REPORT RESUMES

ED 010 517 48
TAGMEMIC AND MATRIX LINGUISTICS APPLIED TO SELECTED AFRICAN
LANGUAGES. FINAL REPORT.
BY- PIKE, KENNETH L.
MICHIGAN UNIV., ANN ARBOR
REPORT NUMBER BP. -5 -1301 PUB DATE NOV 66
CONTRACT 0Er-4-065
EDRS PRICE MF-$0.45 HC-$11.88 297P.

DESCRIPTORS- *AFRICAN LANGUAGES, PHONOLOGY, LINGUISTICS,

*STRUCTURAL ANA,YSIS, *LANGUAGE PATTERNS, LANGUAGE RESEARCH,
*LINGUISTIC PATTERNS, *GRAMMAR, TABLES (DATA) : ANN ARBOR,
MICHIGAN

THE APPLICATION OF TAGMEMIC AND MATRIX TECHNIQUES IC

SOME PROBLEMS IN AFRICAN DESCRIPTIVE LINGUISTICS, AND Thc
ILLUSTRATION OF THIS APPLICATION WITH DATA SOURCES (RESTATED
IN TAGMEMIC TERMS) WERE PRESENTED. (TAGMEMIC AND MATRIX
TECHNIQUES REFER, RESPECTIVELY, TO THE STUDY OF (1) ONE-WORD
OR MULTIWORD GRAMMAR UNITS AND THE FUNCTIONS OF THESE UNITS
WITHIN SENTENCE CONSTRUCTION AND (2) LANGUAGE ELEMENTS WITHIN
SENTENCE STRUCTURES BY USING TABULAR RELATIONSHIPS.) PROJECT
STUDIFS CENTEREP ON 18 NIGER -CONGO WEST AFRICAN LANGUAGES,
COVERING THE SUBJECTS OF CLAUSES, CLAUSE CLUSTERS IN
SENTENCES, SENTENCE CLUSTERS IN PARAGRAPHS (NARRATIVE AND
DISCOURSE), NOUNS AND NOUN PHRASES, VERBS, PHONOLOGY, AND
PEDAGOGICAL PLANNING. PAPERS WERE INCLUDED IN THE REPORT ON
THE MBEMBE CLAUSE SYSTEM, SERIAL CONSTRUCTIONS IN KASEM,
INDEPENDENT CLAUSES OF DEGEMA, GRAMMATICAL PROSODIES, NOMINAL
AND VERBAL GROUP MATRICES FOR KASEM, NOUNS OF ETUNG, VERB
STRUCTURES IN ETUNG, PARALINGUISTICS IN MBEMBE. AND
PHONOLOGICAL DATA OF AGBO. (GC)
4

1.14 S: DEPARTMENT OF HEALTH, EDUCATION AND WELFARE

of Education
(figs document has ; -..-::,)ducer.: exactly as receiyed from the
person or organ -Jr -gating it. Po:nts of view or opinions
stated do not necessarily represent official Office of Educatioll
position or policy.

FINAL REPORT
Contract No. 0E-5-14-065
4f. 5-150 I

Tagmemic and Matrix Linguistics

Applied to Selected African Languages

November 1966

U. S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE

Office of Education
Bureau of Research
 TAGMEMIC AND MATRIX LINGUISTICS
APPLIED TO SELECTED AFRICAN LANGUAGES,

Contract No. 0E-5-14-065

Kenneth L. Pike

November 1966

The research reported herein was performed pursuant to a

contract with the Office of Education, U. S. Department
of Health, Education, and Welfare. Contractors under-
taking such projects under Government sponsorship are
encouraged to express freely their professional judgment
in the co: duct of the project. Points of view or opinions
stated do not, therefore, necessarily represent official
Office of Education position or policy.

The University of Michigan

Center for Research on Language
and Language Behavior

Ann Arbor, Michigan

 ii
TABLE OF CONTENTS

Page

INTRODUCTION 1

CHAPTER

I: CLAUSES 12

1.1. Clause Typology 12

fl
1.1.1. Basic (Kernel, Nuclear) Clause Types 12
U
1.1.2. Derived (Marginal) Clause Types 18
1
il 1.2. Clause Contrasts 26

1.3. Clause Variants 31

1.3.1. Variation by Presence of Nuclear Tagmcmes 32

1.3.2. Variation by Optional and Marginal Tagmemes 32

1.3.3. Variation by Order of Tagmemes 33

N 1.3.4. Variation by Simple Substitution 34

1.3.5. Variation by Substitution of Manifesting Constructions 34

1.3.6. Variation by Agreement (Concord) 35

a 1.3.7. Variation by Occurrence in Clause Clusters 36

1.4. Clause Distribution 38

1.5. Some Clause Components of Bobangi (Bantu) 38

1.6. Some Clause Components of Hausa (Chad, Afroasiatic) 47

II: CLAUSE CLUSTERS IN SENTENCES 55

2.1. The Sentence as Setting for Clause Distribution 55

2.2. Clause Clusters (Serial Clauses) 57

2.2.1. General Restrictions on Sequence Types in Clusters 57

2.2.2. Specific Restrictions on Sequence Types in Clusters 60

1 2.2.3 Restrictions on Tagmemes within Clause Clusters 62

 I
TABLE OF CONTENTS (continued)

CHAPTER Page

2.2.4. Agreement Restrictions within Clause Clusters 65

2.2.5. Development of Clause Subclusters 66

2.2.6. Development of Auxiliary Verbs from Clause Clusters 71

III: BEYOND THE SENTENCE 79

D
3.1. Sentence Clusters in Paragraphs 79

3.2. Discourse Structure 84

3.2.1. Narrative 84
Q
3.2.2. Indirect versus Direct Discourse 86

IV: NOUNS AND NOUN PHRASES 93

a LA

4.1. Types of Noun Phrases 93

fl
4.2. Development of Compound Nouns 99 Li

4.3. Syntactic Markers Developed from Nouns 104

4.4. Irregularities within Noun Classes as Distortion of

a
Field 106

4.5. Noun Concord 108

4.5.1. Ranking in Singular-Plural Prefix Pairs 109

V: VERBS 114

5.1. Vowel Fusion in Matrix Display 114

5.2. Tone Pattern in Verbs

5.3. Hausa (Afroasiatic) Formatives in Person-Aspect

Markers 120

VI: PHONOLOGY 125

6.1. Quasi-Isochronic Units of an Accentual Hierarchy 125

6.2. Tone 132

6.2.1. Basic Tone Heights 132

-iv-
4
TABLE OF CONTENTS (continued)

CHAPTER Page
1

6.2.2. Extra-High Tone Developed from Lost High 133

6.2.3. Lowered Key (Terrace) Developed from Lost Low 135

6.2.4. Terrace Tone Developed from Low-Replaced-by-High 136

6.2.5. Indications of a Developing Downstep in a Three-Level

System 137

6.2.6. Overlap of Conditioned Allotones 139

6.2.7. Towards Internal Reconstruction of Tone Chains 141

6.2.8. Theory of Phoneme Types as Item, Process4and Relation 144

a 6.3. Segmental Phonemes 145

6.3.1. On the Phonetics of Vowel Harmony 145

6.3.2. General Phonetics 145

6.3.3. On Data Collection Preparatory to Workshop Analysis 146

6.4. Intersection of Voice Quality with Gesture 147

VII: IMPLICATIONS OF MATRIX APPROACH IN PEDAGOGICAL PLANNING 150

VIII: SUMMARY 154

j fl
-- APPENDICES

I: Notes on the Mbembe Clause System - A preliminary analysis,

Kathleen Barnwell 156

0 II: Preliminary Notes on Serial Constructions in Kasem, Kathleen

Callow 182

III: Preliminary Paradigm of Some Degema Independent Clauses, Elaine

Thomas 187

IV: Grammatical Prosodies ??, John T. Bendor-Samuel 192

V: Nominal and Verbal Group Matrices for Kasem, J. C. Callow 197

VI: Nouns of Etung Classified by Their Singular-Plural Prefix Pairs,

Eileen Edmondson 206

-v-
 TABLE OF CONTENTS (continued)

APPENDICES Page

VII: Preliminary Description of Some Verb Structures in Etung, Tom

Edmondson 227

VIII: Preliminary Report on Para-Linguistics in Mbembe (E. Nigeria),

P. M. Revill 245

IX: An Interim Workshop Report on the Phonological Data of Agbo,

Klaus and Janice dpreda 255

X: Some Articles in Preparation 289

 INTRODUCTION
Goal: The purpose of this study is (1) to discuss the application of tagmemic
and matrix techniques to some problems in African descriptive linguistics, and
(2) to illustrate this application (a) with data from published sources, re-
stated in tagmemic terms, and (b) with hitherto unpublished data, gathered prin-
cipally by members of the Summer Institute of Linguistics.
By a tagmeme we mean a unit of grammar, such as the predicate or the ob-
ject of a sentence; the tagmeme is viewed as having (a) function in a construc-
tion--"object," for example--and (b) a set of forms--e.g., The boy, John, some-
body-- which can come in the appropriate slot. By a linguistic matrix we mean
a table of language elements; rows and columns represent significant properties
of a structural system or subsystem; entries in the boxes of the table signal
properties of the system; and--preferably--the rows and columns are ordered so
as to show in the most effective way possible the relation of the semantic prop-
erties of the system to groups of entries in the table.
It is assumed that most readers of this report will have had initial ex-
posure to the tagmeme concept through--say--the introductory pedagogical treat-
ment of Benjamin Elson and Velma Pickettl; and the matrix concept seen in one
of the available articles.2 A tagmemic bibliography, complete up through 1964
but already outdated, is also available.3 An attempt to bring tagmemics into
relation with some other theories (as of 1964, but also now incomplete because
of rapid developments within other theoretical approaches) is found in bibli-
ographical sections of my larger work4 on tagmemics.
From the first (1954) volume of this latter study, there was an attempt
to locate and specify certain characteristics which were universal to all the
languages of man--and to the structure of his nonverbal behavior as well. It

was urged that all purposive human activity and perception were constrained
by in-b a elements of the human mechanism. Different languages--or sets of
nonverbal behavior patterns--might be learned from one's cultural surroundings,
but each such set was in turn restricted by the nature of man himself. I

sought there to exploit samples of observed characteristics of language behav-

ior to deduce limits on the total kinds of possible language features. Extrap-
olation from these samples aimed at a systematic, "etic" summary and presenta-
tion of this kind of language potential.

-1-
L.

--2-

The kinds of universal characteristics postulated included the presence

of units in all languages. It was affirmed that units, to be presented
adequately, must be described in reference to contrasts between them, vari-
ability within them, and distribution of small units within larger units and
within classes and systems of units.
Inclusion within successively larger units was possible, in turn, only
within a theory such as tagmemics which gave due weight to hierarchical order-
ing of units. Since, however, the borders of unit types did not always coin-
cide, a multiple-hierarchical approach (lexicon, phonology, grammar) had to
be postulated (with analogues in nonverbal areas)--and these features, also,
were set up as universals. Relation between these hierarchies, in turn, in-
volved fusion between units of one hierarchy (e.g., lexicon) in part induced
by their inclusion in--and modification by--another (e. g., the high-level
phonological units such as stress groups).
The emphasis of the approach was heavily heuristic: The scheme of con-
cepts, etic typologies, and unit descriptions, was utilized first of all as
a basis for the search for pattern in language. The discovery of structure
F

within numerous languages was greatly facilitated thereby.

Later (say 1959 to 1965) the portrayal of certain kinds of pattern was
aided by the development of tabular displays (matrices) in which contrastive
U
features of a syntactic system served as the dimensions of the system (much
as articulatory features serve as the dimensions of a phonetic chart). Through
the relationship between such matrices, the relationship between sentences
could be more easily discussed.
Pattern seen through matrix, furthermore, entered the conceptual frame-
work of field, in this over-view of language. Particle, wave, and field pro-
vided diverse perspectives for emphases on static, dynamic, and functional
characteristics of units-ia-system.
Units were seen first of all as particles segmented out of the stream
of speech. When, however, the units were viewed as having indeterminate non-
segmentable boundaries in a dynamic flowing hierarchy of nuclear and marginal
elements, a wave perspective was more appropriate. As points in an intersect-
ing set of contrastive categories, units such as clauses were seen in field
perspective.
 *144.4401., ....0 ' ...........ve..........4a ..........

et

-3--

The extension of the matrix approach to morphological data led to the

development of technological procedures for grouping together bits of mate-
1
rial which, in a classical approach, would be seen as highly irregular. In

a matrix approach some of these appeared to have other kinds of regularity. t,

The field approach probed beneath the morphemic layer of structure--threaten-

ing the classical view of the morpheme, but re-establishing the morpheme as
a special instance of a field formative. (A simple morpheme, in this view,
is a vector formative, a phonological segment coming in each cell of one
vector--one row or one column--of a matrix, and therefore in a one-to-one
relation of phonological shape to matrix category). Irregular morphemic
systems were thereby seen in a different perspective. Irregularity, from
'one viewpoint, might sometimes represent regularity from another.
Meanwhile, the development of transformational grammar had had a heavy
impact on linguistic studies. Tagmemic investigations had not been immune.
It seems highly probable that the implicit transformational question5 "How
=M.
are sentences related?" as further developed by Noam Chomsky6 may have hastened
the development of such approaches as that of matrix multiplication. Earlier
yet, the work of Lounsbury and Goodenough on kinship systems was an encourage-
ment to develop clause matrices. I am happy to acknowledge these influences.
r
Already many tagmemic descriptions are using a transformational component,
encouraging us to hope, as I stated in 1960,8 that the two approaches should
some day 'come to a point of complete overlap.' i
t

The generative approach to grammatical description, as worked out by

Chomsky and his associates, is also beginning to have its impact on tagmemic
discussion. Longacre, for example, has given us a first attempt9 at making
explicit the generative power of tagmemic formulas. On the whole, nevertheless,
tagmemics has itself had a much greater contribution to the making explicit
of heuristic procedimes than to the deductive generative ones. This is not
surprising, in view of the many unwritten languages being studiedl° by people
trained in tagmemic theory and practice. In such circumstances, the search
for pattern must precede any descriptive or generative grammar.
Tagmemic interests, furthermore, have continued to broaden. From a
simple dismission--or listing--of types of discourse in 1954, study has now
 -4-
11
widened to include the nature of paragraphs, and the study of principles
of rhetoric itself--or the discussion of poetry. From the early treatment of
nonverbal behavior have grown extensive studies of social community. 12 Peda-
13
gogical principles have also been illuminated.
Even so, however, a severe limitation lay on the theory: Its basis was
a biased sample, with its inductive sources chosen largely from languages of
North America, South America, New Guinea, and the Philippines: The postulated
language--or human--universals needed to be tested in Africa and in the Orient
before the empirical sample was large enough to allow us to rest assured of
adequate coverage. What, specifically, might a study of African languages re-
veal which would either confirm the usefulness--and truth--of the tagmemic
view of human nature and human languages or force its rejection, revision, or
amplification?

Data of the real world are notoriously more varied than the unaided imagina-
tion of man. The philosopher at his desk has never been able to cono.oct a
world as rich as that outside his window. Such a large block of the world as
Africa, therefore, could likely be concealing some surprises. These surprises
might be of several general kinds. One is the addition of new kinds of data
never seen before. Another brings forth data which are "new" in a subtler
sense: It consists of the discovery of an exaggerated form of a known problem- -
a problem which previously was known only in a form so simple that it did not
warrant the development of an elaborate apparatus to handle it--or so simple
that it was not complex enough to allow the researcher to predict the kind of
formula needed for which this known problem was but a special instance- A
third type of surprise is one for which there is less excuse, but which meets
us all: the forcing to attention of data already described by other research-
ers, but in a language in which he has had but little interest, so that the
special solutions already achieved have not had their deserved impact on him.
All of these were found in the African project being reported here, An
early search of the literature, with the help of graduate students who attempted
to restate in tagmemic terms data from various parts of Africa, guaranteed
broader coverage than could otherwise have been achieved. The written sources
included reference to West, East, and South Africa. Study of live data was
confined to West Africa--to Ghana, Nigeria, and Dahomey.
 4

-S-

Of kinds of material new to me, perhaps the most interesting was on dis-
course structure. The intricate "ranking" matrices--or the ordered rules- -
required to describe Bariba use of direct versus indirect quotations were ele-
gant in beauty and totally unexpected. It was almost as if stage directions
for spotlighting characters were translated into language rules.
Equally surprising and elegant--though with less obvious semantic impli-
cations--were the elaborate matrix relations required to give insight into the
subtle traces of broken down concord in Abua verbs. When an independent o'::,ect

occurred in the clause, verb prefixes sometimes underwent a set of changes which
could be understood best as the fusion of forms of an old lost concord matrix.
A further surprise came with the concord elements in Etung noun classes.

Here there were patterned constraints in the relation of singular to plural

forms, determined by a ranking of reconstructed front versus back elements (such
that singular form x could be paralleled by plural y, z, or w; singular y by
plural z or w; singular z by plural w, etc.). These two results were especially
gratifying, in that the study of concord was one of the particular aims of the
project, since prior work on morphological forms had implied that complexity of
this type might be illuminated by this approach. Matrix approaches, therefore,
were justified not only for the micro-analysis of concord relations--as we had
hoped--but also for macro-analysis in discourse structure. Further support to
micro-analysis came from work in Bimoba verbs (where-scattered remnants of reg-
ularity could be abstracted from massive irregularity), and in Kasem noun forms.
Support for discourse analysis in tagmemic terms mile additionally from Sisala
(with time words marking early places in discourse) and from Vagala (with se-
quences of sentences, the first of them marked for focus, comprising emic para-
graphs).

A kind of structure which I had previously observed (in some variety) in

14
New Guinea, but for which I then had no useful solution, was a major, accen-
tuated problem in West Africa: Clauses occurred in a series, or cluster, such
that several could share the subject Gf the first (omitted from the remaining
clauses of the series), the object of the first (or each have its own object),
and some other tagmeme or tagmemes (e.g., location). This kind of grouping
led to several consequences: (1) the whole series formed one entity, as in
Kasem, distributed in a sentence slot; (2) certain standard types of sequences
 -6-

could, by various historical stages, decay into lower-level verb phrases (with
one of the original free verbs becoming averbal auxiliary or particle), as in
Vagala; or, on the contrary, (3) the original series could remain a subcluster,
but become tagmemically specialized (filling a slot in a larger including cluster),
building up higher-level structures.
Probably tone structures, on the other hand, best illustrate the kind of
problem which had been described by other scholars, but not adequately worked
15
into my own thinking. In the middle of this century, William Welmers called
attention to "terrace tones." (This is a phenomenon where a tone which normally
is high after another high is sometimes lowered a bit after a high--but contin-
ues to have following it, at whatever level it then happens to be, other high
tones of the same system. Both the original high and the stepped-down high con-
trast with a low in that position after high. The step-down is occasioned,
historically, by the loss of a low tone, between the two original highs, which
lowered the second of the two by--then--non-phonemic conditioned variation.)
I have enjoyed experimenting with treating the "downsteppingness" as a "process"
phoneme--rather than as a segmental one.
The extensive morphophonemic changes, furthermore, require a special handling
16
of the relation of tone patterns to words as wholes rather than to stems or
affixes as more-or-less discrete parts. Work on Etung and Bimoba highlighted
this problem at the interface between lexicon, grammar, and phonology.
In phonetics proper, the most interesting item was the study of vowel quality
17
underlying the vowel harmony of Twi, a study undertaken withDr. John M. Stewart,
(who emphasized the linguistic reasons for needing a--Firthian--prosody for the
"close" subset) and Dr. Ruth M. Brend of the University of Michigan and Michigan
State University (who with Charles Peck of the University of Michigan provided
the instrumental analysis). Instead of treating /i/ and /I/ as differing by
either tense and lax, or close and open features, it appears best to treat the
'open It set as basic, with the other set modified by having the root of the tongue
thrust forward_.

Various other analytical problems (e. g., of negative as a clause type; or

of paralinguistics and its relation to gesture in Mbembe) and pedagogical ones
(e.g., the techniques of planning a course of study, before the analysis was
complete, so one could simultaneously carry on further research, and learn the
language) added to the complexities of the project as a whole.
 -7-

Before reaching Africa I was able to consult with various scholars who very
kindly suggested areas of concern, warranting investigation, and gave me valu-
able bibliographical references. Among these scholars were Professor Jack Berry,
Nortliwestern University; Professors H. A. Gleason and W. Samarin, Hartford Semi-
nary Foundation; Earl W. Stevick, Foreign Service Institute; Prof. A. E. Meussen,
Musee Royal de L'Afrique Centrale, Tervuren, Belgium; Dra. Bertha Siertsema,
Free University of Amsterdam. From some old (1954) lecture tapes of Professor
William Welmers, Univ. of California at Los Angeles, I had access to data pre-
viously made available to members of the Summer Institute of Linguistics. Be-
fore reaching Africa, also, I had available some preliminary tagmemic restate-
ments of work by other scholars, made by graduate students Krohn (Shona), Sherman
(Grebo), Rensch (Bobangi), Kappler (Hausa), Nicklas (Tswana).
The time in Africa was divided into two parts. In Ghana, various members
of the Summer Institute of Linguistics gathered at Accra with their informants
from November, 1965, through the middle of January, 1966.
The personnel, and the languages involved were:
D John and Kathleen Callow, Kasem
Jack Kennedy, Dagaari
Gill Jacobs, Bimoba
Monica Cox, Basare
Marjorie Crouch, Vagala
Ron and Muriel Rowland, Sisala
From the middle of January until the end of March a second workshop was
held at the University of Nigeria, Nsukka.
Members of the Summer Institute of Linguistics, with the languages studied
at the workshop were:
Ian and Amelia Gardner, Abua
Paul and Inge Meier, Izi (and Dr. John Bendor-Samuel, Consultant)
Katherine Barnwell, and Patricia Revill, Mbembe
Richard and Nancy Bergman, Igede
Klaus and Janice Spreda, Agbo (and Dr. John Bendor-Samuel, Consultant)
Thomas and Eileen Edmondson, Etung
Elaine Thomas, Degema and Engenni
Member of the Sudan Interior Mission, Miss Jean Soutar, Bariba
 -8-

Members of the Assemblies of God Mission:

Irene Crane and Ruby Peterson, Bette.
Member of the Evangelical Lutheran Mission (not official members of the
workshop, but making data available to us):
Herbert Stahlke, Yachi
In each instance, I first ran through the available published and unpub-
lished data of members of the workshop, watching for clues to material which
might lend itself to the testing of tagmemic and matrix approaches. Some of
the analyses were already arranged in such a way as to allow immediate partial
transfer to the desired framework. Other material was obviously accessible to
the members of the workshop (each of whom spoke the language concerned) through
their informants.
Data, as already available, or as gathered for this purpose, were presented
to the whole group, discussed, revised, and amplified. In several instances
the data were then discussed in fuller form, in materials co-authored by the
present writer, so as to allow maximum relation to the theoretical problems
fl
of broader relevance to the project as a whole.
Special assignments to particular individuals, futhermore, filled in vari-
ous lacunae in reference to over-all coverage of the phonological and grammatical
hierarchies. The latter were principally in focus. Lexical (or related semantic)
problem were not dealt with at any length, but left for further study.
The following list of Niger-Congo languages of West Africa attempts to place
them in some ki:id of genetic order--following Greenberg's groupings (shown by his
code symbols such as IA3) occasionally supplemented by unpublished data and cal-
culated guess where necessary, from Bendor-Samuel, who also supplied me with
crude estimates of numbers of speakers of these languages.
NIGER-CONGO LANGUAGES (after Greenberg) a
Country Estimated Population

Gur (Voltaic) IA3

Grusi IA3c
Kasem Ghana 45,000 a
Vagala Ghana 10,000
Sisala Ghana 60,000
Dagaari IA3d Ghana and 200,000
Upper Volta
D
 -9-

Bariba IA3f Country Estimated Population

Dahomey and 100,000
Nigeria
Gurma IA3g
Bimoba Ghana 50,000
Basari Ghana 40,000
Kwa IA4
Akan-Twi IA4b Ghana
Degema IA4e Nigeria 10,000
(related to Greenberg's Bini)
Engenni IA4e Nigeria 25,000
Igede IA4f (?) Nigeria 10,000 (?)
Yachi IA4f Nigeria 12,000
Izi IA4g Nigeria 200,000
Benue-Congo IA5
Abua Nigeria 20,000
Agbo IA5C (?) Nigeria 15,000
(related to Mbembe)
Bette IA5C1 Nigeria
(related to Bekwana)
Mbembe IA5C3 (?) Nigeria 50,000
(related to Ukelle; rather
than to GreenbergisiMbembe)
Bantoid IA5D
Etung IA5D Nigeria 10,000

Concordances made from texts in Vagala and Sisala were produced at the
University of Oklahoma by the Linguistic Retrieval Project for Aboriginal
Languages, partially supported by National Science Foundation grants GS-270
and GS-934. Input data on magnetic tape are archived at the University of
Oklahoma.
 -10-

FOOTNOTES
1
An Introduction to Morphology and Syntax (Santa Ana, Calif.: Summer Insti-

tute of Linguistics) 1962.

Or the fuller treatment by Robert E. Longacre, Grammar Discovery Proc,,Jures
(The Hague: Mouton and Company) 1964.
2
For example, for syntax, Kenneth L. Pike, "A Syntactic Paradigm," Language,
39.216-30, 1963; or K. L. Pike "Dimensions of Grammatical Constructions," Lan-
guage, 38.221-44, 1962; for morphology, K. L. Pike, "Non-linear Order and Anti-
redundancy in German Morphological Matrices," Zeitschrift fur Mundartforschung,
32.193-221, 1965.
3K. L. Pike, "A Guide to Publications Related to Tagmemic Theory," Current
Trends in Linguistics3: Theoretical Foundations, (T. A. Sebeok, Editor) (The
Hague: Mouton and Company)1966, pp. 365-394.
4Kenneth L. Pike, Language in Relation to a Unified Theory of the Structure
of Human Behavior, Second Edition (The Hague: Mouton and Company) in press.
(Preliminary edition Vol. I, §1-7, 1954;. Vol. II, §8-10, 1955; Vol. III,
§11-17, 1960.)
5See Zellig S. Harris, "Discourse Analysis," Language, 28.1-30 (1952).
6For example, in its early form in his Syntactic Structures, (The Hague:
Mouton and Company) 1957.
7Floyd G. Lounsbury, "A Semantic Analysis of the Pawnee Kinship Usage,"
Language, 32.158-94 (1956); Ward H. Goodenough, "Componential Analysis and
the Study of Meaning," Language, 32.195-216 (1956).
8See my Language..., III, p. 36.
9Grammar..., pp. 24-34. The need for grammars to be able to generate
sentences, however, has long been implicit in the presentations such as that
of Doris Cox, "Candoshi Verb Inflection, "International Journal of American
Linguistics, 23.129-40 (1957) (where verb structures are given in detail)
and that of William Wonderly, "Zogue I: Introduction and Bibliography,"
0
International Journal of American Linguistics, 17. 1-9 (1951) where the need
for productivity of formulas beyond the limits of a closed corpus is insisted
upon--and found necessary for persons who wish to do grammatical analyses as
a basis for creating a written literature in a language.

Or
 10
The Summer Institute of Ling-4.stics has work going on in more than 350
languages. The researchers of these languages have all been exposed, to some
degree, to tagmemic concepts. When a component of the theory would not work
in these contexts it was revised.
11For example, K. L. Pike, "Beyond the Sentence," College Composition and
Communication, 15.129-35 (1964); and Alton Becker, "A Tagmemic Approach to Par-
agraph Analysis;" College Composition anet Communication, 16.237-42 (1965).
12Cf. Philip Bock, The Social Structure of a Canadian Indian Reserve,
Harvard University Ph.D. Dissertation, 1962; and Philip Bock, "Social Structure
and Language Structure," Southwestern Journal of Anthropology, 20.393-403 (1964).
13K. L. Pike, "Nucleation," The Modern Language Journal, 44.291-95 (1960).
14See now, however, Joy McCarthy, "Clause Chaining in Kanite," Anthropological
Linguistics, 7.5.59-70 (1965).
15In "Tonemics, Morphotonemics, and Tonal Morphemes, "General Linguistics,
4.1-9 (1959). I had not known of this phenomenon when I published my book on
tone languages in 1948.
16
Versus my skepticism concerning this possibility expressed in my Tone
Languages (Ann Arbor: University of Michigan Press) 1948, p.11 fn 22.

a 17See his "Tongue Root Position in Akan Vowel Harmony" (forthcoming).

a
 CHAPTER I: CLAUSES
In this chapter clauses are in view, and the central question is: Can
clause typology, structure, and dynamics for an area as a whole be conveniently
discussed in a tagmemic framework? Can we, that is to say, study the selected
basic likenesses--or differneces--in a substantial number of languages (whether
closely related, distantly related, or unrelated) without being overwhelmed by
detail? In this chapter we attempt to show, on preliminary African data, that
the answer to each of these questions is definitely yes. More detail must await
the publication of articles and monographs now in preparation.
1.1. Clause Typology 0
The data on which this chapter is based have been gathered only in a
preliminary form. One special caution grows out of this: The reader must not B
argue from silence. Whereas any particular datum reported is present in the
language concerned (although, even here, later revision may in some instances
yet cause changes), the absence of a clause type in the tentative description
has far less probability of continuing to be confirmed.
Precisely here a prime heuristic value of the preliminary cross-language
comparison is in evidence: Languages of a particular area, from the same lan-
guage family, often share many grammatical characteristics; an item reported
in one suggests a search in texts (or, with due care, with the help of an in-
formant) for the comparable type in the second. If the type does not occur in
a large body of text, and if the informant will not create such a form (as a
translation of an English eliciting model), or rejects suggested forms created
by the analyst on an analogical basis, the absence then appears to be more
significant, and may enter the permanent description as a null cell in the
clause matrix.
1.1.1. Basic (Kernel, Nuclear) Clause Types
1
The first generalization for the region of West Africa: within the
nucleus of their clause systems (i.e., within the kernel, and so basic or cen-
tral to the system) the languages have the clause set:
Intransitive ('without object)

Transitive (with optional direct object)

Ditransitive (with optional--or obligatory--indirect object)

-12-
 ......._

......,40- ..............

