu-ri hoolaa bit-ee+]

ST_nom. sheep buy-pf.
'Did Tulluu buy [a] sheep?'

Now, going back to the question of where complementizers are found in

D-structure, we may take both initial and final positions as possible sources.

Though there are instances where a complementizer may occur preceding the

verb or its complement, as in the examples in (68 b-c) , it does not seem
 25(

possible to argue for a clause internal base position. All positions in

the projection line of verbs are argument positions. Such positions must

be filled by arguments such as NPs, PP5, etc., but not by complementizers

since the latter are not arguments; they are adjuncts and as such they

must be generated in adjunct positions. One such position is the node

which branches down from S.

A clause internal position might be possible only if we assume that

complementizers are particles attached to verbs as (68-69(c)). This has

some parallel in structures of simple declarative clauses, where the mood

marker /hin-/ appears as a prefix of the verb. And this seems plausible

given the fact that Oromo does not have cyclic movement which would otherwise

necessitate a clause initial comp. position for intermediate traces, as in

successive wh-movements in languages like English, for example. But

whereas particles are tightly bound to some category in a clause, the

complementizers /akka/ 'that' and /kan/ are not. They move freely towards

the VP or the V of their clause. The downward movement, which is less

permissible compared to an upward movement, may be avoided if these elements

are treated only as verbal particles, optionally moving to clause initial

positions.

Without denying the plausibility of this line of argument, I shall

make the choice between the initial and the final positions for the purpose

of this study. Since complementizers can occur in both these positions,

the choice of one over the other may be made arbitrarily. However, if we

assume that complementizers are heads of S, then the fact that other

categories including Ss are head-final (cf. Chapter Four) may induce us to

favour the clause-final analysis. But taking this position as basic and

other positions as derived would mean that the complementizers /akka/ and
 257

/kan/ have to undergo an obligatory movement rule, since such structures

as (68 d) and (69 c) where they appear in final position are ill-formed.

This is not appealing from the point of view of economy.

The clause-initial analysis does not involve any such movement. The

question particle I-reel, and the element /-f/ 'for' which we have considered

as a complementizer of non-finite claus are both optional. That a clause

is a question can always be deduced from the intonation. If it were the

case that Oromo used wh-words to introduce such clauses and that the

position of such words were in clause-final position, then the clause-final

analysis would have been preferred. But it does not use wh-elements as

question words. Such forms are used as wh-NPs. Observe the following,

for example.

75(a) [Tulluu-n maal bit-el?

S
T-nom. what buy-pf.
'What did Tulluu buy?'

(b) [eeu-ttu hoolaa bit-e[?

Swho_js_jt sheep buy-pf.
'Who is it that bought a sheep?'

In short, there is no obligatory movement with the clause initial

analysis. It may even be argued that the element i-f,! 'for' is a

postpositional clitic and that the infinitival clause in Oromo may be said

to have no complementizer. In other words, the comp. node in such clauses

is not lexically filled. If the complementizer is 0 and if the question

particle /-ree/ is optional, then the only obligatory elements in the

position of comp. would be /akka/ 'that' and /kan/ 'of(?), and since these

occur only in the position as shown in the grammatical structures of

(68-69), the clause-initial position appears attractive. Further support

 258

comes from movement. As shown earlier, pronominals like /isaa/ 'him'

move to this position, which means the comp. node can serve as a landing

site only if it is in this position.

If the arguments forwarded in favour of a clause-initial comp. are

sound, then the categorial rule that expands clauses in general should be

as follows:

76. S > comp. S

Before we wind up this section, we need to mention the Inf 1. node

and make a few points about its position in relation to the type of

categorial rules we have discussed in Chapters Three and Four. In surface

structures, aspectual and agreement features appear as affixes of verbs.

These features are believed to have the role of assigning case to the

external arguments of verbs. According to Chomsky (1981:170) nominative

case is assigned to the subject (external argument) of a verb under

government. This means that AGR or ASP should be in a position from which

it can govern the position of subject NPs. As affixes attached to verbs,

they cannot govern this position in the sense of the definition we had for

government. Hence AGR/ASP should appear as realizations of an abstract

category Inf 1. This category branches down as head of a clause(S) from

the same node that dominates the subject argument. In other words, we

have a categorial rule of the type 1n(77).

77. S > NP VP INFL.

But notice that (77) violates a constraint of the X-bar theory which

dictates that in the expansion of a maximal X", the head must be in the

same category as X but with one less bar. This is not the case in (77)

for there is no such categorial similarity between the maximal category S

and any of its daughter nodes.

 259

If we assume with Jackendoff (1977:47) that S and S are instances of

V''', again violating in part the above-mentioned constraint, since that

would also mean a category dominating itself, (77) would have the form

in (78)

78. v''1 (=S)

comp. V''' (=S)

NP V'' INFL

Although (78) appears to have solved the problem (77) faced (though

still only in part) it is not itself free from other problems. According

to Jackendoff (1977), and the discussions we have had in the previous two

chapters, v u ' is the head of v'' 1 . The subject NP has been argued to be

a specifier. The complements of v'' are some sentence adverbials. Infl.

(= MIX in his terminology) is simply regarded as a daughter of V'' 1 and it

dominates (T)ense and (M)odal (p.50). Tense is believed to be part of

the V''' specifier and Modal part of the V'' specifier, which means

implicitly that AUX, from which both elements branch down, is a specifier too.

If this is the case, then nominative case cannot be assigned to the

subject NP by AUX ( Inf 1) because in order to assign case, the case-assigner

must be a head. In (78) it is V' 1 rather than Inf 1. ( =AUX) which is the head

of V''' 1 which means that v'' should assign case to the NP in subject

position. This is contrary to the theory of case developed by Chomsky.

According to his theory, only lexical (minimal) heads are case-assigners.

V' 1 is a maximal category, and as such its role is to 0-mark the position

where subject arguments are expected. As stated above, only Infi. or

AGR/ASP in Inf 1. assigns case to the NP in such positions.

 260

One way out of this problem is to dispense with the idea of

nominative case assignment under government as hinted at earlier, and

assume, instead, that case is a concomitant of agreement between the subject

NP and the verb. In this way, we can also do away with the Infl.node

altogether (cf. Bouchard l984).

This may appear simple and nice. But it leads to a situation where

we lack generalizations. It may work only in affirmative clauses, in which

the verb displays features of agreement. In negative structures of the

perfect type, such features are not available in Oromo verbs (as discussed

earlier in this chapter), which means that the subject argument cannot

receive case even under agreement. But the position is case-marked and

hence a lexical subject is licensed to occur. Consider (79) below.

79. [Tulluu-n farda hin-bin-n-e]

ST_nom. horse neg-buy-neg-pf.
'Tulluu did not buy [a] horse'.

The situation seems to force us to believe that in such cases,

nominative case is assigned as a result of the presence of the feature

ASP. But ASP is not a nominal feature like person, gender and number.

Rather it is a verbal feature. Hence it is difficult to believe that

case is assigned as a result of agreement for there is nothing in the verb

in (79) which relates to the subject, Tulluu. This means we have to

reconsider the assumption that nominative case is a result of agreement,

and look for another alternative.

We may maintain the claim that there is an Inf 1. node carrying either

or both features of AGR and ASP, and that this node governs the subject

position and assigns case to the subject argument. In order to do this,

however, Infl. has to be treated as the head of S in (78). For this

reason, we have to disagree with Jackendoff about his treatment of both S

 261

and S as being part of the projection of V or as instances of V' s, as

stated earlier. Instead, we may follow Hornstein (1978:137), Chomsky

(1981) and Hoekstra (1984:72) in assuming that the projection of V ends

at V''('), and that S is an independent major category.

Extending the notion of head to clauses, we may treat Inf 1. as being

the head of S. If so, then Inf 1. may assign nominative case to the

subject NP under government. In other words, we have a structural

configuration of the following type:

80. S

comp. S

N''' V't Infi.

Notice, however, that (80) violates the same constraint (78) did above,

since both Inf 1. and comp. fail to have the same category label as their

dominating categories. Furthermore, the subject NP and Inf 1. are not

adjacent to each other as required by the adjacency condition on case

assignment. 9 In order to avoid this problem we may extend the use of

bar notations to both Inf 1. and comp. as in Chomsky (forthcoming) to a

double-bar level, parallel to the categories of adjectivals and

adpositionals, which, as we have discussed, are also characterized by

two bar levels. Accordingly (80) has to be modified to (81).

 262

81. (C'')omp.

Spec. C

C (l'')nfl.

[3 //
N'' 1'

V'' 1 Infi.

AGR/ASP

This analysis differs from that of Jackendoff (1977), since: (1) it makes

a distinction between V''' and 5, and (2) it treats V''' and N''' as

independent categories instead of as single constituents with a head and

specifier relationship. His analysis fails to hold in cases when certain

syntactic rules such as gapping and postposing operate on NPs and VPs.

In the case of gapping, for example, the rule deletes only the VP leaving

the NP, its specifier, intact. If it were the case that the two form a

unified constituent, the rule would have deleted the NP as well and if it

had done that, the structure would have meant something entire different

from the intended interpretation. Consider (82) for example.

82(a) [Tulluu-n hoolaa bit-el. [Fayyiisaa-n [hoolaa bit-el]

ST_nom. sheep buy-pf. SF_nom. sheep buy-pf.
'Tulluu bought [a] sheep. Fayyiisaa bought [a] sheep'.

(b) [Tulluu-n hoolaa

bit-el . [Fayyiisaa-n-iss[ I]
T-nom. sheep buy-pf. F-riom. too
'Tulluu bought [a] sheep. So did Fayyiisaa'.

The same is true in VP preposing or NP postposing. In either process,

the movement of one does not include the other. In other words, NP

postposing moves the NP only leaving the VP intact. On the other hand,
 263

the same rules delete or move the specifiers of NPs along with their heads.

If we, therefore, accepted Jackendoff's analysis as valid, we would have

to explain the reason why the same rule operated only on the head in one

category, and on both the head and its specifier as a unit in another

category. In other words, we would have a case where a rule failed to

apply across categories as required by the X-bar convention itself.

The problem may not arise if we distinguish N''' and V''' and treat

them as independent maximal categories related only in thematic terms, as

arguments and predicates, in other words, if we follow Chonisky's analysis

as indicated in the tree structure in (81). One problem which we may have

to face by following Chomsky's analysis is that the subject NP in (81) is

riot governed by Infl.(1), as required by the definition of government of

Chomsky (1981). The governing category here is the intermediate Inf 1. (1')

since only it is adjacent to the NP. This problem seems to have been

noticed for distinctions have recently been made by Hoekstra (1984:78),

for example, between structural and e-governors. Inf 1. belongs to the

former whereas t-N] categories belong to the latter. The relation between

Inf 1. and the subject NP is a simple structural configuration, whereas

the relation between, say, a verb arid its strictly subcategorized argument

is a relation of theme-assigning and theme-receiving elements as well.

The verb has the lexical property of assigning thematic roles to its

internal argument. This involves the notion of government by a lexical

head. The relation between Inf 1. and the subject NP on the other hand,

is not based on the assignment of thematic roles, hence Inf 1. may not

have to be a lexical (minimal) governor.

 261

5.4 Summary

In this chapter we have been concerned with the internal structures

of two types of clausal complements. These are the modifying complements

of riominals and the argument complement of verbals.

Like any other clause, such clauses may be tensed or non-tensed.

The former are characterized by lexical subjects whereas the latter are

associated with an abstract pronominal element. We have, however, noticed

instances of non-tensed clauses with lexical subjects. In such cases,

it has been argued that the Inf 1. node has the feature AGR. Since this

feature is a case-assigner, the subject NP can receive case in order to

escape the case filter. Because of this we have recognized two types of

non-tensed clauses: those with and those without the feature AGR. We

may call them finite and non-finite infinitival clauses. The finite

infinitival lacks the feature (ASP)ect, which is understandable given the

fact that aspect is a verbal feature.

In both tensed and finite infinitival clauses, the subject can be

dropped since it is always identifiable from the feature AGR in Inf 1.

However, such is not the case in structures of perfect negatives. In such

clauses, AGR is not available in Inf 1., though a lexical subject is possible,

which suggests that in such structures only ASP may assign case to the

subject NP, but this NP cannot be dropped. In the light of this situation,

we have argued that the Inf 1. node should be recognized as having or

lacking both or one or none of these features. This determines the type

of NP a clause may have as its subject.

The akka-clause is limited to object positions, whereas the infinitival

clauses can occur in all positions associated with nouns. This asymmetry
 265

seems to have to do with case and its morphological realizations. Any

NP in subject position should be characterized by the nominative case-

marker I-ni, in other words, it should be case-marked. Akka-clauses

appear to be either case-resistant or that their verbal heads fail to show

case morphologically; case affixes, like other nominal affixes being

exclusively features of the category nominal, and not of verbal. On the

other hand, the heads of infinitival clauses are nominal and as such

categories are case-recipient, they can occur with the nominative marker

/-n/ in any case-marked position.

Regarding adnominal clauses, we have argued that such clauses do not

have relative pronouns. The role of such pronouns is assumed by

independent personal pronouns. What has often been treated as a relative

pronoun, that is, the element /kan/, is analysable as a complementizer.

As in languages like English where wh-elements in relative and

interrogative clauses move from their base positions to the position of

comp., the personal pronouns in Oromo relative clauses also move to the

position of the complementizer. However, such movement is optional; but

whenever it takes place, the moved pronoun attracts the case form of its

antecedent, which is the head of the relative clause; it also fails to

observe one of the conditions of chains-case uniqueness.

In both tensed and infinitival clauses NP movement is possible. But

the type of movement is again optional not only in adj unction but also in

substitution. Whenever a category is moved, it observes subjacency and

the ECP (empty category principle). But there is one problem with such

movements as raising to subject, with regard to case. The trace is case-

marked since the movement is out of a governed position. That is why it

is optional. We have noticed that in such cases the chain, which is formed
 266

by the moved element and its trace is associated with two possible sources

of case. Furthermore, the trace is both free and bound in the sense of

binding relations. This is problematic.

Regarding the notion and position of compelmentizers, we have made

distinctions between subordinate clauses of various types, of adverbial

functions and of subordinate clauses of nominal function. The elements

which introduce the latter types of clauses have been argued to be

different from those associated with the former. Hence a category

complementizer has been recognized. Its function is to introduce such

clauses as finite or non-f inte or as declarative as opposed to interrogative.

Concerning the position they occur in in D-structure, we have argued

in favour of a clause-initial position. The elements which occur in

clause-final position are optional, whereas those in initial position are

obligatory. A clause-internal position does not seem plausible given

the fact that all positions within a clause are argument positions and that

only arguments such as NPs and PPs can occur in such positions. Since

complementizers are not arguments, they cannot occur in any such positions

without this leading to a violation of some other principles. It might,

perhaps, be argued that they may occur as adjuncts adjoined to arguments.

And this may appear quite possible given the fact that in surface

structure such elements do appear in clause-internal positions adjoined to

either the verbal head or to any of its complements. But this would lead

to another problem, namely that movement rules of the type seen in

relative clauses would have no landing site for the extracted category.

In other words, comp. would fail to attract moved categories if it were

clause-internal. Furthermore, preposing of categories to which comp. is

adjoined would not take place, since the resulting structure would be
 267

ill-formed. Hence, pending the discovery of facts to the contrary, a

clause-initial comp. has been adopted.

Finally, we have considered the abstract category of Inf 1. and argued

that it constitutes the head of S. As a head, it governs the subject

position when it has the feature AGR or ASP. In doing this, we have

argued against Jackendoff's claim that both S and S-bar are extensions of

the projection of verbs. This would cause problems in the assignment of

case and also in the application of certain syntactic operations. We have,

hence, made distinctions between the projections of V on the one hand, and

that of Inf 1. on the other. Similarly we have recognized NPs and VPs

as independent maximal projections of their lexical heads, N and V, and

that their relationship in predicate structures is that of argument and

predicate as opposed to head and specifier.

Regarding the contents of Inf 1., it has been observed that in case-

assignment, it is not only AGR which assigns nominative case to the subject

of a clause. We have noticed that in perfect negatives AGR is blocked and

that nominative case is assigned by ASP. In infinitival clauses,

nominative case can be assigned by AGR alone. On account of this

situation, we have made the suggestion that Infl. should be treated as a

complex of features, and that what determines the case of the subject NP

in a clause is the presence or absence of one of the two features, namely

[ASPI and [AGRI.

 268

NOTES TO CHAPTER FIVE

1. This excludes the modifying complements of nominals and adjectivals.

2. These are positions of adjunctions such as (comp)lementizer and

topic.

3. According to Chomsky (1981:101), the 0-criterion 'requires that

each argument bear one and only one 0-role and that each 0-role
be filled by one argument...'

4. The definition of government according to Aoun and Sportiche (1983)

is as follows: a governs $ if f they share all the same maximal
projections. a is a governor if a is X in the X-bar system, i.e.,
lexical catetory, [+TENSE} and [+AGR}.

5. A governing category, according to Chomsky (1981:211) is a category

containing a, a governor of a and a SUBJECT accessible to a.
a is accessible to B iff B is in the c-command domain of a and
assignment to $ of the index a would not violate the i-within-i
constraint (Chomsky 1981:212). i-within-i is formulated in
Chomsky (ibid.) as --- S ---1. The idea here is that ô( = an
anaphor) in the domain of y (= NP subject) cannot be coindexed with
AGR since AGR is already coindexed with its y.

6. a c-commands $ if f the first branching node dominating a also dominates

$, and a does not dominate B nor$ a (Radford 1981:314).

7. y is a domain-governing category for aiff (a) y is the minimal category

with a subject containing the c-domain of a and a governor for the
c-domain of a and (b) y contains a subject accessible to a (Manzini
1983:433). A c-domain (constituent domain) is the S containing PRO.

8. This referss to akka- and f-clauses only.

9. Travis (1984) argues that this condition does not hold for SOy
languages. The condition works for the assignment of other cases
within VP.

10. Inf 1. is not a lexical governor for English. The reason is that
it is not rich enough to identify the empty category. In Oromo
Infi. may be a lexical governor since it is rich enough to let us
know the content of the empty category.
 26

CHAPTER SIX

SPECIFIERS

6.0 Introduction

In the last three chapters we have been dealing with the various types

of complements of the major lexical categories at every level of their

projection lines. In this chapter, we shall take up the class of specifiers

associated with each category and give a syntactic-semantic account of them

in the light of the discussion we had on complements.

Before we go into that, however, a definition of some sort is in order

for the term specifier. Chomsky (1972) considers specifiers as syntactic

categories, that is, as categories with the potential for maximal projections,

occurring preceding a head in constituent structures in English. Jackendoff

(1977:37) argues that there is no evidence that specifiers function as

syntactic categories since they do not move or delete as units and, unlike

normal constituents, no part can be designated as a head'. Both

Jackendoff and Chomsky use the term as a cover name for all the material

preceding the head of a constituent structure. In this sense it might be

said that the term is used as a convenient device to refer to material in

pre-head positions, without this having any theoretical import.

According to claim (ii) of the X-bar convention, which we have stated

in Chapter One (1.2.2), syntactic categories are expressions of lexical

categories. In this sense, specifiers cannot be syntactic categories as

assumed by Chomsky, since there is no lexical category-specifier to which

all the material which can occur before a head of a constituent uniformly

belong, arid which like other categories, develop into maximal projections.
 270

This suggests that the term simply refers to concatenations of elements

occurring in the position stated above for English, as Jackendoff argues.

As we have stated in Chapters Three and Four, Oromo is not uniformly

headed. The category [+N-V] is left-headed, whereas all other categories

are right-headed. This means a definition of the type given for English

which is based on the notion 'preceding the head' cannot be fully adopted

to apply for Oromo since (1) such material follows the head in the category

[+N-VJ and precede it in all others, and (2) complements, like specifiers,

also occur preceding or following their heads, and, hence distinctions

between what is a specifier and what is not, may not be easily made in

terms of the definition given above.

What we may perhaps say about Oromo is that in all categories specifiers

assume peripheral positions, and as stated in the chapters on complements,

complements occur in between them and the heads. A possible definition

in the light of this situation is therefore one which runs: material which

occurs in adjunct positions following complements in noun phrases, and

preceding complements of heads in constituents of all other categories and

their derivatives. This is a distributional definition. A functional

definition would include a statement that specifiers define the heads of

the constituents in which they occur in terms of their entity (what they are)

and/or quantity or intensity. This involves the use of articles,

quantifiers and intensifiers respectively. Excepting some quantifying

expressions, the rest are characterized by their lacking the potential

for maximal projections of the type discussed in connection with the major

lexical categories. In other words, specifiers do not occur as heads

having other categories as their complements. This distinguishes them

from complements. The latter have this potential as discussed throughout

Chapter Four, and also occur in argument (A) positions.



27

6.1.1.1.1 Definite Articles

These include deictics and genitive NPs. Both refer to what may

be called in pragmatic terms the 'old information' in discourse. The

following are structures with such elements.

1(a) [ [fard-i-iJ [duf-eji

S NP
horse-sgl-nom. come-pf.
'The horse came'.

(b) [ [fard-i-i kun(-i)] (gurraaa.]J

SNP
horse-sgl-nom. this-nom. black-is
'This horse is black'.

(c) [ [fard-i-kooj [adii-da]]

S NPho white-is
'My horse is white'.

(d) [ [fard-i kam-(it)} [du?-eJ]?

SNP
horse-nom. which-is it die-pf.
'Which horse died?'

The NPs in (1) are normally used in non-discourse initial contexts,

that is, after the head has been introduced into discourse and become part

of the shared knowledge of the participants. This pragmatic process may

be grammaticalized either morphologically as in (1 a) or syntactically

as in (1 b-d)

In (1 a) /-i-/ is an inflectional affix. The noun inflects for

the feature singulative and/or determinative, in (1 b-'d) the head is

made definite by the element following it, which would be base-generated

in situ. The fact that /-koo/ 'my' (1 c) is attached to /fard-i/

'horse-norn' may make it look like an affix, in which case it may be treated

as an inflectional element like /-i'-/ - probably showing genitive case.

But such an approach would lead to a problem, since the noun is already in

the nominative case indicated by the affix /-i/, and according to Chomsky
 271

On the discussions that follow, it is anticipated that all the major

categories have certain elements which meet the definitions given in both

Chomsky and Jackendoff and adopted here with the qualifications as noted.

Nominals select deictics and quantifiers, whereas all other

categories take intensifiers and quantifiers of some sort. It will be

noted that both intensifiers and quantifiers belong to the same level of

hierarchy across all categories. Furthermore, in nominals and verbals,

unlike in the other two major categories, the treble-bar projection, which

we have argued for in the sections on complements, will be confirmed here

by the type of specifiers occurring at each level.

It will also be argued that what have been called demonstratives,

interrogatives, and indefinite pronouns (cf. Gragg 1976:178) may be analysed

as forming a subclass within the class of specifiers in general. Likewise,

what may be called indefinite quantifiers may be treated as belonging to the

category of adjectivals.

Finally, Some cross-categorial rules will be proposed to show the

various levels to which the various specifying elements may be attached.

6.1 Nominal Specifiers

These may be loosely defined as elements that restrict the referential

range of the head of constituents in which they occur. One may argue

that this is a definition which may also be used in defining modifying

complements, which is also true. There does not seem to be any significant

semantic difference between the two. Both are elements which restrict the

head in some sense. However, as stated earlier, there is a formal

difference separating one from the other. Specifiers do not have the
 272

potential for maximal projections, whereas complements do. In other

words, complements but not specifiers can be reduced to lexical categories

which are characterized by inherent properties of selecting certain other

categories as their complements. In short, specifiers are not lexical

categories in the sense nouns and verbs are. As Jackendoff (1977) has

noted, they are concatenations of elements and the name specifier is an

abbreviatory term used to cover all such elements. Assuming this to be

a valid difference, we shall make further distinctions between types of

specifiers as we go on.

6.1.1 Simple Nominal Specifiers

These include all elements which occur in constituents of NPs the

heads of which are simple nominals 1 whose referents are either identifiable

as single entities or quantifiable as a mass of substance. Accordingly,

their specifiers may be understood following Lyons (1977:664) as entity- or

quantity-denoting forinatives. The former may be referred to by the term

article, while the latter may just be called quantifiers. Each will be

dealt with in turn.

6.1.1.1 Articles

As stated above, the term article is a general label for all entity-

denoting elements. These include demonstratives, interrogatives and

singulatives, 2 for which the narrower term deictics may be employed in

contradistinction to indefinite pronouns and genitive NPs.. 3 On the basis

of the type of information, old or new, they are associated with in

discourse contexts articles may be definite or indefinite.

 274

(l98l:l7Off) an argument can have only one case assigned to it by the

lexical governor with which it is in configuration. The analysis of

/-koo/ 'my' as a genitive case-marker would, therefore, mean that the head

is both in the nominative and in the genitive cases at the same time. And

adopting such analysis goes contrary to our goal of having a highly

constrained grammar.

The alternative is to consider such elements as /-koo/ 'my' as possessive

pronouns having the role of specifiers in a manner characteristic of

genitive NPs. In other words, they may be analysed as base-generated

elements, like the deictics, but undergoing a process of encliticization

at some stage of the derivation of the phrase.

There are examples which may give support to this analysis. Consider

the following.

2(a) [ [fard-i [Tulluu}] [duf-el]

S horse-nom of T. come-pf.
'[The] horse of Tulluu came'.

(b) [ [mull-i [fard-iaa]] [adii-da]]

S leg-nom. of horse-sgl. white-is
'The leg of the horse is white'.

In (2) the elements in the innermost brackets are genitive NP5

functioning as Specifiers. 4 Like the deictics in (1 b) and (d) , these

elements are also independent and their position is as labelled. Parallel

to this, /-koo/ 'my' may be treated as being base-generated and later

encliticized on to its head. This alternative gives us a neat

generalization about genitive pronominals and genitive NPs as being base-

generated specifiers. Furthermore, it enables us to avoid the case

conflict that we referred to above, if such pronominals were to be

treated as genitive case-markers.

 wI

Following Jackendoff (1977:104), we shall put a semantic constraint

on the maximal number of specifiers a head may have. If we limit the

number to one quantifying and one defining/referential element, structures

such as those in (3) will be easily excluded.

3(a)* [[fard-i kun(-i) kam(-it)] ldu?-e)]?

S horse-nom. this-nom. which-is-it die-pf.
'Which this horse died'.

(b)* [[nam-oon-ni bay?ee sadii] [du?-an-i]]

S man-pl-nom. many three die-pl-pf.
?'Many three men died'.

The structures are excluded as being semantically anomalous for the

reason that the specifiers are contradictory in terms. The presence of

one excludes the other.

Following the constraint we have assumed, we shall determine the

position of each specifier type in the projection line of the head.

Before we go into that, however, we need to determine the status of /-ic?'-/.

In the examples in (1 a-b) earlier, we have stated that it is an

inflectional element indicating the singulative. We have also included

it in the list of deictics which means that it is also a determinative

element. This is also deducible from its gloss as 'the' in the structures

in which it occurs.

If /-i-/ is a determinative element, then structures like those

below should be excluded as being ill-formed, on account of the constraint

we assumed above. For here the head is made definite by two defining

elements one of which is /-i'-/.

 27€

4(a) [[fard-i-i kun(i)] [adii-da]]

S horse-sgl-nom. this-nom. white-is
'This horse is white'.

(b) [[sang-ia-i kam(it)] [du?-e]]

S bull-sgl-nom. which is it die-pf.
'Which bull died?'

Cc) [[mu-i-i Tulluu] [bareedaa-dal]

S child-sgl-nom. of T. beautiful-is
'The child of Tulluu is beautiful'.

However, these structures, unlike those in (3), are perfectly grammatical,

which means that either the constraint we have suggested is not workable, or

that /-ic-/ is not a determinative element.

Assuming the constraint to hold, we shall argue that this element is

primarily a singulative marker. Moreno (1939:129) calls it 'forma

individuante' - the singulative form. What it shows is that the noun

with which it is associated refers to a single referent, as opposed to two

or more. In other words, it is a singular marker. If this is correct,

then it should be limited to nouns which are [+ COUNT] but [-P1.]. In

other words, it should not occur with mass or collective or plural nouns,

since such nouns do not refer to individual entities. Let us observe

the following structures.

5(a)(i) [Tulluu-n [[nam-i-aI [arg-e]]]

ST_nom. man-sgl. see-pf.
'Tulluu saw the single man'.

(ii)* [Tulluu-n [[nam-oot-i-a) larg-e]J]

ST_nom. man-pl-sgl. see-pf.

(b)(i) I[horii hor-siis-uu] [nan-3'aal-add-a]1 (Gragg 1982:214)

S cattle raise-cs-to I-cm. like-mid-impf.
'I like to raise cattle'.
 277

5(b)(ii)* [[hor-ia1 hor-siis-uu nan j'aal-add-al

S cattle-sql. raise-caus-to I-cm. like-mid-impf.
?'I want to raise one single cattle'.

(c) (i) [[kurtummii-n] biaan keessa hin-'ira-a]

S fish-nom. see inside cm.exist-impf.
'Fish live in [the] see'.

(ii)* [[kurtummii-i-i] biaan keessa hin-5ir-a]

fish-sgl-nom. sea inside cm.exist-impf.

(d) (i) [Tulluu-n [aannan] dug-el

S
T-nom. milk drink-pf.
'Tulluu drank milk'.

(ii)* [Tulluu-n [aannan-i3a] dug-el

S
T-nom. milk-sgl. drink-pf.