-13-

This at first seems either trivial, because expected human nature requires-
on occasion--reference to an action by someone without reference to another
dramatis persona (or element) involved; action by one person or element on
another; and action by a first party on a second party or item, in reference
to a third. These same data seem important if one is searching for language
universals, since the human necessity for the presence of these relationships
by no means requires that they be formalized in every language in the same way.
Precisely here some elegant formal surprises creep in: A kind of arbitrar-
iness, overriding universal probability,becomes formalized. Normally in Twi,
2
n for example, J. M. Stewart has shown (1) that when two pronouns are expected
after a verb, the first as indirect object and the second as direct (parallel-
ing the normal occurrence of two nouns in this position) the construction is
a not allowed to occur, because of a restriction which allows the pronoun to
come only directly after the verb, and not after another pronoun. The desired
Q communication effect, therefore, is achieved by replacing the construction with
a more complex form (which will be further discussed in §2.2).

D The etically expected--but unallowed--

* 0 - femm me no
* 'He-lent me it'
D
(analogous to the acceptable
O - femm me ne p6nka no
0 'He-lent me his horse that' )

is replaced by the acceptable

D O - de n6 femm me
'He-take it lent me'

a paralleling the optional--also acceptable

.. ..
o - de ne ponko n6 femm me
'He-take his horse that lent me'
D
Stewart also shows that a further restriction applies [with ditransitive
verbsl. With'give,' for example, the possibilities for the direct object not
0 only exclude the pronoun (after the indirect object) but also exclude a defin-
ite nominal object in that position; only an indefinite nominal is allowed there.

.. ,
 -14-

Thus the definite

* o - de me sika n6
*'He-gave me money the'

must be transformed into the more complex

o - de sikg n6 maa me
'He-take money the gave me'
Whereas the indefinite
o maa me sikg
'He-gave me [some] money'
does not need to be transformed. But this restriction, like the one on pronoun
objects, could not have been predicted by etic probabilities--i.e., by human
cultural universals or near-universals
If, furthermore, one wishes to say
'He cut the meat with a knife'
one finds (still using Stewart's data) that no simple clause will permit it.
A complex construction is necessary for the expression of the instrumental re-
lation, even when an indefinite noun phrase is used:
o - de sagn twaa ngm n6
'He-take knife cut meat that'

Here a universally expected etic relation, expected with a high degree of prob-
ability within simple clauses, is realized only through a different kind of
construction. The result is that the list of non-complex clauses will appear
to have an etic gap at this point; and complex constructions (clause clusters,
see §2.2) will be forced to occur more frequently and to be more varied in
kind, than one might have guessed.
The implications of the preceding data for an etics and an emics of
clause analysis are at least three:
(1) A compilation of known clause types is needed, classified in some con-
venient way. Preferably, experience shows us, this should not be a mere random
list of types, nor even a hierarchy of successively more inclusive types only,
but rather (a) some kind of dimensioned display (like a phonetic chart) with
intersecting categories--which allows the reader to abstract classes, at will,
from various viewpoints (from rows, or from columns, for example, rather than
forcing, by a tree structure, some One predigested set), and (b) some device
 -15-

for cutting unwieldiness (when the chart gets too large), through 'multiplica-
tion' of basic matrices by simple elements or by other matrices (or by some
other kind of transformational device--e.g., by transformational commentary,
if the complexity is little enough to make the matrix devices appear to be
more cumbersome than needed).
(2) Emic analyses of clauses are needed, one language at a time. (a)
The sum of these emic clauses becomes the list for a general--universal--etics
under (1). (b) The specific emic clauses need to be studied for variation,
as seen for the Twi constructions containing pronoun forms and definite object
forms discussed above. (c) In order to differentiate these variants from the
emic clauses themselves, contrast between emic clause types must be specified.
(3) Since, (a), it is apparent in the Twi data that specific tagmemes or
even manifesting variants of tagmemes within a clause--e.g., the direct object- -
may affect the possibility of the occurrences of that clause, and since (b) a
concept such as instrumentality, which in other languages may be often found
manifested by a tagmeme within a clause but here is realized, rather, on a
different hierarchical level, it is important, before analysis is assumed to
be complete, that the analysis based on--for example--clauses abstracted from
text be supplemented by eliciting techniques (with the safeguards necessary and
appropriate to all eliciting techniques) one must see how a particular language
expresses those various kinds of concepts reported in other languages, expressed
in some of them within clauses, and assumed to represent human conceptual univ-
ersals.
On the other hard, it is important that the clauses of each language be
studied in reference to its own system, for their distinguishing, contrastive,
formal and informal features; for their kinds and ranges of variability; and
for their structural and transformational relations to, and distribution within,
higher-level or lower-level structures. If one then chooses to do so, comparison
of one area--e.g., West Africa--can be made with other areas of the world. This,

however, is a goal beyond the purview of our present study--but within the lar-
ger goal of which this study has been but one part.
With these elements now more explicitly in mind, we return to the listing
of some clause types observed in West Africa--a list begun, above, with mention
of the intransitive, transitive, and ditransitive.
 -16-

Probably most, if not all, of these Niger-Congo languages have also a

clause type which may be called Locative (or Directive). (Reported, in our
data, for Dagaari, Vagala, Kasem, Bariba, Mbembe.) In Mbembe, for example,
a small number of verb roots are classified as directives: yin come, go,
fetch', and to 'go' being among the most common. They are often followed
directly by a noun indicating a place (e.g., epya 'market', or a preposi-
tional phrase made up of a preposition plus one of these nouns:
5 to elgyd

'He-goes market'
or: o-ta sa Cpyg n
j
'He-goes to market'
See, also, Bimoba:
n saa daak nie
I am-going market in
II am going to market'
3 ,
Cinsiderably uncertainty has come when analysts have attempted to treat
the locational noun as a direct object of locational-directive verbs such as
these, since the locational clauses contrast with transitive clauses not only
by the potential expansion of locative noun to locative prepositional phrase,
but also in that they canner. undergo certain transformations allowed to the
transitives:

Ote O'ci eten

'Father eats meat'
or ate 51t6ga eten bcf (cf., above, similar trans-
'Father takes meat eats' formation with indirect object)
but not *3t6 5't5ga gpyg 5.ta
*'Father takes market goes'
nor may the directive--as different from transitive--function as the first
element of a result clause composite.
In Twi (continuing with Stewart's data) there seems to be a restriction
such that locative noun phrases may follow certain locative verbs, but the same
locative expressions may not follow--where we might expect them to--certain
other verbs plus object; we cannot say, there a
*kofl guu ntoma n6 g-It no mu
*"Kofi put the cloth in the box'
 -17-

even though it is acceptable to say

ntomg n6 gu ada kg n6 mu
'The cloth is in the box'.
Mbembe, however, does allow a locative phrase, after direct object, and also
the corresponding complex form:
6-yilga ij5:g fa Ckpa
'He put snake in bag'

o-t5ga ij5:g 6-yig6 fia Ckpa

'He-took snake put in bag'
An alternativ Tothesis which might well be explored for Mbembe (and
Twi?) would be to set up a pair of clause types:
Locative-intransitive
and Locative-transitive
(with 'go' and 'put' as typical of their verbs) contrasting with each other and
with the regular intransitive and transitive. I do not know how this would
work out in view of the whole system.
In most of these languages (e.g., Dagaari, Vagala, Kasem, Bariba, Igede,
Mbembe, Degema,) one finds some kind of:
Stative.

For Dagaari note:

a zong
'He is-blind'
For Vagala, note the alternative forms:
wgys6
'He is-good'
0 or u dd weyr
'He is good'

Occasionally this area of form and meaning is divided by the authors into
Equative (or Copulative) and Stative, in which the equative has a predicate nom-
inative, and the stative a special verb plus predicate adjective, or a verb which
carries both the affirmative and the quantitative (as in the Vagala example just
given). For Degema, compare:
51115-y5 ae5
'[The] child-emphatic is-there' (stative)
6m5-375 5-yin 5m5-m5e1
 -18-

(child-emphatic 3rd-sg.-be child male)

0
'The child is a male,'
A futher pair of emic (contrastive) clause types posited for some of these
languages (Dagaari, Vagala) is:
Descriptive versus Demonstrative.
In Dagaari, for example, special verb forms differentiate these from each other,
and from statives and intransitives. In addition, the subject of the demon-
strative clause is an emphatic type, and demonstrative tagmeme following the
verb may be manifested by a demonstrative particle--and other items--not found
in the slot following a descriptive or stative verb.
waa n/ng-kpowng
'He is person-king'
and lg a d55
'He(emph) is the man'

Less clear, but probably necessary to add to the list of numerous West
African languages is the
Impersonal.
In Kasem, some clauses with the subject ku 'it' differ in function from intran-
sitive clauses which superficially look like them. In a clause cluster (or
series, see §2.2), for example, the impersonal pronoun subject may occur within
the series which has a different pronoun subject, where Kasem otherwise would
require Uniformity. Note the 6...6 series with intervening kd:
...5 vim 5 nowri ku mgngi di sdwla fintb o gya ba
he go he beg it approx. with shillings thirty he take come

'He went out and obtained thirLy shillings by begging and

brought it back.'
In Bimoba the impersonal, as distinct from the intransitive, cannot be
followed by other basic clause types in a series ( §2.2.2).

1.1.2. Derived (Marginal) Clause Types

If a clause system is assumed to have nuclear (as in §1.1.1.) and marginal

elements--i.e., if the system itself is viewed as having a "wave" form--then
the marginal kinds of clauses are those which can be treated as in some sense derivabl
from--or peripheral to, or transformed from--the nuclear system. Several kinds
of peripheral clauses occur in various of the Niger-Congo languages:
 -19-

Modal: Interrogative, Imperative, Subjunctive

Qualitative: Negative
Emphatic
Causal
Benefactive
Dependent: Subordinate, Relative
In Dagaari, for example, all basic types can be transformed to interrogative
and imperative modes, as well as to negative and emphatic types (data not avail-
able concerning causative or benefactive).
Mbembe is set up by Barnwell, tentatively, as having basic intransitive,
transitive, ditransitive, copulative, directive, motive (i.e., a directive
after a verb which takes an object, as in 'Father sends [the] child [to fetch]
meat'), stative, initiative (as in 'Father begins meat to-eat'), independent
introductory, dependent introductory. Each of these can be transformed into
imperative, subjunctive, interrogative, subordinate,"and relative--though some
of the forms have thus far been found only in elicited data, and not in text.
Contrastive features of form and meaning differentiate the types. Imperative,
for example, has obligatory absence both of independent subject (with one
exception in the data) and of person and tense subject prefixes; certain imper-
ative verb prefixes (ma-plural); and imperative tone patterns.
Contrast:
A
ote O'ci eten
'Father eats meat'
with ci eten
'Eat meat'
The negatives in Mbembe--and various other West African languages--are so
different from the positives that analysts sometimes prefer to set them up as
separate clause types, derived from the positive ones. Mbembe negative imper-
fect indicative moves the predicate to the end of the clause, adds a negative
verbal prefix -m, utilizes a distinctive tone pattern ("); and utilizes a
Group 1 vowel for the final vowel of the verb. Compare:
bte oci dten sa 5s5:m
'Father eats meat in house'
\
*ate eten sa 5s5:m moci
'Father meat in house won't-eat'
 Qr
-20-

The perfect indicative contrasts with the imperfect by having the predicate
follow the object (but precede the locative margin); with object (or predicate)
preceded by clitic k'; with different tone patterns; and with verb-suffix -a.
From this the corresponding negative has predicate as clause final; negative
verb prefix k'; a further distinctive tone pattern; and a Group 2 vowel as
final in verb:
btg kidten bcg sa 6s3:m
`Father meat ate in house'
I
ote gten sa 5s5:m k 6ci
'Father meat in house hasn't-eaten'
Imperative and subjunctive, etc., have further changes for negative.
Some scholars feel that the extensive changes for aspect, as seen above
for Mbembe, make it desirable to treat these as transforms also. Kennedy, for
Dagaari, first sets up a matrix to show basic types (with possibility of trans-
forms to emphasis, mode, and dependency--see Dagaari Matrix I; note gaps- -
signalled by hyphens--for *imperative stative, and *imperative demonstrative,
as well as limitation of locative and demonstrative forms to positive qualities
only(Negative locative concept is expressed by a special negative verb--not
by a regular verb negated by a particle--and that negative verb must be followeu
by a locative word bgy 'there', as in a chg by 'He is-not-located there'.)
DAGAARI MATRIX I: Occurrence-Matrix of Basic Clause Types with Some
Permitted Transformations

Transformed to: Tense Emphatic Interrogative Imperative Dependent

Basic
Form (in Perfective
or Imperfective
Aspect)

Ditransitive x x x x x
Transitive x x x x x
Active x x x x x
(Intransitive)
Positive
Locative only x x x x
Stative x x x x
Descriptive x x x x x
Demonstrative Posi- Subj. x - x
tive only only
 -21-

Each cell in the Dagaari matrix can be filled, in turn, by a submatrix of

clauses or clause variants. (See the accompanying Matrix II, of Tense and
Quality Transform Types). The implication here is that, tentatively, negative
and tense forms are treated as contrastive types. Note, as illustrative:

da 26w
"He did run'

u do bl zow
'He did not run'

u na 26w
'He will run'

kown 26w
"He will-not run'
DAGAARI MATRIX II: Permitted Tense and Quality Transform Types

Tense
+ Past + Future

Positive + da 'past' + na

dgg 'far past'

zaa 'yesterday'

Negative + [as above] + bg +kown

Even with these differences of form, however, many investigators (like

Crouch for Vagala in §1.2) would undoubtedly prefer to treat clauses marked
for tense as etic variants (optional expansions) of the basic clauses--which
4
already include, in the verb stem, signals for aspect (perfective or imper-
fective). Thus negative changes might be treated differently from tense.
Tagmemic theory focuses on etic versus emic differences but--in spite of
R. Longacre s 5 dual structural criterion some intermediate situations such
as the negative remain doubtful. Whether this is due to the permanent indeter-
minate nature of the data, or to lack of adequate theory or method, is not known.
 -22-

Emphasis in Dagaari leads to classes of emphatic clauses, differing as to

the included tagmeme which is emphasized. Non-subject tagmemes may be emphasized
in any basic clause type, or any derived by the modal and tense-or-quality
clauses just discussed, by shifting the emphasized tagmeme -to a position be-
r
fore the subject--and adding to it ka which in turn is preceded optionally by
lg. Subject, when emphasized, is followed by lg; a nonemphatic pronoun is re-
placed by an emphatic one.
For Bariba, several kinds of clauses are reported, by Soutar, which have
not been seen in the preliminary data concerning the basic clause types of
Dagaari, Vagale, and Mbembe. (Note the causal and benefactive elements.)
For the Bariba, Soutar lists certain clause types, expanding the list with
sub-matrices to show some inner-layer transformation potential, for derived
benefactive, reciprocal, and passive types.

I. Bariba Basic Clause Types -- Independent

A. Intransitive:
Reciprocal-
Norm Benefactive Benefactive

Norm 1 3 5

Causal 2 4 6

B. Directive:
Numbers 1 to 4 in above chart
C. Transitive:
Norm Passive Reciprocal

Norm Benef. Norm Benef. Norm. Benef.

Norm 1 3 5 7 9 11

Causal 2 4 6 8 10 12

D. Ditransitive:
Numbers 1-5 and 9-12 in above chart.
E. Stative:
F. Equative:
1. Subjective complement
2. Objective complement
II. Basic Clause Types -- Dependent
A. Introducer
 -23-

These are now illustrated, giving first a detailed list of the lexical
items to be used, followed by a citation paradigm with examples numbered accord-
ing to the clause list, followed in turn by tagmemic formulas for these clause
types.

List of Lexical Items

na - 'I' mam 'me'

u - 'he' wi / num 'him'

'we' bh - 'them'
ba 'they'

wi 'he' emphatic

kuke (St. kukii) - 'hide' -sia - 'causative'

kpiina - 'lie down' -na 'reciprocal'
di (St. daa) go
'go' -ra 'passive'
wa (St. waa) - 'see' Tone change
nb (St. no5) - 'hear' -a/-wa/-ya - 'benefactive'
=
we (St. wEE) - 'hand over'

ki (St. ktt) - 'give'

wes/i - 'give back'
bir5 - 'be sick'
stoma - 'to be'
soku - 'call'
k5 (St. koo) - 'do'
dwe (dwi) - 'buy'

gber; - 'farm' teere - 'blame'

gobi - 'money' 0115 - 'person'
tasu - 'yams' be - 'those'
gari - 'words' / 'matter' feku - 'common cold'
bil - 'child' shn5 - 'chief'
yihni - 'master'
 -24-

Citation Paradigm
I. Independent
A. Intransitive
1. u kukba 'He hid'

2. u nth kukU - s15 'He made him hide'

3. ba mam kpun5 (tonal morpheme for benefactive)
'They lay down for me (prostrate themselves before me)'
4. u man nuth kpuna - sla (tonal morpheme for benefactive) D
'They made him lie down for me'
5. ba kukU - na (tonal morpheme for benef.)
'They hid for each other (from each other)'
6. u hu kukil - na-sla (tonal morpheme for benefactive)
'He made them hide from each other.'
ri
B. Directive:
1. u gbaru da `He went to farm'
2. u mam gbaru daa-slA 'He made me go to farm'
3. u mam gbaru da5 (tonal morpheme) 'He went to farm for me'
4. u mam bu gbaru daa sia (tonal morpheme)
'He made him go to farm got me'
C. Transitive
1. na bu wa 'I saw them'
2. na bu gobi wgg-sla 'I made them see (obtain) money'
3. na bU tasu wgg-wa 'I found yams for them'
4. u mam bu gobi wad.- sia (tonal morpheme)
'He caused them to find money for me'
5. tasu waara 'Yams were found'
6. ba gari no5 - ra - sia 'They caused the words to be heard'
7. s5bura man koo - ra (tonal morpheme) 'The work was done for me'
.

8. ba man gari no5 - ra - sia (tonal morpheme)

9.
'They caused the words to be heard for me'
ba waa - n5 'They saw each other'
ai
10. na bu waa - na - sia 'I caused them to see each other'
11. ba yabanil dwi-a - na (tonal morpheme) 'They bought shirts for each other'
12. na nufi wim bibu waa - na - sia (tonal morpheme)
'I caused his children to see each other for him' a

'-'-'-'7-7-77T7V777777777--
 -25-

D. Ditransitive:
1. u man gobi wT. 'He gave me money'

2. u mam bti gobi wee - sa 'He made them give me money/

He made me give them money'
3. u mam bti gobi - ya 'He gave them money for me'
V
4. ba man wi COM be gobi wee - sia
'They made him give those people money for me'
5. bli sun kii ra 'A child was given to us'

6. ba taarg wee - na 'They gave each other blame

(They blamed each other)'
7. sa bti gari wesig na sla
'We caused them to give words back to one another
(to discuss the matter)'
8. ba sun (gari) wesig - na (tonal morpheme)
'They discussed the matter for us'
9. ba sum bh gari wesig ng - slg (tonal morpheme)
'They made them discuss the matter for us'
E. Stative:
1. na bEf5 'I am sick'

2. na ftku bar5 'I have a cold (I am cold sick)'

F. Equative:
1. u sad suns 'He is chief (a chief)'
2. wf - ye suns 'He is the chief'

3. wi guho wg 'He is a chief'

II. Dependent:
A. Introducer:
1. u na,.... 'He said,...'
2. u bh soowa 'He told them
This material from Bariba in Dahomey seems so different from Dagaari, Vagala,
and Kasem of Ghana and from the Mbembe and Degema of the lower part of Nigeria,
that I re-checked, after these pages were written, with Crouch concerning Vagala.
Somehow, the cultural universals of causation and benefaction would have to be
expressed in them also. Had the Bariba type of data been overlooked in these
other languages, or did it in fact not exist?
 -26-

Here, once more, the semantic features found their expression in clause
clusters ( §2.2, or subclusters, §2.2.5)--rather than in single clauses with
extra tagmemes within the verb (i.e., as affixes) or outside theverb but
within the single clause.
Compare the following Vagala set for simple normal transitive, with the
corresponding causative transitive cluster and the corresponding benefactive
cluster:
U .k6yog 0
'He did this'
gil h k6yng
'He made him do this'
U 6 U t6 A
'He did it gave me'
'He did it for me'
!

U sd igyo to u 1)&71

he came danced igyo-dance give his village

'He danced the igyo-dance for his village'
We have, then, a major difference in "grammatical style" between the two
languages. The Vagala uses certain clause clusters where the Bariba might
use single or complex clauses. The Bariba, at this point, looks typologically
much more like Bantu languages--as we shall see in §1.5 for Bobangi--than does
Vagala. The structure and frequent use of clause clusters such as these of
Vagala, furthermore, is one of the most striking--and typical--of the character-
istics of many of the West African languages. We will discuss these complexes
more specifically in §2.2.

1.2. Clause Contrasts

First, however, we must discuss the structural markers of contrast between
simple clauses and the kinds of variation they undergo.
For Vagala, we take from Crouch a syntactic paradigm in two parts: First,
a chart (or matrix) in which independent, basic clause types form the rows,
whereas specific tagmemes or (for the predicate) classes of tagmemes form the
columns. If a tagmeme may (or must) occur as part of one of these clause
syntagmemes, it is included in the row, at a place of its most frequent or styl-
istically normal occurrence. Free variation of order is not signalled here.
 -27-

TENTATIVE TAGMEMIC-NOTATION PARADIGM FOR BASIC CLAUSE TYPES IN VAGALA

Nuclear Tagmemes Marginal Tagmemes

+ + + + + +
Ditrans.: + Subject - Tense* + Dit.Pred. +(- I.Obj. - Object) - Manner - Location - Time
NP future, DitVP NP NP adverb locative temporal
1(all) la la,aa,d la(all)
time, NP loc. noun NP
NP l
NP subjunct. 2 NP 2a NP NP 2c
2a c
particles 3 NP2a NP NP NP 3a NP
3c
NP NP 3b NP
4a 5a,b
NPSa,b NP
NP
5a,b
5ab
, 5d NP
NP 5c
NP 6
6

+ + + + +
Trans.: +Subject - Tense + Trans.Predicate - Object - Manner - Location - Time

same same TrVP same same same same

la
2
3

+ + + + +
Intrans.: + Subject - Tense + Intrans.Pred. - Loc.-Acc. - Manner - Location - Time

same same InVP NP same same same

la 3a
2 -

+ + + +
Locative: + Subject - Tense + Loc. Predicate + Loc.-Act. - Manner - Location - Time

same same LocVP same same same same

la
-NP
2 3b
3

+ + +
Stative 1: + Subject - Tense + St Predicate - Manner - Location
I ±- Time
same same SOT same same same
la
-3b
2
3

Stative 2:
+ + + +
+Subject - Tense + St Predicate + Adjective - Manner - Location - Time
2

same same du adjective same same same

numeral -3b
NP (+num.)
laa

Descr.: + + +
+ Subject - Tense + Des.Predicate + Complement - Manner ± + Location - Time
same same DesVP NP same
la la,aa,b same same
2 NP -3b
3
4a
NP
5a,b,c,d
NP
2a

Demons.: + +
+ Subject - Demonstrative - Tense
Emphatic NP NP restricted time
la(all) la,aa,b
1I
NP NP and adverbial
2a 4a,b particles,
NP NP subjunctive
4a 5a,b,d
NP NP 2a
5a,b
locative
la

* The order of manner, location, and time tagmemes is not rigid.

Some tense and time elements have co- occ-'rrent
relations not shown here.

r.-3 BM e7r",
 -28-
(See, however, the footnote to the chart, indicating that the order of periph-
eral tagmemes cr nanner, location, and time is not completely fixed. Often,
in tagmemic formulas of this kind, arrows from one place in the string to
another show free variation of position.) Contrastive change, for emphasis,
has been discussed in 51.1.2. When a tagmeme is obligatory, it is preceded
by a plus sign: if optional, by +.

Note that, in each instance, any two of these emic clause formulas differ
both by their predicates (or absence of predicate in demonstrative) and by one
other tagmeme (presence or absence of object, adjective complement, etc.),
whether obligatory or optional. The demonstrative differs from other clauses
by its lack of marginal tagmemes.
Underneath each tagmeme is given in a small subcolumn a preliminary list
of the classes of items or constructions which can fill that tagmemic slot, and
which, along with the function named by the tagmemic label, comprises the tag-
meme as a whole--i.e., the function set. Abbreviations for the constructions
include:

NP Basic noun phrase

1
NP Relative noun phrase
2
NP Locative noun phrase
3
NP Possessive noun phrase
4
NP Coordinate noun phrase
5
NP List noun phrase
6
1; VP Simple verb phrase
1
VP Pre-modified verb phrase
2
VP Post-modified verb phrase
3
Subscript letters to the classes (e.g. NP ) indicate further relevant subsets
la-
of nov--3 (or verbs). These do not interest us at the moment--but such classes
are important for the productive use of language. To whatever extent this
tentative listing proves to be erroneous, it implies the possibility of forms
not permitted by the system. This compacting display provides restrictions on
occurrence of phrase type by the context of clause type. The generative power
of a tagmemic presentation works, in part, by this type of predictive formula.
The (±0 ±0) form indicates that one or the other--or both--of the
objects must be present; but if indirect object is present, direct object is
not required, and vice versa.
 -29-

Two locatives may occur in certain of these constructions--the first

(locative-accompaniment) as part of the nucleus, and the second within the
margin. The nuclear locative is obligatory to the locative clause; the mar-
U ginal one is optional there:
u du Sawla u mad' by
he is-at Sawla his mother vicinity
'He is at Sawla with his mother' (Sawla, main locative;
with his mother, marginal locative)
In the locative and intransitive clauses, the nuclear locative (like the
object of a transitive verb, or the indirect object of a ditransitive verb)
is moved to the position before the verb when the clause is negative in non-
future time (but not shifted with negative future). This characteristic,
among others, leads to treatment of the first locative (or locative-accompani-
ment) as nuclear.
The same marker ni used for locative may, after some verbs, be used in
the same position with meaning of accompaniment:
u ba u ni bowl
he came it with village

D 'He came with it to the village.'

The two come together in a sequence of locative-accompaniment and locative in:
1 ,
n ding *la i ni bawl
I will go you with village
'I will go with you to the village.'
We also give, from Crouch's data, a citation paradigm in matrix form illu-
strating the basic types of clauses. In order to keep to space requirements,
only nuclear tagmemes are illustrated. This kind of citation display allows
quick access to contrastive illustrations since no more variety of lexical
selection is used than is necessary. The reader is to assume that each lexical
change is caused by a change of emic structure (or occasionally by etic agree-
ment characteristics, etc.)
Added to this matrix are a column of imperatives, and three columns of
dependent clauses--for normal, purpos!ve, and conditional.
When two inner (tagmeme) contrasts cannot be seen in a citation pair, then
a further contrast may be expected, either (a) in the potential occurrence of
optional expansions (the optional occurrence of a tagmeme--say, object--which
 INDEPENDENT DEPENDENT
Regular Imperative Regular Purpose Conditional

Ditr. IDi IImDi DDi DPDi DCDi

5 to a kabfla. to 6 kabila ri to n kabila 6 t6 n kabila di u to a kabila...

He gave me fufu. Give me fufu! When he gave me fufu In order to If he gives me fufu...
give me fufu

Trans. IT IImt DT DPT DCT

A! *.
16 kabila lo ka bile 16 kabila... ti 16 ka bile di u 16 kabila...
He pounded fufu. Pound fufu! When he pounded fufu... in order to pound If he pounds fufu...
fufu

Intr. IIn IImIn DIn DPIn DCIn

A ! A A !
lg yau wa la yaziwg lg yau wa 6 lg yaAwg di 5 16 yaii.wg
He went market. Go to market! When he went to market... In order to go to If he goes to market...
market

Loc. IL IImL DL DP L DCL

A A! A A A A!
u du Sawlg du Sa wla u n du Sawla... u du Sa wla di u dti Sawlg...
0
He Is-at Sawla Be-at Sawla! When he was-at Sawla... in order to be-at If he is-at Sawla...
Sawla

Stat.1 IS DS DCSi
1 1
A A
u weyso ri weysg... di u weysc5...
He is good. When he is good... If he is good...

Stat.2 IS DS DPS DCS

2 2 2 2

u du weyr 5 a d6 weyr... 6 d6 weyr di u d6 weyr...

He is good. When he is gccd... in order to If he is good...
be good

Descr. IDs IImDs DDs DPDs DCDs

5 4 ngbamg e nabiimg ngbOmg... 6 6 nab6m6 di 5 6 n6b6m6...
He is an elder. Be the elder! When he was the in order to be If he is an elder...
elder... an elder
Demons. IDm DCDm
wAbil 116bOm6 di wgbii ngbamg...
He is the elder. If he is the elder...