The comparison between the grauunatical and the corresponding

ungrammatical structures in (5) shows that /-i-/ cannot occur with plural

nouns as in (au) , collective nouns as in (b ii)-c ii) or mass nouns as

in (d ii) . As predicted, it only occurs with singular nouns, and as

such functions as a specifier showing singularity or individuality, that is,

the noun is specified as a single or individual entity, as opposed to a

collection of entities or an agglomerated mass. From this may follow

that it is a quantifying rather than a defining element. This explains the

reason why structures like those in (4) are not ungrammatical. In such

structures, /-i?'-/ specifies the head of the phrase indicated in brackets

in terms of its quantity, just as in the same way the deictics restrict it

in terms of its entity, which again means that the constraint we have

proposed earlier has not been violated.

If we are correct thus far, then a further possible conclusion that

emerges from this situation is that [+ COUNT] nouns in Oromo are inflected


278

for both singular and plural number. In other words, the language has both

singular and plural marking devices. These are /-id-/ 5 for the former,

and /-(Q)ota/ for the latter. (See Webster (1960) for other plural markers.)

In the same way [-COUNT] nouns may be divided into [±MASS]. [+MASS]

nouns include those like /aannan/ 'milk' whose amount can be specified in

terms of units of measurement as we shall see later. Those like /horii/

'cattle', which are [-MASS], may be specified in terms of the collection

of individual elements forming the whole unit, that is, as a special class

of collective nouns.

The one problem that the analysis of /-i-/ as a quantifying element

faces is the determinative interpretation it also acquires in the structures

in which it occurs as a singulative marker. This is again observable from

the following examples.

6(a) [[mu-i-i] raf-e]

S child-sgl-nom. sleep-pf.
'The child slept'.

(b) 1[naxn-i-i] deem-e]

S man-sgl-nom. come-pf.
'The man came'.

(c) [[gabr-i-i] kolf-e]

S slave-sgl-nom. laugh-pf.
'The slave laughed'.

In such structures the function of /-i-/ does not seem to be just

that of a singulative, but also that of a determinative. Nouns with this

affix do not constitute new information in discourse. They always refer

to old information. New information is introduced either by use of the

generic or by use of the indefinite form of a noun, as in (7 a) and (b)

respectively.
 279

7(a) [[fard-i] horii Iflana-ti]

S horse-nom. animal-of house-is
'Horse is a domestic animal'.

(b) [[fard-i tokko] duf-e]

S horse-nom. one come-pf.
'A horse came'.

Once a noun is introduced into discourse in such a manner, subsequent

references to it employ the use of /-i-/, if it is masculine, and /-itt--/,

if it is feminine or diminutive. In such cases, the interpretation of the

noun is not that it is single or just one, but also the same one mentioned

earlier in the discourse.

Now the question is whether we should recognize two /-i-/s; one

for the singulative and another one for the determinative, or whether we

should recognize only one /-i-/, which is primarily singulative, and argue

that it has developed the determinative function incidentally. If we

recognize two types, then the determinative /-i-/ should occur with all

types of nouns, and that only the singulative one should be limited to

1+ COUNT] nouns. In other words, we should be able to find distinctions

of syntactic usage between the two of them. This does not seem to be

possible, however, because no noun other than those with the feature

(+ COUNT] allows /-i-/. This is what we noticed in the structures in (5).

The situation is, therefore, one which forces us to accept only one /-i-/,

indicating singularity, and maintain that its determinative use may only

be incidental. This may gain support from the fact that Oromo, unlike

Amharic (and perhaps certain other languages in Ethiopia) does not have a

determiner per se. What it does have are deictics, as mentioned earlier.

Hence it might be the case that it extends the use of /-ia?-/ to show what

in other languages is expressed by a determiner. It might even be said



28C

that this is perhaps by analogy with Amharic, which has both a determiner,

I-u-I, and a singulative marker /4yy-/. Since Oromo has only a

singulative marker, it seems that it extends its use to include the role

of the Axitharic I-ui as well. This is a likelihood given the extent to

which the two languages have been in contact, and have influenced one

another.

We have mentioned I-itt-i as the feminine counterpart of /-i-/.

Its use may be said to be basically the same as that of /-i'-/. It occurs

only with [+ COUNT] but singular nouns. Like /-i-/ it has the extended

use of showing diminutiveness. The following are some examples.

8(a) [lfard-ittii-n] duf-t-e]

S horse-sgl-nom. come-f-pf.
'The horse (mare) came'.

(b) [(mu-ittii-n] duf-t-e]

S child-sgl-nom. come-f-pf.
'The baby girl came'.

(c) [[len-ittii-n] deem-t-eJ

S lion-sgl-noxn. go-f-pf.
'The lioness went'.

In all such examples, the idea expressed by the morpheme is

singulativeness. The idea of definiteness or diminutiveness seems to be

secondary. The situation is parallel to what we have seen with regard

to /-i'-/ as a singular marker and its extended use as a determiner.

This again is parallel to the Amharic /-itu/ which shows both the feminine

singulative and/or the diminutive.

Assuming this to be on the right track, we may maintain the claim we

hinted at earlier that /-i-/ is basically a quantifying element indicating

a single entity, in opposition to /-oota/ which shows multiples of entities.



28

Like the latter it is an inflectional element and may hence be accounted

for in terms of noun morphology as a specification of the grammatical feature

of number. This leaves us with just deictics and genitive NP5 to comprise

the set of definite articles.

6.1.1.1.2 Indefinite Article

Indefiniteness may be indicated by the presence or absence of

indefinite pronouns. However, unless it is clear from the context, the

absence of an indefinite pronoun may lead to ambiguities between the generic

and the indefinite use of a noun. This is noticeable from examples like

(7 a) above, or (9 c) below.

9(a) [[fard-il horii fe?isaa-ti]

S horse-nom. anomal-of load-is
'Horse is [a] draught animal'.

(b) [[fard-i tokko] duf-e]

S horse-nom. one come-pf.
'A horse came'.

(c) [[fard-i] duf-e]

horse-nom. come-pf.
'[A] horse came'.

In (9.a) , the absence of the pronoun /tokko. 'a/one' gives /farda/

'horse' a generic interpretation. In this sense, it refers to the class

as a whole, as distinct from other classes of animals, such as that of

lions, for example. In (b) this generic reference is reduced to just one

member only. This interpretation is possible because of the presence of

the pronoun /tokko/ 'a/one'. The situation in (c) is slightly different.

As in (9 a) , /farda/ 'horse' is not followed by a pronoun; so a generic

interpretation is possible. But unlike (9 a) , where the structure is a



282

copulative one, as is usually the case in such statements of general reference,

the structure in Cc) is non-copulative. 'n other words, (9 a) is a kind

of definition, whereas (c) is not. In this sense, /farda/ 'horse' in (c)

may be interpreted as being distinct from the generic /farda/ 'horse' of

(9 a) . But such distinction is not possible in structures such as (10)

below.

10. [Tulluu-n [farda] bit-el

T-nom. horse buy-pf.
'Tulluu bought [a] horse'.

Here /farda/ 'horse' may be generic if understood as Tulluu buying

not mules but horses, or indefinite, if interpreted as Tulluu buying a horse.

Either interpretation is possible unless context - either pragmatic or

linguistic - makes it clear.

The point at issue is not, however, one of interpretation, as it is of

the syntactic relation that exists between a head and its specifier. As

shown in the examples above and in those below, the pronouns occur following

the head.

11(a) 1 [yaroo kaxn-iyyuu] eem-ii]

S N?
time which-ever go-imp.
'Go any time'.

(b) [ [biyya eessaa-yyuu] deem-e hin-beek-u]

S NPregion where-ever go-pf. neg-know-impf.
'He has never been to any country'.

(c) [ [muaa-n eeu-yyuu-n] aannan hin-5aalat-a]

SN?
child-nom.who-ever-nom. milk cm. like-impf.
'Any child likes milk'.

(d) [ [wan maal-iyyuu] daga?-uu hin-barbaad-u]

S NPthing what-ever hear-to neg-want-impf.
'I don't want to hear anything'.
 283

In these and other similar structures, the position of the pronouns

is the same one we have noticed in relation to the deictics or the genitive

NPs in the preceding section. Imd in accordance with the constraint we

have assumed, only sbustitution, but not co-ocurrence, is possible between

any two of such elements. Hence, structures like those in (12) are

ill-formed.

12(a)* [[biyya (eessaa-yyuul [kana] deem-e hin-beek-u]
go-of. neg-know-impf.
S region where-ever this

(b)* [muaa-n [eeu-yyuu-n1 [kun-(i)] aaxinan hinaalat-a]

Shild who-ever-nom. this-nom. milk neg-like-impf.

Such structures would be possible if only one of the bracketed elements

followed the head.

From what we have noticed thus far, it appears that the class of

article includes deictics and genitive NPs on the one hand and indefinite

pronouns on the other. The former constitute the subclass of definite

article and the latter the indefinite article.

6.1.1.2 Quantifiers

As stated earlier, these are specifiers of the type which denotes

quantity rather than entity. The elements are related to questions of

'how much' rather than 'what' or 'which'. Like articles, they may be

divided into definite and indefinite quantifiers on the basis of the type

of nouns they are associated with.

6.1.1.2.1 Definite Quantifiers

These are elements which are associated with the class of nouns whose

referents may be counted directly as individual units or measured as an



284

agglomerated mass. They include numerals and what Lyons (1977:461ff)

calls measure and classifier phrases. The latter are analysable as

syntactic categories as we shall observe later on.

6.1.1.2.1.1 Numerals

To the exclusion of /tokko/ 'one', which, as we have seen earlier,

also functions as an indefinite article, all cardinal numerals are quantifiers.

Their distribution is restricted to [+ COUNT] nouns, as can be gathered

from the following examples.

13(a) [ (farda lama] 6 na-n bit-a]

SNP
horse-two I-cm. buy-impf.
'I will buy two horses'.

(b) * [[bi'aan lama] na-n dug-al

water two I-cm. drink-ixnpf.

Corresponding to (13 a) above, (14) below is also possible with an

inflected numeral.

14(a) [ (i3-oolle-n laznaa-nuu] duf-an(i)]

SNP
child-pi-nom. two-the (?) come-pl-pf.
'The two children came'.

(b) [Tulluu-n [nam-oota sadee-nuu] arg-el

S NP
T-nom. man-pl. three-the(?) see-pf.
'Tulluu saw the three men'.

From the glosses it appears that the suffix /-nuu/ is a determinative

element equivalent to 'the' in English or I-ui in Amharic, and that its

relationship is with the numeral.

Such structures as those of (14) are used when the head of the phrase,

namely the noun, is restricted in reference to a group of men whose identity

is already made known. n other words, when the head is definite and

plural. In such cases, the nunieral seems to agree with the head by


28

exhibiting this affix, and whenever this is the case, the head may even

be dropped as in (15) derived from (14).

15(a) [[lainaa-nuuj duf-an(i)]

S two-the come-pl-pf.
'The two came'.

(b) [Tulluu-n [sadee-nuu] arg-e]

ST...nom three-the see-pf.
'Tulluu saw the three'.

The fact that /-nuu/ occurs with numerals which are in configuration

with a definite plural head makes it an agreement phenomenon, which means

that it is basically a feature of the nominal head, and that the numeral

acquires it for reasons of concord. But nouns do not allow it since

structures like (16) are unacceptable for the intended interpretation.7

16(a)* [[farda-nuu] nan-bit-a]

S horse-the I-cm. buy-impf.
'I will buy the horse'.

(b)* [[farda-nuu] duf-e]

horse-the come-pf.
'The horse came'.

The situation seems to suggest that /-nuu/ is an inflectional element

characterizing numerals (and other modifying elements as we shall observe

soon).

If it were the case that /laxnaa-nuu/ 'the two' or /sadee-nuu/ 'the

three', etc. could occur without a preceding NP, we could perhaps argue

that numerals are nominals or a subclass of them as they are in Somali (cf.

Saeed 1982) a language to which Oromo is both genetically and geographically

related. But the examples in (15) show that they occur only as derivatives

of structures like those in (14). In other words, numerals cannot occur in

discourse-initial contexts, contexts in which they could have appeared if

they had been nominals.

 286

To substantiate this argument further, we may cite some examples of

structures in which numerals - but not nominals - are excluded. Such

examples include the following:

17(a) (i) (nani-oo-ni (kaleessa (kan) duf-an(i)1]...

man-pl-nom. yesterday comp. come-pl-pf...
'The men who came yesterday...'

(ii)? [lainaa-nuu (kaleessa (kan) duf-an(i)]]

two-the yesterday comp. come-pl-pf.
'The two who came yesterday'.

(b)(i) [Tulluu-n [nam-oota laxnaa-nuu] arg-e]

ST_nom. man-pl. two-the see-pf.
'Tulluu saw the two men'.

(ii)? [Tulluu-n [lamaa-nuu] arg-el

5T-nom. two-the see-pf.
'Tulluu saw the two'.

(c)(i) (nam-i-i duf-e]

Sman_sgi_nom. come-pf.
'The man came'.

(jj)* tokkoo-nuu duf-e

one-the come-pf.
'The one came'.

The structures (17(a ii)-(c ii) in which the numeral appears to

have substituted the nominal are dubious, in each case, for the reason that

they have no independent references. Their occurrence without a preceding

nominal head is only apparent, for underlying there is the NP itself

realized as (e]. In other words, such structures with the numerals

occurring as 'heads' of phrases are derived, not base-generated.

The ust plausible conclusion that we can draw from the observations we

have made thus far is that numerals are not analysable as nominals, although

like the latter, they may display some defining elements.

 287

Before we definitely draw this conclusion we need to make some more

observations. Earlier we mentioned that /-nuu/ also occurs with other

modifying elements. Such elements include adjectives. It might, hence,

be possible that they are adjectives. If this is the case, then they

should occur in all the positions associated with such items. Two such

positions are: following a nominal, and preceding a verbal head in

structures of NPs and VPs respectively. And in both positions numerals

also appear as the following examples demonstrate:

18(a) [[nama lama] arg-e]

S man two see-pf.
'[He] saw two men'.

(b) [nam-oôn-ni [lama tur-an-il]

man-pi--nom. two exist-pl-pf.
'The men were two'.

In this case, they may be treated as belonging to the category of

adjectives. But there are two other positions where only adjectives but

not numerals are found. Consider the following examples.

19(a) Ci) [arakee-n [ [korra-af] gaarii] -da]

S A' PP
wine-nom. cold-for. good
'Spirits is good for colds'.

(jj)* [arakee-n ( [korra-af] tokko] .-da]8

S .. A'PP
spirits-nom. cold-for one is

(b) Ci) [farsoo-n I [korra-af] hamaal -da]

S A' PP cold-for
beer-nom. bad is

'Beer is bad for colds'.

(ii)* [farsoo-n [ [korra-af] tokko] -da]

A'
beer-nom. cold-for one is

The examples show that only adjectives but not numerals subcategorize

adpositional phrases.
 288

The other position where adjectives, though not numerals, are found

is the position following a degree word. The following is an example of

this.

20(a) (bay?ee deeraa]

AP
very tall

(b)* [bay?ee tokko]

very one

The fact that numerals occur in two of the positions where adjectives

are found may lead us to conclude that they are a subclass of this category.