I= c=v t: saC) tz:= C:=1 t=3 c=1 c=) C:=1 am) t= r c=i) r E=3
 -31-

may occur in a more extended sample of the one clause but which may not occur
in the other); or (b) in differences of potential for transformation to another
clause type (or in differences, that is, in their potential for coming in
same cells in matrix multiplication), or (c) in differences in occurrence of
the clauses, as wholes, in still higher-level slots--as, for example, in the
slot: for answer to a question versus reply to a statement, etc. Relevant emic
differences include not only the differences of tagmemes within a construction,
but also differences of the clauses, as wholes, comprising restrictions as to
higher units within which the clauses may themselves be found. (Or, one may
say, differences of internal and external distribution are both relevant; or
emic contrasts include features drawn from different hierarchical levels.)
Specifically,
to n kabila
'He gave me fufu-food'
contrasts with
t6-A ka.bila
'Give me fufu-food'
not only by the lack of subject of the imperative (and by tone) but also by the
fact that 'Give me fufu-food' but not 'He gave me fufu-food' can come in the
context:

Comment Reply
to A kg'bila mag to I

'Give me fufu-food' 'I will not give'

Tagmemic theory requires that attention be given to such questions. It
does not allow one to ignore, permanently, the significance of etic versus
emix units, constructions, or levels--i.e., of contrast as against variation-
at any point in the system.
1.3. Clause Variants
We turn, therefore, from the study of contrastive clause types to a dis-
cussion of types of variation within clauses of West African Niger-Congo lan-
guages. Several types occur: (a) optional presence of nuclear or marginal
tagmemes, (b) optional orders of tagmemes, (c) optional kinds of fillers for
tagmeme slots, and variant 'oims conditioned by occurrence in certain larger
structures.
 -32-

1.3.1. Variation by Presence of Nuclear Tagmemes

In the formulas of the first Vagala matrix, above, the diagnostic locative
tagmeme is marked as optional in the nucleus of intransitive clauses. One can
say not only
A lgngb
'He went'
but also
A lg
'He went market'

The diagnostic locative-accompaniment tagmeme (manifested only by noun phrase

NP ) plays its part in contrasting intransitive from transitive (whose optional
3a
object tagmeme has a different set of noun phrase manifestations). Whereas the
optional presence or absence of the nuclear locative tagmeme of Vagala leads to
free variants of a construction, the obligatory omission in Twi of the second
of two pronoun objects (see §1.1.1.) leads to conditional gaps in the systematic
pattern of ditransitive clauses. So, too, does the nonpermitted definite object
after indirect object.
1.3.2. Variation by Optional and Marginal Tagmemes
On the other hand, the same Vagala matrix shows that manner, location,
and time tagmemes are all optionally present with any of the basic clause types
except the demonstrative. Thus they are not diagnostically useful to separate
the ditransitive from transitive, intransitive, locative, stative, or descrip-
tive (but it does help separate these, as a class, from the demonstrative).
Since these tagmemes, relative to the class containing them, are both optional
and non-diagnostic, they are treated as marginal. Compare:
Transitive:
Manner: yag 16 mgrfa kgyng
we shot gun thus

Location: ...di gya 16 fang.fgni A nyud maa

and there hit soap her head all
1 1

...and there worked up a soap lather all over her head'

Time: kung diy A sogmfg zinaa
things ate his corn today
Descriptive:
Location: A fa g n4zrylig n4 A b6wl bilyni
I past was man-big my village inside
'I used to be an important man in my village'
 -33-

Other languages have different lists of optional tagmemes following the nucleus:
Dagaari is reported with location, time, degree, manner--and instrumental-accom-
paniment. (Contrast, for the latter, discussion of lack of basic-clause instru-
mental for Twi, §1.1.1.). For Igede: adverbial, onomatopoeic (ideophones),
time, and perhaps others. Degema: manner, location, time. Bariba (which,
above, was very different from these other languages) agrees in having manner,
location, and degree following the nucleus: but allows cause and time preced-
ing it, and instrument or accompaniment tagmeme within it.
1.33. Variation by Order of Tagmemes
The early placement of the time tagmeme before the nucleus occurs in
Degema and in Sisala, but as a variant from the postnuclear position. In

Sisala, the placement has relevance to discourse structure ( §3.2.1.). (See

also, contrastive--emic--early placement for emphasis discussed for Vagala

in §1.2.).

In addition, as footnoted in the Vagala matrix, there is, on rare oc-

casions, another kind of variation--a free variation of order among the mar-
ginal tagmemes themselves.
We have earlier seen ( §1.1.2) that Bariba differs substantially from our
other West Africa samples, in the syntax structure of kinds of basic clause
(versus clusters) it contains. So, too,it differs in relation to changes of
tagmeme order within a clause.
Within Bariba causative and benefactive clauses, two or three "objects"
may occur--direct-object-as-goal, indirect object, object- as- actoi, Lenefactee.
Only two personal pronoun objects can occur in a single clause; but a third
pronoun may occur if it is nonpersonal. Within the constructions, furthermore,
certain normal sequences of object (object-as-actor, and indirect object; ob-
ject-as-actor and direct object; benefactee and object-as-actor; indirect ob-
ject and direct object; benefactee and object-as-goal) will most frequently
be reversed if--ignoring singular and plural--it is necessary to do so to get
them to conform to a ranking system such that first person pronoun precedes
second; second precedes third; third personal precedes third nonpersonal pro-
!noun. Tagmemic function (indirect versus direct object, etc.) is ignored in
determining this ordering. Infrequent orders sometimes carry special functions:
when one repeats a clause, to answer a yes-no question, he is likely to change
; 0 the order of pronouns.
 -34-

1.3.4. Variation by Simple Substitution

It may also be assumed that emic clauses have etic variants (etic mani-
festations differing only by the particular element or elements filling one
or more of their slots. This may be seen in three types. The first is the
mere substitution of one member of a class for another member in that same
slot, with no further relevance to the structure. One may say in Vagala:
!

a ,:.376 wo

'He ran

or a zgnwo
'He jumped'
1.3.5. Variation by Substitution of Manifesting Constructions

A second -more interesting--type of variant is the substitution of members

of different constructions as fillers of the same slot. For fillers of the
subject slot in Dagaari, for example, Kennedy gives us:
(i) Pronoun Phrase: ± article + head ± quantifier D
the pron numeral
Num P a
attribute
num
e.g. A tly mlney
a
the we some
'(some of us)'
(ii) Noun Phrase : ± art ± possessive + head
+
- quant
B
simple
a pron noun as above
noun II [
NP
allos: (A) Occurs in possessive slot 11
± poss + head ± quant
tily d6b6 bgta
D
our men three
e.g.: a tiy dobo bata zuw kanga
the our men three head certain
a
('the head of one of our three men')

a
a
ii
--------111,_-- -- .
 -35-

(B) Occurs elsewhere

± art ± poss + head f quant
I
e.g.: a tiy ningkpowni bata
the our persons-big three
'(three of our big men)'
(iii) Noun Phrase : + NP/PronP + link + NP
coordinate . .

as above ani 'and' as above

//
e.g.: biiy 'or'
dobo bata ani pogbo banaar
men three and women four
, .
(iv) Numeral: bata
three
(v) Numeral Phrase: + Num link + Num
bata biiy banaar
three cr four
(vi) Clause Dependent Transform of Basic Clause Matrix

1.3.6. Variation by Agreement (Concord)

Variation in clauses is also caused by the mutual requiring of particu-
lar members (one subset) of a class by particular members (a subset of another
class, according to some formal structural or categorical pattern of agree-
ment or concord.
None of this kind of variant occurs in the preliminary data of the lan-
guages so far discussed. It does occur extensively, however, in the data
given us by Thomas and Eileen Edmondson for the Etung materials of Nigeria,
which in this respect differ markedly in typology. In the Etung, the pre-
sence of different noun classes makes sharp differences in forms of numerous
tagmemes within the clause. Note, for example:
okop eye ani yit
box his that one
biti ebe ambi bit mbi
stick histhat one which
. ..!, .
nni a-gboe nyen nni
while it-fell I-have-seen it
.1.
bi-gbc'e n-yen mbi
it-fell I-have-seen it
 -36-

Perhaps, if we had available moLe syntax data on this language, further

6
typological elements of intereqf would emerge.

1.3.7. Variation? by Occurrence in Clause Clusters

Some clause variants are conditioned by their occurrence as a serial
(or secondary, non-first) member of a clause cluster (or clause series).
A close-knit series of clauses functions as a unit cluster within the
sentence ( §1.4). Within the cluster, linkages tie the clauses together. The
8
formal linkages involve the sharing of tagmemes. Two clauses in a series
containing the same subject delete the second, sharing (noncontiguously) the
first. The second object, when same as the first object, is also deleted.
9
Stewart has made these details explicit for Twi:
akoroma n6 kyeree ak6k5 n6
hawk the caught chicken the
akoroma n6 wee ak6k5 n6
hawk the ate chicken the
but

akoroma no kyeree ak6k5 no wee

hawk the caught chicken the ate
'the hawk caught the chicken and ate it'
The word wee of the last example may be seen, from one point of view, as com-
prising an entire serial clause, as a variant cf an independent clause con-
ditioned by coming in an emic cluster.
From a second viewpoint, the object belongs simultaneously--as a "port-
manteau" tagmeme--to both clauses. This latter approach has the advantage
of making it simple to discuss the "sharing" of a tagmeme (as I did a few
paragraphs back) but the disadvantage of implying noncontiguous sharing of
subject--or noncontiguous sharing of object.
A third view treats the cluster as a single complex unit. The advantage:
The cluster, not the separate clause, is said to contain the "shared," non-
repeated, elements. The disadvantage: One cannot as easily discuss clause in
relation to basic and conditioned-variant forms. I shall leave the theoretical
situation here indeterminate, and utilize that particular viewpoint (or combin-
ation of viewpoints) which is momentarily useful.

--
 -37-

The non-Africanist, however, needs to be alerted to the extremely impor-

tant role that clause clusters play in these West African languages. Not only
do they occur as part of the inventory of available language apparatus, but
0
they comprise part of the system necessary for the expression of various etic
concepts such as instrumental (see §§1.1.1., 2.2.3.), and become an intermedi-
ate state in the dynamics of developing new forms of grammar as we shall see
presently (§2.2.5).

An extensive illustration of suppression of shared independent subject

(but differing from Twi by the retention of pronominal subject) and object is
seen in Mbembe (where Barnwell refers to the domain over which a shared element
is relevant as the 'wave of reference' of that element):
o
a lkwanbo
woman
5ka"ba
she-fetches water
/ b-wona / b-gwei / 6-nioa gwbr
she-pours she-drinks she-sits seat
'The woman fetches water, pours it, drinks it'
The underlying sentences would be:
ikwanbo 5-kaba a's1
woman fetches water
O-wtina

pours
6-gww6
drinks
Onioa Neor
sits seat
B The wave of reference for subject would be the entire cluster; for the object,
the cluster minus the last serial clause.
8 Once it is clear that at least some tagmemes may be suppressed 10 --or
FI
shared--in a clause cluster, we want to know the limits of this sharing. Can
a all peripheral tagmemes, for example, be shared? And if so, in what position
do they come in the cluster?
Other variations of clause structure also occur, but I shall postpon,
discussion of them until §2.2.3 (for tagmeme limits in clause clusters),
§2.2.4. (for agreement factors in clause clusters), and §3 (for changes and
restrictions caused by discourse structure).
 -38-

1,4. Clause Distribution

We might well turn, now, from the occurrence of variants of clauses to
their distribution without reference to modifications of them. Clause distri-
bution is affected both by limits imposed by the structure of clause clusters,
and by limits imposed by the structure of sentences.
Since, however, the distribution of the clause types is relevant both to
the description of clauses, and to the description of clause clusters and of
sentences (as contributing contrastive characteristics of the higher-level
units), we shall defer discussion of clause distribution until we come to these
higher units in §§2.1. 2.2.1, 2.2.2. We only emphasize here that no unit is
well described until a statement of its distribution is included.
1,5. Some Clause Components of Bobangi (Bantu)
11
I now wish to show some of the clause types of a Bantu language. The
12
data on Bobangi come from work by Professor Malcolm Guthrie, and are re-
stated in collaboration with Dr. Calvin Rensch. The choice of these data is
dictated by the fact that Guthrie's brief study is one of the very few on
African languages which explicitly attempts to exploit slot and class relations,
and hence can be rephrased in terms which make it possible to compare it with
our workshop materials.
Rensch and I suggest a simple matrix, with tagmemic formulas and a cita-
tion paradigm accompanying it, to allow one to see a restatement of some basic
clause types of Bobangi which have differences in concord requirements when
they are independent or embedded. The simple matrix implies a larger matrix- -
with accompanying specific hypothecated items (marked with asterisk). If

these could be shown to be either possible or impossible--by the original

author's comment, or by attempted elicitation from an informant--our deriva-
tive understanding of the system might be to some degree confirmed or corrected.
Professor Guthrie discusses a general approach to syntax, illustrating
it, in part, by data from Bobangi. Even though he makes no attempt to give a
complete presentation of the data--it is an 'avowedly incomplete application'
of his technique (p. 16)--nevertheless they lend themselves to tentative re-
statement in terms of tagmemic matrices.
 -39-

Even the gaps (due to his criteria for selection of somewhat isolated
illustrations for an article designed to illustrate theory) are instructive
in suggesting questions which an outsider would now like to ask of the au-
thor or his informants (whereas a professional scholar of Bantu would prob-
ably know these answers). Field investigation can move rapidly only when one

can ask questions to test a specific hunch about systemic structure. Wrong

hunches are almost as valuable as correct ones--if by their careful checking

and repudiation they open the way for revised hypotheses.
The languages discussed by Guthrie in this work each include at least
the following general types of sentences (p. 1):
Neutral (Independent; can initiate conversation)
Response
Question
Command
Wish
Dependent, embedded (pp. 13, 16)
Contrasts between Nuclei of Neutral Clauses
Only the first of these--the neutral type--is treated for Bobangi. It

divides into transitive and intransitive, and may be modified to become bene-
EI
factive or causative--see the labels on the left column and top row of Matrix I.
We see that from the simple types--transitive and intransitive--can be
derived either benefactive clauses or causative ones. Unanswered, however, is
the question whether both benefactive and causative may simultaneously occur.
Only negative samples are given. The positive contrastive forms are un-
available - -arid therefore the negative signal in the verb complex (which in-
cludes the function of ka at the end of the clause) cannot be identified here.
Dependent clauses are sometimes embedded within the listed clause types, with
a few samples to be discussed below.
Illustrations can now be brought together in a citation paradigm to illus-
trate the three transitive-intransitive pairs of Matrix I--(words involved are
clEngE 'youth', elike-nde '?', olinga 'to like, want', lotomo 'work', ka '?',

o-pim-a 'to go out', o-ten-ela 'to cut', mpomba 'elder', njete 'trees', o -'kon-
isa 'to plant', masangu 'maize'.)
 e6

o644.4mci.,6;dA.4s.stio.,1" Aet.' .A14_

Negative Neutral

Clauses Simple Benefactive Causative

Transitive (1) S + tP - 0 +ka S + +

(3) - btP(el) +B ±0g +ka (5) - S - P(is) +(-+ 0 -0 ) +ka
[c] ac [c] [c] ac [c] [c] cau [c] ac g

+ i
Intransitive (2) 8 + iP +h.q. *(4) -[c]Sac + biP(el) +
+13 -41:1 (6) - S + +
ciP(is) -0 +ka
[c] ac [c] .P-
(c] [c] cau [c] ac CD
I

BOBANOI MATRIX I

Negative Neutral Clauses in Bobangi, with contrastive sequences of nuclear

tagmemes indicated in the cells.

t=1 C.= t= =3 Cla =z11 C013 C=f CZ3 C21=1 c:= Crz CZ, C:74)
. -^
 -41-

1. Transitive, Simple:
elenge eliki-nde olinga lomoto ka [see (2) pp. 13, 14] 'The youth
did not like the work.'
2. Intransitive, simple:
elenge eliki-nde opima ka [see (2b) p. 14] 'The youth did not go out.'
3. Transitive benefactive:
elenge eliki-nde otenela mpomba njete ka [see (2c) p. 14] 'The youth
did not cut down the trees for the elder.'
4. *Intransitive benefactive:
*(elenge eliki-nde opimela mpomba ka) [constructed, by analogy, from
information from the chart at the bottom of p. 15] *('The youth
did not go out for the elder')
5. Transitive causative:
mpomba eliki-nde okonisa elenge masangu ka [see (2d) p. 14] 'The
elder did not make the youth plant maize.'
6. Intransitive causative:
mpomba eliki -ride opimisa elenge ka [see (2e) p. 15] 'The elder did
not make the youth go out.'
Within the cells of Matrix I we have placed symbols (quite different from
those of Guthrie) to show the points of structural contrast between clauses:
The transitives differ from the intransitives (a) by a different list of
verbs (so far as available illustrations go) in the transitive versus intransi-
tive predicate slot (tP versus iP). In addition (b), the transitive clauses
each optionally(-) have a direct object (0 ) which functions semantically as
the goal of the action of the main verb (if simple or benefactive) or of the
secondary verb (if in a causative clause), whereas the intransitive clauses
have no nuclear object functioning as goal of the main verb.
The benefactive clauses differ from the simple ones (a) by the affix con-
tent of the verb (-el versus zero) and (b) by the specific obligatory (+) ad-
dition of the benefactive tagmeme (B).
The causatives differ from the simple and benefactive clauses (a) by the
affix content of the verb (-is versus zero or -61), (b) by addition of the
13
causal role (the causing of someone's action, not the performing of the action
itself) in the optional subject slot (S ) of the clause, and (c) by the
cau
transformation of the optional subject-as-actor tagmeme (of the simple and
benefactive clauses) into an optional object-as-goal tagmeme.

vrtcomm.
77.777.
 --42-

Concord between Subject and Predicate

Both subject types (S ,S ) share the requirement that they be marked
ac cau
(by prefixes) as being in concord with the second part of the predicate (not
with eliki-nde, but with olinga, for example). It is precisely such concord,
plus comparable position preceding predicate, which makes S and S com-
ac cau
prise a formal class of tagmemes. These two tagmemes differ from each other,
however, (a) in transform potential, (b) in role relation, ar4 (c) in re-
lation to the permitted occurrence of other tagmemes (e.g., Oac) in the clause.
The subscript [c] before [c] 0 acl and 0 and tP(e1), etc., symbolizes
[c] cau' [c]
the concord requirement. The absence of such a subscript on the tagmemes 0 ,
0 and B, specifies that no such concord links them to the predicate. Any
ac
concord internal to a phrase would be separately symbolized when the interior
structure of the phrase is itself under discussion.
A nominal phrase may fill subject, benefactive, or object slots. We de-
duce, as a basic formula:
+ Itemh:lotomo- Deictic:/lonya > [c]Attributive:osisc>
lcJ lci
for lotomo lonya losiso 'that other work' (p. 9).
The plus (+) symbol before the tagmemic slot called Item implies that it
is obligatory; the subscript h implies that it is the head of an endocentric
phrase. The colon tells us that the class filling the slot is next to be
given. The pointed brackets represent the class (presumably here a noun) by
a typical member lo-tomo. The deictic lo-nya 'that' and adjective (?) lo-siso
Tod- -' are symbolized as in concord with it. (Contrast the concord of ElengE
esiso 'the other youth' (p. 8).) It is the control of the concord of -siso by
lotomo and clengc which leads us to treat the latter two as head and the others
as modifiers.
F-

The separate modifiers can themselves, however, fill the subject slot with-
out their head. On this basis Guthrie treats all three tagmemes as 'of equal
status' (p. 8, fn.), and as not comprising con_rastive tagmemes in the internal
chain of the nominal phrase. To us, however, the concord requirements, plus
difference of role (and--perhaps--the impossibility of arbitrary, meaningless
change in the ordering of the three, such as *[losiso lonya lotomo]?) imply the
presence of a zero manifestation of a dominating head in subject (or object)
, slot when the deictic or attributive appear to occur by themselves.
 -43-

Dependent ClausEs within Noun Phrases

Included in a noun phrase, attributive to its head, one may find a depen-
dent (relative) clause. The total phrase serves as filler of an 0g tagmemic slot-
and possibly other slots where noun phrases occur. In the following formula
note that the verbal element of the dependent clause is in concord with its
'logical object', the preceding noun which is simultaneously head of the noun
phrase, but is not in concord with the nominal following it (with no other tag-
meme allowed between) as its logical subject:
[ +
( +mdtP + dS)]

Compare lotomo lolakisi moninga (p. 9) 'work-showed-friend' i.e., 'work which

his friend showed him.' Note that the formula shows the Item-as-goal is in con-
cord with the dependent transitive predicate but not--in contrast with independ-
ent clauses--with the (dependent) subject. It is the goal item which determines
the concord prefix of the dependent verb (p. 16). Note, further, that the Item
tagmeme is simultaneously serving in two roles--one as head of the noun phrase,
and one as goal of the dependent verb; the vertical stroke in the subscript
after Item symbolizes this double tagmemic function.
The entire noun phrase, however, serves as filler of the goal slot of the
main clause, in the sentence: elenge eliki-nde olinga[0 :NP I (lotomo lolakisi
g rel
moninga) ka (see [4], p. 7) 'The youth did not like the work his friend showed
him.' Here we have placed [0 :NP outside the parentheses to show that the
g re1]
role of goal has a function on two levels at once--on the level of independent
clause, mnd separately but simultaneously with the included relative dependent
clause. Note further, as indicated by the phrase formula and made explicit by
Guthrie (p, 9) that the dP and dS are mutually obligatory (+...+...), but that
the presence of the combination with its noun head is optional (±[ +... +...1)--
but may not occur without that noun head (fItemblip). (The ±[+... +...] section,
apart from the preceding noun which is its simultaneous head and goal, is called
by Guthrie an F clause.)
A second type of dependent clause (Guthrie's type K) may also fill the 0
tagmemic slot in an independent clause. Here, again, a noun may - -but is not
required -to precede the included verbal element as its object:
[+Og +dP2]
 -44-

This modal nomino-verbal phrase (for terms, see p. 16) differs from the
noun phrase with attributive relative clause, in that this second dependent
clause type (a) has an included predicate which is not in concord with its
preceding object and (b) hence differs at least this much in internal structure
from the relative predicate; (c) has its goal optional, so that the predicate
may be itself fill the object slot of the independent clause; (d) is not accom-
panied by a dependent subject; (e) may not, in its form containing the object,
fill the subject slot of an independent clause. For this dependent clause in
context, note:
elenge eliki-nde olinga 0 :NP = (lotomo bomeki) ka (see [6], p.7)
nom-vb g
'The youth did not want to try [to] work'
Allo-Forms--Variants--of the Independent Clause
From the examples provided us, the meaning of neither eliki-nde nor of
ka can be determined. We assume that eliki-nde has some temporal significance,
however, since we are told (p. 13) that if it is omitted, the sentence as a
whole refers to the'time implied in the context'--but that in such an instance
the subject is obligatory, not optional as shown in Matrix I. If these inter-
pretations of the article happen to be correct, then as an allo-construction
of the independent simple transitive clause (1) we have:
(1) - [ + S - Time + tP 3 - 0 + ka]
ac
in which tP is tP of (1) minus the temporal element mentioned. This conclusion
3
is cast in doubt, however, by other data regarding temporal elements which
lead to expansion variants of (1). Preceding S in (1) we optionally find
ac
nambisa 'afterwards' (p. 12) as a Time tagmeme--or, instead, it may follow the
0 g, and in turn be followed by an Instrument-Cause tagmeme, e.g., by naepamba

'with a knife' or (same tagmeme?) by naewala-embula 'because of the rain.'

The implication for a more complete formula for the independent simple transi-
tive is then:
+ ± + + + + +
[- T S ± tP -0 -Inst+ ka] [- S + tP -0 -Inst +ka]
[c] ac [c] [c] ac [c]
Some Unanswered Questions
Even a study as brief as Guthrie's allows for an initial understanding
of first approximations towards many basic components of a system--and suggests
hypotheses for checking not only with informants in the same language, but with
other closely related languages.
 -45-

Some of these questions we now list:

1. Can further subdivisions of nominal phrases, or of dependent clauses,
be found which can fill subject and object slots?
2. What is the explicit set of restrictions on all slots of an indepen-
dent clause when it is transformed to one of the dependent clause types?
3. Can every clause type of Matrix I be transformed into a relative or
into a modal dependent clause? If so, what is the total dependent matrix of
clauses?
4. Do all clause types of the independent matrix have variants, like
the first type, in which subject is obligatory if the pre-predicate temporal
is omitted?
5. Is this an isolated kind of variant, or one of a pattern of variants
in which deletion of other clause tagmemes is permitted if the behavioral con-
text--or the dialogue context -- specifies it?
6. Are such variants part of a set discourse--conditioned variants which
would include reference to discourse-initiating (p. 1) versus non-initiating
clause types or variants?
7. What other clause types or tagmemes need to be added to expand the
matrix? (Note, for Kongo--a related language--data, later in the article:
numerals in noun phrases, certain particles, special word orders (p. 17);
restrictions by special lexical lists of stems, special clause complements
implying indirect object, location, possession (p. 18); copula in concord
with following element (p. 19); subject nomino-verbals requiring same stem
morpheme as main verb; emphatic versus nonemphatic orders of tagmemes, double
concord--with nreceding and with following elements (p. 20); non-clause linked
by hiatus with clauses--differing as to whether the prehiatus item does or
does not have the same referent as a potential subject of the post-hiatus ele-
ment, tagmeme order variants conditioned by deleted tagmemes (p. 21); paren-
thetic--non-concording--dependent clauses, expanded nominals, quoted questions
embedded in relative clauses (p. 22); clause restrictions with copulas (pp. 23-
24); equational clauses with a variety of contrasting [independent-marked ver-

sus dependent-unmarked] nominal predicates [which control concord with the

accompanying ncminalsj, with contrasting orders for emphasis (pp. 25-26).)
 -46-

8. Must a desiderative clause type be set up in contrast with all those

of Matrix I--inasmuch'as olinga 'want' contrasts with okela 'do' and opima 'go'
(p. 14)?--or is this possibility already covered by handling 'want' as a main
verb, with modal dependent clause as its object (see [6], p. 7--and see dis-
cussion of dependent clauses above.
9. Can causative and benefactive be combined in the same sentence? If

so, how does this modify Matrix I, both as to general possibilities of sen-
tence types, and as to order of tagmemes within them? And can desideratives
be added to such hypothetical combinations, whether to main or dependent verb,
or both?
Stated in terms for checking with an informant, for example, we are curi-
ous to know how many--if any--of the following sentences are possible--or pos-
sible with change of order; or how comparable semantic components are handled
where these forms a7e incorrect:
(The asterisk before the parentheses means, here, that we have in-
vented these sentences. Their tentative justification or rejec-
tion can be obtained by informant elicitation. Firm judgment,
however, must rest upon finding analogous sentences in running
text. This checking is comparable to the standard checking,
against uncontrolled text, of morphological paradigms--but is
more essential for syntactic material since biased word order
in sentences is more likely to be introduced by elicitation or
by translation than is biased morpheme order, or occurrence,
within words.)
a. *(cicnge eliki-nde opimela mpomba ka) 'The youth did not
go out for the elder'
b. *(clengc eliki-nde opima bomaki ka) 'The youth did not go
out to try'
c. *(cicngc eliki-nde olinga opima ka) 'The youth did not want
to try'
d. *(cicngc eliki-nde olinga otenela mpomba njete ka) 'The
youth did not want to cut down the trees for the elder'
e. *(clengc eliki-nde olinga opimela mpomba k..) 'The youth
did not want tc. go out for the elder'

tt
 -47-

f. *(mpomba eliki-nde olinga okonisa elenge masangu ka) 'The elder

did not want to make the youth plant maize'
g. *(mpomba eliki-nde olinga opimisa acne ka) 'The elder did riot

want to make the youth go out'

h. *(mpomba eliki-nde okonisa olingisa elenge masangu ka) 'The elder
did not make the youth want to plant maize'
i. *(mpomba eliki-nde olingisa opimisa elenge) 'The elder did not
make the youth want to go out'
j. *(mpomba eliki-nde opimisa elenge olingisa) 'The elder did not
make the youth go out, wanting to'
k. *(mpomba eliki-nde olinga opimisa bomaki elenge ka) 'The elder
did not want to make the youth try to plant'
1. *(mpomba eliki-nde olinga elenge okonisa olingisa masangu ka)
'The elder did not want to make the youth want to plant
maize'

In comparing, now, this Bantu material with Niger-Congo languages studied

in West Africa, I would point out the Bobangi benefactive and causative compo-
nents in the verb, which lead to clause structures quite different from those
reported for--say--Vagala and Dagaari (of Ghana) or Mbembe (of Nigeria). The
1-1riba (of Dahomey) comes closer, typologically, to the Bobangi in this respect,
with benefactive and causative suffixes.
1.6. Some Clause Components of Hausa (Chad, Afroasiatic)
Miss Gisela Kappler has attempted to abstract some of the data needed for
14
our particular typological interests from Abraham and to restate it in a for-
mat which would facilitate our comparisons.
Based on Kappler's formulas, as applied to Abraham's illustrations (except
where otherwise stated) but with numerous uncertainties, incomplete statements,
and--perhaps--errors, we see the following clause positions and filler alter-
natives:

Intransitive Clause (with marginal tagmemes in parentheses):

(- Time) - Subj + Pred (-Instr. - Modal - Loc )
[c] Lc] intr
NounPh. NPh IntrVPh PrepNPh mNPh LocN
TempPrep NPh Asp Pr PrepPrind Prep NPh LocNPh
LocPrepNPh

_A
 -48-

Here the initial optional temporal tagmeme (with noun phrase, or noun phrase
preceded by preposition and temporal marker), may alternatively. come at the
end of the clause (with same fillers, or with a temporal particle).
The independent subject tagmeme is in concord with the predicate per-
sonal-aspect pronoun of the predicate verbal phrase, agreeing with the pro-
noun as to person, number, and gender. fl

The predicate tagmeme is composed of the verb proper (which in part

determined the basic clause type--e.g., intransitive, transitive) preceded
by the person-aspect pronoun which carries the remaining functions of per-
son (number, gender, aspect), or is composed of the aspect-pronoun by it-
self.