There is no harm in drawing such a conclusion, since some other quantifying

elements which, as we shall observe later, also belong to this category.
What is important is the function they perform which, as we have noticed
throughout, is one of specifying the heads of those phrases in which they

occur, and with which they agree, for such features as definiteness and

number. It is in such cases as when the head is a definite plural that

the numerals show up with /-nuu/.

6.1.1.2.1.2 Measure Phrases

As we have observed in the preceding section, nouns which are (^ COUNT]

may involve numerals in specifying their amount. There are, however, other
sets of nouns whose quantity or amount cannot be figured out by direct
counting as such. Such nouns are (-COUNT] and their specification involves
measure phrases. The following are some examples:

21(a) [ [daadii [birille sadiil] dug-e]

SNP NP
mead bottle three drink-pf.
'(He] drank three bottles of mead'.

(b) [ (taaf ii [kunnaa

S • lamal] nan-barbaad-a]
SNP, NP
t ef basket two I-cm.want-impf.
'I want two baskets of tef'.
S
 289

21(c) [ (lafa (kalaadii lama]] nama tokko-f hin-ga?-u]

S land rope two man one-for neg-suffice-impf.
(Literally), 'Two rope lengths of land is not enough for a man'.

The constituents in the innermost brackets, which are noun phrases

themselves, are the measure phrases which specify the amount of the head in

the outer bracket. The use of numerals alone in the position of the measure

phrase will lead to the following ungrammatical structures.

22(a)* [[daadii sadii] nan-barbaad-a]9

S mead three I-cin.want-impf.

(b)* ([taafii lama] nan-barbaad--a]

S tef two I-cm.want-impf.

In the same manner, the head of the measure phrase alone cannot

function as a specifier since structures like those below are ill-formed.

23(a)* • [[daadii birille] 10

nan-barbaad-a]
S
mead bottle I-cm. want-impf.

(b)* [[taafii
• kunnaa] nan -barbaad-a]
S
tef
a basket I-cm. want-impf

The situation suggests that it is neither the numeral nor the head

of the measure phrase but the two as a structural unit which function as

a specifier. Furthermore, the head of the measure phrase is not just any

noun, but a noun which is associated with certain standards of measurement.

The same kind of restriction can be put on the nouns referring to the

things to be measured, since they too are limited to a set which is high

up in the hierarchy of relevance for human life. In other words, only

nouns referring to things which, because of their significance for life,

are quantified, allow measure phrases.

 290

6.1.1.2.1.3 Classifier Phrases

Like measure phrases, classifier phrases are noun phrases. They

are used within phrases in which the head is one of a set of collective

nouns. Such nouns are formally identifiable by their inability to take

a plural marker, their reference being to the class as a whole rather than

to the individual members forming it.

Reference to individual entities forming the whole may be made by

using one of a set of classifier phrases as denristrated by the following

examples.

24(a) ( [warakata [baala sadii]] fid-i]

SN? NP
paper leaf three bring-imp.
'Bring three pieces of paper'.

(b) [ (aeeta [mataa lama]] na-kenn-i]

SN? NP
green corn head two me-give-imp.
'Give me two heads [cobs] of corn'.

Because of their reference to individual entities, such phrases have

often been called individuating or enumerating elements (cf. Lyons 1977:462).

As in measure phrases, a classifier phrase should contain a numeral. This

is noticeable from the ungrammatical structures in (25).

25(a)* [[warakata baalal f id-fl

S
paper leaf bring-imp.

(b)* ([aeeta mataa] na-kenn-iJ

S
greencorn head me-give-imp.

Again, like the head of a measure phrase, which, as stated before,

is a type of noun associated with units of measurement or their equivalents,

the head of the classifier phrase is limited to a set of nouns whose

referents are similar, literally or figuratively, in shape to the thing to

be measured. Further restrictions are also made on such nouns in terms of

 291

such selectional restriction features as [± ANIMATE], (± PLANT], etc.

Hence, /baala/ 'leaf' and /mataa/ 'head' cannot substitute for each other

without this leading to some degree of anomaly of the type in (26).

26(a)* [ (warakata [mataa sadii]] f id-i]

SNP NP
paper head three bring-imp.
?'Bring three heads of paper'.

(b)* ((aeeta [baala lamaa]] na-keen-i]

S greencorn leaf two me-give-imp.
'Give me two leaves of corn'.

Furthermore, it does not appear to be the case that every collective

noun allows a classifier phrase as its quantifying expression; rather, it

is only those nouns which are high up in the hierarchyof salience in the

sense of Comrie (1981:182) that allow such quantifying expressions. Hence,

structures like those in (27) are excluded on pragmatic grounds.

27(?) [ [burtukaanii [mataa sadii]] na-kenn-i]

SNP NP
orange head three me-give-imp.
?'Give me three heads of oranges'.

It is therefore only nouns like /aeeta/ 'greencorn' but not others

like /burtukaani/ 'orange' with which classifier phrases form structural

relationships without there being any anomaly.

What we have considered so far is the specification of nouns whose

referents are quantifiable either by direct counting or by measurement,

the latter involving the two quantifying phrases. However, there are

other types of nouns which cannot be specified either by direct counting

or measurement.

6.1.1.2.2 Indefinite Quantifiers

The nouns referred to above as being unspecifiãble in the manner

discussed include those with the feature (- CONCRETE]. Nouns like /rakkoo/
 292

'difficulty', /rafiitii/ 'sleep', etc., belong to this type. Such nouns

enloy the use of forms like /tinnooee/ 'little', /guddaa(saa)/'abundant',

or /bay?ee/ 'much', not much as a measure of quantity but as an expression

of the intensity or the seriousness of the thing the terms designate.

Hence we have phrasal structures like (28).

28(a) [rakoo guddaa(saa)]...

NP
difficulty great
'Great difficulty'.

(b) (rafiitii tinnoo(ii) .. .1

NP
sleep little
'A little nap'.
Such quantifying elements are not, however, limited to [- CONCRETE]

nouns since structures like those in (29) are also possible, It is only

the nouns themselves which are sensitive (only) to them.

29(a) [nam-oota guddaa(saa)1...

NP
man-pl. great

'A large number of men'.

(b) (farda bay?ee]...

NP
horse many
'Many horses'.

(c) (biaan tinnoo(ii)]

NP
water little
'A little water'.

Only some, such as /tokko tokko/ (literally) 'one one', i.e., 'some',

are restricted to [+ COUNT] nouns. Hence (30 a) but not (30 b)

30(a) ('aara tokko tokko]

NP
people one one
'Some people'.

(b)* [bisaan tokko tokko]...

water one one


293

In all such structures, the quantifying elements (perhaps with the

exception of /tokko tokko/ 'some') may be treated as adjectives for the

simple reason that the position they occupy is the same as that where the

former are found. Further instances of such distributional similarities

are also noticeable from structures like the following.

31(a) [nam-i-i [guddaa] -da]

Sman_sgl_nom. big is
'The man is big'.

(b) [biaan(i) [tinnoo(ii)] -cia]

S is
water little
'(The] water is little'.

(c) (rakkoo-n [guddaa(saa)] -cia]

S. is
difficulty-nom. great
'The difficulty is serious'.

(ci) (nam-oon-ni [bay?ee] -da]

Sman_pl_nom. many is
'The men are many".

Furthermore, such forms, like adjectives proper, do allow a preceding

degree word. The structures in (32) are examples of this.

32(a) [nama [bay?ee guddaa] arg-e]

S AP
man very big see-pf.
'(He] saw a very big man'.

(b) (biaan (bay?ee tinnoo] dug-e]

Swater APvery little drink-pf.
'[He] drank very little water'.

(c) (hool-oota [,guddaa bay7ee] arg-e]

S sheep_pl . APbig many see-pf.
'(He] saw very many sheep'.

One other structural device often used to distinguish adjectives from

other categories is comparison. And as expected, the forms in question

do occur in such structures, as the examples below show.

 294

33(a) [Tulluu-rra [Fayyiisaa-n t guddaa-dal]

ST_from 1 SFnom 1 big-is
(Literally), 'From Tulluu Fayyiisaa is big',
'Fayyiisaa is bigger than Tulluu'.

(b) [hundurna-rra [Fayyiisaa-n t guddaa-daJJ

1 S 1
all-from F-nom. big-is
(Literally), 'From all Fayyiisaa is big',
'Fayyiisaa is the biggest of all'.

(c) [biaan kana-rr [biaan(i) sun-(i) t bay?ee-a]1

5water this-from 5water-riom. that-nom. much is
(Literally), 'From this water that water is much',
'That water is more than this water'.

(d) Ihunduma-rra [biaan(i) kun-(i) t tinnoo-da] I

5 1 S 1
all-from water-nom. this-nom. little is
(Literally), 'From all this water is little',
'This is the least of all the water'.

From what we have seen thus far regarding indefinite quantifiers, with

the exception of /tokko tokko/ 'some', which we may consider as an indefinite

pronoun, the most plausible conclusion to draw would be one which treated

them as adjectives.

If we are on the right track about the claims we have made regarding

specifiers as entity- and quantity-denoting elements, and with the argument

presented to support this claim, then the next thing we have to do is

determine their relative position in structures of phrases in which both

types occur. In order to do this we have to consider structures like the

following.

34(a) [i3'-oollee sana]

NPchjldpl these
'These children'.

(b) [i3'-oollee lamaa-n sana]

NPhildl two? these
'These two children'.
 295

34(c)* [i-oollee sana laina]

child-pi. these two.

35(a) [fard-oota Tulluu]

NP
horse-pi. of T.
'Horses of Tulluu'.

(b) [fard-oota sadeen Tulluu]

NP
horse-pi. three of-T.
'[The] three horses of Tulluu'.

(c)? [fard-ota Tulluu sadeen]

NP
horse-pi. of-T. three.

From the ungrammatical structures of (34-35) it is possible to infer

that the position of the quantifying element is after the head, but

preceding the deictic /sana/ 'these' or the genitive NP, of-Tulluu.

From the fact that specifiers are optional elements it follows that
11
they cannot be sisters of N, for N allows only strictly subcategorized

argument complements. This makes them either a sister of N' or N'' in

a noun phrase. The structures in (34-35) show that the quantifying

elements follow the head, and that the deictics come last in the string.

This suggests that the former branch down as a sister of N' and the latter

as a sister of N''. And according to the semantic constraint we have

postulated, an NP can at most have one article and one quantifying element,

which means that we have no other elements to consider as specifiers.

Hence the structural relationship of the elements in question in relation

to the head must be as shown in the tree below.

36.

N'' Art.

N' Q
I
sana
those
/ (
N laina
two
ioollee
children
.'Those two children.'
 296

This analysis accounts for the structure of phrases involving

deictics and quantifiers, and it also observes the constraint we have

postulated. But there are structures which it cannot account for.

Such structures include the following.

37(a) [ [muaa-n Tulluu kun(i)1 gadee-da]

SN?
child-nom. of-T. this-nom. rude-is
(Literally), 'This child of Tulluu is rude'.

(b) [ (muaa-n Tulluu kaxn(iti)] du?-eJ?

SNP
child-nom. of-T. which-is-it die-pf.

(Literally), 'Which child of Tulluu died?'

In (37), the head noun /muaa-n/ 'child-nom' is followed by the

genitive NP 'of-Tulluu' and the deictic element /kun(i)/ 'this-nom'.

Both belong to th subclass of article, and their occurrence following

the same head violates the semantic constraint we postulated earlier.

The presence of both elements does not seem to have any effect on the

head as far as the degree of definiteness is concerned. The head could

have been made definite by either one of them.

Given the concatenative nature of specifiers, structures like these

may be analysed in the manner shown below.

38. N'''

j/\tl
N'' NP Art.

muaa-n Tulluu kun(i)

child- of- this-
nom. Tulluu nom.

'This child of Tulluu'.

In other words, both the genitive NP and the deictic /kun(i)/ 'this

man' branch down from the same node, and their linear relationship is as


29?

indicated. The problem mentioned earlier with regard to the analysis

in (36) and the counter-examples in (37) consists in the violation of

the constraint. We may disregard this as a tenuous issue. But if we

add a quantifying element to structures such as (38) the situation becomes

entirely different. Consider (39).

39(a) [ [i-oolle-n Tulluu lamaa-n sun(i)] gadee-dal

SNP
child-pi-nom. of-T. Two ? those-nom. rude is
'Those two children of Tulluu are rude'.

(b)* [ [ i 5'-oolle -n lama Tulluu sun(i)] gadee-da]

S NPhildl two of-T. these-nom. rude is

In (39), it is the genitive NP 'of-Tulluu' which is closest to the head

contrary to the analysis in (36) where the quantifier is shown as being the

closest. According to (39) then, the genitive NP has to branch down from

N'' as sister of the quantifier /lama/ 'two' in the manner shown below.

40. N'''

N'' Art.

N' NP Q sun(i)
Ithese

ijoole-n Tulluu lama

child- of-T. two
pl-nom.

'These two children of Tulluu'.

The comparison of (40) and (38) shows that the genitive NP has two

possible positions. It may branch down from the maximal node as in (38),

or from the intermediate node as shown above. The question is which

position is basic? Or could it be the case thatOromo is characterized

by three types of specifiers instead of by two as we have been arguing?

 298

If we maintain the idea that the specifiers are basically of two

types and that their positions are as shown in (36), we shall have to

account for the analysis of (40) as being derived from a corresponding

structure in which the genitive NP is sister of Art. In other words,

(40) is a transform of the structure in (41).

41. N'''

N'' NP Art.

N' Q of-Tulluu sun(i)

/ i-oolee-n
1
lama-n

'Those two children of Tulluu'.

According to this analysis, the genitive NP 'of-Tulluu' moves from N'''

in (41) to N'' in (40). And this appears quite possible given the fact

that the genitive relation is indicated by the possessor NP coming next

to the possessed NP. In such structures as (41), the possessed NP, that

is, the head is separated from the possessor by the quantifier llama! 'two'.

In order for the structure to be grammatical, the possessor NP has to move

down to N'' and be immediately next to the head, which it possesses.

If Oromo had a genitive morpheme similar to the Imharic /y-/ 'of',

for example, then the movement might not be necessary. The possessed-

possessor relationship could still be maintained while the genitive phrase

remained in situ, i.e., under N'''. It is the way in which Oromo shows

such relationships that makes the movement obligatory.

Alternatively, as we have hinted at earlier, it might be possible to

argue for a three-way specifier system for Oromo. A nominal head may be
 299

said to be restricted in terms of its possessor, spatio-temporal context

and in terms of its quantification; in which case its projection potential

in terms of bar levels would be quadruple instead of treble. And

accordingly such structures as (39) would have to have the analysis given

in (42), instead of, or rather than the one in (40) and which was derived

from (41)

42. N''''

N''' Art.

/
N'' Q
I
sun(i)
these

N'NP
I
i 3'ooiien Tulluu lamaan
children of-Tulluu two

'These two children of Tulluu'.

Whereas this analysis avoids the transformation which the genitive

NP had to undergo in (40) and hence appears economical, it is not free

of problems itself. As we have noted in Chapter Four, a degree of cross-

category generalization has been achieved with regard to the value of n

in This generalization, which X-bar syntax wants us to look for

and maintain, is now endangered because n is going to be four for nominals

instead of three as has been claimed until now.

In order to maintain the generalization we need to do away with this

analysis and look for a better one, or else maintain the analysis shown

in (40). This is particularly necessary given the fact that the presence

of both the genitive NP and the deictic element has no significant effect

on the head in terms of the degree of definiteness. This means then that
 30(

the analysis in (40) may be adopted in favour of the three-way specifier

system we have been considering as an alternative.

A last alternative would be to consider genitive NP5 not as part of

the specifier of a head but as part of its complement. As discussed in

Chapter Four, genitive NPs of various functions have been treated as

complements either of N' or N''. It might not be impossible to include

genitive possessives as complements of N' along with genitives of 'source',

for example. This alternative is appealing for three reasons. Firstly,

it enables us to maintain the cross-category generalizations we have

otherwise managed to achieve. Secondly, it enables us to make yet

further generalizations about genitive NPs. As discussed in Chapter Four,

other genitive NPs are complements. It was only the present type of

genitive which had been left aside as a specifier. This would, in effect,

mean making distinctions between a complement genitive and a specifier

genitive. The distinction is vague since the definitions of both

specifiers and complements as 'concatenations of elements' are equally vague

and appear to be two terms that are used simply as devices without any

theoretical import (cf. Jackendoff 1977). If we include genitive

possessives in the list of complements, we can make a better generalization

and the class of specifiers will be limited just to deictics and quantifiers.

Thirdly, this alternative gets rid of the transformation we would have had

otherwise to apply to (41) if we had adopted that option. In addition to

these advantages, this alternative leads to a situation where the genitive

NP fits the distributional definition of complements we made previously.

We have stated that complements always occur immediately next to the head

and that specifiers assume peripheral positions. This is noticeable

from the structure in (39 ) , where the genitive occurs after the head
 301

and from the ungrammatical counterpart in (39(b)), where the genitive is

separated from the head by the quantifier /lama/ 'two'.

Since this alternative avoids the problems the previous two approaches

have faced, while at the same time lending to further generalizations, it

may be adopted as a better option.

6.1.2 Derived Nominal Specifiers

What we have considered so far is the class of specifiers associated

with simple nominals. In this subsection we shall examine the type of

specifiers that derived nominals are characterized by, in the light of

the arguments presented for the former.

As stated earlier (cf. Note 1), we shall restrict the notion of

derived nomirials to infinitivals. This restriction is necessary for

the distinction which is to be made between them on the one hand and

other non-infinitival nominals on the other. The latter are subject to

the same kind of analysis we have argued for for simple nominals.

Regarding the former, consider the following structures:

43(a) 1 [deem-uu-n] gaarii-da]

S
5 go-to-riom. good-if
'To go/going is good'.

(b) !deem-uu-n-isaal gaarii-da]

S Sgo_to_nom_lhisl good-is
'His going is good'.

(c) I !Tulluu-n (deem-uu-n-isaa]] gaarii-dal

S ST_nom. go-to-nom-'his' good-is
'Tulluu's going is good'.

The structures in the innermost brackets in (43) are infinitivals.

As we have observed in the preceding chapter, they are indicated by the

deverbative element /-uu-/.

 302

Within the theoretical framework we are following, infinitivals

are clauses and are as such subject to the same kind of analysis other

types of clauses are characterized by. In the structure in (43), the

subject of the head is Pro.in (a), PR0.in (b), and /Tulluu/ in (c).

It is PRO in (a) in accordance with the Extended Projection principle

of Chomsky (1982) which stipulates that all clauses have subjects. It

is Pro. in (b) because the head has the pronominal element /-isaa/ 'his'

which refers to the possible subjects in this case to a subject which is 3ms

and which can be dropped as in this particular case or maintained as in (c).

Now if we assume with Jackendoff (1977) that subjects of both NPs

and clauses are specifiers of their respective heads, /Tulluu/ in (43 c)

may be treated as being both a subject and a specifier of its infinitival

head /deemuunisaa/ 'his going'. Its position in the hierarchy is parallel

to that of articles in the simple nominals discussed in the preceding

section. In other words, it branches down from the maximal node N'''

as sister of N'' in the manner shown below.

44. N'''

N''' N''

I f

I r
Tulluu-n deemuu-nisaa
T-nom. going-his

'Tulluu's going'.

One difference between simple NP5 and infinitivals or gerundives like

/deemuu-n-isaa/ 'his going' is in the position of the head. As shown in

the preceding section, simple NP5 are lef-headed, whereas gerundives are


30:

right-headed. It does not, therefore, seem possible to make a general

statement about the two types of NPs with regard to the notion of head.

In fact, gerundivesarevery much like verbals both in the position they

assume as heads, and the type of specifiers and complements they select.

This will become more explicit when we get to the section on verbals.

Now we shall consider a point that makes gerundives and simple NPs look

alike. This emerges in structures uk (45).

45(a) [ !deem-uu-n Tulluu] gaarii-dal
ss
go-to-nom. of-T. good-is
(Literally), 'Going of Tulluu is good'.

(b) [ ! uf-uu-n Tulluu] gaarii-daj

S Scome_to_nom. of-T. good-is
(Literally), 'Coming of Tulluu is good'.

What we have in (45) is a gerundive NP which is left-headed. This

is identical to the situation we find in such simple NPs as:

46(a) [ [fard-i Tulluul guddaa-da]

S NPhorse_nom. of-T. big-id
(Literally), '[The] horse of Tulluu is big'.

(b) [ [hoolaa-n Tulluu] adii-da]

S NPh of-T. white-is
(Literally), '[The] sheep of Tulluu is white'.

In both (45) and (46), the head is on the left, and again in both the

non-head categories are genitive NPs. The situation is, therefore, one

which makes the distinction between the two types of nominals very tenuous,

and also induces one to treat them alike. To treat them alike means to

assume that all NPs - gerundive or simple - are basically left-headed

and that structures like /Tulluu-n deem-uu-n-sa&. ./ 'Tulluu's going' in

(43 c) or (44) which are right-headed are derivatives of the former.

 304

This is problematic. As stated earlier, infinitivals or gerundives

are clausal, and in order to analyse them as such, we need to assume a

base rule of the type: S > NP VP. If we adopt the idea that NPs

are basically all left-headed, then subjects of infinitivals such as

/Tulluu-n/ in /Tulluu-n deem-uu-n-sa4!in (43 c) , for example, would

have to be base-generated in post-head position and then undergo some

transformation(s) to assume their surface positions. This in effect

would also mean having to write two base rules for clauses: one for

finite clauses with the usual NP-VP order and another for infinitivals

with the opposite order. This goes against our goal of achieving a

highly constrained grammar.

In order to avoid this, we need to assume that although the structures

/Tulluu-n deem-uu-n-isaa/ 'Tullu's going' and /deem-uu-n Tulluu/ 'going of

Tulluu' are paraphrases, and so may be assumed to be derivatives, the

direction of the derivation must be the opposite of what has been assumed

to be the case above. In other words, /deem-uu-n Tulluu/ 'going of

Tulluu' must be derived from /Tulluu-n deem-uu-n-isaa. 'Tulluu's going' by

an optional rule of agent 'postposing'. Hence, the similarity of

structures like /deem-uu-n Tulluu/ 'going of Tulluu' and /farda Tulluu/

'horse of Tulluu' is superficial, and that in terms of headedness the two

are different, the former look very much like verbals rather than nominals

as stated before.

Notice that the derivation of /deem-uu-n Tulluu/ 'going of Tulluu'

from /Tulluu-n deem-uu-n-isaa/ 'Tullu's going' has a number of morphological

or perhaps morphosyntactic effects, other than or in addition to the

'postposing' of the subject NP Tulluu. Firstly, Tulluu is no more in the

nominative case for such structures as (47 b) are ungrammatical.



305

47(a)* [ [deem-uu-n Tulluu-n] gaarii-da]

S Sg0_0_0•
T-nom. good-is
(Literally), 'Going of Tulluu is good'.

(b) [ [deem-uu-n Tulluu] gaarii-a]

S
of-T. good-is

Secondly, the pronominal /-isaa/ 'his' is no longer on the infinitival

head, since again structures like (48) are ill-forirted.

48* [ ! deem-uu-n- isaa Tulluu-(n)] gaarii-dal

S
5 go-to-nom-his. T-(nom) good-is
?'Going of Tulluu is good'.

The questions which follow from a consideration of this state of affairs

are: why is the 'postposed' subject without nominative marking, and why is

the infinitival head without the pronominal element /-isaa/? These

questions need attention because in simple clauses postposed subjects

retain their nominative marker as the example in (49) shows.

49(a) [Tulluu-n deem-el

ST_nom. goOpf.
'Tulluu went'.

(b) [ t [ [deem-el Tulluu-n] I

S 1 VP VP
go-pf. T-nom.
'Tulluu went'.

(c)* [t [ [deem-el TulluuJ I

S VPVP
go-pf. T.
'Tulluu went'.

The answer to the questions seems to be linked to the category to

which the postposed NP is adjoined, and to the type of structural

configuration which results from this relationship. In order to have a

closer observation of the process involved, let us observe the structure

in question i.e. (43 c) in the form of a tree. (Details have been avoided.)
 306

50.

\ gaarii-4a
good-is

comp. S

I
e NP VP

deem-uu-n-s aa

In (50), if Tulluu moves and gets adjoined to the VP or even higher

to the S in S, the structure would still be one of predication. In

other words, Tulluu would still be the subject and it should, therefore,

retain its nominative marker /-n/. But such structures are excluced

as ill-formed, as shown in (48). This kind of movement seems to be

possible only in simple clauses like (49 b) for example. The movement

indicated in (50) must, therefore, be adjunction to the NP of the matrix

clause. In other words, Tulluu has moved out of its clause and forms a

new structural relationship with this NP. In this position Tulluu

functions as a specifier of the infinitival head /deemuu-n/ 'going-nom'.

As is observable, Tulluu is not in the nominative case in this position,

it is in the genitive case. And it seems that it receives this case

from the infinitival head /deemuu-/ 'going' since Inf 1. which would

otherwise have assigned nominative case to it (Tulluu) is not available in the

new configuration of /deemuu-n Tulluu/ 'Tulluu's going'.

 307

The movement does not violate subjacency since Tulluu crosses only

one S. It does not seem to violate any of the principles that determine

the relation between the moved element and its trace. It does show,

however, a minor violation of the projection principle, since the derived

structure fails to show Inf 1., and with it the feature AGR, which is the

nominative case-assigner. In other words, the movement has resulted in

the loss of the category Inf 1. This is why the moved NP (Tulluu) has

to get its genitive case from the infinitival head.

It might perhas be argued (as an alternative) that some infinitives

are like simple (ordinary) nouns, and that NPs like Tulluu may be base-

generated in post-head positions as specifiers of such infinitives. In

other words, /deemuu Tulluu/ 'Tulluu's going' may be treated as a simple

(non-clausal) NP with Tulluu functioning as a specifier generated in situ.

This may avoid having to recognize a violation of the said principle. But

it has its own problem. How do we know which infinitives are simple and

which ones are clausal? There is no formal distinction between any

infinitives. Besides we have to make further distinctions between such

infinitives and simple nominals with regard to their subcategorization and

other formal properties. For this reason, the movement analysis is adopted

as a less problematic solution.

6.2 Non-nominal Specifiers

In this subsection of the discussion we shall argue that the arguments

presented with regard to nominal specifiers may be extended to the other

categories, allowing for some degree of variation.

6.2.1 Verbals

Like nominals, verbals select certain types of quantifiers as specifiers.

 308

Consider the following structures.

51(a) [Tulluu-n I [yaroo sadii [kuf-e)]]]

S V'1 N''' V'
T-nom. time three fall-pf.
'Tulluu fell three times'.

(b) [Tulluu-n I [yaroo sadii [biaan dug-e]]]1

S V'1 N' 1 ' VI
T-nom. time three water dririk-pf.
'Tulluu drank water three times'.

In both structures the quantifier phrase, which is an NP, is structurally

related to VI, which is the head of the larger category V''. As the

labelling shows /yaroo sadii/ 'three times' is outside V'. It cannot be

inside V' because only complements are allowed in this position since they,

but not specifiers like /yaroo sadii/ 'three times', are expressions of

the inherent properties of lexical heads. Hence, their position is always

peripheral, as stated earlier in this chapter.

This situation is parallel to the position of quantifiers in nominals.

There we have seen that their position is outside N' for the same reason

as that stated above.

Some types of verbs select degree words such as /bay?ee/ 'very'

rather than NP quantifiers of the type shown in (47). The following

are some examples with such verbs.

52(a) [dubartii-n kun(i) [bay?ee [iiyyit-e]]1

S V'' VI
woman-nom. this-nom. very yell-f-pf.
(Literally), 'This woman cried very',
'This woman cried a lot'.

(b) ETulluu-n Ibay?ee [aba-at-e1]1

T-nom. very strong-mid-pf.
'Tulluu got very strong'.

In both structures /bay7ee/ 'very', like /yaroo sadii/ 'three times'

in earlier examples, forms a syntactic unit (V'') with the head in the


309

inner brackets. That this is so is verifiable by any constituency-

testing device such as gapping or VP preposing. In (52), for example,

V t alone cannot be preposed without the resulting structure being

ungrammatical. This is what (53) below demonstrates.

53(a)* iiyyi-t-e [dubartii-n kun(i) Lbay?ee [t]]]]

S S V'1
yell-f-pf. woxnan-nom. this-nom. very

(b) * aba-at-e (Tulluu-n [bay?ee [thu

S V''
strong-mid-pf. T-nom. very

The position of the quantifier /yaroo sadii/ 'three times' or the

degree word /bay?ee/ 'very' has to be in V'' if we argue with Jackendoff

(1977:48) that v''' is the position where the subject branches down as a

specifier12 parallel to the articles in simple NPs and infinitives. The

structural relationship of the quantifier in (51) or the degree word in

(52) in relation to the head, V', is as shown below.

54. V''' (= S)

V'1

Tulluu-n N'"/A'' V'

f yaroo sadii ) biaan 4ug-e

times three/ water drink-pf
( bay?ee
I. very
f aba-at-e
J strong-mid-pf.

As mentioned earlier, the infinitives fit this analysis. Like verbals

they are left-headd; they may have quantifiers of the same type, and as

already discussed, they may even select a lexical subject. Compare the

following for example.

 310

55(a) [Tulluu-n [yaroo lama [gara Gimbii deem-el]]

V.,, V,1 VI
T-nom. time two to G. go-pf.
'Tulluu went to Gimbii three times'.

(b) [Tulluu-n [yaroo lama [gara Gixnbii deem-uu-n-isaa...]l]

N''' N'' N'
T-nom. time two to G. go-to-nom-his
(Literally), 'Tulluu's three times going to Gimbii...'

The difference between (55 a) and (b) is a difference of category.