The instrumental slot is filled by prepositional noun phrase, or pre-

position with independent pronoun. This tagmeme may optionally occur in
other positions--such as following the locative.
The modal tagmeme may be manifested by a modal noun or noun phrase,
with or without certain prepositions.
The locative slot may be filled by noun, noun phrase, or prepositional
noun phrase, marked for locative function. The modal and locative tagmemes
may come in reverse order in respect to each other.
zo" (Predicate containing aspect-pronoun and verb)
'I-have come'

yg. zO" da x-6:fly (with temporal prepositional noun phrase in

'He-has come by day' final position)
iya'lln she'n sun sauka la"flya° (with modal noun final; ex-
'Tamily-of Shehu they-have arrived well-being' ample from Hodge)
clan zaki yg shigg cikin bilkka (with locative phrase final; ex-
'young lion he-has entered inside-of hut' ample from Kraft)
kan blyd. ma. 01'14 (with initial temporal prepositional noun
'at two we-shall finish' phrase)
Transitive clause (nuclear tagmemes, only, shown here; marginal tagmemes
in some degree similar to intransitive):
+ +
S + P ± ID 0
[c] [c] tr
NPh VPh NPh NPh
tr 1. 0

Pr Pr
10 o dep
 -49-

Subject agrees with predicate, as it does for intransitive; predicate

slot has transitive-verb filler (or, as an alloconstruction, verbal noun with
other co-occurrent restrictions).
0 Indirect object (=benefactive) tagmemic slot is filled by a marked noun
phrase, or by an indirect-object pronoun, The direct-object slot is filled

o by (noun or) noun phrase, or by a dependent object pronoun

Yg ggya mita sa'kem da dare=

o 'He-has told to-her [a]message at night' (with indirect

object 'her', direct object 'message'; example from Hodge)
nA harb4 shi da bindiga
D
'I-have shot him with gun' (with dependent object pronoun,
followed by instrumental)
TraLlsitive-Alternate 1:
+
+ IO + 0
-[c]S +[c]P tr-1
VPh-1 PrepNPh
tr
PrepPr
ind
D One alternative for transitive requires that certain verb phrases
(trVPh-1)
be accompanied obligatorily by a direct object tagmeme marked by a preposition:
A 0

D na rabti da Itg (with marked independent pronoun)

'I-have left (prep) her'

yang ricie da mgshi

D 4
'He-is holding (prep) [a] spear'
Transitive-Alternate 2:
B +
[c]Ptr-2 + 0
[c]S +
VPh-2 NPh
II tr
.
10

. Pr
10
A second alternative for transitive requires that certain verb phrases
D (VP-2) be unaccompanied by indirect object, but obligatorily accompanied by
direct object with fillers of that form which would--in the basic transitive
type--be indirect object:
yg cfm ma zki
n 'He-has encountered lion'
.
za'ki 37. to sam masa
4
'[A] lion he-has suddenly-attacked him'
(Note ma in 'him'.)
 -50-

Ditransitive Clause (nuclear tagmemes):

±0) "' I
+ P +(±0anim S + P
[c] [c]
[c) ditr [c] [c] ditr.
VPh N . N
ditr arum Pr.
dep (1
Pr. NPh
dep
Vb.N Pr..
ndep
Here Kappler sets up a clause type with a further contrastive verb-stem
list (accompanied, as above, by a preceding aspect-pronoun in the verb phrase),
along with special object co-occurrence restrictions. At least one of, two
objects must follow the verb of the two types, the first requires an animate
noun (or a dependent pronoun, or a verbal noun?), the second has noun or noun
phrase, independent pronoun (varying to dependent pronoun if it directly fol-
lows the verb), or verbal noun (?):
U
shi ita
'I-have given him it'

bAuci. t6' f6 kani5 girma

'Baud she exceeds Kano as to size'
Under some circumstances a particle da 6r she comes between predicate
and animate first object:
na sha -sh6 shi rdwS'
'I-have given-particle him water'
Some clauses may lack a normal verb predicate. A verbal noun phrase (in-
cluding its object) may be followed optionally by an aspectpronoun (in agree-
ment with the object of the verbal noun) and obligatory prepositional phrase
as modal complement:
dibgr muse rdwa' }Tana' dA wayac
drawing of for-them water it-being with difficulty
'It is difficult to draw water for them'
An equational clause:
+
[c] Descrip. + Pred
[c-g] equa
NPh NPh
Pried Pr
poss
Here the descriptive taguleme may come on either side of the predicate (though
shown only preceding it, in the alloconstruction given--with possessive pro-
noun observed in the pre-predicate position but not--in Kappler's formula--
 -51-

post-predicate); the descriptive noun is often translated by English adjective

(see Hodge). Predicate has concord with subject (or sometimes descriptive sub-
stantive?) limited to gender; otherwise these equative verbs do not change in
form--except that their verb tones differ according to the tone of the preced-
ing noun:
mace c4

woman is

ja'kfn nen qaqgne n4'

donkey this small is

ygrg ne' dO'gwa'ye" diA sti

boys are tall with them

'The boys are very tall'
Clauses related to these include information questions (with added particle,
and with clause-final falling tone), yes-or-no questions (differing from affirm-
ation by clause final tone changed to falling), commands (with special verb forms- -
including tone and 1<mgth under certain circumstances):
kddl ngwa ne
money how-much is (-it)?
ya* sa.mi ja'ki
did-he get donkey?
d6ki ne'

horse is?

The negative involves striking changesdifferences large enough to lead

Kappler to tv-fat them as types emically different from their corresponding
affirmatives. For (the various kinds of) transitive and intransitive state-
ments, the negative tagmeme is discontinuous ba.,.bg, with the first part
preceding the verb phrase, which in tura begins with special aspect-pronoun
elements, which, then, determine the tone and length of the negative ele-
ment preceding them. In equational clauses, the first ba precedes the subject
tagmeme. The second bg usually comes at the end of thy: various clauses (but
is absent from clauses with progressive aspect). Command clauses use kgek
instead of ba...bg:
kwA.na' !akii hen ggn shl bg
days three not-have-I seen him not
kgda ka saver
don't you sell-W
 -52-

From this preliminary sample, inadequate though it may be, several like-
nesses and differences are seen between the 14froasiatic) Hausa and the Niger-
Congo languages of West Africa.
As for likeness, both have a quite similar order of tagmemes in transitive
clauses (e.g., subject, predicate, indirect object, direct object, manner, loca-
tion, time (see the Vagala, above). Both have intransitive, transitive, ditran-
sitive and equative types--though this; list of possibilities may be due as much
15
to some degree of universal constraints on human nature and language communi-
cation as to any other factor.
The complexity of the negative clause in its relation to the affirmative
is an interesting typological parallel.
On the other hand, some striking differences appear: (1) The Hausa seems
to have a much wider variety of transitive subtypes, with co-occurrence rela-
tions between a selected set of verbs and the manifesting forms of indirect and
direct object. (2) Hausa has an instrumental tagmeme in its basic transitive
clause. This would lessen the need for some of the special kinds--or frequency
of use--of serial verbs which are so characteristic of the West African Niger-
Congo languages. (3) Hausa allows two pronouns in sequence, in object posi-
tions; the lack of this freedom leads, in Niger-Congo types such as TwL, to
further forced serial constructions. (4) The obligatory, intricate aspect-
16
pronoun complex cf the Hausa verb phrase is quite different from the usual
simple verb (with pronominal prefix and tone changes) of many of the Niger-
Congo languages.

War

;Nr
 -53-

FOOTNOTES
1
Based, of course, on our limited sample--a restriction which we shall
not repeat each time, but which may lead to modifications later. The Niger-
Congo languages of Ghana and Nigeria which we refer to as our primary sources
for clause data are largely Baraba, Dagaari, Kasem, Vagala, Degema, Igede,
Mbembe--and, from secondary sources, Twi.
2
In an article entitled 'Some Restrictions on Objects in Twi,' Journal
of African Languages, 2.145-49 (1963).
For Twi, cf. Stewart, op. cit. p. 149: ' "come, bring, send" ...can be
either intransitive or transitive, but which paradoxically never have a direct
object even when transitive.'
4
Imperfective is derived from perfective by a set of rules, phonologically
conditioned, involving added vowel length, sometimes with r or n, and with vowel
harmony.
5
See his Grammar Discovery Procedures (The Hague: Mouton and Co.) 1964.
Note some restrictions--such as in reference to concord, or agreement, where
two differences do not seem to justify an emic contrast. For concord rer.

strictions,see K. L. Pike 'Dimensions of Grammatical Constructions,' Language,

38.221-44, 1962.
6
In morphology, Etung has concord systems reminiscent of the more extensive
Bantu concord. We shall in §1.5 turn to published Bantu data to show some
syntactic characteristics -- which in part are more like the Bariba of Dahomey
than the Dagaari or Vagala of Ghana, or even the Mbembe of Nigeria.
71 have been encouraged, in dealing with the distribution of clause vari-

ants, by recent descriptive approaches by Robert Longacre used in dealing with

Trique (Mexico) clauses, "Trique Clause and Sentence: A Study in Contrast,
Variation and Distribution," IJAL, 32. 242-52 (1966).
8
For semantic linkages of subclusters, see §§2.2.5., 1.1.1.
©p. cit.
10
A detailed description of some of these phenomena from a transformational-
st viewpoint (but without the attempt of studying emic--vs. etic--differences)
is seen in Kay Williamson, A Grammar of the Kolokuma Dialect of Ijo, West African
Language Mont 2 (Cambridge: The University Press) 1965.
 -54-
11
Greenberg's classification--not accepted by all scholars--places Bantu
in his section IA5D, which implies that Bantu is more closely related to Etung
than to any other language within the direct purview of this report.
12
Bantu Sentence Structure, School of Oriental and African Studies, Uni-
versityof London, 1961. Page references refer to this article. Preceding the
page numbers, the numbers in-parentheses refer to his numbered sentences on
those pages.
13
0ur use of role, here, is related to Guthrie's mention of 'logical sub-
,

ject' and 'object' in fn. 1, p. 16. Our use of labels--rather than Guthrie's
letters and Roman numerals--for tagmemes and tagmemic slots allows the reader
easier insight into the relations involved. Guthrie's work, however, is val-
uable in another fashion, in demonstrating the distributional validity of
groups and sequences, by using formal labels which lack semantic overtones.
14
R. C. Abraham, The Language of the Hausa People (London, 1959). She
uses some supplementary samples for C. T. Hodge, An Outline of Hausa Grammar,
Language Dissertation, #23 (1947); and Hodge and I. Umara, Hausa, Basic Course
(Washington, 1963); C. H. Kraft, op. cit. A Study of Hausa Syntax, Vol. 1-3,
Hartford Studies in Linguistic 8-10 (1963).
15
Precisely here is one need for wider sampling of languages--a sampling
forwarded a bit by this report. Once a large enough sample is available, more
certain generalizations can be made.
16
See chapter 5 for this data.
 CHAPTER II! CLAUSE CLUSTERS IN SEITENCES

Investigations of the workshop dealt only briefly with materials concerning

the sentence level as such. Studies concentrated more on clause materials,
and clauses in (less-than-sentence) clusters. Nevertheless, the study of sentences
proved useful. On the one hand it was needed to separate clauses and clause
clusters from larger structures, and, reciprocally, it was essential for differ-
entiating emic paragraphs and discourse from smaller units. The sentence proved
to be relevant as the distributional setting for clauses, but in sequence certain
sentence structures comprised paragraphs.
2.1. The Sentence as Setting for Clause Distribution
Crucial observations about Kasem sentence structure were supplied to the
members of the workshop by Dr. John Callow and Mrs. Kathleen Callow. From their
data, the accompanying Kasem chart may be constructed. In Kasem Chart 1, Row A
lists the simple sentence as itself a unit as a whole. Row B shows the impor-
tant included structural sequence of optional dependent clause (or clause cluster),
then the obligatory nucleus of the sentence, which is an independent clause
(or clause cluster), and last the final, optional, dependent clause (or clause
cluster).
Row C shows the gross potential internal sequence with each of-the slots
of Row B. Each slot of the simple sentence may be filled by a sequence of
clauses--a clause cluster. The first member of the cluster (whether the cluster
as a whole is dependent or independent in relation to the sentence as a whole)
begins with a primary clause (the nucleus of the cluster, as over against the
higher-level nucleus of the sentence), and is followed (with certain restric-
tions to be discussed in §2.2) by one or more secondary (serial) clauses.
The three elements of Row B are differentiated from each other not only
by their place and function in the sequence within the sentence (i.e., by their
external distribution), but also by their internal structure. For the latter,
see Kasem Chart 2, where NP, VP are nominal (or pronominal) and verb phrase:
subscript r implies 'realized' aspect, such that VPr = + waha/ta + Vnf,
in which the verb is n(on)f(uture); subscript nr implies 'n(on)r(ealized)'
aspect, with VP + in which, in turn, VP = ± ma/maa/ta + V
nr Vf /Vs /dip' f'
and Vf is f(uture) verb; and VP. = + yf ± to + Vip, with V as i(m)p(erative)
ip ip
verb.

-55-
 Kasem Chart 1

A [unit]: SIMPLE SENTENCE

1
1
+ +
B [sequence]: Initial Dependent Clause + Independent Clause or Final Dependent Caluse or
or Clause Cluster Clause Cluster Clause Cluster
I 1 1

I I 1 I
I
+ + +
C [sequence]: + Primary - Secondary + Primary - Secondary Clause + Primary Secondary Clause
Clause Clause Clause (or Clauses) Clause (or Clauses)
I(or Clauses)
I

1 t
t

D [class]: Neutral {- Secondary]

Nonterminal [+ Secondary]
CI

__Terminal [- Secondary]

E [class]: Intransitive Auxiliary Equation

Transitive Abilitative Locative

Ditransitive Causative

ICS en =3 =I CMC3 C2111 Ce21 1:223 C=:3 CM CZ) CZ:3 C:23 C=3 C=3 CZZ C:=
IIITII
 Kasem Chart 2
CONTRASTIVE NUCLEAR STRUCTURE OF PRIMARY CLAUSES

Dependent + S:NP + + ?red: VP

Independent +. S:NP + Pred:VP

r/nr

Final
Dependent +sl/pg +S:NP + Pred:VP
nr

0
Note, therefore, that these clauses differ internally by special particles,
and by the range of tense and mood which is allowed in the verb phrase.

0 Contrasts between the primary clauses of Row C of Chart 1, that is, are differ-
entiating features of elements of Row B.
2.2. Clause Clusters (Serial Clauses)
When, however, our focus shifts to the clause clusters themselves, we
are interested in the relationships between the p :imary and secondary clauses
which respectively comprise the nuclei and the serial members (or margins)
of the clusters. Row D of Kasem Chart 1 shows some of the constraints on this
relation.
2.2.1. General Restrictions on Sequence Types in Clusters
Specifically, certain Kesem clauses (shown in Row E of Kasem Chart 1 to
be the equative and locative clauses) cannot be followed by a secondary, serial

fl clause when they themselves are the nucleus of the cluster (i.e., when they
are primary). This set, therefore, is by the Callows called terminal, in rela-
tion to the cluster. On the contrary, however, nonterminal primary clauses
are required to have a secondary clause (or clauses) following them. This set,
seen in Row E, is made up of a special set of auxiliary, abilitative and causa-
tive clauses. The other type is neutral in respect to this requirement:
it may be followed by a secondary clause, or it may itself close the cluster.
Neutral clauses, as a class, include intransitives, transitives, and ditransi-
tives.

These observations concerning primary clauses apply, whether in reference

I
to the independent clauses of Row B, or the dependent ones.

:
 -58-

The Callows demonstrate that the members of the neutral, nonterminal

and terminal primary clause sets differ not only by their contrastive distri-
bution in relation to the occurrence or non-occurrence of secondary clauses
with them, but also in relation to their internal structures. Members of the
first set, the neutral one, have internal differences in relation to object
(intransitive, none; transitive, optional object; ditransitive, obligatory
occurrence of either of two objects, or both of them), and by different verb
stems. The nonterminal set likewise differs in terms of stems (auxiliary, with
fourteen verbs such as dgari 'to do next', kwgani 'to do with effort and suc-
ceed% kowri 'to continue doing'; abilitative, two verbs, wgni 'to be able',
and wari 'to be unable'; causative pa 'to give'), but differs also in that no
r-
added object, complement, or related marginal tagmemes are allowed. The ter-
minal clauses, on the'other hand, have special complements or margins as well
as stem differences (equational, with nominal or adjectival complement; loca-
tional verb with a locational marginal tagmeme as well).
Note that one might have expected here a description in which the auxil-
iary verbs were treated as auxiliaries to the main verb, and in the same verb
phrase with them. This is the analysis implied in much of our work in §1.
There is a difference in Kasem, however. Here the auxiliary has its
subject, and the semantically main verb also has its own subject, making it
simpler for the Callows to treat the subject-auxiliary combination as a clause
in its own right. In order for the reader to see this facet of the structure,
we add a partial syntactic paradigm. Verbs here were chosen by the Callows
to demonstrate the differences between clauses. Perfect aspect is used when
possible (not possible in nonfuture equative and locative, for example).
Kasem, Nonfuture, Independent Clauses:
Intrans. 6 kytiwga 'he jumped down'
Trans. 6 nyOgi ng 'he drank water'
Ditrans. 6 biri ni sOngo kdm 'he showed me the house'
Auxil. 6 kwgani -6 kytiw 'he jumped down with an effort'
Abil. 6 wgni o kydw 'he was able to jump down'
Caus. 6 p-6 -6 kydw 'he made him jump down'
Equat. 6 yi babia 'he is/was brave'
Locat. 6 wu Offigo ni 'he is/was in the house'
 -59-

Kasem, Future Independent Clauses:

Intrans 6 wit kydw 'he will jump down'
Trans. 6 wit ny6 na 'he will drink water'
Ditrans. o wu biri ni songo kum 'he will show me the house'
Auxil. 6 wit kwaani o kydw 'he will jump down with an effort'
Abil. 6 wit wani o kydw 'he will be able to jump down'
Caus. 6 wiz pa 6 kyilw 'he will make him jump down'
Equat. o wit to yi babia 'he will be brave'
Locat. 6 wit to wi sbngo ni 'he will be in the house'

Kasem, Consecutive Independent Clauses:

Intrans. -6 ma kydw and he jumped down'
Trans. 75 Ilia nyO na and he drank water'
Ditrans. -6 Ma biri ni sOngo kum 'and he showed me the house'
Auxil. U ma kw6ani -5 kydw 'and he jumped down with an effort'
Abil. -5 ma w6ni -5 kyiiw and he.was able to jump down'
Caus. 75 ma pa U kydw and he made him jump down'
Equat. U maa yi babia and he was brave'
Locat. -6 mil wu sOngo ni 'and he was in the house'

Kasem, Imperative Independent Clauses:

Intrans. 6 kydw 'he should jump down'
Trans. o ny6 na 'he should drink water'
Ditrans. 6 biri ni sOngo kdm 'he should show me the house'
Auxil. 6 kvaani 6 kydw 'he should make an effort to
jump down'
Abil.

Caus. opa kydw 'he should make him jump down'

Equat. o to yi babia 'he should be brave'
Locat. 6 to wu songo ni 'he should remain in the house'

Kasem, Initial Dependent Clauses:

Intrans. 6 na kydwgi td... 'when he jumped down...'
Trans. 6 na ny6gi na td... 'when he had drunk some water...'
Ditrans. 6 na biri ni sOngo kum td... 'when he had showed me the house'
Auxil. 6 na kwg.ani o kydw td... 'when he had jumped down with
an effort'
 -60-

Abil. b na wAni 3 kytiw tii... 'when he had been able to jump down' 0
Caus. 6 na p-e- 3 kydw tu... 'when he had made him jump down'
Equat. 6 na yi bara tit.. 'as he is brave...'
Locat. O na w-i1 sOngo ni tti... 'when he was in the house...'

Kasem, Final Dependent Clauses:

Intrans. CT kytiw 'to jump down'
Trans. ,..si o nyO na 'to drink water'
Ditrans. -6 biri ni sOngo ktim 'to show me the house'
Auxil. ...si o kwaiani o kytiw 'to make an effort to jump down''
Abil. ...si -6 wki o kytiw 'to be able to jump down'
Caus. ...si -6 pi -6 kytiw 'to make him jump down'
Equat. ...si o to yi babia 'to be brave'
Locat. ...si o to wu sOngo ni 'to remain in the house'
2.2.2. Specific Restrictions on Sequence Types in Clusters
Studies were started to specify in more detail the particular clause types
from language to language which might be primary or secondary--or tertiary,
etc.--in the clusters.

Jacobs, for Bimoba, prepared for me a co-occurrence matrix, with clause

types listed at the left, and the same types along the top. If the clause
in a column could follow a clause indicated by a row, a check was put in the
cell at the intersection. Preliminary results--see accompanying Bimoba
Matrix--showed: (1) That neither introducer nor stative clauses could occur
as the first of a cluster (note that all cells are empty in fifth and sixth
rows but not in fifth and sixth columns). (2) The stative clause could not
follow a demonstrative clause, (3) nor could an equative follow itself.
(4) Impersonal clause (which I have added to the chart because of her written
data accompanying it) never entered into a cluster at all (except in certain
adverbial uses), either as first or as second membersec, empty cells in last
row and column.

t.

f-'"",""7777-77=1111111M"-'""""'"7"'-`727-"'"' -
 -61-

Co-Occurrence Matrix of Bimoba Clauses in Clusters

Secondary
Clause Intrans.2 Trans.2 Ditrans.2 Equat.2 Introducer2 Stat.2 Impers.
2

Primary
Clause
Intransitive x x x 1C x x
1
Transitive x x x x x x
1
Ditransitive x x x x x
1
Equativel x x x x x

Introducer
1
Stative -
1
Impersonal].

In addition, Jacobs specified a number of general considerations relating

to clause clusters: (1) When the cluster threatened to become too long (exact
lengthnotspecified), with too many clauses in it, it would be broken into two.
(2) If spoken so slowly that two breath groups were involved, the series would
be broken into two clusters. (3) If any one verb were accompanied by too
many phrases, the containing clause was not followed by further members of a
cluster. (4) Some clauses--e.g., impersonal, with the exception of the
adverbial uses suggested above--never entered any clusters.
On the other hard, fast speech, especially when there was shared object
(plus shared subject), often led to joining clauses in a cluster. Similarly,
when two verbs seemed to be involved in representing a single action, these
also were often built into a unit cluster (see instrumental in §1.1.1 and see
§2.2.3).
Kennedy, meanwhile, had been developing his description of Dagaari clauses
with special reference to their contrast, variation, and distribution. For
the latter, he likewise prepared a matrix showing co-occurrence restrictions
2
on clause distribution within clusters. His preliminary matrix--which see --
indicated a regularity and lack of restriction for intransitive, transitive,
ditransitive, locative, stative, and descriptive. Each could be first in a
cluster and each could be second. The demonstrative, on the other hand, did
3
not appear in either position, in series with these others.
 -62-

Co-Occurrence Matrix of Dagaari Clauses in Clusters

Secondary
Clause: Intran. Trans. Ditr. Loc. Stat. Des. Dem.

Primary
Clause:
Intransitive x x x x x x
Transitive x x x x x x
Ditransitive x x x x x x
Locative x x x x x x
Stative x x x x x x
Descriptive x x x x x x
Demonstrative - - - - -

A much more extensive study of distribution of clauses in clusters is now

underway by Kathleen Callow, with two developments beyond the work of Jacobs
and Kennedy: (1) She adds to a Kasem matrix those clause types which are non-
terminal (see reference to material in §2.2.1 above), and certain others which- -
perhaps- -are derived. (2) She is obtaining from Jacobs and Kennedy similarly-
arranged extended matrices, going beyond their early work in Bimoba and Dagaar!,
and is thereby able to make comparison of important likenesses and differences
between the three of them. (Kasem, for example, has much less freedom of
occurrence of clause types--only intransitive, transitive, ditransitive occur
reciprocally in both first and second positions--than does Dagaari.)
2.2,3. Restrictions on Tagmemes within Clause Clusters
4
In §1.3.7 we saw that for most of the languages of West Africa the subject
of the second (or third, etc.) clause must be deleted when it is the first
subject of the first clause of a cluster. A repeated object in a clause cluster
is also deleted (or the two clauses of the sequence may be said to share the re-
tained subject or object).
Once this phenomenon was clearly in view, I wanted to know what happened
to other tagmemes of the second clause of a cluster. What would happen to the
tagmemes of location, time, degree, manner, when these (like subject or object)
were the same in two clauses?
For this purpose, our most useful set of illustrations came from R. Bergman's
work on Igede.
First we give a set to show the sharing of cluster-initial subject and cluster-
medial object (or, alternatively stated, deletion of like-subject and like-
object from second clause):
 -63-

AhI hu 610
we take load
plus ahI chi5 616
we put-on-head load
yields ahi hu 61i; chv

we take load put-on-head

Next we show that a locative peripheral tagmeme, shared by both clauses
of a cluster, occurs at the end of the cluster rather than--like object--at
the end of the first clause (or, one may say, the first of two like locatives
is deleted).
Ahi hu 6lC7 i-ihi
we take load in-market
ahI chu au i-ihi
we put-on-head load in-market
ahI hu Olj chU i-ihi
we take load put-on-head in-market
The time tagmeme acts like the locative:
AhI hu (51A3 a1

we take load to-day

AhI chT) 6117 Alg

we put-on-head load today

AhI hfi 616 chT) Alg
we take load put-on-head today
The manner tagmeme acts similarly:
AhI hu 61A-5 Inylnyl

we take load similarly

ahi chT) (51.(3 Inylnyl
I.

we put-on-head load similarly

am MI 61(3 chv Inylnyl
we take load put-on-head similarly
Compare an "adverbial" modifier:
AhI hu 61E, wfia

we take load all

AhI chT) 61Z w.da

we put-on-head load all

ahi hfi 61(7, chU wuu
we take load put-on-head all
 -64-

Compares also, the onomatopoeic adverb ("ideophone"):

A

;he chI ate h5 wiriwiri

sky cut drops do (of sprinkling)

'It was sprinkling'
Similarly, note the cluster-final place of an aspect particle:
S Aux Pred Do Pred Do Pred Do Asp

5 ka da 5mU hd TpWa
he will do start take thing to be house completive

'He will start to take the things from the house'

(Various possibk, relations still need investigation. In the text studied
by Bergman, for example, no two ditransitive clauses were f'. 'd in a clause cluster.
Is this accidental? Or a structural restriction?)
All such special placements, and shared elements, contrioute to the tying
together of two or more clauses (or clause fragments) into a clause cluster which
as a whole comprises a structural unit.
It is to see unity in the cluster that one emphasizes that sharing or
placement is found in relation to the whole cluster. From this point of view the
cluster is a single wave, with the primary clause as its nucleus, but with no
possibility of clean-cut segmenting between the clauses without distortion.
If, however, one wishes to study separate clause types in relation to their

variants--the alloclauses--distributed in the (higher-layered) cluster (as in

§1.3.7 in relation to contrast, variation and distribution), then one must accept
the risk of segmenting-distortion, assign each tagmeme to one clause (or assign fl
the tagmeme twice, once to each). This leads to a particle perspective for the
units and their allos.
Wave and particle views interlock, however, if a process (wave) view is
used to express "deletion" of tagmemes before the allos are described in terms
of items and their arrangements (distributions). (One can, however, use a seg-
menting approach, describing distribution of allos of clauses without implying
a dynamic change, if an unmixed item-and-arrangement approach is desired.
That levels of a hierarchy interlock,. furthermore, is demonstrated by the
relation between §1.3.7 and this one: Whereas in the former the cluster was a
distributional matrix for clause allos, in the latter the clause variants signifi-
cantly help to characterize the cluster. (So, also, allophones conditioned by
the stress group may help to characterize the form of the stress group.)
 -65-
Some tagmeme occurrences can prevent the formation of a cluster. In
Vagala, for example, two or more clauses may not fuse (by the relevant deletions)
to a cluster (1) if the following clause begins (a) with a time word, (b) with
5
a tagmeme front-shifted for emphasis, (c) with a clause marked for focus (see
§3, where such a clause begins a new "paragraph"), (d) where the first clause
is du 'to be there', (e) where the original subject is expanded or changed in
any way. Other constraints are under study.
2.2.4 Agreement Restrictions within Clause Clusters
In addition to the presence of tagmemes as wholes being affected, and their
placement, there are further restrictions concerning tagmeme variants. Certain
allotagmas of one clause must be accompanied by the requis;te allotagmas in the
other clauses of the cluster. In the first clause, for examp16, a tense particle
(e.g., dang 'will') may be used--but further tense markers will not occur in the
following clauses of the cluster:
37.14 clang iyzi di vOwl di nA
we will get-up and walk and see
a clang kpA nail la to kpA
he will take cow the give him he take eat

'He will give him the cow to eat'.

B Similarly, nonfuture--usually unmarked in the first clause--will continue

unmarked. See fa 'past':
kpang 6(1 fa la pir' di ga na wia gan di ba. sal di...
hunter used past to-go bush and go see things much and come roof and...
'The hunter used to go to the bush, see many things, and come home, and
A negative cluster, likewise, does not repeat the negative marker:
I'
wa lAnge gA nA A maa
11
he-not go go(out) see his mother
'He didn't go and see his mother'
If however, the second clause is to be negated, whereas the first clause
is not, then the cluster is broken, an adversative added, and a subject (some-
times fused with negative particle) is reintroduced to the second clause:
A lAngO kA wa u maa na-we
he went but he-not his mother see
'He went but he didn't see his mother'
Contrast the double positive:
A lAnge) ga na a maa
he went went saw his mother
 -66-

On the other hand, the clauses may have mixed aspect, as perfect or imperfect:
ee nil di ra tá wii'zi hOplz1 bily maa
he perfective water and imperfective throw-to fetish day every
(did) (continuation)
'He took water and sprinkled it to the fetish daily'

A cluster restricted--say--to future would be an alloconstruction, as one etic

member of the emic construction with its varied list of potential manifesting
types.

In Kasem, all clauses of a regular cluster must agree as to aspect and

mode. If the series is in perfective (non-continuous) aspect, the mode will
be consecutive! if aspect is imperfective (continuous) the mode throughout will

1
be non-future.
In the Kasem occurrence of subject pronouns in secondary clauses, the
pronoun must agree with the subject of the primary clause in person, number, and
class: its tone is mid in non-continuous clauses, low in continuous ones.
Occasional irregular clusters occur in Kasem--as when the object (instead
of the subject) of the primary clause has the same referent as the pronominal
subject of the secondary clause--see 'mouse it' in the following sentence (which,
normally, would be broken into ?arallel regular clauses or sentences):
mu o ne tutwdy di gyey di soli gOli yi dl nabiyli D
focus he saw mouse it sitting it stirring porridge and its tail
sin zwin
washing calabashes
eft

'He saw a mouse sitting stirring porridge while its tail washed calabashesli

2.2.5. Development of Clause Sub clusters

In §1.1.1 we pointed out that in many West African languages no instrumental
tagmeme could be found within the basic, single clause, but that the expression
of the instrumental concept required the usage of a clause cluster: 'He take
knife cut meat that', i.e., 'He cut the meat with a knife'.
The necessity for using a complex construction for this relation forces us
to attribute that relation to some part of the resultant structure. On the B
surface, one can see only ordinary clauses within a routine cluster. Where, then,
is the relation signalled? Carried by what units?
 -67-

My approach is to treat this construction as a clause subcluster having

two tagmemes relevant to that layer of structure. The first is an instrumental
tagmeme, the second an action tagmeme:
+ Instrumental + Utilizing-Action
Filling the instrumental slot is a special clause which has formal features
accompanying the instrumental semantic implication.
Using, now, data and description worked out with Crouch for Vagala, I wish
to demonstrate the difference between a (close-knit) subcluster and the more
regular (loose-knit) clause clusters.
The ordinary serial cluster can be viewed as the sum of the component clauses
of the cluster. The total meaning of the series is the sum of the included clauses.
But in the close-knit series, the subcluster, the meaning of the series is not
the sum of the meanings of the included clauses. Rather, the total function is
something above and beyond that of the individual elements.
We begin with the illustration from Vagala which parallels Stewart's Twi
example in §1.1.1:
kpg klyzee mong Owl
he took knife cut meat
'He cut the meat with a knife'
Here 'knife' is the direct object of 'take'. Thus 'He took knife' is a simple
regular clause so far as the transitive predicate plus direct object is concerned.
Similarly '[He] cut meat' is a regular secondary clause (with regular subject deletion)
which includes a predicate with its regular direct object. The first clause- -

'take knife'--has the meaning 'instrumental'.

This instrumental function is not marked directly by any one morpheme;
nor is it marked by any one particular ordered sequence of morphemes. It has

to be deduced from a complex of factors. The general overall situation (non-

linguistic setting, or setting in discourse) lets one see that the knife is the
instrument of the cutting; and restrictions on the seauence show that it is,
however, special. Inasmuch as the two included clauses are both transitive, however,
and inasmuch as the verb kpg may occur as the first transitive verb in an ordinary
sequence, it would appear that one could expect some ambiguity. We now show
not only how this ambiguity can occur, but the means by which the ambiguous in-
strumental type is structurally differentiated from the regular sequence.
 -68-

An elaborated formula for the instrumental restricted series is:

+ Instrumental + Utilizing Action
+Subj +Predylinstr +Obj +P +Obj
tr
INP Ikpa INP Iv INP
'cake' 1 tr 2

Here we have preserved, on the first level of symbolization, the special two-
tagmeme close-knit series, of an instrument set forth and then utilized. Within
each of the tagmemic slots (shown by vertical lines) is an included construction,
tagmemically symbolized, filling the slot. For each of its slots, in turn, the
fillers are given. Here note that the only verb which is allowed as a filler of
the instrumental slot is 'take'. Similarly, only the transitive verb is allowed
as the second of the series--and its object must be different from the object
of the instrumental member of the series.
In contrast with this, the regular series of a sequence type has various
allo-constructions which allow a great deal more flexibility. We give typical
samples here, involving transitive versus intransitive verb, and presence or ab-
sence of object:
Alloal: (+S) +P +0 +P +0
tr tr
IV NP IV NP
tr 1 tr 2

Allo : ( +S) +0 +0
a2 +Ptr +Ptr
IV INP IV I -NP
tr 1 tr 1

Allo : (+S) +P -0 +P
a3 tr tr
fl
V EJ
tr IVtr INP2 E-NP11
Allob: (+S) +P. +P
intr i ntr
vintr v
intr
Allot: (+S)

Vntr
+Ptr
tr
±0
+1)i
v
intr
Alio (+S) +P. +Ptr ±0
d. intr
vntr v
tr
These allo-constructions are designed to show certain co-occurrence restric-
tions. Allo-Set a begins with a regular transitive verb and is followed by a
direct object--with the direct object of the second transitive verb deleted
(Allo a-2) if it is the same as that for the first. The third allo of Set a
with transitive verb deletes the first direct object if it is in a sequence of
sentences, such that the context implies clearly what this object is. Allo b
includes intransitives only; whereas c and d show transitive with intransitive,

111715.7,11,1e, um,
 -69-

9 and intransitive with transitive respectively. (Further allos occur when emphasis
leads to shift of certain tagmemes to the front of the construction--or when pe-
ripheral tagmemes are involved. These are not of concern to us at the moment.)
What we wish to point out is that ambiguity can occur with the instrumental
and the sequence series only when the sequence is of such a type that the first
verb of a sequence is kpa and the second verb is transitive with a noun-phrase
object which differs from that of the object of the first noun phrase. Such a
circumstance can occur with
kpa bikAwne di ee kylAl
he took round-stone and made [did]. blood
That is 'He "xid" and (he) "y'd".'
The second of these clauses, as indicated by the formula, can have the
first verb replaced without destroying the overall relationship of the parts-
for example, 'He saw a stone and made blood'.
On the other hand, the same morpheme sequence might conceivably manifest
the instrumental subcluster: 'He, with "x", "y-ed", 'He with a stone made blood'.
But in that case the kpa 'take' is replaceable by no other verb; and the second
noun cannot be replaced by the first (*'He take stone made stone') and then be
deleted(*'He take stone make') according to the formula for the regular cluster,
Allo a-2; nor can the second verb be replaced by an intransitive (*'He tike
stone ran') as for allo-c of the regular cluster.
The instrumental series is a close-knit subcluster of the grammatical hier-
archy coning between the simple clause and a regular--loose-knit--clause cluster,
and constituting a complex unit in its own right, with its own contrastive tagmemic
sequence.

In our view, furthermore, these unit subclusters with their special tagmemes
appear to be somehow "new" to the language. The language is in a state of
transition. Here the transition is toward a more complex structure of clause
clusters. Only a dynamic view of the total system can do descriptive justice
to such data.
Although we have used the instrumental subcluster for illustrative purposes,
Vagala seems to have various other types of subclusters currently relevant, or
in process of development:
I3enefactive:

In the benefactive, the verb to 'give' works with the main verb, which
precedes it. The object of to is the benefactee of the action of the main
 -70-

verb. In this construction to takes on a slightly specialized meaning which we

would translate in English as 'for'. There are restrictions on the objects
which to may take in this construction:
wa sg igyo
gyo te u bowl
he tame danced igyo-dance give his village
'He danced the igyo dance for his village'
Indirect Object:
Certain verbs may take two objects--a direct and an indirect object.
But in terms of frequency- if the verb is to 'give' it will usually resort to a
two-verb construction--kpa 'take' plus the direct object, and to 'give' plus
the indirect object. In this construction the ditransitive verbs seen so far
are limited to to and bagli. There are further restrictions on the objects,
since the first object must be different from the second, and the classes of nouns
filling the second object slot are limited:
I

kpa 'hii 'bagli

he took yam showed me
'He showed me a yam'
compare: U bagli n hii
he showed me yam
Accessory:

The verbs kpa 'take' and laU 'take hold of are often used as accessory
verbs to a main verb. The main verb may occur alone and take an object, but some
verbs rarely occur without the accessory verb. The accessory verb precedes and
takes the object (if there is one) and the main verb follows, without an object,
AIN
since the object must be the same in this construction, and, therfore, not repeated: [I

a kpa 'nmag dU
he tocY rope put-in his bag
'He put the rope in his bag'
! .
compare: U du nmeng u huwr
he put rope his bag
Accessory motion:
Preliminary counts suggest that motion-verb combinations are more common
than any other type in the language. The verb iyzi 'get up', for example, in
one set of texts, occurred alone 45 times, and in combination with another verb
180 times. In many instances the meaning had become specialized or slightly changed.
 -71-

Another verb kaall. 'to leave' occurred alone 45 times and in combination 60 Limes.
(It occurred alone more often in questions and commands.) When it occurred in
combination it often had a slightly different or specialized meaning.
U I kpglkeeli
he took went
'He went'

2.2.6. Development of Auxiliary Verbs from Clause Clusters

It should not appear surprising, in the. light of the preceding section on
subclusters, that auxiliary verbs--joining main verbs to make complex verb phrases- -
should develop out of clause clusters. Once the regular full cluster is weakened
to a subcluster with special verb and special distributional restrictions, the
subcluster would look like a phrase as soon as a few remaining components (e.g.,
object of the first-verb-turning-auxiliary) were also lost.
The weakening process, once under way, does not always stop there. The
auxiliary verb may itself lose so many of those characteristics identifying it
as a verb, that it is more convenient to treat it as a particle. New parts of
speech, or new classes of forms may be created in this process--or a class may
occur, nonuniform in type, in a transitional state.
It is this dynamic (wave) component of synchronic description that I now
wish to illustrate. I continue with Vagala data, and in collaborating in its
presentation with Crouch.
Several sets of verbs occur: transitive, such as 16 'hit', which may op-
tionally take an indirect object; intransitive verbs such as sow 'sit' which are
never allowed to take an object; equative ee 'is'; etc. Many of these verbs can
occur in clause clusters in a fairly free manner. (These have often been called
"serial verbs".) See §2.2.
Note for example:
U sowg6 kg.f ra ni di a 16 kg'f oile
she sat chair on and imperf. pound fufu [food]
U ge bd ko6wri di re sow ba se.ey

he is their chief and imperf. sit their front

'He is their chief and sits in front of them'
In a sequence of two such verbs the first--the primary one--often has a
specially-marked form which differentiates it from others in the sequence after
 -72-

it. Note, for example, that the verb sow 'sit' has low stem tone and a suffix
when it occurs in the primary clause, but it is high and does not have the suffix
when it occurs in a cluster:
!

sOwgO kg rg ni a le ka'rg di sow

he sat chair on he got chair and sat
Note, also, that preceding either the primary verb or the secondary verb there
can be the particle ra or a 'imperfective' (see examples above).
fl
Now, however, we note that preceding the optional imperfective marker
there can be a modifying morpheme--modifying the main verb by an indication of
motion. Note, for example:
wa a .16 kg.bila
he came imperf. pound fufu
u ga a 'sow Vara ni
he went imperf. sit chair on
Our intent here is to call these two elements, wa and ga verbal auxiliaries- -
a subset of verbs--and the primary or secondary verb as main verb. Auxiliary
plus particules plus main verb, etc., make up the verb phrase, whether primary
or secondary.
The problem: These two auxiliary morphemes are not allowed to have the
kind of affixes which occur with the main verb of a verb phrase. With the main
verb dggif 'cook', the final i may be dropped and replaced by 6 'perfective'.
When no object follows the verb, this ending is restricted to occurrence with a
positive verb. If an object follows, the i form is used; or the i may be replaced
by e 'perfective' if the verb is negative. (This ending is also used in various
other situations, as, for example, when emphasis is on the subject. The further
problerl here are irrelevant to our discussion.)
Note, for example,
ba dggli dO'zi
they cooked soup
ba daglo
they cooked
ba wa dggle
they not cook
The wa and ga auxiliaries, however, are never allowed to have these suffixes.
In this sense, therefore, they are sharply different from ordinary verbs. The
 AN

-73-

bk.

verb which follows them, however, will be marked not only with one of these suffixes
but all the other evidences of a main primary verb of a regular series.
In addition, four other characteristics make the wa and gA differ from main
verbs: (1) They cannot be preceded by the imperfective ra. (2) wa and ga are
not allowed to occur alone as verbs; they occur only in this position modifying
a main verb. (3) The tone rules which apply to wa and ga are a different set
from those which affect ordinary verbs: The tone of wa and ga may be raised by
a high tone on a preceding word in nonfuture constructions; in the future con-
structions (including purpose clauses), wa and ga are never raised by the preceding
high tone subject or particle, whereas other low verbs are. (4) Unlike other verbs
discussed, whether they are primary or secondary in a cluster, in a series these
auxiliaries are not allowed to have those complementing tagmemes such as location
or adverb which sometimes accompany ordinary verbs.
If this were the complete picture, one might feel that it was not worth-
while calling these morphemes verbs in any sense--but that it was preferable to
set them up as some kind of particle (even though in a neighboring dialect of
Vagaln the wa and 0 act as fully regular verbs the dynamic breakdown has not
gone as far 1n that dialect).
Our reasons for treating them as verbs: (1) In the same kind of 'auxiliary'
function there are three further stages of morphemes transitional between these
highly dependent morphemes and the regular independent verbs. It is convenient,
therefore, to treat wa and 0 as verbs--of a special auxiliary type--in the same
way as we want to treat the morphemes which are in a lesser stage of breakdown
in the same class, The morpheme kuari 'make, fix', for example, has a slightly
greater degree of freedom. It can, as a matter of fact, come alone as a verb or
as primary in a cluster. Note the following illustration:
U kugrd
'it is-fixed'

(2) In these circumstances it can have peripheral tagmemes, as can any regular
verb. When, however, kuari comes in the preposed auxiliary position where wa and
ga can occur, several changes occur. It takes on a special meaning of 'again'.
Note, for example:
U kuari la dig
he again went home
'He went home again'
 -74--

(3) It is no longer allowed to have a perfective ending which was possible for
6
it earlier (see illustration under [1]).
(4) In contrast to verbs after wa and 0, a verb following kuari will be treated
like the secondary verb of a cluster.
One could, presumably, refuse to identify the auxiliary kuari with the
separate verb kuari. If one adopted this course, then one would merely move the
homophonous auxiliary kilari into the class with wa and ga. We, however, much prefer
for purposes of this article, to treat the two kuari forms as still the same
morpheme. We are attempting to bring forth the evidence which shows that the system
is in a state of transition--and it is precisely in terms of residues such as this
one that incompleteness of change can be observed. The form kuari, in our view,
is in the process of becoming a particle. If all the members of the class were
like wa and ga, on the other hand, we would assume that this change of the class
from verb to particle were already complete.
A third stage of transition is even closer to the regular verb than is
kuari. In this sub-set, also, we have just one member, bir 'to turn'. With
this same meaning, the ,=r1) occurs as a regular member of a cluster--either
primary or secondary. When it is the primary verb, the secondary verb following
has the serial form.
bitO ba
he turned came
'He came baa'
u izo di bir u har
he got up and turned his back
In these regular positions of the cluster, as main primary or main secondary
verb, it (like other verbs) can have the normal sets of peripheral tagmemes.
Note, for example, the following:
bir leng

'He turned here'

bir manang
'He turned quickly'
When, however, bir occurs in the auxiliary tagmemic slot, then two restric-
tions remain: (1) No peripheral tagmemes are allowed, and (2) a special meaning
is giver to it--the meaning of 'again'--as, for example, in u bir ng6 He again
said'. When carrying this meaning, the word is not allowed to serve as an isolated
 -75-

verb--nor as a main verb in a cluster. The developing tagmemic slot has forced
on it a semantic change.
We come now to an even more difficult decision. Should weyr 'be able',
ktiti 'do purposely', ai.d futiri 'do unintentionally' be treated as auxiliaries
or as regular verbs? In favor of treating them as main verbs we note:
(1) They take the perfective endings:
.
weyto 16 ka-bila
he is-able pound fufu
'He is-able to pound fufu [mush]'
(2) The verb which follows them has the secondary form, not the primary form:
ktit6 bir
'He purposely turned'
(3) They can stand alone:
weyto
'He is-able'
On the other hand, we choose to treat these as auxiliariesor, more
accurately, as just moving into the auxiliary function--for the following reasons:
(1) If, in rare instances, these do stand alone, it is usually in a cultural
context in which a following main verb is understood, as for example in the fol-
lowing situation:
Question: i weyto ee gung '[Are] you able [to] do that?'
Response: ft weytii 'I am-able'

(2) When occurring with another verb following them (cien though the following
verb is in the secondary form as just indicated above) no peripheral tagmemes
are allowed to occur with them. (There can be no adverb, or locative, etc.
with them.) In this respect, therefore, they have already begun to receive the
specialized function which characterizes the restrictions on the preceding three
subsets.

In concluding this section, therefore, we ask: How does a new part of

speech develop? What kind of synchronic evidence would provide data which would
show such a part of speech in process of emerging? We suggest: (1) shorter
forms of items may develop with special restricted distributions. (2) Special
minor rules--as for tone--may be brought to bear on them. (3) Special meanings,
reflecting something about the positional usage may be added. Phonologically there
may be special fused forms, pro-clitic forms, tonal characteristics. Lexically,
the semantic component may be modified. Grammatically, restricted distribution
 -76-

is involved. The developing class--or developing part of speech if one wishes

to call it that--involves the entire system.
We turn now to a further set of morphemes where it would appear that this
development has run its full course. We refer specifically to a set of morphemes-
in distinction to the auxiliary verbs which we have mentioned--which we can call
adverbial particles: (1) These are allowed to have no peripheral tagmemes
attached to them. (2) They do not occur in isolation. (3) They do not have
inflection. (4) They cannot be head of a verb phrase. (5) They are involved
exclusively in modifying--in some sense--the verbal phrase which they introduce
(if they occur at all). (6) They are in the first position of the verbal phrase.
(7) There is no obvious relationship between these and other free verbs. Never-
theless, because of the pattern of verbal sequence which is involved in the verb
phrase, it would appear likely that these have a verbal source. Evidently here
they have been subject to special restrictions. The source-verbs from which these
may have developed are unknown or uncertain.
Having dealt with pre-verb auxiliaries end their intermediate status between
particles and full verbs, we can consider a similar approach to possible post-
%
auxiliaries. One candidate for such a position is the word hOnzi 'to fail'.
(1) This verb never occurs alone: it must be preceded by another. (2) It

shares any peripheral tagmemes with the preceding main verb (never has any which
refer only to it.) Example:
u kudr-O hunzi
he fix it fail
'He wasn't able to fix it'
Some other verbs, e.g., tOng 'finish', kiing 'tire', ping 'be satisfied' also
seem to share peripheral tagmemes with the preceding verb. Examples:
pato tung
he hoed finished
'He finished hoeing'
path kOng
he hoed tired
'He hoed until he was tired'
!

di ywo ping
'He ate until he was satisfied'
full

.0192--,I.
 -77

(2) The verbs ha and ta, both meaning 'throw' frequently have a slightly different
meaning when occurring with other verbs:
gil u ha
he left him throw
'He deserted him'
U le u ha
he got him throw
'He saved him'

nmeng U ha
he cut-off it throw
'He cut it off'
(3) When used as post-auxiliaries, these verbs may have peripheral tagmemes only
if these tagmemes apply both to the auxiliary and to the main verb:
hUnzi deelA
he fix-it fail yesterday
'He wasn't ablei[yesterday] to fix it yesterday'
So, also, tung 'to finish', kung 'to become tired', ping 'to be satisfied':
U pato tUng creel:a

he hoed finished yesterday

'He finished hoeing yesterday'
U pato kung deela
he hoed tired yesterday
'He got tired hoeing yesterday'
diywO ping &e'en
he ate satisfied yesterday
'He ate until he was satisfied yesterday'
 -78-

FOOTNOTES
1
Dependent clauses, whether initial ones which may optionally precede the
sentence nucleus, or final ones, have been but little studied for this report.
But see some earlier work: Structure of Subcrdinate Clause Groups in Kasem, a
paper presented by Kathleen Callow to the Fifth West African Languages Congress
(April, 1965).
2
I have changed the order in which he listed them to parallel more closely
that of Jacobs.
3
Kennedy suggests further restrictions of distribution in relation to
some special clause sequences, involving benefactive, desiderative, and comparative,
and following auxiliary. It is not clear to me from his data, however, how these
would be related to clause subclusters--see §2.2.5. A complete distributional
statement must await further research.
4
Kasem is the one exception listed in our materials, where a secondary
clause must begin with a pronominal subject. There the identification of a
clause sequence as a clause cluster is determined by other criteria, such as
relation to slots in the sentence ( §2.1) and agreement restrictions in the series
( §2.2.4).
5
Y. Callow states more generally, for Kasem, that in the secondary clause
no preverbal tagmeme ever occurs other than the [pronominal] subject--so that
in secondary clauses there is obligatory absence of the introductory and tem-
poral tagmemes, and of all preverbal particles, including negative ones. (Nega-

tion of secondary clauses must be accomplished by dropping the cluster form, adding
a conjunction, and producing a complex sentence.) Consequently, fifty percent of 13
their observed secondary clauses were minimal--pronoun plus verb (whereas many- Q E-

most- -West African languages would also drop the pronoun).

6
There is some possibility that the loss of this ending is historically
conditioned by its occurrence after a preceding particle ka which would suppress
the ending. This speculation we are unable to check at this point.

a
 CHAPTER BEYOND THE SENTENCE

3.1. Sentence Clusters in Paragraphs

In §2.2. extensive data were given to show that clauses entered clusters
and subclusters, forming units larger than the clause but smaller than the sen-
tence. Now I wish to show that there is some evidence that sentences sometimes
enter into formal units larger than the sentence but smaller than the narrative
or discourse. The evidence, still in a very tentative form, is again drawn
from the work of Crouch on Vagala.
What we wish to accomplish, therefore is to find: (1) Some kind of for-
mal marking of a primary independent sentence, versus a secondary sentence,
following this independent one. (2) We must then show that the second ele-
ment of this sequence is indeed a sentence, and not merely a clause. Otherwise,
the independent-dependent combination could be merely a clause cluster of the
type already discussed in previous sections. (3) We then wish to show that
following the independent one there can be more than one dependent sentence-
there can in fact be a sequence of dependent sentences, all of them marked as 1

other than independent. Finally (4) we wish to suggest that the relationships
between the independent and the dependent elements in these paragraphs may lead
to sets with semantic significance.
1
(1) There is a clear formal marking, in at least some of the primary in-
dependent sentences versus the secondary dependent sentences. They are char-
acterized by a contrastive complicated pattern of tone differences, occasionally
supplemented by special suffixes. In some instances, for example, the dependent
verb will be mono-syllabic high--as for she 'carried [on head }' -- whereas the in-
dependent form has the first syllable in low tone, and an added suffix g6 with
high tone. Note the following contrastive pair:
Independent: u kylgo nif ba
she carried water came
Dependent: u kyf nif ba
she carried water came
The independent sentence of the paragraph can often be thought of as a
variety of 'topic' sentence--or one with the verb marked for focus of attention.
For a contrastive pair in context, note the high tone of ya 5 'buy' in the first
sentence but also in the second sentence the low tone yail 'buy' (the translation
could be in past tense, by implication of the context, not by formal marking):
 -80-

h ng6 di u you hia yell -we

'He said that he buy yams market.'

u n la yetilwe u yaii hia dey
'He when go market he buy yams there'
i.e., 'He said "Go buy yams in the market" [And so] when he
went to the market he buy yams there'
The preceding illustration was chosen from Command-Response paragraph struc-
ture in order that the same verb would occur in the two parts of the paragraph
and therefore the focus and out-of-focus markings could be clearly seen. In

order, however, to see the focus versus out-of-focus differences where the verb
changes, we use a pair of paragraphs. In the first paragraph, the primary sen-
tence will have the verb in focus, but in the secondary sentence a verb out-of-
focus; in the second paragraph the first sentence will now have the latter in
focus and the former first verb will become the second, but out of focus. Note
the verb kyag 'wash' in focus in the first paragraph, versus nyer 'sew' out of
focus; and nyer in focus, but kyag out of focus, in the second sentence pair:
haang la kyag wegyezi u gyaari gey nyer tagta
woman the washed clothes her husband contrastive sewed shirt
But now note neyr in focus, with kyag out of focus:
bal la nyer tagta u haang gey kyag wegyezi
man the sewed shirt His wife contrastive washed clothes
Sometimes, as in a clause cluster, tenses are given only in the primary sen-
tence, and not in the secondary sentence 'elaborating' the first:
6. ? 6. I 6. A A I 6. 6. I 6. .
n nu n haang ne de ra naa n haang kyag
I and my wife emph yesterday imperfect fight, my wife insult
(tense)

A kyage 11341.

me insult one
'Yesterday I and my wife had a fight (quarrel). She insulted me
(with a certain insult)'
(2) We next wish to show that the structure which we have called para-
graph is not, in fact, merely a clause cluster of types such as we have had
earlier. One might have assumed that all such paragraphs could be defined as
simple clause clusters with the following rule: In the clause clusters here
called paragraphs, the second clause has a subject different from the first,
 -81-

and therefore the second subject is retained; but in the clause clusters pre-
viously defined ( §2.2) the second subject happens to be identical with the
first and is automatically deleted. This at first appears to be an attractive
rule, since relationships semantically between elements of a clause cluster
are often similar to the relationships between the clauses of the sentences
of a sentence cluster. Note the paralellism of structure in the following
two illustrations, in which the first has two sentences, but the second two
clauses of which the second has deleted subject:
dgglf dolzi di diy di ping
she cooked soup and ate and was satisfied
dagli do'I zi ba diy di ping
she cooked soup They ate and were satisfied.
How, we then ask ourselves, can one prove that the first set is indeed a se-
quence of sentences--rather than one of clauses?
It appeared to us that the crucial differential evidence to separate these
as belonging to different levels of the grammatical hierarchy would reside in
studying more complex sentences. We will recall that from the Callows' mate-
rial ( §2.1) a single complex sentence may have an optional dependent clause,
an obligatory independent clause, and a following optional dependent clause.
Furthermore, the Callows demonstrated that in each of these three positions
within a sentence there could be a clause cluster. Note, therefore, that
there is a crucial element in such a complex sentence--each component of the
sentence, each dependent and independent clause complex, can be either a single
clause or a clause cluster. It follows, therefore, that a complex sentence
with subordinate and independent clauses is not the same as a sequence of
clauses in a cluster (since it is essential that one be able to speak of sev-
eral clause clusters within a single sentence which is complex in this way).
If, therefore, one can show that, in the focus versus out-of-focus sequence of
independent versus dependent sentences the dependent sentence slot (or the in-
dependent slot) may be filled by a complex sentence which is itself composed of
dependent and independent elements which are clause clusters, it follows that
the sequence of independent dependent sentences must be on higher level than
that of the clause clusters included within one of the major parts of one of the
other of the sentences.
 -82-
D

Notice, therefore, that in the second part of an illustration given ear-

D
lier, there are dependent and independent parts to the sentence:
t ii lg yatilwg a you hig dgy

he when went market he bought yams there

'When he went to the market he bought yams there'
Here the first dependent clause is 'when he went to the market,' and the E
second is 'he bought yams there.' Note, furthermore, that each of these parts
can be expanded into a clause cluster:
a a lg yatilwg di ng hig a yad ghin6 ba a ni dia
he when went market and saw yams he bought some came them with house
a
Note that in this expansion of the sentence type there is a clause cluster
in the dependent part of the sentence ('went market and saw yams') and another
in the independent part ('bought some, came'). Thus it is clear that this ele-
ment as a whole, including the temporal dependent clause, is not a clause se-
quence but a genuine complex sentence, each part of which may, in turn, include
a clause cluster as already defined.
With this decision reached, it now must be clear that the full complex, in-
cluding the two sentences, must be on a higher level. Note, for example, the fol-
lowing expanded paragraph:
,! , ..!
u ng6 di d ya hig you wa ba a ni . a a la. yau wg
he said quote he buy yams market bring them with. he when went market
t

di na hig u you ghing ba a ni .di.

and saw yams he bought some came them with house

'He said he should buy yams at the market and bring them home.
When he went to the market and saw yams, he bought some and
brought them home'
In this expanded paragraph there are two sentences each of which includes
clusters of clauses. Nevertheless the sentence sequence does not include two
sentences which are on a par with each other, inasmuch as the first is marked
for independence and focus, and the second is marked by the respective tonal
elements as being dependent and out-of-focus.
(3) We now turn to the next element of the argument: In the dependent
slot of the paragraph, there can be more than one dependent sentence. There
B
is a parallelism between clause cluster and sentence cluster, on a higher level
of the hierarchy. To the preceding illustration, for example, the following
sentence may be added:
 -83--

[paragraph contined]...t h hey1 u dig di

D zini

he when reached his house and entered he

A P
kyi dungee di yir u biey
stood [in]courtyard and called his child
This third sentence is marked as being dependent. (Contrast it with the inde-
simple sentence:
u kyig6 di yir u biey
(he stood and called his child)
Note, furthermore, that in this third sentence of the sequence there is a
dependent and independent section, which, in accordance with the Callows'mate-
rial, demands that it be treated as a complex sentence rather than as a serial
clause; and note, finally, that in each part of this third complex sentence
there is indeed an included clause cluster which once more shows that the levels
are not the same.
(4) Just as one can have contrastive clauses (such as transitive'versus
intransitive versus equative) so one can have different kinds of paragraph
13 structures. The analysis of these is not yet complete. How many there will
ultimately prove to be, and the full analysis of the contrastive markings and
their justification in terms of such internal components plus contrastive dis-
tribution in still larger structures, is not yet clear--and a running text can-
not yet be cut into such paragraphs without residue or indeterminacy. Here we
merely wish to suggest the direction in which the research is moving, and re-
serve analysis for later study.
In general, we will label the independent element of the paragraph the
Proclamation, and the succeeding dependent sentence or sentences in series,
as Commentary:
Command plus carrying out of command: (see first paragraph illustration,
above.)