The head /deem-/ 'go' in (a) is [^ V - NI whereas /deem-uu-n-/ in (b) is

[- V + N].

6.2.2 Adjectivals

Adjectives do not seem to permit quantifiers. Structures with

quantifier phrases are ill-formed as the following examples demonstrate.

56(a)* [Tulluu-n [ [yaroo sadii [4eeraa]I] -da]

S N''' A'
T-nom. time three tall is
?'Tulluu is three times tall'.

(b)* [Tulluu-n A'' N''' lama [gabaabaa]]] -da]

T-nom. time two short is
?'Tulluu is twice times short'.

Such quantifying phrases may, however, occur with verbs of adjectival

origin in structures of comparative clauses as in (57) below.

57. [Tulluu-rra Fayyiisaa-n [ [yaroo sadii

S V'' N'''
T-from F-nom. time three
[(hin) deer-at-a] 1]]
v'
cm. tall-mid-impf.
(Literally), 'From Tulluu Fayyiisaa "talls" three times',
'Fayyiisaa is three times taller than Tulluu'.

Although they do not allow quantifiers, adjectives do allow degree

words.
 311

58(a) [Tulluu-n [bay?ee [deeraa]] -da I
ST_nom. A''very A'taii is
'Tulluu is very tall'.

(b) [naxn-ic!-i [bay?ee [furdaal] -da]

S A'' A'
man-sgl-nom. very fat is
'The man is very fat'.

The position of the degree word may be argued to be the same as that

of the quantifiers in nominals and verbals. Examples of structures

containing adjectives derived from transitive verbs may give support to

this, since in such structures, the degree word appears preceding the

complement of the adjectival head. Consider, for example, the following:

59(a) (i) [aaltuu-n [bay?ee [sirba sirbi-tuu]I tur-t-el

S '.' A'' A'
ç-nom. very song singer-fern. be-fern-pf.
(Literally), 'aaltuu was a very song singer',
'aaltuu was a "nice" singer'.

(ii) ?[ c aaltuu-n [sirba [bay?ee sirbi-tuu]] tur-t-e]

S, A'' A'
L-nom. song very singer-fern. be-fem-pf.

(b) (i) [Tulluu-n [bay?ee [nama sodaataal] tur-el

ST_nom. Avery Arnan fearful be-pf.
(Literally), 'Tulluu was very man fearful',
'Tulluu was too afraid of people'.

(ii) ?[Tulluu-n [naina [bay?ée sodaataall tur-e]

S A''
T-nom. man very fearful be-pf,

The (ii) structures may not be totally excluded as ungrammatical, but

would be marked by a relatively longer pause following the nouns /sirba/

'song' or /nama/ 'man'. This might suggest that the nouns have been

displaced from their base positions. In fact, the occurrence of the

degree word /bay?ee/ 'very' following the noun and preceding the adjective

may lead to some type of ambiguity if there is not such a pause. The

degree word /bay?ee/ 'very' is homophonous with the quantifier /bayee/

 312

'many/much'. The syntactic difference between them is the position they

assume relative to their respective heads. The degree word /bay?ee/

'very' occurs preceding its adjectival head, whereas the quantifier

/bay?ee/ 'many/much' follows its nominal head. This means that in

structures such as (58 b) , the degree word may be treated as being part of

either of the preceding noun, in which case it is a quantifier, or as

part of the following adjective, in which case it is an intensifier element.

The following trees may make this clearer.

60(a) A''

Deg A

nama bay?ee sodaataa

man very fearful
'Very fearful of man.'
 313

60(b) A''

I
: Np A

NP Q

7 .
nama bay?ee sodaataa
man many fearful

'Fearful of many men'.

It is the structure in (60 a) which we have said is (and should be)

marked by a pause in order for its interpretation to be distinguished from

the one associated with the corresponding structure in (b), where /bay?ee/

'many' is part of the complement NP.

For the degree word /bay?ee/ 'very' to be unambiguously part of the

adjectival head, it should branch down from the maximal node A'' as shown

in (61) below.

61. A''

Deg. A'

7N
bay? ee
I 7
nama sodaataa
very man fearful
 314

It was claimed in Chapter Four that adjectives have only two levels

of projection; in other words, there is no A'''. And this really does

seem to be the case, for the category is not characterized by articles

which would otherwise have branched down from such a node, parallel to

the situation in the other two categories we have observed.

6.2.3 Adpositionals

The situation with adpositions is nre like that of adjectivals with

respect to their potential for maximal projection. Like the latter, they

are limited to a double-bar projection level. As regards their choice of

specifier, they are not any different from the other categories in general.

They are characterized by both the quantifying and intensifying (degree)

phrases. The structures below are illustrative of this.

61(a) [ [lookoo lama [gara fuuldura}J deem-il]

S Adp'' Adp'
metre two towards front go-imp.
'Go forward two metres'.

(b) I [tarkaanf ii sadii [gara-na]] koott-u]

SAdp'' Adp'
stride three towards me come-imp.
'Come towards me three strides'.

In both structures the quantifying NP5/looko lama/ 'two metres' or

/tarkaanf ii sadii/ 'three strides' is structurally related to Adp', that

is, to the structure in the innermost brackets. This is easily verifiable

from structures of co-ordination like the following:

62(a) ' [lookoo laina gara fuulduraa] (-f)

Adp''
metre two towards front (and)
(lookoo sadii gara duubaa]
Adp'
metre three towards back
'Two metres towards the front and two metres towards the back'.
 315

62(b)* [lookoo lama gara fuuldura] (f) [lookoo sadii]

Adp'' N'''
metre two towards front (and) metre three.

Given the fact that co-ordination is possible only between parallel

constituent structures, the grammaticality of (62 a), and the

ungrausnaticality of the corresponding (62 b) prove that /lookoo lama/

'two metres' and /gara fuuldura/ 'towards the front' form a single

structural unit.

Structures like (62 a) show that the quantifier and the adpositional

phrases are in configuration. They do not tell us whether the quantifier

phrase is exclusively a specifier of the adpositional phrase Adp' or of

the verb phrase v''. The fact that the latter is also characterized by

the same type of specifier makes this question very important. The

argument that /tarkaanf ii sadii/ 'three strides' is part of Adp', would

be strong only if it were possible for such structures as (61) to be

grammatical with a second quantifier phrase. In other words, we need to

have structures like (63) to support our claim.

63. [ [yaroo tokko (tarkaanfii sadil [gara-na]J

SV'' Ad'' Ad'
time one stride three towards me
[koott-u] I I
come-imp.
(Literally), 'Come three strides towards me one time (once) '.

The structural configuration of (63) is as shown in (64). (Details have

been avoided.
 316

64

yaroo
tokko
Adp'' V

NP Adp' koottu

tarkaanf ii NP P
sadii
\
-na gara

In (63) as represented in (64) /tarkaarlf ii sadii/ 'three stridest

is part of Adp'' which is the complement of V t . The argument that this

quantifier phrase might be part of V' t along with /yaroo tokko/ 'one

time' does not seem to hold, because in that case, the structure should

remain grammatical when gapping operates on V'. But this is not the

case as we can gather from structures like (65 d)

65(a) (Thlluu-n (yaroo sadii [tarkaanf ii sadii gara fuuldural

S V'' Ad"''
T-nom. time three stride three towards front
deem-e]]
go-pf.
'Tulluu went three steps forward three times'.

(b) [Fayyiisaa-ri [yaroo lama (tarkaanf ii sadii gara fuuldura deem-el

S V'' Ad''
F-nom. time two ' strides three towards front go-pf.
'Fayyiisaa went three steps forward two times'.

(c) [Tulluu [yaroo sadii [tarkaanf ii sadii gara fuuldura

T-nom. time three strides three towards front
deem-e,]l (Fayyiisaa-n immo [yaroo lama ( 1]
S VI' time two V'
go-pf. F-nom. also
'Tulluu went three steps forward three times, and Fayyiissaa
two times'.
 317

65(d)* [Fayyiisaa-n ium [yaroo lama tarkaanf ii sadii [ 1]]

S VI' VI
F-nom. also time two strides three
'And Fayyiisaa two times three strides'.

(d) would have been grammatical if /tarkaanf ii sadii/ 'three strides' were

part of V'. But it is not, which means that the phrase is part of V'1

and within V 1 it is a part of the complement Adp'' as shown in the tree

in (64).

Like adjectivals, adpositionals do also allow the same degree word

/bay?ee/ 'very'. Consider (66) for example.

66. [Tulluu-n [bay?ee [rakkoo-rral] 'ira}

s p11 p1
T-nom. very difficulty-on exist
(Literally), 'Tulluu is very in difficulty',
'Tulluu is seriously in difficulty'.

The structural relationship of both the quantifier and the intensifier

in relation to the head is as shown in (67) below.

67. Adp''

c Q 1•. Adp'
1 deg.)

This is identical to what we saw in the other categories. The

quantifier and/or intensifier branches from the X I ' node.

6.3 Summary

From the data presented and the discussions that followed throughout

this chapter, the following general statements may be made:

a) All the major categories have some kind of elements, phrasal

or lexical, which restrict them in terms of their identity,

quantity and/or intensity. Nominals have articles and

 318

quantifiers. Adjectivals have intensifiers only.

Verbals and adpositionals have both quantifiers and

intensifiers.

b) In all the categories the quantifiers or intensifiers

branch down from the same node level, as predicted by

the theory of X-bar syntax.

c) Apart from simply nominals which are left-headed, all

other categories are right-headed, meaning that they

are preceded by their specifiers.

d) Regarding the specifiers of simple nominals, we have

argued that demonstratives, interrogatives, and

indefinite pronouns may be treated as belonging to the

sub-class of articles within the broader class of specifiers.

The element /-i-/ has been treated as a quantifier which

mainly shows singularity, in opposition to /-(o)ota/ which

shows plurality. Its use as a determiner seems to be

incidental and perhaps analogical.

Genitive NPs of possession may be treated as specifiers. They

branch from the same node that dominates articles. However, in structures

of NPs with quantifiers, they must undergo a movement rule which adjoins

them to the node that dominates the quantifiers. An alternative analysis

which we have proposed to avoid the movement is one which treats them as

complements. According to this alternative they may be base-generated

in situ as the sister of N', parallel to other genitive complements.

Finally, the following phrase structure rules may be proposed for

each category:
 319

/N
68(a) v''1 Cs) (b) N'''

N''' VI, Art

N'Q

Cc) A'' Cd) Adp''

Q Adp'

(where Q = quantifier and/or intensifier).

With regard to the position of Q, the following cross-category

generalization may be made:

69. x'' >QX'

(where X [±

This distinction is not a substantive one. Rather it is a formal

or positional one holding between the category of nouns on the one hand

and all the other categories on the other. The former is left-headed.

In fact, at the level of assumption the distinction may not be relevant,

so that (69) may be recognized as characteristic of Oromo.

 32(

NOTES TO CHAPTER SIX

1. These include all nominals, derived or not, which have no deep subject.

2. This is in reference to their use as determiners. But since we shall

provide an alternative analysis of them, their classification as
articles here is tentative.

3. Heads of NPs with a genitive NP 'specifier' may not be old information.

It is the head of the genitive NP which appears to be the old information.
For example, one may start discourse with:
[hoolaa-n TulluuJ du?-e
sheep-nom. of T. die-pf.
'The sheep of Thlluu died'.
/hoolaa/ 'sheep' may not have been previously mentioned, hence it may be
indefinite. It is Tulluu who is known to both participants. In
other words, it is through Tulluu that /hoolaa/ 'sheep' is made definite.

4. An alternative analysis will be proposed later on.

5. According to Andrzjewski (1960) only nouns with the feature [+ HUMPN]

are characterized by the singulative affix in the Boranaa dialect of
Oromo spoken in Southern Ethiopia. This is not the case in Mec
as every [+ COUNT) noun is characterized by it.

6. A quantified NP may not be marked for the plural.

7. Possible to mean 'even with a horse'.

8. The problem here could be seen as semantic or pragmatic. But there

are no instances where a numeral subcategorizes a category as its
complement.

9. Possible only in restricted contexts where the measure phrase is

understood to be there underlyingly.

10. Possible only as a reduced measure phrase.

11. This is a very general statement, the complement of simple nominals

may be optional.

12. According to the alternative analysis proposed in Chapter Five, such

specifiers are out of the projection line of the verb. Their
relation with the latter is in this case that of argument and
predicate, rather than of specifier and head.
 321

CHAPTER SEVEN

CONCLUS ION

In this chapter we shall recapitulate the major points discussed

throughout the preceding chapters. This may help us to put the whole

study in perspective and evaluate it in the light of the claims and

predictions of the theoretical framework we have adopted.

The central theme of the study is the identification of the major

lexical categories that constitute the Oromo lexicon and the specification

of the potential of these categories for maximal syntactic projections in

terms of bar levels, following the theoretical assumptions and claims of

the X-bar convention. In short, it is the identification of X and the

value of n in Jackendoff's schema of X > --- X- which we have

discussed in Chapter One.

The schema presupposes that Oromo belongs typologically to the class

of configurational languages. Configurationality is a feature which

distinguishes languages with a fairly rigid word order from those with no

such order. The former are believed to be hierarchically structured, whereas

the latter are 'flat'. We have argued in Chapter One that Oromo belongs to

the former. The argument has been substantiated and developed further in

subsequent chapters. We have observed that non-head elements, that is,

complements and specifiers, are tightly bound to their heads. Hence any

syntactic rule that operates on any category X does not leave the complement

or the specifier intact so as to operate only on the head or vice-versa.

Rather the rule takes the entire string as its domain of application.

We have observed instantiation of this throughout Chapters Four, Five and Six.

Regarding lexical categories, we have (on purely formal grounds)

recognized four major and two minor categories. These categories are:
 322

nominals, verbals, adjectivals and adpositionals. The minor ones include

specifiers and particles.

The major lexical categories have the potential to develop into

maximal syntactic categories. This potential is a concomitant of their

inherent property for subcategorizing certain maximal categories as their

complements and certain others as their specifiers. The categorial

component and the lexicon are the two domains where such potentials are

expressed. We have observed this in Chapters Three, Four and Six.

As stated in Chapter One, Jackendoff (1977) claims that the maximal

value of n in X is treble bar. In other words, all major lexical categories

have the potential for a uniform three-bar projection. The types of

categories they select as complements or specifiers are also believed to

be similar across the categories with the implication that the rules which

operate on such categories at any one level are also similar. Hence, in

order to capture such cross-categorial similarities by a general rule, the

use of syntactic distinctive features has been proposed in the manner

discussed in Chapter Two.

It has been one of the objectives of this study to test the validity

of Jackendoff's claims with data from a language which is in no way related

to the one on which the theory was originally based. We have observed

throughout Chapter Four that not all the major categories have a uniform

treble-bar projection. Only verbals and nominals realized this potential;

the categories of adjectivals and adpositionals are one bar short.