Request plus carrying out of a request:

n da
N A 'A -
nyiyngi 1 kuari n Ite n. u kuari
I imp. want you fix my house give me. He fixed
u
.A
1

te u

his house gave him

'I want you to fix my house for me. So he fixed it for him'
 -84-

Statement of intent plus carrying out of intent:

ti ng6 ti rang IlA ga peezi h mee. h h
'He said he was going to ask his father. So he
lenge h peezi
went and asked his father'
Principal action initiating sequence of actions plus subsequent action:
u kpa kpanf6 di la plyr. a h v614 gaing
he took axe and went, bush. He when walked long he
wa heyl da-zeyng lbel di kpe kpen'fO nmeng.
came reached tree-big one and took axe chopped,

h kyOwl
It fell,
Initiating action plus result of action:
u kpa daazi du nung nfyng lau daazi la

he took wood put in fire Fire caught wood the

di faiwn a maa tong
and burned them all finish
a
3.2. Discourse Structure
Just as a simple phrase like big John can be treated as a construction wave,
with John as nucleus and big as margin, so also a narrative may have nuclear sec-
tions and marginal ones. The use of such wave patterns is a universal.
When margin and nucleus are seen in discourse--whether via conversation,
narration, oration--one often finds, therefore, some kind of introductory sec-
tion, a basic body section--or a series of them--and some type of concluding
element. These may be considered as tagmemes of a discourse construction.
3.2.1. Narrative
Within the workshop Ron Rowland first developed this area for us, in Sisala.
In the process of exploring clause structure he noticed that various kinds of
time expressions came sometimes at the beginning of clauses, sometimes preceding
the predicate, and sometimes final--or at other places in the clause. When he
investigated, it became apparent that the different placements were not random,
but were in part conditioned by their place in discourse--and, reciprocally, in
part the placement could be viewed as a contrastive feature of the discourse
pattern.
 -85-

Various types of discourse differed from each other in ways barely investi-
gated. In narrative, for example, Rowland failed to find certain words of a type
common in conversation--words such as 'day before yesterday,' 'yesterday,' 'to-
day,' 'tomorrow,' 'day after tomorrow.' Compound time words and time phrases,
however, were found in narrative.
Within narrative structure a sequence of tagmemic slots seemed to occur:
± Preview ± Narrative Setting ± Sequential ± Focus change...±Climax
-Comment by Narrator Summary Application
In general, each slot in this structure has respective identifying formulas
II composed of the kinds of variants of transitive clauses occurring there.
Preliminary formula for transitive clause in slot for narrative setting:
ma

TP + S:NP ± Part + Pr 4- [?]0:NP - LP [?]AdvP

1 tr
Here the superscript f implies fixed order, in this discourse slot (the time
phrase in that slot of narrative setting comes specifically here); subscript
1 labels the permitted subclass of noun expressions allowed there.
The sequential slot has transitive allos in a contrasting order:
+ f +
+ S:NP + Pr - [?] O:NP - LP - [?] Adv P fPart/TP
tr 1/3

Here time phrases, or particles, are rare, and limited to clause final posi-
tion; time phrases exclude the types mentioned for conversational style, but
include those for narrative plus element T3 which is a particle fa meaning
'pre-present'--i.e., of historical relevance.
In the focus-change slot, only one sentence occurs at a time. Its struc-
ture is similar to that of narrative except that particles do not occur between
subject and predicate; and the clause may optionally end with wa/ya 'emphatic.'
The Climax slot has a formula like the setting slot, minus the initial
time tagmeme, but adding optional final emphatic particles.
The summary slot (and the initial preview slot) has the time particle fa --
and no other--occurring optionally between subject and predicate, or between
subject and particle + predicate.
Results from Vagala research--even in a most preliminary form--reinforce,
for West Africa, conviction that structures like that hinted at for Sisala may
be widely found. Special preliminary markers, in addition (like 'a long time
ago', for English) may open the story:
 -86-

h mdr td kdyng

my story emph. (like) this

'This is my story'
or I aim kiing la n bg ya zi ni bOwl
you know thing the rel-marker come salt with village
'Do you know what brought salt to the village?'
Or, in summary, close the story:
! 4..

nang dar u kp a ningzu ma di na u hing Bala bey

spider he took sense and saved himself elephant from
'Spider saved himself from Elephant by his cleverness'
Or, in a sign-off slot:
tono
I finish
3.2.2. Indirect versus Direct Discourse
In the previous section I showed that discourse could have a tagmemic
structure--a grammar above the sentence level. I now add discussion of a
problem in Bariba in which the cultural setting, the focus of attention,
and sequence of pronominal reference are heavily interwoven in the structure
of direct versus indirect discourse. The data are provided by Jean Soutar;
D
she also collaborated closnly in their presentation.
When, in Bariba, should one use direct discourse in a quotation and when
indirect? Constraints are imposed by the total discourse situation, by parts
of sentence, and by choice of quotation type.
Primary, Secondary, and Tertiary Statement Patterns
We need to be able to refer to statements which contain no quotation- -
2
i.e., nonquotation, or primary statements; to those with an included quo
tation- -which for convenience sake may be called a secondary statement;
and to those statements with a quote within a quote, which may be called
tertiary. For our present purposes, a primary statement is viewed as com-
posed of one part; the secondary of two; the tertiary of three.
The primary statement has a nonquotation filling its total area--e.g.,
'The man saw the boy', etc.
The secondary statement has an initial quotation indicator, containing
a verb of saying, reporting, or the like, and a second part, representing the
quotation reported: e.g., He said to the man that the boy would come. This

quotation is called a first degree quotation. The quotation indicator has

-
 -87-

two diagnostic pronominal (or nominal) parts whether explicit or implicit,

which identify the f2faker within the quotation indicator (e.g., 'He [said)')
and the audience within the indicator ('...to the boy').
The tertiary statement pattern begins with a quotation indicator, as does
the secondary statement. It then, however, follows the quotation indicator
with a second quotation indicator which serves simultaneously as a first degree
quotation.
Following the second quotation indicator, the tertiary statement then has,
in its third position, the second degree quotation:

(3) y; nag na n nun s3 Woru, siba nec nen dumA ba

3 4 5 6 7 8 9 10 11

koo ka yakuru ko (Words--minus some names-

12 13 14 15
are numbered to key into the translation.)
'So I came in order that I tell you Woru, they [those
1 2 3 4 6 5
people)7 said8 [to me] it is my9 horse10 theyil willi2 makel5
a sacrifice14 with13.'
This plan can now be diagrammed as in Figure I.

FIGURE I: STRUCTURE OF BARIBA STATEMENTS

SHOWING DEGREES OF QUOTATION
Non-Quotation Statements with un-
Primary restricted predicate
Statements and object
'The boy will come'

Quotation, Quotation First Degree

Secondary Indicator Quotation
Statements 'He said [to you] "I will come"'

Tertiary Quotation Second Quotation

Statements Indicator Indicator
as

First Degree Second Degree

Quotation Quotation
'You said [to me] "He said [to I would come"'
you]

iy
 -88--

We turn to the distribution of direct versus indirect speech in quotations,

predictable by various criteria.
Indirect Status of Second Degree Quotations
In the third slot, that of the second degree quotation of a tertiary state-
ment, all quotations are indirect.
(When the particle ma 'that' precedes a quote, we know--as in the English
translation--that the quotation is indirect. Unfortunately, for the ease of
our analysis, the ma is optional, leaving some ambiguity. Various items other
than ma occasionally give hints indicating directness or indirectness of dis-
course.)
Alternative Forms of First Degree Quotation
In the second slot--the quotation--of a secondary statement of Figure I,
direct discourse is sometimes used for the quotation, and sometimes indirect.
Off-Stage First Degree Quotations, as Indirect
In a narration, the setting of the scene before the principal action it-
self begins, may be called "off-stage". Quotations within such introductory
off-stage parts of a narration are in indirect discourse.
On-Stage Quotations with Ranking Matrix
The main part of a discourse, including the action parts of a long dis-
course, are "on-stage," in contrast to the introductory (or concluding) off-
stage elements. An on-stage quotation may be direct or indirect. Several
factors control the choice of the indirect or direct form of a first degree
on-stage quotation. One is its relation to the dramatis personae (DPs) of the
quotation indicator (Q. I.) preceding IF. In Matrix A, this factor is charted.
The character (first, second, third, or fourth [less important], whether sin-
gular or plural) chosen by the narrator to be speaker in the quotation indicator,
serves as marking one of the rows of the matrix. The character chosen by the
narrator as explicit or implicit audience to the subject of the quotation in-
dicator serves to label one of the columns. At the intersection of the rele-
vant row and column, the cell is filled either with the symbol I, meaning
'indirect discourse will be used for the first degree quotation, after the quo-
tation indicator'; or the cell is filled with D for direct discourse; with
I/D for sometimes one and sometimes the other; or with A when the forms are
expected to be permanently ambiguous; or with question mark when data are
lacking.
 -89-

MATRIX A: OCCURRENCE OF A FIRST DEGREE QUOTATION AS DIRECT

OR INDIRECT, IN ACCORDANCE WITH THE DRAMATIS
PERSONAE OF THE PRECEDING QUOTATION INDICATOR.
DPs of the Audience in Q. I.
Quotation
Indicator 2 1 3 4

Subject of Q.I.
2 ? I

1 A I/D I/D

3 I I/D 9 I/D
n 4 1 I/D ?

The matrix has several quite unexpected characteristics: Note, firgt

that the rows and columns have been specially changed (permuted) from the
routine order of 1, 2, 3, 4, to the order 2, 1, 3, 4. Only by this permuta-
tion is its special characteristic clearly seen--that whenever 2 (singular or
plural) is involved at all in the quotation indicator, whether as speaker or
audience, the succeeding quotation will be indirect. The presence (the in-
volvement) of 2 overrides the presence of any other person. I call this a
ranking'3 structure.
For an illustration of the 3-2 cell, note the following where the Q.I.

n is 'They would say [to you]' and the quotation is indirect in that cell of
Matrix A (i.e., M.A ):
32
b4r1 tii vet ba tunuma ba koo bikia amona mba ka mba
1 3 4 5 6 7' 8 9 10 11 112

kan man14 sikual5

'When2 they3 arrived4 they5 themselves]. would5 ask7 [you] how8 you12

buried me and with what '

15 14 13 11
On-Stage Quotations in Focus
Leaving now the quotations preceded by an indicator containing 2, we
study only those quotes varying from I to D, and where the indicator has
only 1,3, or 4.
The pattern found for general conversation, on-stage, in focus is re-
presented in Matrix B. Off-diagonal cells are all direct. Tha parentheses
indicate infrequent forms; it is rare to have the general conversation in
focus when the conversation is between the narrator (1) and people other than 2.
 -90-

MATRIX B: ON-STAGE GENERAL CONVERSATION IN FOCUS

DPs of the
Quotation
Audience in Q.I.
Indicator
1 3 4

Subject
1 A (D)

3 (D)

4 (D)

For an illustration note:

kpal Woru2 na3 u4 nee5, Sabi, sd6 h7 de8
yam9 m110 dp3511?
'Then Woru came and said5, "Sabi, aren't we ever going
1 2 3 7 6 8 11

to that place ?"' (Here the Q. I. is 'Woru said [to Sabi]';

10 9

the direct quotation is from cell M.334.)

It is unusual for one character to be in focus when it is neither the
narrator (1) nor audience (2). It is usual, however, when one character is
the narrator, for the narrator to be in focus (if 2 is not involved) and to
speak with direct quotes; and the other to speak to him with indirect quotes.
Focus is likely to be placed on a chief, if a chief and another person (3 and
4) are involved in conversation.
In the reported speech of one character to another in a given on-stage
section of a discourse, on the other hand, one of those quotations (one
utterance of that character) may be in focus, pinpointed within its Indirect
context, and made Direct, for example, if it represents a quotation which be-
came the source of the name of a person or place:
nee2, a3 doo4 a5 n6 dal daare8 m{9
Y6Y610 Yall ka12 Daaril3
m114... w13415 ban m517 Diiril8 yeruma
19

'They]. said2, "(You3) go4 and stay (drop off)8 there9." So itil

was called Danri13 there14....It-was-he15 they16 called17

Danri's oldest-son ' (Q.I. 'they said [to him]'; Quotation

18 19
3-19: Direct, quotation in focus)
fl -91-

On-Stage Quotation Out of Focus

Related criteria affect unfocussed quotations on-stage. When general
conversation is out-of-focus, for example, all quotations are indirect.
An incident from history, where the action is more important than the
conversation represents this kind of circumstance.
ye Gera koo tabu ko5, u Sabi Nena somo gorima wi koo
1 2 3 4 6 7 8 9' 10 11

tabu ko a wi s-etnu ke
12 13, 14 15 16 17

'When Gera was going to make war4, he sent a messenger

1 2 3 5 6 9 8
to Sabi Nena (saying) that he was going to make war12,
7 10 11 13
so would he (Sabi Nena) give him arrows ' (Q.I.: 'Gera
14 17 15 16
said to Sabi Nena'; Quotation: 10-16: Indirect, general con-
versation, out-of focus.)
Of several characters--but usually one of a pair--one may be out-of-focus.
The character out-of-focus is quoted indirectly--whereas the character in fo-
cus would be quoted directly. In a string of direct quotations, with the
string as a whole considered to be in focus, however, extra attention within
the larger attention span may be achieved for one direct quotation by making
the preceding quotation indirect. (Much as in an italic--emphasized--para-
graph, one word may be re-set in Roman type for higher emphasis.)
Although the speech of a chief is normally quoted directly, and the reply
to him directly, the following illustration reversed this to highlight the
source of the place name:
(a) u nee tamaa u koo ba gura a ka doona
1 2 4 5 6 7 8 9 10 11
(b) win tonu u nEE Domma a6 dab bu9 mo10?
1 y6ya
1 2 3 4' 5 7 9 10' 11
ba ka m5 More
12 13 14 15

(a) 'He said he thought he would gather them and

1 2 3 4 5 6 8 7 11
take them away.'
(b) 'One of hisi people2 said4, "When5 did you6 not7 haven

them9?" Soil they12 called14 (chat place) More15.'

(Q.I., Sentence 1.: 'He said (to his people)';

Quotation: 3-11: Indirect, on-stage, one quotation
out-of-focus, reversal. Q.I., Sentence 2: 'One of
his people said to him'; Quotation: 5-10, Direct,
one quotation in focus)
 -92-

FOOTNOTES
1
Others are ambiguous where the expected contrasts are neutralized.
2
Primary and secondary are used here in reference to relation to state-
ments. This use differs from that of Section 2.1, in reference to clauses
in a clause cluster.
3
First pointed out in matrix form in the submorphemic (or morphemic)
"formatives" of a complicated morphological structure. See K. L. Pike and
B. Erickson, 'Conflated Field Structures in Potawotomi and in Arabic, '

IJAL 30.201-12 (1964).

Note: Since this analysis was finished (covering text from seven
sources, 660 quotations) a second body of text, which included approximately
C3
1,000 quotations, was studied in the light of the hypotheses here. About
fifty quotations did not fit the rules. Ten of these were commands. Four
were a repeat of a quotation. Five instances of direct form occurred
unexpectedly in a second degree quotation; two of these are speeches making
up the important point to the story, so that emphasis or focus brought
directness into the second degree slot.
Other things being equal, quotations which are written, or are true, are
found in indirect form. Quotations which are oral, or fiction, are given in
direct form.
 CHAPTER IV: NOUNS AND NOUN PHRASES

We have already had occasion to list some kinds of noun phrases, as fillers
of the subject slot of clauses (for Dagaari, in §1.3.5) and of other slots (for
Vagala, in §1.2). We now wish to see more detail of this type, as well as the
structure of the nouns themselves.
4.1 Types of Noun Phrases
For Mbembe, Barnwell gives as the most common types of nominal phrases the
following (symbolized in terms of their manifesting classes rather than by their
comprising tagmemes):
+ noun - (+ demonstrative particle): eten ndo sa
meat that there
!
+ noun + possessive: aten CE

meat his
+ noun
+ (-
+ min:a + numeral): anbr) min:a afa

people about two

+ noun - qualifier + numeral: 31130 kpengokpn afa

person every two

Less common is a Mbembe time phrase (in appropriate clause slots) with a
time word as its head, with optional expansion by a relative clause:
awil 'tido c makwu ma
day that which they-came here
or a prepositional phrase, with preposition plus optional noun phrase or pronoun,

plus further optional noun phrase:

k ekwb:r 1 6s3:m

to tortoise to house
'to tortoise's house'
The most common Mbembe phrase composites are noun phrases in appositional
sequence (or pronoun in apposition to a phrase, or vice versa): or with a coor-
dinating link: or one subordinate (e.g., a relative phrase) to another.
In order to show in more detail the relation of noun phrase types to their
distribution in tagmemic slots, however, we return to the Vagala material of
Crouch. Classes of phrases are given here. Their distribution in clauseswith
some margin of error--was indicated in the chart in §1.2.

-93--
 --94-

NP a-- - Pre-Demonstra- + Head Numeral Post-Demons-- Qualifier

1
tive trative

pre-demonstra- noun numeral post-demons- qualifier

tive pronoun trative

zal la maa a bala maa

hen the all the elephants all
'the whole hen' 'all the elephants'

diini &Le la

houses two the

'the two houses'

as -- + Head Numeral Post-Demons- - Qualifier

Itrative

noun same same same

pro.
[baal la] b6w1 [kalangku la] b6r la
[man the] village [peanut the] place the
'[the man's] village' '[the peanut] place'

[ft] tagzi la maa [ft] haan-nananana bane

[my] medicine the all [his] wives-good two
'all of [my] medicine' '[his] two good wives'

[i] haana bane naa

[your] wives two these
'these two wives [of yours] '

b-- + Head - Demonstrative

adverbs same

gong la [hing ning ee]

like that [you will do]

'how [you will do]'

gang la [ban fi ning dad u]

how [elephant rel. will put him]

'what [elephant will do to him]'
 -95-

c-- - Predem. + Head Numeral - Post-Demons- Qualifier

same time noun same same same

A sal-1ga sanga la
the (that) time time the
1

a ha 'bgl a ha riga
a day one day that
'one day' 'that day'

d-- + Head

numeral

baling [dti 'dgy] kahtirbzi aniy-dane

some [are there] [his hats are] seven

[a ga kg] gne
[they went left] two
'there were two left/they left two'

NP a-- + Relative Head + Relator - Post-Demonstrative

2

NP I dep. cl. post-demonstrative

la

wif la A dti u zed

thing the rel. is its bottom
'the reason for it'

dig la n ky6w16 la
house the rel. fell the

'the house which fell'

+
b-- + Relative Head + Relator - Post-Demonstrative

NP dep. cl. post-demonstrative

lb

[yag ng] Ong la niighing bigy n &ice

[we saw] how dog child rel. is
'[We saw] what the dog's child is like'
f.
 -96-

c-- + Relative Head

+
+ Relator - Post-Demonstrative

NP dep. cl. 1 post-demonstrative

lc

a h61 dnglA il bg1e la

a day the other rel. came the
'the day the other one came'

NP a-- + Head + Relator

3

directional nouns
NP la

[u zuii] nii 14 bilyni

[he enter] water the inside
i
'[he went] into the water'

[bA wA 116371] kyo6wmg b6w1 'ug

[they came reach] rabbit village mouth
'[they reached] the entrance to rabbit's village'

b-- + Head + Relator

1
n. sub-class directional-ni
t

[grze wA iyzi] fAngg ni

[don't come get up] strength (with)
'[don't get up] vigorously'

c-- + Head + Relator

NP directional-ni
lc

[koowri nyAng lyz6] kyo6zi ni

[chief again got-up] morning (in)
'[the chief got up] in the morning [as usual]'

d-- + Head + Relator

adjectives directional-ni

RI dd] sum ni ne
[it is] sweet emph.
 -97-

NP a-- + Possessing Head...n* + Possessed Head

4

NP NP NP
la laa laa

kekiwri biey [sitiwo] hang ntig

chief child [died] his wife mouth
'the chief's child [died]' 'his wife's mouth'

b-- + Possessing** Head + Possessed Head

locative NP
laa

[ma mini] dey nmemee

[I-neg know] there road
don't know] the road of that area'

c-- + Possessing Head + Possessed Head

time noun NP
laa
time nouns (?)

kere kyciozi

tomorrow morning

d-- + Possessing Head + Possessed Head

adverb NP
laa

[galinggga wa] gang Mere [fewe] .

[crow didn't] such thoughts [think]

'crow hadn't thought like that'

*n here means repeatable without structural limit.

**The terms here are awkward. The noun-noun relation shows parallelisms
leading to their being grouped here.
 -98-

NP a-- + Coordinate Head + Conn. + Coordinate Head - Conn.... - n

5

NP conn. NP conn. NP
la la la

NP NP NP
4a 4a 4a

NP3a

h niy h hgang nly blyzi [ba diy'w6]

he and his wife and children they ate (something)

dngla h ng 'd di fgyli ng6 146 ba sgni sidwo nly

the-one rel. saw her and first said that their lover died and

dngla h kpg 'bg zan ba la, nly dngxa h kpa

the-one rel. took them flew come dem. and the-one rel. took

dlyfig di 1g1(1) .6 la...

tail and raised her dem.

`the one who saw her (in his mirror) and first said that their
lover died, the one who brought them flying (with his wings)
or the one who had raised her from the dead by striking her
with his tail [which one should have her;` B

+
b-- + Coordinate Head + Conn. + Coordinate Head - Conn.... n

numeral conn. numeral conn. numeral

11
NP NP NP
ld ld ld

baniy di ban&
five and two
'seven people'
+
c-- + Coordinate Head + Conn. + Coordinate Head Conn.... n

temporal conn. temporal conn. temporal

NP NP
lc lc lc.

NP NP NP
2c 2c 2c

NP NP NP
3c 3c 3c

zinga nfy Ikgrg

today and tomorrow

rilMormir, 7-

-_
 -99-

CI

d-- + Coordinate Head + Conn. + Coordinate Head Conn.... n

locative t conn. locative conn. locative

NP NP NP
4b 4b 4b

NP + List Head + List Head + List Head... n

6

NP NP NP
laa laa laa laa

[deyni kpgwo ra
l la, di] wokozi nly kpownazi, vibazi...
then he took cont. go, and yams and yams yams...

'He went and came upon yams [type one], and yams [type two], and
yams [type three]...' (Here niy is optional, and no,t indicated in

the above formula.)

Bariba, which differed substantially in clause structure.from these other

Niger-Congo language., has a quite similar pattern of noun phrase in subject slot:
Possessive Noun Adjective Numeral Demonstrative

an naa baka tia 376

my cow big one that

'that one big cow of mine'

siin56 dli baka tee t6

chief house big one that

'that one big house of the chief's'

(Compare, above, Vagala type NPlaa.)

4.2. Development of Compound Nouns

We saw in §2.2.5 that subclusters of clauses could develop from clause
sequences. Special distributional restrictions, coupled with phonological and
semantic specialization led to new complexes in a synchronic view of the dynamic

system. To noun phrases a related dynamic process applies, leading to 'subclusters

of nouns'- -i.e., noun compounds. As was true for the clauses, these cannot easily
be treated as a single, uniform structure, since the fusion is not even at every
point in the system.
In order to show how this material functions, we move to Igede. The remainder

of this section is quoted directly from the report of Mrs. Nancy Bergman, A Note
3

on Degrees of Noun-Plus-Noun Fusion in Igede.

 -100-

The Problem:
In Igede, certain nouns in sequence were seen to be fusing, with the fused
form sometimes having a specialized meaning. Were these fused forms simply fol-
lowing fusion rules which were normal to syntactic phrases, or were they a more
closely bound morphological unit? A practical problem was also involved: How
should these noun complexes be written--as two words or one?
Ube + utoji > iiblitTijI 'clinic'

room medicine
(Notice that the final vowel and tone of the first noun, Ube 'room',
completely elides.)
The Aim:
In view of the setting up of a new orthography, the aim, necessarily, was to
determine the difference (if any) between cwo nouns which are simply following
normal fusion rules and two nouns which are more closely bound together as a
single unit.
The Approach and Research:
The first step was to locate the different slots in which each noun and
each fused form could occur. In other words, in what slots can Ube 'room', titT5g.
'medicine', and Ubtitoji 'clinic' be distributed? Are they found in the same slots

or in different ones? This was done to determine the restrictions, if any, on

the occurrences of the two forms--the single noun and the fused form. I tested
each and found that there was no difference in the slots in which they occurred.
Ube nyaI date qula Ube
room our different he to-repair room
'our rooms are different' 'he repaired the room'
8
utoji hligu alubek66 lg utoji
medicine good white-man to-have medicine
'the medicine is good' 'white men have medicine'

IgElaa hti -nyT kA

clinic be-there in-that-place to-take child to-go clinic
'the clinic is over there' 'take the child to the clinic'
 -101-

The following is a list of attested forms:

Igede Noun + Noun Forms
Noun Noun Fused Form
1 2

Ube 'room' titEsji ' medicine' Ublitaii 'clinic'

T.
'hunger' enyl 'water' lmanyi 'thirst'

Uhye 'above, on Mpa 'bamboo rack' UhyMpa 'rack for

top of for storage) storage'
Ube 'room' TTO 'writing' Ubl7p6 'school'

We 'god' alubye 'it up above' oholuhye 'God'

abw6 'hands' ela 'matter' abw-ild 'character'
apwa 'house' 6.01 'ground' Tpweji 'village'

Zgb-5 'train' aluhye 'it up above' Tigbalaye 'airplane'

*env 'afternoon sun' + Gturu 'morning' Tntlf5r17 'day'

IY-6 'meat' -enyi 'water' IyanyI 'fish'

ihu 'year' omd ? Ihwomil 'last year'

MIT 'year' enyI ? IhwgnyI 'this year'

*(anU, vturu, and -e-ntit-Jril are all time words so they have certain restrictions

in the slots wherein they may occur)

Once again, notice that in noun, the final vowel and tone are dropped and are
completely overridden by the initial vowel and tone of the second noun.
The next step proved to be very interesting and quite helpful. A test of
internal expansion was tried for each of the fused forms. In a noun + noun re-
lationship, only nyd 'for, of (possessed by)' can occur between two nouns.
Not only do Igede speakers say:
fl
UbutOji
room-medicine
'clinic'

but they also say:

Ube ny(d) litaff

room for,of medicine

'clinic'

withOut changing the insaning.

In applying this test to the other fused forms, it was discovered that some
B could be expanded (by ny&) and some could not. Table I sows this. (The x in-

dicates tested permitted occurrence.)

B
 -102-

TABLE I: Expansion and Non-Expansion of Igede Fused Forms

Fused Form nya 'for, of' -nya 'for, of

n
Li
iibutOji 'clinic' x

imenyi 'thirst x

iihynpa 'rack for x

storage'

iibUpT) 'school' x
611-61uhye 'God'

'character'
Tpweji 'village'

TigbOliihye 'airplane'

Entit17)rT) 'day'

i yenyi 'fish'

ihwomu 'last year'

ihwgnyi 'this year'

Now we ask: Why this difference? What is the controlling factor here?
It seemed apparent that there were differences in degrees of fusion in the
Igede noun + noun relationship. Following this hypothesis, four degrees of
noun fusion were set up: (a) obligatory absence of fusion, (b) optional fusion,
(c) obligatory fusion]. (with identifiable morphemes), and (d) obligatory fusion
2
a
(some parts unidentifiable).
(a) Obligatory absence of fusion. The loosest type of noun + noun rela-
tionship in Igede is found with certain nouns which cannot fuse and which have
the obligatory particle nya occurring, which usually denotes a genitive re-
lationship.
onyI nya Job ry oje ny(a) ;1e6
child of Job bicycle of man
'Job's child' 'the man's bicycle'
(b) Optional fusion. This type of noun + noun relationship is made up of
a set of nouns which fusr but which also can be expanded without changing the
meaning.
imenyl r"\J Imi ny(a) enyi
hunger-water hunger for water
'thirst' 'thirst'

Lam_ T
 -103-

Ub5p5 N Ube ny(&) tip6

room-writing room for writing

'school' 'school'

In fact, 'room for writing' occurs more frequently than 'room-writing', which
may be an indication that the form with nyd used to be the prevalent form,
but through the years fusion has become more dominant.
(c) Obligatory fusion,. In this type of noun + noun relationship there
is a very strong degree of fusion; obligatorily so, in fact. These fused
forms cannot be expanded, but the morphemes constituting the forms are still
recognizable and definable by informants (see also, the preceding list for
the full forms of the fusing nouns).

0 awT1A TPweji
hands-matter house-ground
'character' 'village'
There is wider divergence from the meaning of the source morphemes when there
is a tighter degree of fusion. That is, when two nouns fuse, the meaning of
the whole is often quite different from the meaning of the nouns as separate
forms.