Furthermore, the complements of the categories at every level of projection

are similar only if the categories are derivationally related. Hence only

adjectivals or nominals which are derived from verbal sources can have

complements which are similar to those of verbs. We have observed, for

 323

example, that nouns such as /buddeena/ 'bread' optionally select genitive

NPs of purpose or source at their minimal level of projections whereas at

this same level, verbs such as /bit-/ 'buy' are characterized by NPs

which function as patients. On the other hand, both /bit-/ 'buy' and

its derivativ/bituu-/ 'buying' or /bitaa/ 'buyer' are characterized by

similar complements all along their projection lines. The same may be

said about adpositionals since their complements are similar to those of

verbals, though their potential for projection is a bar short of that for

the latter.

At the level higher than their minimal projections, all the major

categories are characterized by restrictive modifiers. In the case of

riominals these are: relative clauses, adjectival phrases and some genitive

NPs having locative and temporal functions; in the case of verbals: all

adpositional phrases and clauses of various adverbial functions; and in

the case of adjectivals and adpositionals: the various degree phrases and

clauses. At the maximal level, we have observed appositive relative

clauses and sentence adverbials occurring as complements of nominals and

verbals respectively. As Jackendoff (1977) has noted, the complements at

the minimal levels are the functional arguments that the heads require

to form their main (nuclear) predicate structures. Those complements at

the intermediate and maximal levels of projection are restrictive and

appositive modifiers respectively. They are believed to be peripheral

to the main predicate structure.

Regarding specifiers, we have observed throughout Chapter Six that all

the major categories have certain quantifying or intensifying phrases at

their intermediate levels and that only verbals and nominals are characterized

by articles and certain NPs at their maximal levels.

 324

From the positional relationships of the head and non-head terms,

it may be said that Oromo is a head-final language. Taking this into

account and abstracting away from minor surface differences that some

categories exhibit, the discussions on complements and specifiers may be

roughly captured by the following tree structure.

/N
(SP) (CP) X''

(sP) (CP)

where: SP = specifier (max)imal

CP = complement max.

Although (1) shows the general picture of the projection of the major

categories, we have argued in Chapter Five that this projection should not

include Ss and Ss. This is contrary to Jackendoff (1977), who argued

that S and S could be treated as projections of verbals. Their inclusion

leads to problems when considered in relation to other principles which

interact with the X-bar convention itself. Such principles include those

of the case and binding theories of Chomsky (1981; 1982). Both are deep-

rooted in the notions of government, which in turn presupposes structural

configurations of some type between head and non-head terms. The

assignment of nominative case by Irifi. may not be effected as required if

the subject NP which receives this case is a projection of V, for according

to the theory, V does not govern this NP and hence cannot assign case to

it. For this reason, and following the literature, distinctions have been
 325

made between the maximal projection of V on the one hand, and of Ss on

the other. S has been considered as a projection of the category Inf 1.

and S as that of comp.; both Inf 1. and comp. constituting the head in

each case.

In Chapter Five we have argued that Inf 1. should be treated as a

complex of the features [± AGRI and [± ASP]. This consideration is a

result of a background discussion on types of complement clauses and their

properties with regard to the theory of case. We have argued that

infinitival clauses may be [+ AGR - ASPI and that whenever this is so, a

lexical subject is always possible in the position, which is otherwise

filled by PRO. We have also observed that in some tensed clauses Inf 1.

may be characterized by [+ASP - AGRI. In such cases the subject can have

case, but it cannot be dropped because AGR which licenses its dropping is

not available in Inf 1.

Regarding the comp. node and its lexical realization, it has been argued

that in complement clauses it is realized as /akka/ 'that' and /kan/ 'of(?)'.

The latter is found mainly in relative clauses. In either case, a clause-

initial base position has been preferred to both final and clause-internal

positions. A clause-internal position is intuitively appealing, since

in surface structures the complementizers may appear in this position.

In relatin to the internal structure of both types of complement clauses,

we have argued that Oromo does not have relative pronouns. The element

/kan/ which has been believed to be the relative pronoun, is a

complementizer. The clause satisfies the projection principle by having

a (pro)nominal argument in the relativized position. This pronominal

argument moves to the position of the complementizer. Whenever this

takes place, the pronoun takes the case form of the head of the relative
 321

clause. This has some impact on Chomsky's theory of chains and their

properties, since the chain formed by the moved element and its trace is

characterized by two cases: the case of the trace and the case the moved

element takes from the head of the relative clause. Although the chain

is an A-chain, since the moved element is in comp., which is an A-position,

and that Chomsky leaves such chains aside, and concentrates only on

A-chains, others have argued that A-chains should be treated in the same

way A-chains are treated as a mapping principle. (See Koopman and

Sportiche 1982.)

An alternative analysis, which generates such pronominals in topic

position, has been suggested. This alternative avoids the chain and with

it the problems that the movement analysis has encountered. Furthermore,

according to rules of movement to comp. the pronoun cannot land in this

position because comp. attracts only wh-elements. The pronoun is not

such an element, though the movement is like that of a wh-elexnent. This

is not surprising given the fact that Oromo does not have wh-movement in

its syntax. It may be argued that the constraint that only wh-elements

should land in comp. may not apply to languages like Oromo, which have no

such movement per se. Such languages may use comp. as a landing site

for elements which correspond to wh-elements in languages which allow

wh-movement.

In predicate nominal clauses, we have noticed that Oromo has NP movement.

The movement is not only adjunction to comp. or to topic position, it also

includes raising to subject. This position is empty in D-structure and

in S-structure this position may remain empty, since Oromo does not have

non-referring (pleonastic) arguments. Equally, it may be occupied by the

subject of the complement clause, as in corresponding structures in

languages like English.

 327

The problem with this movement in Oromo is that the clause from which

the extraction takes place is finite, which means that the movement is not

motivated by case. The chain, which is an A-chain since the landing site

is in an A-position, is going to have two cases, one from the complement

clause and one from the matrix clause. In languages like English, the

trace of a raised NP has no case since the movement is triggered by the

absence of a case-assigner in the complement clause. In Oromo the trace

can receive case since there is a case-assigner. Further problems arise

in relation to the binding theory. The trace is an anaphor, and as such

should behave like any other anaphor. Such elements should be A-bound

(argument-bound) in their governing categories. The trace here satisfies

this condition since it is bound to the antecedent, but it is also free

since it can get its reference from AGR in Inf 1. in its own clause. In

this sense it behaves like a pronoun rather than an anaphor. In other

words, the trace is like a pronoun and an anaphor at the same time. This

is a property of PRO. This problem has been left unresolved and needs a

fresh start.
 328

APPENDIX

The following are a sample representative of the lexical categories

established in Chapter Two. The [-NI categories are indicated along with

their complements. The [-s-NI categories are not so indicated because

their complements are not obligatory. It is believed that the inclusion

of this appendix may provide data for anyone wanting to do further research

on this language.
 329

APPENDIX

abaar- (NP—] 'curse'

abboom- [NP—] 'obey'
abdat- [S — 'hope'
adams - (NP—] 'hunt'
agar s ii s - [NP NP—] 'show'
aaa?- stink'
aj'j'ees [NP—] 'kill'
aman- (NP/S — 'believe'
arain- [NP—] 'weed'
araars- [NP—] 'reconcile'
arg- [NP/S — 'see'
ariifat [PP — 'hurry'
aifat- 'sneeze'

'disappear'
bad-

bakat- [PP - 'flee'

ba7- [PP — 'go out'

bal ? is- [NP—] 'widen'

bar- [NP/S - 'learn'

barbaad- [NP/S —] want'

bars us - (NP NP/s — 'teach'

baat- [NP—I 'carry'

bit- (NP—] 'buy'

bobees- [NP—J 'burn'

bu ?- [PP - 'descend'

bul - [PP — 'spend a night'

abs - [NP—I 'break'

aal- [NP—] 'excel'

uf- [NP—I 'close'

dabal- [NP — ] ' increase'

daèaas- (NP — ] ' fold'
 330

daddaf- [PP - 'hurry'

daab- 'be tired'
damf- 'boil'
danda?- [S - 'be able'
dubbat- (PP - 'speak'
dukkana?- 'get dark'
'die'
(NP/AP/S - 'be'
dab - [NP-] 'lack'
daga? - [NP/S - I 'hear'
dal- [NP—] 'give birth to'
deebodd- [NP—] 'be thirsty'
deekkam- 'be angry'
di ?aat- [PP - I 'get near'
dils- [NP—] 'leave'
doks- [NP - 'hide'
do?- 'burst'
duf- 'come'
dug- [NP-I 'drink'
dukkubsat- 'get sick'

ebbis- [NP—] 'bless'

elm- [NP—] 'milk'
erg- [(PP) NP—] 'send'

fakkaat- [NP/AP - I 'resemble'
fak]caat- - I 'seem'

faaas- [NP—] 'sow'
fayyiis- (NP PP - I

save'

f jig- [(PP) - I 'run'

fina?- [NP—] 'urinate'

fi- [NP - 'finish'

fo ? - [NP—I 'choose'

fudat- [NP—] 'take'
 33j

gaafat- ((PP) NP—] 'ask'

ga?- [PP — 'suffice'
gal- (PP — enter'
gainmad- 'rejoice'
gargaar- (NP—I 'help'
god- (NPNP—] 'make'
gog- dry'
gowwaams - [NP—I 'fool'
gubat- (NP—] 'burn'
guddat- 'get big'
gunguum- 'murmur'
gurgur- [NP—I 'sell'
guut- [PP NP] 'fill'

haaddat- [NP—] 'shave'

haf- [S - 'remain'
haam- [NP - 'mow'
haxnmumat- 'yawn'
haasa ?- [PP - I 'talk'
hat- [NP—I 'steal'
hid- [NP—] 'tie'
hiik- [PP - I 'move'
him- [PP NP - I 'tell'
hir- (NP—] 'distribute'
h irinat- [PP NP — 'share'
ho6d- [NP—] sew'
ho 'et- [NP—] 'work'
hokkol 'limp'
hook- [NP—] 'scratch
hubat- [NP—] 'understand'
hukkat- 'get thin'

iaar- (NP—] 'build'
irraanfat- [NP/s — forget'
'cry'
iyy-
iyyom- 'get poor'
 332

3'aal lat- [NP—] 'love'

a1]ab- [NP/S - start'
5ed- [S - say'
5 ibb- [NP/S - 'hate'
[PP - 'exist'

ka?- [(PP) - 'stand up'

kaa?- [PP NP - 'put'
kaat- 'run'
kenn- [PP NP - 'give'
koif- (PP - 'laugh'
kutat- (NP—] 'cut'
lçab- [NP-I 'grasp'
[NP—] 'plough'
kuf- [PP - 'fall'
Jçufa?- 'cough'
çun is [NP - 'peel'

lakkaa?- [NP—] 'count'

1 ahab- [NP—) 'preach'
1i)eess- (PP NP - 'lend'

maaa?- [(PP) - 'get drunk'

mild- [NP—] 'hurt'
mii- [NP - 'wash'

naat- [NP—I 'eat'

oobaas- [NP—] 'water'

ool - (PP - 'spend a day'

raf- [PP - I 'sleep'

rakkat- 'be troubled'
raas- (NP—] 'shake'
roob- 'rain'
rukut- [NP - 'hit'
 333

sodaat- [NP—] 'fear'

ta? - [NP/AP - 'become'

teessis- [NP-I 'seat'

uf fat- [NP—] 'dress'

utaal- (NP - I 'jump'

waam- [NP—I 'call'

waraan [PP NP - 'pierce'

wayy- [PP - I 'be better than'

Yaad- [PP - 'think'

yadaat- [NP—] 'remember'

N
abba father'
adaadaa 'aunt'
adeemsa 'hunter'
aduu 'sun'
adurree 'cat'
afaan 'language'
aaga 'fortune'
ag&buu 'hunger'
(a) ala(a) 'outside'
albee 'knife'
akaakayyuu 'grandfather'
akkoo grandmother'
amartii 'ring'
andako 'hen'
ankaakuu 'egg'
aannan 'milk'
arba elephant'
arakee 'alcoholic drink'
areeda 'beard'
 334

arraba 'tongue'
aseeta tripe corn'
ayyaana 'grace'

ba ?a load'
balbala 'door'
bara 'year'
beekuma 'wisdom'
b ineensa 'animal'
birrii 'Ethiopian currency'
biaan 'water'
bitaa 'buyer'
biyyaa 'country'
bofa 'snake'
booru 'muddy'
bar 'tomorrow'
bokkaa 'rain'
buddeena 'bread'
buna 'coffee'

daaa 'valley'

daadii 'mead'

dafinoo 'Monday'

daggala 'forest'

damma 'honey'

dubartii 'woman'

durba 'girl'

dadaa 'butter'

dagaa 'stone'

deena 'food'

dugunia 'truth'

dukkuba 'illness'

farda 'horse'

far soo 'beer'

fayyiisaa 'saviour'

foon 'meat'
 335

ganna 'winter'
ganda 'village'
gaangee 'mule'
gaara 'mountain'
gara 'stomach'
garbuu 'barley'
gatii 'price'
gurra 'ear'
gurbaa 'boy'
guyyaa 'day'

haaduu 'knife'
haaa 'mother'
haafa 'dirt'
halkan 'night'
(h)amma 'now'
har? 'today'
harree 'donkey'
harka 'hand'
V.
hojii 'work'
hoolaa 'sheep'
horii 'wealth'
humna 'power'

ibidda 'fire'
ij a 'eye'
i'oolee 'children'
ilk aan 'tooth'
intala 'girl'
irbaata 'supper'

j aldeessa 'monkey'
ara 'people'

kara a 'road'

kalaadii 'measure of land'
 336

kamadii 'wheat'
keereens i 'leopard'
]il leensa 'wind'
konna 'farm'
koraari 'ir
kuba 'finger'
lurlumli 'fish'

laf a 'land'
laga 'river'
laagaa throat'

makaa 'name'
mana 'house'
marga 'grass'
miillaa 'leg'
muaa 'child'
muka 'tree'

nadeen 'woman'
nagaa 'peace'
nama 'man'
naata 'food'
niitii 'wife'

obboo 'mister'
obbolees sa 'brother'
od uu 'news'

rifeensa 'hair'

sagalee 'voice'

sang aa 'ox'

sa?a 'cow'

s aree 'dog'
 33?

s iib iila 'steel'
sirba song'

siree 'bed'

soba 'lie'

warakata 'paper'

warkii 'gold'

wayaa/wayyaa 'clothes'

yeroo 'time'

A
aadii 'white'
afaalaa 'clever'
ar5aa 'generous'

ba?eessa 'fine'
bal?aa 'wide'
bareedaa 'handsome'
bay?e 'much'

aalaa cunning
imaa 'strong'

dalaa 'grey'
diiinaa 'red'
deeraa 'tall'

fagoo 'far'

gababaa 'short'

gaarii 'good'

gurraacca 'black'

guddaa 'big'
 338

hamaa 'evil'
happii 'thin'
haraa 'new'

iyyes sa 'poor'

abaa 'strong'

naafa 'am

sooressa 'rich'

taliila 'clear'
tilloo 'little'
tinnoo 'small'

Adpositionals
akka [ - NP/si 'as'

booddee [NP - after'

duuba [NP - 'behind'

duura [NP - 'before'

erga [ - NP/SI 'after'

-f [NP—] for'

gara [ - NP] 'to'

gidduu [NP - 'between'
gubba [NP - 'on'

(h)amina [—S/NP] 'until'

 339

irra [NP - 'from'

-itt (NP - 'from'
-itii (NP - 'towards'

'ala [NP - 'under'

malee [NP - 'without'

-n (NP—] 'with'

otuu [—SI 'while'

wain [NP - I 'with'

yolnmuu/yommu ( - SI 'when'

Pro nominal s

nom. acc.

an i (a)na I

ati si. OU

inni isa he

iseen ie she

nuy flu we

isiin isiin you

isaan isan(i) 'they'

wan/tokoo 'anyone'

hundaa 'all'

eennu 'who'

maali mal 'what'

 340

kam(i) kam 'which'

kun(i) kana 'this'

sun(i) sana that

asi 'here'

ai 'there'
 341

REFERENCES

Abera Nafa. 1982. 'Oromo Verb Inflection', Institute of Language Studies,

Addis Ababa University, ms.