It was suggested that the vowels of these fused forms be considered in

hope that this would give a clue to a controlling factor in optional fusion
and obligatory fusion. This was done and the results are in Matrix 7.
13
MATRIX I: Relationship of Vowels to Optional Fusion and
Obligatory Fusion
Final Vowel Initial Vowel
+
of Noun of Noun - nya - fly&
1 2
13
-e u-
- 1 e- x
- e

- e

- e o- 40. x

-5 E_

-a 6-

-0 o-
-u 13-

-o e-

-u 5-
- u g-
 -104-

Of the nouns considered in this Matrix, optional fusion occurs only when
the final vowel of the first noun is a front vowel. This occurs, as is seen
in the upper section of the matrix, in the fused forms, which can be expanded.
Obligatory fusion occurs when the final vowel of the first noun is a
central or back vowel knotice the one exception, which is front vowel plus
low, back vowel; it is shown between the two main sections of the matrix).
Also notice that these forms are the ones which can nct be expanded.
(d) Obligatory fusion2. Finally, the tightest degree of fusion in the
Igede noun + noun relationship is considered, where:
(i) there is no separate occurrence of the morphemes, and
(ii) the semantics of some morphemes cannot be determined by most
informants (but source in noun + noun pattern is assumed).
0,

ihwomu ihwcnyI
year-? year-?
'last year' 'this year'

The informant tried to attach a meaning to cnyI and omu but he was incon-
sistent, particularly with omu; and since we have not gotten any meanings from
previous informants, my conclusion is that no one is really sure of the mean-
ings any more. This could also be an indication that change is taking place
in Igede towards not only a dominance of fusion but also towards obligatory
fusion. In the first three types (obligatory absence of fusion, optional fu-
sion, and obligatory fusioni), the morphemes are still recognizable and de-
finable--but in the last type this becomes extremely difficult.
Conclusion.
My conclusion and solution to the problem, therefore, is:
(1) Types (a) and (b) are phrases and should be written as sequences of
separate words.
(2) Types (c) and (d) are compounds and should be written as single words.
4.3. Syntactic Markers Developed from Nouns
Just as independent verbs sometimes developed into restricted auxiliaries
or particles ( §2.2.6), furthermore, so regular freely-distributed nouns may
develop into particles or into special word classes marking other syntactic
functions. For illustrating the synchronic discovery and functional descrip-
tion of this dynamic state, we tux_ back to Vagala, with data from Crouch:

4
D -105-

Li
In Vagala a large list of nouns may occur with or without accompanying
modifiers or other compounds of noun phrases. Note, for example, dla 'house'
in the phrases:
dia la din{ ane la
house that house two those
'that house' 'those two houses'
...- .1
dia la u n ma'we la
house that he which built that
'the house which he built'
A locative element may optionally follow a noun phrase with locative-noun
head (e.g., dia 'house') in locative slot; but is required in lc-ative phrases
with nonlocative head (e.g., maa 'mother'). Compare the phrase after transi-
tive verb na 'see', il na il maa /a 'he saw his mother that', with a locative
phrase after a verb of motion la 'go: il la il mag la begy 'he went [to] his
1
mother's vicinity'. Note that the marking locative can also follow a fuller
noun phrase: i

1 %
n la dini Ana la seey
I went houses two those front
'I went to the front of those two houses' i

We are assuming that the locative. marker is in immediate constituent relation 1

to the rest of the phrase. I

Whereas the demonstrative and numerals are con-
I
sidered to be modifiers of the noun phrase head, the locative is considered to
be in exocentric relation to the phrase as a whole, even though it is optional
as seen by the preceding footnote.
Three observations must now be made about the list of locatives:
(a) In general this list is made up of body parts: niia 'mouth', harbO6
'back', sgey 'front', bilyni 'inside', zee 'bottom', bang 'middle'.
(b) The locative marker picks up its special meaning from its tagmemic
2
function.
(c) In addition, however, there are two or more morphemes in the class,
which have locative meanings, and serve as locative markers in the same con-
struction but which do not occur as free nouns elsewhere. Unlike 'mouth',
'back', etc., which can occur as subjects or objects in ordinary clauses, these
.4 .... . ...,
other morphemes are fount here alone. Thus one can say that n nua ra wil

.
 -106-

'my mouth hurts' or h 16 h nild 'it hit my mouth', buc one cannot use in these
positions either ni 'on, vicinity', or by 'to, presence'. Nevertheless one
finds n la diva 1a by 'I went house that presence (I went to the house)' with
bey, and u sowgo ka ra la ni 'he sat chair that vicinity (he sat on the chair)'
with ni. We conclude, then, that there is a specialized subset of nouns, iden-
tified by the fact that most of them occur as nouns in subject and object slot.
But they are a specialized subset in that they have a semantic central meaning,
in general (body part) with a tagmemic overlay of meaning (direction). As
members of this class, however, there are two elements to be analyzed as nouns,
on the basis of analogy with the other members of the class, even though they
occur here alone. ci
The resulting assumption is that these two morphemes by and
ni should be treated as nouns with restricted distribution; and that one should
be able to find them as regular nouns in some related dialzts.
4.4. Irregularities Within Noun Classes as Distortion of Field
Turning now to the structure of a nominal system, we find in a large num-
ber of Niger-Congo languages a difference in stems according to vowel harmony-
3
in which the vowels are paired into 'upper' (close) or 'lower' (open) sets.
The upper set, under certain circumstances, determines the selection of a mem-
ber of the comparable set in an accompanying word or affix.
The relations between the variant forms may be expressed by morphophonemic
rules or by prosodic formulas of a Firthian type, or by rules.
4
For this project, however, John Callow (who had previously explored three
alternatives) tested a presentation of Kasem data via matrix display. A small
sample of bis results are given here. (For further material, see Appendix.)
The first sample displays a simple relation between a chart of patterns
of Kasem noun stems, chosen because they comprise a class ending in d, but
arranged in two halves. The upper half of the chart includes stems which con-
tain upper vowels, placed in the order the vowels would come on a vowel chart.
The pattern of arrangement of the lower half of the chart as a whole is re-
peated, but with stems containing lower members of the paired vowels. The stem-
initial consonant is irrelevant to the structural relation, and is represented
by C. To the right is the word form, composed of the respective stem shape plus
the 'singular' morpheme, with allomorph /a/ after upper-harmony stems, but the
allomorph /a/ after lower harmonies.

'pee?.
77,
 -107-

U
KASEM MATRIX I: Nominal-Group C, with d-final Stems

d-final stems d-final stems, Sg. form

Upper
Harmony Cid- Cud- Cide Cude

0 Cad- Cod- 0 Cede Coda

+ cen.V
'singular'
Lower CLd Ctda Coda
Cod-
Harmony 0 Cada Coda
0 Cad- Cod-

The 0 signifies systemic non-occurrence, i.e., /e/ and /e/ cannot occur
in stems of this general shape.
No startling results come from one simple chart of this kind. But it lays the
groundwork for compacting rules via matrix patterns. If, for example, the
upper left were to be written as Nu-d (noun of upper harmony, final consonant)
u u
versus NQ -d, the rule N -d + {e}N -d+a NZIld + a points toward a device in

which units become symbols of matrices--with a potential compacting value.

Secondly, when the pattern does get complicated, and crisscrossing, a geo-
metric display via matrices allows (for analyst and reader both) much faster
and easier insight into the pattern. Note, for example, Callow's matrix Pule
for the plural. Here, however, the vowel-harmony character of the preceding
matrix is assumed; capital letters of stem and affix represent either upper or
lower harmonics, as the stem may dictate.

KASEM MATRIX II: Plural, Nominal Group C, Consonant-final Stems

CId- CAd- COd- CUd- CIdI CAdI COdI CUdI

CI1- CA1- C01- CUl- CI1I CA1I COlI CUll
CIn- CAn- COn- CUn- +fr. V CInI CAnI COnI CUnI
CI0- CA0- C00- CUo-
Cl/I CE CwE i CwI /I
CIg- CAg- COg- CUg-
CI CE CwE ' CwI
i

6 5 6

By use of the braces, pairs of sub-groups can be shown: (1) the set of stems
ending in alveolar consonants; these have no variants in the plural; versus (2)
sterns ending in velar consonant; the consonant is lost in plural. (3) Stems
 -108--

ending (a) in velar consonants and (b) containing rounded vowels; the vowel is
replaced by hal, versus (4) stems ending in front or central vowel; the vowel
is lost. (5) Stems (a) ending in velar consonants, (b) containing low vowels
a, o (central or back); the plural allomorph is the low front vowel (in open
or close harmony); versus (6) stems (a) ending in velar consonant but (b) with
f
high vowel; the plural allomorph is the high front vowel. (7) Stems (a) ending
in velar nasal consonant; (b) included in previously-mentioned set (6), (i.e.,
i

with high vowel); the suffix vowel alternates freely from oral to nasal quality;
versus (8) stems (a) ending in velar stop, and. (b) included in (5) and (6)
above; no further criterion; hence those paired members of (7) and (8) are am-
biguous except when (in 7) the freely-variant nasal vowel differentiates them.
Some Kasem examples:
bide 'wall' bidi 'walls'
jioa
u
'hand, arm' jLijr 'hands, arms'
digs 'room' di 'rooms' 1

bekede 'boy' bekedi 'boys'

beoe 'beam' be 'beams'
yaga 'market' YE 'markets' D
kode 'voice' kodi 'voices'
cone 'path' cwe 'paths'
koga 'back' kwe 'backs'
tule 'granary' tuli 'granaries'
zooa 'calabash' zwL/r 'calabashes'
The total structure of such a system of relations may be considered as a
field, or subfield. Regular groupings lead to a regular field structure. Ir-
regularities, such a:s these of Kasem, may be viewed as distortions of the regular
field. A complicated matrix structuring is seen in the verbs--for which see the
Appendix.
4.5. Noun Concord
In the preceding section we saw that noun classes differ in some Niger-Congo
languages, according to their forms in singular versus plural, determined by the
Q
shape of stem, shape of the singular versus plural suffixes, and morphophonemic
fusion rules within the word.
D

B
 -109-

4.5.1. Ranking in Singular-Plural Prefix Pairs

When, now, we turn to the Etung, of the Bantoid5 group of West Africa, we
find not only singular-plural stem and affix differences, but two further exten-
sive phenomena. (1) The concord-agreement between the noun and other words in
the immediate noun phrase, and between the noun and some other tagmemes in the
clause; (2) a system--well-known for Bantu--in which (a) some particular prefix--
1 6,
or formative --(e.g., /en, used for singular, forms a pair with a correspond-
ing prefix (e.g., /bi/) used for plural with that same stem; whereas (b) on a
different stem, the same prefix used earlier for plural may now appear as marker
for the singular, now paired with a different prefix to mark the plural of this
second stem; etc. (c) The morphophonemic changes or suppletive replacements
(or correlative formatives of a concord set) undergone by any one prefix through-
out the concord patterns is the same whether it applies to its use as singular
or its use as plural, or whether paired with one or more shapes for different
stems respectively.
The following analysis and presentation of the Etung concord is prepared
in collaboration with Mrs. Eileen Edmonson. (For her extensive list of noun
stems according to overlapping singular-plural pairings of prefix shapes, see
Appendix.)
A striking insight of Edmonson added a further characteristic to the list
of those already mentioned: (d) There are patterned constraints (a ranking)
on the permitted pairings of prefix shapes: Some prefixes are allowed to occur
much more freely than others; and the occurrence-possibility forms a 'ranking'
series such that--to some degree--if the first prefix pairs with the second (in
singular to plural relation), or with the the third, or fourth, then the second
may be expected to pair with the third and fourth, while the third may be ex-
pected to pair with the fourth. The system is not as clean-cut and regular as
such a model implies, however, so the data are presented in Etung Figure 1.
Arrows connect the prefixes--with the arrow head indicating the plural form
of the pair. Arrows above the line refer to countable nouns, which may be found
with numerals in the plural. The dotted shaft shows movement form right to left
in the chart--with only two pairs, both involving o. An arrow completely over
a prefix indicates that certain stems use the same prefix in singular and in
plural. Arrows beneath the prefixes have nonmodified plurals as indicated--but
when accompanied by numerals, the plural words use the prefix which otherwise
 1E1E11 OIOM 1E1E11 MEI OMNI. 11=1 111111111 11=1 IMOD - MIMI IMO 11111111
$

)
e N bi a Iba

Etung Figure 1

CZ3 C=3 co
ilrir"mr.tut MrtmgrA, loormropprommon
would be singular. Certain mass nouns do not occur in numerical plural contexts;
the contrast between singular and plural is neutralized; a prefix which occurs
with a set of these nouns is surrounded by a box in the figure.
A crucial question: What controls the ranking order--and the gaps? The
vowels from left to right in Figure 1 (from A to J) divide roughly into two
halves--the vowels /i/ and /e/ which are obviously front versus the vowels /o/
and /a/ which, in contrast to /i/ and /e/, may be called back. The ranking move-
ment from singular to plural is in general from front toward a position farther
back. There are some departures from this condition, as /e/ outranking /i/, and
as /o/ is involved in the only instance of a back vowel singular paired with a
front vowel plural. The pair {0} and /ba/ are outside the basic system, used only
with certain loans.

A second problem involves the identification of the vowel coloring of the

three homophonous nasal prefixes. In this instance, the vowel color is found
in other elements of the concord series. In Etung Table 1, rows are labelled
from A to J (from Figure 1), but the order has been changed to bring closer
together concording sets which share various tonal, vocalic, and consonantal
characteristics.

ETUNG: TABLE I
Concording I-have
Noun his that one which it-fell seen it
Sentence
'X'
Noun Class
,
/-..A - A..% . ;
D 'wife' fikge owe ano yot nno a-gboe n-yen 37&
v ,-1 .!, . ,-,
B 'box' akOp eye ani ylt nni a-gb.oe n-yen nni
H 'bucket pokit eye ani y ±t nni a-gboe n-yen nni
. .., I. . . -.,
A 'yam' eyu eje
''^ anji jlt nji e-gbO e n-yen nji
.01. .
C 'stick' biti ebe gmbi blt mbi bl-gbo e n-yen .nbi
.3, .0!
E 'broom' 8fak 8b& ambi blt mbi o-gbo e n-yen mbi
..v . i . . .
F 'pestle' ncok ame ama mat mma rii-gb(Se n-yen mma
. ..
G 'needle' aga ame ama mat mma g-gbOe n -yen mma
... . .. . I .. .. ...

J 'buckets' bapokit ame ama --- mma a-gbo.e n-yen mma

Note, in the first row (D) that the vowel /o/ occurs in the second, third, fourth,
and fifth columns--and implies the /o/ color to the simple nasal AI/ in the first
column. Row F, on the other band, has /a/ in the comparable places of the same
columns, and in Column 8 as well, giving /a/ color to prefix /ri/ of F. In Row B,
/i/ occurs in Columns 3, 5, 8 (with front /e/ in 2.)

4.
 -112-

Other phonetic elements--or the front-back contrast recurring--enter deeply

into the structure here, or in related data. Discussion of them must wait for

a fuller paper.
This much, however, has alerted us to watch for source of these phenomena,
or reflexes of them, as we study related languages.

Q.

D.

I
 -113-

Footnotes

1
Unless one knows whether a verb is trasitive or intransitive, certain
ambiguities may develop. Certain verbs of motion may be followed by an unmarked
locative, as u lA di 'he went house'. The subset of unmarked locative phrases
[1
does not concern us further here.
2
A very similar typological situation has long been known elsewhere.
rote, from Mixtec, a similar set ('Analysis of a Mixteco Text,' International
Journal of American Linguistics', 1944, 10, 113-138). As in Vagala, the Mixtec
[1 data includes a couple of items ('road' for 'direction toward' which do not
fit the body part class.
3
See §7.7 for discussion of phonetic analysis of this difference.
4
In 117.asem Yominals, A Study in Analysis,' The Journal of West African
fl Languages, 1963, 2, 29-35.
Ci
5
See Greenberg's Class IA5d, as in Introduction, above. David W. Crabb,
in Ekoid Bantu Languages of Ogoja, Part I (Cambridge, 1965), asserts with
more detailed data that the Etoid group is part of the Bantu family.
6
I use the term formative when I wish to emphasize that there is identity
of phonemic shape, but when I do not wish to commit myself to morphemic identity
in the - lassical sense of morpheme as a relatively stable form-meaning composite.
Thus, the same formative can be either singular or plural.
7
Note that the phonologically-alternate manifestations of the formative
must be treated in the concord series, whether or not the formative is considered
as the same morpheme both as singular and as plural.
0
 CHAPTER V: VERBS
A crucial problem in describing verb structures of some of the Niger-
Congo languages of West Africa lies in the control which the prefix complex
exerts over the shape of the verb as a whole, combining with an intricacy of
fusion of elements within the prefix itself. Into a monosyllabic prefix are
jammed clues to a variety of tenses (or aspects), modes, persons, and contrasts

for positive versus negative. I shall choose two sets of data from the work-
shop in order to highlight, in turn, problems of vowel fusion (in Abua), and
tone spans (in Etung). Then I shall add discussion of some comparable problems
of a Semitic language (Hausa) of this general geographical area.

5.1. Vowel Fusion in Matrix Display

Ian and Amelia Gardner have provided us with data on the structure of

some fifteen prefix forms of Abua. Matrix I lines up the data in order to
bring together into single blocks, insofar as is feasible, like vowels of the
respective prefix complexes.
Note that the vowels, in contrastive columns, differentiate fl
B
MATRIX I: ABUA VERB PREFIXES
(when no independent pronominal object
follows the verb)

Subject Pronoun -- ,T 2 P1 QP13 Sg Sg 2 Sg 1 P1

Nonhuman
Qual- 1)14 yoor yira
rung idi idi na mi
Tense ity4,

B Future e e e/a a a a 0
C Future e e e/a a a a

D Present e e e/a a a a o 0 0 t-
F Present e e e/a a a a O 0 0 r
G Completive e e e/a a a a 0 0
H Perfect E E E/A A A A O 0 0

J Impezative E E/A A u- 0 0

K Participial r. E E E/O O 0 0 O 0 0

L Stative E E E/O 0 0 O 0 0 0

M Stat./Perfect k E E/O O 0 O 0 0 0:

N Stative /past i i/0 0% otf

P Conditional + m i i/ui U u u u u u

Q Conditional m i i i/u: u u u u u u

R Imperative k e i ,i/u u a I u u u u

S Past -4- I I I/A A U li U U

-114-
 -115-

some of the persons: e characterizes second plural and third plural nonhuman,
in "tenses" (rows) B-M; the vowel i in tenses N-S (but with i, e, reversed

in second plural J and R). The vowel a is found in tenses B-H, versus o in
K-M, u in P-0 (and mixture in J, N, R, S) for third singular human, and for
first and second singular. Similarly, o and u occur with 3 plural human,
and first plural inclusive and exclusive:, whereas third singular nonhuman
has alternate forms for e/a, e/o, i/u, and i/a.
Several other elements must be understood on the matrix before illustra-
tions can be matched against it: 'Quality' refers to positive (+) (i.e.,
affirmative) and negative (-); this affects the consonant and tone frequently,
but the vowels (as in J, R) less often. The consonants of the prefix are
often contrastive for tense (e.g., B versus D). The vowels sometimes differ-
1
entiate "tense" (e.g., N-S versus B-M), as well as simultaneously differenti-
ating person (e.g., the vowel i signals rows N-S, and columns 2p1 and 3p1 non-
human). All vowel letters are morphophonemic symbols, representing vowel har-
mony with vowels of the verb stem. Vowels in capital letters are predictive
morphophonemic symbols indicating that the following consonant of the stems
of Class 2 verbs--but not the vowel itself--will be changed. In a Class 2 (but
not Class 1) verb, the stem-consonant will be replaced by w after I, U, but by
zero and vowel length after E, A, 0; p replaced by t by r, k by g,
We leave for the moment the tone contribution to the verb so as to continue
studying the vowel patterning. Abua Matrix 2 shows what happens to any one
vowel quality of Matrix 1 when the verb is accompanied by an independent object
pronoun. (In positive sentences, the pronoun object usually follows the verb;
in negative sentences, it usually precedes it.) Vowel harmony continues to
operate between verb prefix and stem. Capital letters, in Matrix 2, again pre-
dict consonantal morp!ophonemics.
There is an astonishing degree of regularity in the fusion of the old in-
cluded object element to these non-object vowel forms. There are few exceptions
to the implications of Matrix 2:

DP
 -116-

MATRIX 2: VOWELS OF ABUA VERB PREFIXES

(when an independent pronominal object
accompanies the verb)

Object Pronoun 2 pl 1 pl 1 pl 1 sg 2 sg 3 sg 3 pl

Verb prefix vowel Excl. Incl.

from Matrix 1,11 inina iyoor iyirt Vim` nina ngd 14di

I I I I r E

u I U U U u u u

0 I U U U 0 0 0

a U U U A a a

Note the i- prefix in the independent object pronouns 2p1, 1pl incl, 1pl
excl, at the top of Matrix 2: 1-fan .a '2 pl', etc. Only when the independent
DI
object pronoun has this high front prefix does change take place in the fusion
matrix: no change occurs with 2 sg, 3 sg, 3 pl. A similar *i- element seems
to have occurred within the verb itself. There i + i > i; i + e>i; and i +
u,o,a > u except in 2 pl when i + u,o,a > i. The *i-, that is, raises front
and back vowels, and also fronts the central and back vowels except in 2 pl.
This in turn leads in Matrix 2 to the display which shows i outranking
u, and u outranking o and a, resulting in the characteristic L shapes. Re-
turning to Matrix 1, an included ranking structure is seen in the relation of
o to a in Rows B-M of the columns to the right. The source of such ranking
structures, in instances of this kind, seems to be priorities in fusion. The

result is sometimes ambiguity. Ambiguity is not as great as appears from

Matrix 2 alone, however, since the consonant differences of Matrix 1 continue
in force, and the tone differences also contribute to the contrasts.

5.2. Tone Pattern in Verbs

The tone of the Abua verb as a whole is determined by (1) the inherent
tone of the last syllable of verb stems, (2) by the le.gth of vowel in the
prefix (not discussed here), df'cermined by the stem, (3) by the number of
2
syllables in the stem-plus-suffix complex ; (4) by the tone of the prefix
complex (see in Column CT of Matrix 1), (5) by presence of pronominal object
in the context, (6) by position of verb in primary or secondary place in a
cluster:
ci
 -117-

(1) Although stems in Abua verbs differ contrastively by tone, these in-
herent differences are relevant only to the present positive- -Row D--and only
to the last syllable of the verb; e.g., for bula 'to forget' versus tusemi
'to teach'. (All other rows of Matrix 1 of Abua have inherent stem tone com-
pletely overridden by patterns determined by the following features.)
(2) Some stems require a long vowel in the prefix, and a different set
of tone patterns on stem-plus-suffix, conditioned by the prefix,
(3) The tone pattern of polysyllabic stem-plus-suffix is in two parts:
(a) the final syllable and (b) all other syllables. The final syllable car-
ries one tone, and the balance a second tone (with each syllable of the bal-
ance having that second tone repeated). In monosyllabic elements of a com-
parable type, the two tone parts are often fused on the one syllable, often
with a resultant glide, but occasionally the fused form is not the same as
the longer form compressed.
(4) The prefix complex carries its own contrastive tone, as well as
determining the contrastive tone of the verb pattern as a whole.
Some typical patterns of the total Abua complex can be given, keyed into
the rows of Matrix 1. Tones-will be labeled T (top, special tone-in negative
only), H (high), L (low):

TONE PATTERNS OF ABUA VERBS

Tense Short-Vowel Polysyllabic Monosyllabic

Complex Prefix
Medial Final Fused
Tone
Tone Tone Tone

D L varied varied
F T ...L H LH
H H .H L L
M HL ...L H H
S H-L ,..L H H
Stems requiring a long-vowel prefix may have contrastive patterns:
F TL L H H
H H L L L

S HL L
 -118-

Sample verb forms are now given to illustrate both the Abuan tone ele-
ments mentioned, and the vocalic and consonantal elements of Matrix 1:
Tense D, present positive:
nines re-b61 'You are-forgetting'
aniin re-bUla 'The-bird is-forgetting'
odi 'He is-forgetting'
bidi 'They are-forgetting
Tense F, present negative (with " as top tone): 1.;

nina
.
re-bula 'You are not-forgetting'
Tense G, completive positive:
{.

nina e-
e, legheri ni 'You
Tense H, perfect positive:
- .
nina me-ghi 'You have-gone'
Tense Q, conditional negative:
e .
nina mi-ki 'You would-not-have-gone'

(5) Presence of an independent pronominal object in the clause sometimes

modifies the tones further. We have already seen that an *i- comparable to
that of the pronoun seems also to have been fused to the verb prefix complex.
A related fusion appears with the high tone of that *1-, also carried into
the verb prefix. (In addition, although the pronoun nina '2 sg' lacks the
prefix i-, it does have the high tone--and this tone is found fused into the
prefix complex.) Contrast
iga rebula 'The goats are forgetting'
with
Iga riwula liana 'The goats are forgetting you'
in which both prefix tone and prefix vowel are changed--and the final high
stem tone is also lowered. If the verb-prefix vowel is i, however, the tone
change does not occur:
AA
nina fibula 'you did not forget'
-e
nina iwula i yoor 'you did not forget us'
Nor does tone change occur when the verb prefix itself has high tone.
For Etung, materials are abstracted from data and presentation provided
by Tom Edmondson. There is a highly intricate chain of interdependencies
in the tone of the Etung verb. Certain (indicative) prefix tones affect other
(pronominal) prefix tones (preceding them), and the stem tone. Overt tense
 -119-

suffixes affect the tone of the verb stem; the stem-suffix combine is closely
n
U enough integrated to warrant a special term--the verb 'core'. Covert (fused)
tense suffixes are actualized as tone changes of verb stems. (Tone patterns
of the core are distributed over the core, with the last syllable of a core
carrying one tone or one tone complex, and the other core syllables all carry-
ing the other tone of the pattern.) The tone class of a suffix, mediated by
the tone class of the core, may affect the tone of the prefix. The tone of a
prefix occasionally affects the tone of a suffix. The whole verb therefore,
comprises a close-knit but contrastive chain of tones. We can only give brief
illustrations here. Elm: patterned details, see the Appendix.
(Indicative) Prefixes affecting other (pronominal) prefixes:
a-k{ -gure 'He is selling'
! .
a- mo-gure 'He won't sell'
Here a 'he' is raised before the future negative .mo.
(Indicative) Prefixes affecting stems:
a- mo-gure 'He won't sell'
. . . . .
a- mo- bome 'He won't put on'
.
Here the Class A stem gure is
i low-high in tone, whereas the Class B stem has
lowering influence plus high high. All other Class A and Class B stems would
! -
13 act like these, after mo. The indicative prefix may affect a following re-
petitive prefix (as if it were a Class B stem) which in turn affects the stem.
(For this, and patterns after other indicative prefixes, see Appendix.) 1.
Overt suffixes affect the tone of the stem, leading to an over-all pattern
for the core:
a-bome 'He put on'
. , .
a-bom-a 'He puts on habitually'
The CVCV stem b6me loses its final vowel, replaced by the suffix -a; and the
high-low pattern is reversed.
Suffix affecting prefix:
n-su-k 'I have been washing'
. v
n-su-k 'I wash'

The tense labelled by Edmonson past habitual -k contrasts, over the total
verb (seen here in the low prefix), with the present habitual (seen with
high prefix). The tone classes of these suffixes differ arbitrarily but
morphophonemically, at this point, in their effect on the prefix.
 -120-

n
U
Prefix affecting suffix:
A-su-k 'He washes'
0
desd '-k 'They wash'
v
a-su-k 'He has been washing'
The first two illustrations share the present habitual tense, but differ in
9
their pronominal prefix. The phonological contrast arising from change of
prefix, however, is manifested in the core. The third illustration, chang-
ing tense to past habitual, again shows contrast of total word, but with the
pitch of the pronoun actualizing the difference; in the first illustration,
the fused suffix raises the tone of 'a to a.
The interlocking of various parts of the verb with one another, in tonal
relations, raises a question: How can one visualize the mechanism by which
this has arisen? And can an alternate descriptive device aid in understand- n
ing of the synchronic system? I wish to suggest one kind of answer, but
this must be delayed until after certain tone matters have been discussed from
a different viewpoint in the phonology chapter. (See §6.4 where I shall
suggest a left-to-right ordering of rules of fusion, with a morphophonemic
raising influence--as well as down-step phoneme--to handle some of these de-
tails.)
5.3. Hausa (Afroasiatic) Formatives in Person-Aspect Markers
In §1.6 certain clause characteristics of Hausa, a Chad language of
West Africa, were given in order to show how properties of syntax can be
compared. LikeneSses and differences within a region can thereby be discussed,
as over against the family likenesses. Similarly, for the same Chad lan-
guage, I now wish to point out that techniques of matrix arrangement, with
permutation of rows and columns, allow interesting possibilities of abstract-
ing formatives, for contrast or comparison with the Niger-Congo materials.
Here, again, I use the work done for me by Gisela Kappler, based on data of
Abrahams and others (see fn. 14, Chapter 1). Kappler first points out that
the aspect-pronoun complex (seen in syntactic context in §1.6 above) is very
intricately structured, Within it, a bewildering interlocking of components
fuse into irregular patterns of contrastive prefix, suffix, tone, length,
consonants and vowel to signal negation, time, aspect, syntactic function, per-
son, number, gender. In order to show some of the near-regularities buried
within the larger mass, she abstracts a chart (see Hausa matrix of pronoun-
formatives) which gives only the nuclear syllable of the aspect-pronoun
 -121-

complex of the verb phrase (eliminating prefixes and suffixes from the com-
plex, and deleting also length and tone from it, and omitting independent
pronominal forms).

HAUSA MATRIX OF CERTAIN ABSTRACTED PRONOUN-FORMATIVES

1 2f 2m 3f 4 3m 1
pl. p13 p12 sg sg sg sg/pl sg sg

Future MU su ku ki : ka ta a i n

Neg. Future MU su ku ki : ka 1
ta a i n

Neg. Perfective MU su I ku ki : ka ta a i n

Progressive mu su ku ki : ka 1 ta a ya : n
Optative mu su ku ki : ka ta a ya : n
Neg. Progr. 2 mu su i
ku ki : ka to a ya
i .. 1.11-0.
.
Perfective mu su ku ki : ka ta a ya na .
.
.
Relative Perf. mu su ku ki : ka 1
ta a ya na .

Rel. Progr. mu su ku ki : ka to a ya na .
.
.
Habitual mu su i
ku ki : ka ta a ya na .
Neg. Habitual mu su f

ku ki ka ta a ya na :
w .
Neg. Progr. 1 m a
: swa I
k a kYa ka ta a ya na .
w IV
Neg. Indefinite: m a swa k a kYa ka 1 ta a ya na
Future
w .
Indef. Future : m a swa kwa kYa ka to a ya na .

1.

encloses a block with u and u

encloses a block with k
encloses a block with a
Within this matrix, rows and columns have been permuted to bring together
like formatives so that their semantic relations can be more easily studied.
The central block, which is determined by k, covers all second persons--sing-
ular and plural, masculine and feminine; differentiation among them must be
by the vowel formatives, Other consonants signal further persons: nasals for
first person, with m for plural and n for singular (think of the matrix as a
cylinder, to bring the formatives together); s for third plural; t for third
feminine singular, yf.-)i third masculine; # for fourth singular and plural.
Crossing this signal set are the vowels: u for all plurals except fourth-- -
and redundantly helping to distinguish first plural from singular; i"y for
second feminine--with redundant overlap with k as signalling second person in
 -122-

those aspects (first three rows) w!.27.. . is not found as an alternate of y in

third masculine; a as nonplural nonsecond-feminine--but irregularly absent in rd
certain aspects with third masculine (same three rows) and first singular (first
six rows). The vowel i without consonant, however, signals (anti-redundantly) r3
the semantic complex of singular third masculine.
Even with this amount of abstraction, complexity of old fusions is implicit.
In the last three rows internal reconstruction suggests that a suffix *a plus
s.