Aissen, J. 1974. 'Verb Raising', Linguistic Inquiry 5:3, 325-66.

Anderson, S.R. 1982. 'Where is Morphology?', Linguistic Inquiry, 13:4,

571-612.

Andrews, A. 1981. 'The Representation of Case in Modern Icelandic',

in J. Bresnan (ed.), The Mental Representation of Grammatical
Relations (Cambridge, Mass.: MIT Press).

Andrzejewski, B.W. 1957. 'Some Preliminary Observations on the Borana

Dialect of Galla', in Bulletin of the School of Oriental and
African Studies, XIX:2, 354-74.

1960. 'The Category of Number in the Borana Dialect of Galla',

Africa, XXX:1, 62-75.

1966. 'The Role of Tone in the Borana Dialect of Galla', in

Proceedings of the Third International Conference of Ethiopian
Studies (Addis Ababa Haile Selassie I University).

Aronoff, M. 1976. Word Formation in Generative Grammar, Linguistic Inquiry

Monograph One (Cambridge, Mass.: MIT Press).

Backhouse, A.E. 1984. 'Have all the Agjectives Gone?', Lingua, 62:3, 169-87.

Baker, M. 1985. 'The Mirror Principle and Morphosyntactic Explanation',

Linguistic Inquiry, 16:3, 373-416.

Baye Yimain. 1981. 'Oromo Substantives: Some Aspects of their Morphology

and Syntax', Institute of Language Studies, Addis Ababa University, ms.

1984. 'Towards a Definition of Nominal Specifiers in Airtharic',

to appear in Proceedings of the Eighth International Conference of
Ethiopian Studies, Addis Ababa University.

Bender, M.L. and Mulugeta Eteffa. 1976. 'Galla', in Bender et al. (eds.),
Language in Ethiopia (London: Oxford University Press).

Bliese, L.F. 1981. A Generative Grammar of Af far (The Summer Institute

of Linguistics and the University of Texas at Arlington).

Borello, Mario. 1939. Grammatica di Lingua Galla (Torino: Instituto

Missioni Consolata).

Borer, Hagit. 1984. Parametric Syntax: Case Studies in Semitic and

Romance Languages (Dordrecht: Foris).

Bouchard, J. 1984. The Content of Empty Categories (Dordrecht: Foris).

Bresnan, J. 1976a. 'On the Form and Functioning of Transformations',

Linguistic Enquiry, 7:3, 3-41.


4 4 C

Bresnan, J. 1978. The Theory of Complementation in English Syntax

(Cambridge, Mass.: MIT Press)

1982. 'Control and Complementation', Linguistic Inquiry,

343-434.

Bresnan, J. and J. Grimshaw. 1978. 'The Syntax of Free Relatives in

English', Linguistic Inquiry, 9:3, 331-92.

Bresnan, J. and R.M. Kaplan. 1981. 'Lexical-Functional Grammar: A

Formal System for Grammatical Representation', in J. Bresnan (ed.),
The Mental Representation of Grammatical Relations (Cambridge, Mass.:
MIT Press).

Bryan, M.A. 1959. 'The T/K Languages: A New Substratum', Africa, XXIX:1,
2-17.

Chomsky, N. 1965. Aspects of the Theory of Syntax (Cambridge, Mass.:

MIT Press).

1970. 'Remarks on Nominalization', in R.A. Jacobs and

P.E. Rosenbaum (eds.), Readings in English Transformational Grammar
(Waltham, Mass.: Ginn and Co.).

1973. 'Conditions on Transformations', in S. Anderson and

P. Kiparsky (eds.), A Festschrift for Morris Halle (New York:
Rinehart and Winston).

l976a. 'Conditions on Rules of Grammar', Linguistic Inquiry, 2:4,

303-52.

1977. 'On wh-movement', in A. Akinajian, P. Culicover and

T. Wasow (eds.), Formal Syntax (New York: Academic Press).

1980. Rules and Representations (New York: Columbia University

Press).

1981. Lectures on Government and Binding (Dordrecht: Foris).

1982. Some Concepts and Consequences of the Theory of Government

and Binding (Cambridge, Mass.: MIT Press).

1986. Knowledge of Langua Its Nature, Origin and Use

(New York: Praeger).

Chomsky, N. and H. Lasnik. 1977. 'Filters and Control', Linguistic

Inquiry, 8:2, 397-411.

Cinque, G. 1977. 'The Movement Nature of Left Dislocation', Linguistic

Inquiry, 8:2, 397-412.

Comrie, B. 1976. Aspect (Cambridge: Cambridge University Press).

1981. Language Universals and Linguistic Typology

(Oxford: Basil Blackwell).
 343

Comrie, B. and Norval Smith. 1977. 'Lingua Descriptive Series:

Questionnaire', Lingua 42, 1-72.

Da Thiene, P.G. 1939. Dizionario della Lingua Galla (Harar:

Vicariato Apostolico).

Demissie Manahilot. 1977. 'Nominal Clauses in Amharic', PhD thesis,

Georgetown University.

Dixon, R.M. 1977. 'Where have all the Adjectives gone?', Studies in
Language 1:1, 19-80.

Emonds, J.E. 1976. A Transformational Approach to English Syntax

(New York: Academic Press).

Eshetu Kebede. 1981. 'The Verb to be in Oromo', MA dissertation,

Institute of Language Studies, Addis Ababa University.

Foot, E.C. 1913. A Galla-English English-Galla Dictionary (Cambridge:

Cambridge University Press).

Gragg, G. 1976. 'Oromo of Wellega', in M.L. Bender (ed.), The Non-Semitic

Languages of Ethiopia (Michigan: Michigan State University Press).

1982. Oromo Dictionary (Michigan: Michigan State University Press).

Greenberg, J.E. (ed.). 1963. Universals of Language (Cambridge, Mass.:

MIT Press).

Grosu, Alexander and J. Horvath. 1984. 'The GB Theory and Raising in

Rumanian', Linguistic Inquiry, 15:2, 348-53.

Haacke, Catherine. 1970. 'Die Konjunktionen im Galla Eine Syntaktisch-

Semantische Analyse', PhD thesis, Humboldt University zu Berlin.

Hailu Fulas. 1972. 'On Ainharic Relative Clauses', Bulletin of the School
of Oriental and African Studies XXV, 497-513.

Haimanot Moges. 1983. 'Pronominals in Oromo', Institute of Language

Studies, Addis Ababa University, ms.

Hayward, R.J. 1975. 'Middle Voice Verb Forms in Eastern Cushitic',

Transactions of the Philological Society, 203-24.

1976. 'Categories of the Predicator in Affar, with Special

Reference to the Granixnar of Radical Extensions', PhD thesis,
University of London.
 344

Hayward, R.J. 1976a 'A Question in Oromo Morphophonology', Folia

Orientalia 17, 29-40.

1978. 'The Stative Conjugation in Affar', Estratto da Annali

deli Instituto Oriental di Napoli VI, 38, 1-40.

1984. 'A Reconstruction of Some Root Extensions on the Eastern

Cushitic Verb', in Current Issues in Linguistic Theory VOl.28,
69-109.

1984. The Arbore Language: A First Investigation (Hamburg:

Buske Verlag Helmut).

Heine, B. and Wilhelm Mohiig. 1980. Geographical and Introduction,

Language and Society, Language and Dialect Atlas of Kenya,
Vol.1 (Berlin: Reimer).

Higginbotham, J. 1983. 'A Note on Phrase-Markers', Revue Que'becoise

de Linguistigue 13.1.

Hodson, C.M.G. and C.H. Walker. 1922. An Elementary and Practical

Grammar of the Galla or Oromo Language (London: Society for
Promoting Christian Knowledge).

Hoekstra, Teun. 1984. Transitivity Grammatical Relations in Government

and Beinding Theory (Dordrecht: Foris)

Hornstein, N. 1977. 'S and X' Convention', Linguistic Analysis 3, 137-76.

Huang, C.T.J. 1982. 'Logical Relations in Chinese and the Theory of

Grammar', PhD thesis, MIT.

Jackendoff, R. 1972. Semantic Interpretation in Generative Grammar

(Cambridge, Mass.: MIT Press).

1976b. 'Constraints on Phrase Structure Rules', in

A. Akmajian, P. Culicover and T. Wasow (eds.), Formal Syntax
(New York: Academic Press).

1977. X-Syntax: A Study of Phrase Structures. Linguistic

Inquiry Monograph 2 (Cambridge, Mass.: MIT Press).

Jensen, John T. and M. Stong-Jensen. 1984. 'Morphology is in the

Lexicon', Linguistic Inquiry, 15:3, 474-99.

Kabada Hordofa. 1981. 'Nominalization Patterns in Oromo', Institute

of Language Studies, Addis Abada University, ms.

Koopman, Hilda. 1984. The Syntax of Verbs, from Verb Movement Rules
in the Kru Languages to Universal Grammar (Dordrecht: Foris).

Koster, Jan. 1978. 'Why Subject Sentences don't exist', in J. Keyser (ed.),
Recent Transformational Studies in European Languages
(Cambridge, Mass.: MIT Press)
 345

Larson, R.K. 1985. 'Bare NP Adverbs', Linguistic Inquiry, 16:4, 595-623.

Launhardt, J. 1973. A Guide to Learning the Oromo Language

(Addis Ababa).

Lees, R. 1960. The Grammar of English Nominalizations (The Hague: Mouton).

Leiber, R. 1981. 'Morphological Conversion within a Restrictive Theory

of the Lexicon', in Moortgat M.H.L. der Hulst and T. Hoekstra (eds.),
The Scope of Lexical Rules (Dordrecht: Foris).

Lyons, J. 1968. Introduction to Theoretical Linguistics (Cambridge:

Cambridge University Press).

1977. Semantics, Vol.11 (Cambridge: Cambridge University Press).

McCawley, J. 1982. 'Parentheticals and Discontinuous Constituent Structures',

Linguistic Inquiry 13, 91-106.

Manzini, M.R. 1983. 'On Control and Control Theory', Linguistic Inquiry 14,
421-46.

Marantz, Alec P. 1984. On the Nature of Grammatical Relations

(Cambridge, Mass.: MIT Press).•

Matthews, P.H. 1981. Syntax (Cambridge: Cambridge University Press).

Moreno, M.M. 1939. Grainmatica Teorico-Pratica della Lingua Galla

(Rome: A Mondadori).

Newmeyer, F.J. 1980. Linguistic Theory in America (New York: Academic Press).

Nordfeldt, M. 1947. Galla Grammar (Lunth Le Monde Oriental).

Nuessel, F.H. 1979. 'R. Jackendoff, X Syntax: A Study of Phrase

Structure', Lingua, 49:4, 255-58.

Owens, J. 1985a. The Oromo Causative: Lexical Grammar without Lexical

Rules. Reproduced by Indiana University Linguistic Club, Bloomington.

1985. 'Nominal Relations in Syntactic Dependency', ms.

Parsons, T. 1970. 'The Analysis of Mass Terms and Amount Terms',

Foundations of Language, 6, 362-388.

Picallo, M.C. 1984. 'The Inf 1. Node and the Null Subject Parameter',
Linguistic Inquiry, 15:1, 75-103.

Praetorius, F. 1893. Zur Graxnmatik der Gallasprache (Berlin: Verlag von

Reuther and Reichard).
 348

Radford, Andrew. 1981. Transformational Syntax (Cambridge: Cambridge

University press).

Riemsdijk, H.C. van. 1978. A Case Study in Syntactic Markedness:

The Binding Nature of Prepositional Phrases (Dordrecht: Foris).

Riemsdijk, H.C. van and Edwin Williams. 1986. Introduction to the Theory
of Grammar (Cambridge, Mass.: MIT Press).

Rouveret, A. and J.R. Vergnaud. 1980. 'Specifying Reference to the

Subject: French Causatives and Conditions on Representation',
Linguistic Inquiry, 11:1, 97-202.

Saeed, I.J. 1979. 'Copula Reduction: An Argument for the Category

Adjective in Somali', School of Oriental and African Studies, ms.

1980. 'The Syntactic Status of Quantifiers in Somali', School

of Oriental and African Studies, ms.

1982. 'Focus and Topic in Somali', PhD thesis, School of

Oriental and African Studies, University of London.

Sasse H.J. 1977. 'A Note on Wh-movement", Lingua 14:4, 343-54.

Selkirk, E. 1977. 'Some Remarks on Noun Phrase Structure', in Akmajian,

Culicover and Wasow (eds.), Formal Syntax (New York: Academic Press).

1982. The Syntax of Words. Linguistic Inquiry Mongraph 7.

(Cambridge, Mass.: MIT Press).

Serzisko, F. (No date.) 'Collective and Transnuineral Nouns in Somali',

University of Cologne.

Stowell, T. 1981. 'Origins of Phrase Structure', PhD thesis, MIT.

Stroomer, H. 1986. 'On the Borana, Orma and Waata Dialects of Oromo',
paper presented at the International Symposium on Cushitic and
Omotic Languages, Cologne.

Travis, L. 1984. 'Parameters and Effects of Word Order Variation',

PhD thesis, MIT.

Wako Tola. 1981. 'The Phonology of Oromo (Mecca Dialect', MA dissertation,

Institute of Language Studies, Addis Ababa University.

Wasow, T. 1972. 'Transformations and the Lexicon', in Akainajian, Culicover

and Wasow (eds.), Formal Syntax (New York: Academic Press).

Webster, E.J. 1960. 'The Particle in Booraan', African Studies 19, 33-43.
 34?

Williams, E. 1980. 'Predication', Linguistic Inquiry, 11:1, 203-38.

1981. 'On the Notions "Lexically Related" and "Head of a Word",

Linguistic Inquiry 12, 245-74.

1984. 'Grammatical Relations', Linguistic Inquiry, 15:4, 639-72.

Zaborski, A. 1975. 'The Verb in Cushitic', in Studies in Hamito Semitic

(Krakow: Uniwersytet Jagiellonski).