*ku-gave kwa (although here I seem to have two non-identical sets of data, in
D
one of which the w occurs only after the k in these rows); and *ta- plus the
same *-a gave ta. In the early rows of the last column, similarly, the synchro-
nic anti-redundancy is due to n + a> n, and so on.
With the full, non-abstracted data, further layering of nonfusion, partial
fusion, and complete fusion need studying. Note, from Kappler's listing, the
following full forms of the eleventh row (negative habitual), in the order given
6. of.

on the matrix from left to right: bamukan, basukan, bakukan, bakikan, bakakan,
batakan, baakan, bayakan, banakan. Here the addition of ba- 'negative' and -kan
'habitual' (compare takan, 'positive habitual third feminine') seems straight-
forward. But even here the tones are not so easily treated as this would make
a
them appear. The tone of the nuclear element of the complex is high before a
suffix like -kan (hence banakan); but low after a prefix (ba-), when not followed
.
by a suffix, the ba- > ba and the nucleus becomes low, as in ba'su 'negative
progressive2 plural third person'. Yet elsewhere, tone may be contrastive: ka'
'perfective second person singular masculine' versus ka . 'indefinite future
second person singular masculine'. Part of the range of such problems can be
seen by illustrating one column of the matrix, the third plural, in the order
given there: 'future', baz'sil 'negative future', basil 'negative perfect',
sunr progressive , su
'progressive', optative'., ba su 'negative progressive2', sun 'per-
fective', suka 'relative perfective', suke 'relative progressive', sukan
'habitual', basukan 'negative habitual', ba su 'negative progressive', ba so
'negative indefinite future', sa' 'indefinite future'.
In comparison, now, with the Niger-Congo languages Etung and Abua, this
Chad language shares certain typological characteristics: (1) The tone
pattern of the aspect-pronoun complex must be treated as a whole, just as the
9
Etung verb tone (§5.2) had interlocking relations from prefix through suffix.
(2) The complex of elements, with irregular blocks of formatives, is typolog-
B
ically reminiscent of the Abua verb (§5.1).
4
 -123-

D
On the other hand, repeated warning must be kept in mind that general
typological similarity is not evidence for genetic relationship. Details
must be reconstructed.
In thinking of utilizing matrix block-shapes for reconstruction purposes,
one notes (and see discussion in last reference given in fn. 2 of Chapter 1)
that the specific arbitrary shapes in relation to specific category sets must
be compared. The basis of classical comparison is the patterned relationships
across languages of the arbitrary relations of morphemic form (phonological
content) to semantic content (lexical meaning) within a language. An analogous
basis for reconstruction of certain morphological patterns would appear to
B be an equally rigorous application of the comparative method to an arbitrary
U relation of matrix form (that is, shape of its formative blocks) to matrix
semantics (its intersecting categories). General relationship will not do.
It is the highly arbitrary detailed relation of form to meaning which, when
seen as patterns across languages, should be able (1) to carry conviction of
relationship and (2) to be reconstructable.

:1

ta
1

,...--=rai=1=1311==io.
 -124-

FOOTNOTES
1
The dual semantic role of such a formative I have elsewhere called
anti-redundant. See my 'Non-linear Order and Anti-redundancy in German
Morphological Matrides' Zeitschrift fur Mundartforschung, 32.193-221 (1965).
2
The suffixes of Abua are not otherwise relevant to our discussion
here. Several orders of suffixes--and possibly enclitics--occur.
 CHAPTER VI: PHONOLOGY

In preceding chapters attention has been given largely to grammatical mat-

ters (or, in §5.2, to morphophonemics). Here I turn to several interesting
problems of phonology.

6.1. Quasi-Isochronic Units of an Accentual Hierarchy

In both the Abua and the Etung materials ( §5.2) certain rules of tone
placement apply to distribution of tone patterns over entire stems, whether
one or more syllables were involved. It seemed to me improbable that such
control would take place unless a larger phonological unit of some type were
a setting within which these rules could be operative. In Basare, work in col-

laboration with Monica Cox led to the postulation of quasi-isochronic feet which
were in some sense simultaneously accentual units, and an accentual hierarchy
which interlocked with the grammatical_ hierarchy but was not identical with it.
Within a unit substantially larger than a word--a clause, for example- -
replacement of one word by others with respectively different numbers of
syllables left the basic timing and the accentual dynamics of the clause
perceptually unaffected. The stressed syllables (which I shall now call
"accented syllables" in order to avoid the implication of intensity in the
1
term "stress") remained the same in number. A unit of timing was clearly
perceived under these conditions--and may be called a "foot".
Compare:
[1] a "ni It'Ll'wanl tum-,mi-nee 'You know 'thing' which-is-here'
1
... ... . .
[2] a "batilt'Ll'wan I tum- mi-nee 'You carve-up 'thing' which-is-here'
i

[3] a "caa It.ti'wan I tum- mi-nee 'You have 'thing' which-is-here'

I

[4] a "sdicfl It.diwan 1

I
tilm-,mi-ne 'You taste 'thing' which-is-here'
i
4.0

In this set, the units 'ni, fbati,baa, and liakafi ('know', 'carve-up',
'have', 'taste') are perceived as astonishingly similar in length, except
that the length of ni (but not of caa) is partially carried by the condi-
tioned length of the following t (intutwan). The rhythm--clear, and easy
to hear within this arrangement of examples--is unaffected. Accents come
at the same perceived intervals; the accent of tt-Pwan is not delayed, that is,
by the added syllables of the preceding 'dakafi. The short ni versus long .

caa are perceived as phonemically and phonetically contrastive, even while

the rhythm is unaffected; the compensatory length of [t] perserves rhythmic
length without affecting phonemic length. Junctures precede the unaccented
syllables.

-125-
 -126-

Laboratory displays of these recordings were subsequently made for us by

Charles Peck, with analysis and commentary by Dr. Ruth Brend. The duration of
the three feet of illustrations numbered [1] to [4] in two utterances was mea-
sured (in seconds) as follows:
[1]: a?ni tutwan tutminee
(with the [t] calculated as part of the second/foot)
.3 .36 .42

.24 .3 .38

(with the [t] calculated as part of the first foot)

.36 .3 .42

.34 .2 .38

[2]:

.4 .36 .4

.38 .3 .42

[3]:

.34 .32 .38

.32 .3 .4

[4]:

.48 .34 .4

.4 .34 .42

Brend considers that the [t] probably belongs to a shared border between the
first and second foot in utterance [1] and that, taking this into account, the
measured differences between the length of the various utterances of the first
foot in [1] to [4] to be no greater a variation than could be compatible with
perceptual isochrony. (This is especially true in view of the second utterance
of [4].) Other utterances of these same sentences measure as follows:
[1] (with [t] as part of second foot):

.3 .33 .4

.25 .34 .43

(with [t] as part of first foot):

.38 .35 .4

.34 .25 .43

[2] .38 .32 .42

.39 .3 .42

[3] .36 .32 .4

.38 .34 .39

[4] .48 .37 .42

.44 .32 .4
 --127-

0
If the calculation of the utterances of [1], in all four cases above, puts
the juncture in the middle of the [t], then the measurements of the first two
0 feet almost invariably come out as nearly identical.
The final foot proved to be longer than the first two, in some of the
utterances. This was unexpected; perceptually it had not appeared longer. My
assumption, based on prior work with accentual systems, was that perhaps the
lengthening before pause was accompanied by a weakening, which lessened its per-
ceptual impact, and went unnoticed by us. I asked from the laboratory further
measurements which confirmed this hunch. In observing amplitude sections of
these utterances, Brend reported that the over-all amplitude of the final foot,
in each utterance, consistently measures two to three db lower than the ampli-
U tude of the first two feet, and there is, typically, a gradual decrease in
amplitude in the final foot which is not seen in the first two feet.
I had long used whistling by informants to help me perceive tone changes,
and wondered whether it might be helpful in studying dynamic features. On
the basis of laboratory measurements, however, it has since become evident that
informants, whistling forms like these, did not consistently whistle in chunks

a which corresponded either with feet or syllables, but rather seemed to mix
these borders indiscriminately. Nevertheless, they consigtently whistled the
highest tone at places we have marked with double-stress (") to indicate the
peak of the intonation contour as well as the heaviest accented syllable in
the utterance. The frequency measurements here, therefore, would seem to con-
firm the hypothesis that the nucleus of these utterances falls at the places
marked with double stress.
It seemed possible, furthermore, in an area where drums were widely used,
some kind of tapping might reveal or support the analysis of some of the dy-
namic characteristics. We were unable to explore this possibility adequately- -
but tapping on a table or a tin did not appear to give as satisfactory a per-
ceptual relation to speech as did the whistling. Brend, in measurements of the
tapping ,feels that the tapping units correspond more closely to syllable units
than to feet. In instances where several syllables in speech are collapsed into
one spoken foot, for example, the measurements of analogous tapped feet do not
coincide with the groupings of the speech measurements. When, however, there
is a one-to-one correspondence of number of syllables within two feet, these
feet do measure as being very close to identical in length, and to this degree
 -128-

the hypothesis concerning isochrony of feet is confirmed by the tapping, since

the consonant and vocalic structures of the syllables did not change the timing.
fl
For example, the tapping of [1] for the corresponding three feet postulated
U
above measures as: .4 .38 .41. Within a foot, furthermore, une spoken syl-
lables, when measured, reveal a definite pattern of long syllables occurring
as the nuclei of feet, and short syllables as prenuclear. Measurements of the
tapped syllables follow the spoken utterances at this point.
There seem to be contrasts between a foot with stress on the first syllable
with a foot with stress on the second syllable. These differences are easily
overlooked, however, because (a) tone interferes (a high-toned syllable is
likely to sound stressed, a low one unstressed, to English speakers) (b) the
phonetic difference is very slight, perceptually, so that it can easily be
missed; (c) the phonetic character of the contrast may be length, rather than
amplitude-or pitch, (but this particular kind of length difference is awkward
to abstract from syllable length caused by post-syllabic consonants or by bi-
vocalic syllable nuclei)- (d) a short syllable, half-long under accent, must
be differentiated from a phonemic doublet.
[5] ki'jlki 'knife'

[6] kriiO.kifi 'mouth'

[7] ki'bfki 'child'

In [5], vowel qualities and tone heights are kept constant. Here, under con-
trolled conditions, a slight accent carried by half length of vowel seemed
to be perceived.

Pre-nuclear (pre-accented) syllables of the foot were perceived as

especially rapid, helping to separate off one foot from a preceding one.
The laboratory analysis later shows, for [5],[6], and [7], that the con-
trast of tonal heights is confirmed: tone on all syllables of [5] is very
even; tone on the middle syllable of [6] is lower than that on both the
first and third syllables, with the tone on the third syllable being slightly
lower than that of the first syllable; tone on the second syllable of [7] is
considerably higher than that on syllables one and three, the tone on the
third syllable starting at the same point as that on the first syllable, and
then showing a considerable down-glide before silence in one utterance of [7].
As for length of spoken syllables, measurement of spectrograms follows
(with three utterances of each item):
 Q
-129-

B
iii
[5] ki ki
.02 .06 .1

.1 .16 .16

.06 .2 .16

[6] ki 'no ko
.04 .08 .12
t
.11 .18 .2

.1 .18 .21

II [7] ki 'IA' ki
.08 .16 .16
0Li .06 .1 .11
.1 .16 .2

These data confirm clearly that the pre-stress syllable, as expected, is espe-
II
cially short. On the other hand, the final syllable was often as long as--o-
even longer than--the accented syllable, which we had not predicted from our
0 field observations. Here, as above in reference to the foot as a unit, we
would now assume that a final drawled syllable would be heard as part of the
g prefinal phenomena and that this particular length had served as a cue to pre-
pausal position rather than as a signal of placement relative to the nucleus of

El the foot. This necessitated, however, searching for further cues of accent
within the acoustic data. A simple measurement of amplitude, however, does

a not show a one-to-one correspondence with perceived accent. There is a slight

hint, however, that a sharper decrescendo occurs in the final unaccented syl-
lable than in the central accented syllable where the amplitude is more con-
stant throughout the syllable; the shape of the amplitude curve within the
syllable--not just the crude amount--may prove to be relevant here.
We were disappointed to find that amplitude measurements of neither whistle
nor of tap seemed to contribute to the location of the placement of stress
since we were neither able to correlate it with our perceived accents nor to
find in the record regularities which would suggest better hypotheses; our con-
clusion for the moment is that neither whistle nor tap in this instance was a
useful heuristic device for determining the place of stress. I would person-
ally, however, like to see the devices tested with a variety of languages and
ri
informants under the hypothesis that there might be some conditions under which
they would be useful.
D
 -130-

A hierarchy of feet, with smaller groups in larger ones, seemed to be pre-

sent in the Basare data. The contrastive features, however, were not too clear
nor completely consistent. As a hypothesis for further research, however, one
can suggest that feet join into 'phonologicalphrases,' and these into 'phono-
logical clauses'; the nucleus comes early, in each, marked by higher pitch, and
perhaps sometimes by greater length (or sometimes even by intensity):
[8] a "yaa I 'kaa I domin I doo II "an I 'kaa I nanll

Here the intonational pitch seems to drop progressively lower from the begin-
ning to the end, but with major upstep after the first double bar (not, however,
as high as the initial start) and with minor ups and downs caused by the separ-
, ate tones and by the included feet.
Brend suggests that perhaps the most crucial organizing principle for sev-
eral of the levels of the hierarchy arises from length. Perceptually accented
and unaccented syllables within a foot measure long and short respectively (with
the exception of utterance-final syllables, for which see discussion, above);
sequences of feet within an utterance show isochrony, to a remarkable degree
(with the exception of utterance-final feet, for which see above discussion);
while a foot occurring at the peak of the intonation contour turns out to be
slightly lengthened.
Grammatico-lexical considerations affect the hierarchical organization in
various ways. The phonology is partially but not completely independent of
lexicon and grammar, and vice versa.
A pronominal subject, as in [1-4], is often treated as part of the foot
which has its nucleus in the verb. A longer, independent noun subject may com-
prise a separate foot, as in:
[9] .a"liis.an I 'giff-bi 'Alasan chops-them'
The object of a verb joins the preceding foot with verb as nucleus, if the
verb is short--but may comprise a subsidiary foot if the verb is long:
[10] 'doo "yee 1 bl.'nib nin I 'ka fataku 'Formerly I people remote-time I
saw blood'
[11] 'doo"yee Ibi'nib nin I 'dakafi fataku 'Formerlylpeople remote-time I
thought-of blood'

-1
 -131-

Abua, of Nigeria, like Basare of Ghana, also has characteristics of rhythm

dynamics which involve isochronic feet. A hypothesis worth exploring: The
level, medial tones of a verb pattern are those which may be added on to or
after the nucleus of the foot without modifying the timing of the foot. This
hypothesis we have not had time to test. It seems, however, to hold some
promise of illuminating the peculiar pattern of pitch controls which relate
grammar to phonological hierarchy while ignoring--to some degree--syllable
count.

For an Abua set of isochronic units note:

. 40\
[1] [ 9di ] koosagan '[He] has not cooked before'
[2] [ 9d i ] k 9 9 bygolgan '[He] has not drunk before'
[3] [ od i ] kookpogorongan '[He] has not peeled (plantain) before'
Where between k 9 9...gan the medial series of like tones seemed to take per-
ceptually-same time spans, whether one, two, or three syllables.
Imagine my surprise, therefore, when Brend showed me the following measurements
in which the lengths are closely proportional to the number of syllables:
[1] sa -- .28
[2] 144 -- .44
[3] kpOgOrOn -- .6
Since, however, the retrospective perception of isochrony was so strong, I asked
Brend to check the spectrograms again to see what would happen if--instead of
measuring only the unstressed syllables--the entire foot were measured. (This

would leave room for some type of compensatory lengthening such as we had with
the long [e] in the first of the Basare illustrations.) Note the following
measurements, therefore:
[1] kO8'agL -- 1.0
[2] k99eygolgL -- 1.08
[3] kookpogorongan -- 1.18
As we compare these we notice once more that in spite of the difference of
two syllables the respective feet are similar in length. I would conclude, there-
fore, that it is the foot length as a whole that must be studied if one is to
correlate measurements with perception.
One must, if these data prove as a guide, however, not expect that the measure-
ments be identical where the feet contain different number of syllables, but rath-
er that a foot with four syllables will not be twice as long as two, nor one with
five be five thirds longer than one with three. The partial shortening of the
 -132-

whole foot (relative to what it would be if each syllable took the same time)
impresses the hearer as somehow 'rushing to a conclusion' which, in turn, is
interpreted as in some sense 'the same'.

6.2. Tone
Many problems of tone analysis interlock in a fascinating complex: The
placement of tone, as we have already seen, can be viewed as affected by the
morphophonemics of a unit which is prefix-plus-stem-plus-suffix-plus-variant-
D
number-of-E'..eL-syllables ( §5.2); or some of the same phenomena (e.g., a medial
series of like-tone rapid syllables) can be related to accentual dynamics (6.1).
The number of levels of tone in a language can for some purposes be viewed as fl
a.set of classical static contrastive tone phonemes; but for other purposes is
better viewed as a basic set plus a special level limited to negative contexts;
and in other instances of a different kind as a basic set with a special con-
trastive level developed through the loss of low syllables which leave a dynamic
influence behind them. Allotones, under these conditions, bring problems of
recognition and description. Both the theory and practice of treating tone
languages are affected. New phonemes of tone seem to be visible in the process
of development; West Africa seems to be an enormously interesting "laboratory"
for the study of the dynamics of changing tone systems.
6.2.1. Basic Tone Heights
Niger-Congo languages of West Africa vary in the number of their basic
or "normal" tones; e.g., Igede has four, Bette three, Degema two.
In Igede, the four levels of contrast were extraordinarily clear, with the
particular informant brought to the workshop by Richard Bergman. (Tone data are
from him and from Martin Leigh.) Instrumental analysis of utterances--which
were spoken quite normally--gave displays which left the levels so clear cut
that they appeared to Brend almost as if they had been "sung". The whistling
of the patterns by the informant gave a similar result. Very little influence
is seen where one tone affects another.
In the following illustrative set, note Igede contrasts within the frame
5...1E. Symbols are high to low respectively, /'/, //, /'/, /'/ (where /^/
represents a lower-mid level tone, not a glide).
o hu lc 'He has washed'
hu lc 'He has stayed'
_
o hu lc 'He has scattered'
o hu lc 'He has flown'
 -133--

o I

Note also, la 'to miss', li 'to have', la 'to chew a stick', la 'to bulge out'.
For Bette, note the following sets, contrasting both before low and after
low respectively (with some fusion in the second context):
undi ken 'one person'
kuno ken 'one soul'
Q
usi ken 'one kidney'

tya undi 'leave the person'

tya uno 'leave the soul'
..v.
tya usi 'leave the kidney'
(and for further Bette tone heights, see below, 9A6.2.5).
Agbo, also, has three tone levels, as seen in data from Klaus and Janice
Spreda:

leta 'stone'

wadum 'man'

eten 'animal'
On numerous short utterances, frame techniques can be used which compare levels
of preceding and following pitches, without too much interference from 5ntona-
tional downdrift or conditioned variation of tones.
In many of the languages of West Africa, however, two basic tones are in-
volved, plus added phenomena of great complexity--growing out of intonational
downdrift, conditioned variation, and fusion with special residual phenomena
(made up of morphophonemic replacements, newly developed tones, and levels of
pitch which fit easily neither of these two categories). Illustrations will
be found in the next few sections.
6.2,2. Extra-High Tone Developed from Lost High
The Degema and Engenni are very closely related Kwa languages (Delta Edo
group). Yet, of the relatively few differences between them, two are of great
interest in the study of the dynamics of the change of tone systems for West
Africa as a whole. (Elaine Thomas supplied me with these materials.)
In the Degema, the personal prefix carries a regular high tone for sub-
junctive. In Engenni, however, the prefix is lost and the tone of the preced-
ing word is replaced by a new tone level /"/ which is higher than either of
the two normal basic tones and in contrast with them. In addition, all tones
(both in Degema and in Engenni) following the extra-high are a bit lower than
they would be otherwise--the general "key" is depressed. Not only single low
or high, but the entire following sequence is lowered somewhat.

4011111.11wr
 -134-

Degema:
imo eve e-kpein ak,yo
children two should-wash put-the
imo ive kpei akLna
children two-should wash pot-the
'the two children should wash the pot'
and
Degema:
*ploy° o-ta nuvaL

child go to-house
Engenni:
amona to mu
child go house
'the child should go to the house'
I
The negative verbal prefix /o- in Degema already carries the extra-
high tone--from unknown origin - -but also passes it back to the preceding word
in Engenni:
Degema:
omo noonayo o-kpein akLyo
child this not-wash pot-the
Engenni:
amo anona kpei akLna
child this-not wash pot-the
'This child did not wash the pot'
In various other languages an extra-high tone occurs, and may be limited to the
negative. Note, for Abua:
nina ke-ki 'You will go'
nina ke-ki 'You will not go'
nina re-bula 'You are-forgetting'

nina re-bula 'You are-not-forgetting'

_e
nina mi-ki 'You would-have-gone'

nina mi-ki 'You would-not-have-gone'

pog 'Look!'

ke-pog 'Do not look!'

 -135-

Gardner and I seem also to hear, for this special Abua tone, an added intensity,
or "tight" voice quality. Note, further, the high replaced by low following it,
in the first two illustrative pair's. Brend's laboratory findings clearly con-
firm the special height with regard to measured frequencies. The extra-high
tone is consistently and regularly higher than the normal high tone.
6.2.3. Lowered Key (Terrace) Developed from Lost Low
2
It has been less than a decade since Welmers called attention to the lin-
guistsof this century the importance of ' downstep' or 'terrace' tone, to tonal
3
studies. Much earlier Cristaler (who had studied with Lepsius) had devised
an elegantly simple orthography to cover accurately the transcription of Twi
tone with its downstep; and--as Gleason pointed out to me--some published ma-
terials showed, in retrospect, that a downstep of a high tone could be corre-
4
lated with evidence of a lost low tone. The reconstructed low (sometimes
optionally present) by conditioned variation lowered a preceding high; with
the loss of the low, the lowering effect on the high remained,-with three fur-
ther results: (1) The key (general height of all following tones in the phrase,
including high tones) was lowered accordingly. (2) Contrast developed between
a normal-high (i.e., one not thus conditioned downward) after high, and a step-
ped-down-high after high and both contrast with low after high. (3) The result-
ant system of contrasts contained an imbalance: After any one morpheme with
final low tone there could follow it only two contrastive phonetic levels--low
and high; after any one morpheme with' final high tone, a three-way contrast of
phonetic levels could occur between high plus high, high plus stepped-down
high (from a reconstructed low plus high morpheme), and high plus low. Thus
(with /*/ representing the downstepping of the tone immediately following that
symbol; and with *LH having high slightly lower phonetically than in *H):
(1) i) H + H
ii) H + 'H < *H + LH
iii) H + L
but
(2)- i) L + H
ii) L + L
U
In phonetic graph (with hyphens representing relative height in the pitch
envelope):

*
(3)

H L H H ! H
 -136-

The place of the morpheme boundary--whether preceding versus following the

low--affects radically the kind of phonetic differences one can find in frames.
If the morpheme boundary precedes the *L as in (lii), the three contrasts (HH,
HL) can be found following the first morpheme. When, however, the *low is
part of the first morpheme, as in
(4) H.H < *HL + H
then only two contrastive types can follow it:
(5) i) H. + H (high followed by lowered high)
ii) H. + L (high followed by low, which is unaffected
by the loss of the preceding low)
since a normal high without downstep (see li) cannot occur there.
Once a high has been lowered by *L, the contrastive types and the phonetic
lowering repeat,until the end of some kind of larger phonological phrase, pro-
ducing a characteristic terracing effect,
The downstepping influence may also sometimes occur in the middle of a
morpheme--and other complications occur.
The most extensive and illuminating discussion of the problem may be found
5
in a series of exchanges between Stewart, Welmers, and Schachter. Here Stewart
proposes the influence /*/ as itself a 'separate phoneme'; Schachter (p. 37)
assumes that 'no strictly phonemic analysis'of the data is possible--and attempts
to deal with them 'on the morphophonemic level'; Welmers (p. 55) has a writing
system and analysis--which is 'not morphotonemic' but one showing 'phonemic con-
trast', such that one postulates (p. 56) somewhat cautiously either a system
having 'three "tonemes" '(high, drop, iow) or 'two tonemes plus a phonemic
"downstep" '; in which, for example (1 i, ii, iii) could be written as RH, HD, HL. n
The theoretical implications of the problem we leave for §6.2.8, in order to
U
add certain kinds of detail.
6.2.4. Terrace Tone Developed from Low-Replaced-by-High
The major attention to terracing has been given to a downstepped high tone
resulting from loss of a preceding vowel which had carried low tone (which had
previously caused the lowering, nonphonemically, of the high following it).
In this section I wish (1) to emphasize that the downstep sometimes clearly
develops not with the loss of a vowel carrying low tone, but by the morpho-
phonemic replacement of a low tone by a high tone and (2) to show that this
phenomenon sometimes has an extremely limited distribution, controlled in part
by"its place in a larger phrase.
 -137-

The clearest data from the workshop, for these purposes, comes from Elaine
Thomas, in a comparison of two closely-related languages--the Engenni and the
Degema of Nigeria.
In Degema, three of the major tone classes of nouns have a downstepped '4

tone which corresponds to a low tone in Engenni. Note, for example:

. !-
(6) 1) Degema ak 'pot'; Engenni akt
. !

ii) Degema umem e 'faith'; Engenni ume

!atu
.
iii) Degema Esaru 'mosquito'; Engenni us
(Only a few nouns of the class *HL remain in Degema; most have
become H H,)
Within Degema there are numerous further evidences of H < *L in syntactic
positions. Note the verb plus object when verb ends in high:
(7) omon *uko > omon
,
uko 'He saw a canoe'
For morphology, note:
, .!.
(8) Engenni ota-nt'he went; Degema ota an
On tl,e other hand, the Degema changes of *L > .1.1 in the noun seem to be
extremely limited in distribution within the larger utterance. They occur
(with possibly a few exceptions) only on morphemes directly preceding an
utterance-final juncture. Compare:
!

(9) i) omon uko 'He saw a canoe

ii) omon uko-yo 'He saw the canoe'
I.
iii) ogE LICt. 'He looked at pots'
iv) ogcn tict, yo 'He looked at the pots'
v) oyt. a-5.0 'He is there'
vi) a-bo muvat. 'He is in the house'
I.
In (9i) the *uko >.uko ; in i9ii), however, low is replaced by normal high
(*uko >uko) before a further morpheme.
6.2.5. Indications of a Developing Downstep in a Three-Level System
We have shown 06.2.4) downstep in Degema, where only two regular levels
occur (but an extra-high level is found in negative--see §6.2.2). Now I wish
to show a downstep system occurring in a language with three regular levels.
The data, from Bette, come from the work of Ruby Peterson and Irene Crane.
Contrasts of the three Bette levels were shown in §6.2.1. We first, in la-c,
repeat a set of contrasts after a low tone. Reconstructed forms are given with
an asterisk when it is desired to show pre-fusion forms, or forms before tones
have undergone morphophonemic change.
 -138-

(la) *(tyao /tya 'leave the person'

(lb) *(tyao / ty4 un3 / 'leave the soul'
Avf Av.
(lc) *(tyao us) / tya us i / 'leave the kidney'
In (2a-2d), we establish that the first word is now a high tone or a sequence
of high tones. Stresses are marked, perceived on the last high syllable of words,
by a small vertical stroke.
(2a) /kfte't61 kan/ 'one basket'
(2b) Wye k6n/ 'one mother'
. . . .
(2c) *('he la undi) /be luundi/ 'come with the person'
(2d) /kigtAl'gc5 ken/ 'one hill'

A sequence of high-tone words is thus given in (3a). The pitch graph

above the phonemic writing shows that a step down occurs after the syllable
marked with stress (giving the phonetic impression of a down-step system).
(3a) *(1'3r6 '1)6 ha klgt06) /1'ya 'a 'be igu'go/ 'mother she came
to the hill'
Then in (3b) a mid tone is introduced into the sequence, to show high tone
stepping up after it. In (3c) a low tone is similarly introduced to show that
the three-way contrast of levels still exists between two high-tone phonemes,
fl
the first of which is stressed.
I(3b) *(iye' aba he ubo he klgtigO) Wye 'e 'I; Uboihe IgtOg6/
mother came to the main road on the hill'
.7 V
(3c) *(iya aba la kitatal isi) /i'ya a ' a e ita' tel l's1/
'mother came with a basket of kidneys'
Illustration (4a) shows a low-high word which with a preceding low in
(4b) combines into a high high-low sequence. In (4c), it is seen that the
a
change to (4b) first took place (by an ordered rule) and then was treated as
beginning high in (4c), stepping down regularly after a stressed high.
(4a) *(kat6 k;n) /kata kan/ 'one market'
(4b) *(hg kate) /beta/ 'to the market'
-1
(4c) *(kitatal kiba he kata) /kitattal itbatri 'ta/ 'the basket came
to market'
fl

B
o
 -139-

The system described here (three regular phonemic levels plus a nonphonemic
downstep after a stressed high) could very easily develop into a typical phonemic
downstep, plus three phonemic levels. If, for instance, fusions as in (4c) were
to lead to a different (nanfinal-in-word) placement of stress while retaining
placement of downstep, the downstep would probably have become phonemic.
On the other hand, if other investigators (a) did not on these same data re-
cord stress as in the way we have done, and (b) if word boundaries were ignored,
their analysis would probably show downstep as already phonemic. Inasmuch as
stress analysis in this language is by no means obvious--or certain--the alternate
analysis should be kept in view.
After this analysis in the field, laboratory work by Peck and Brend gives
general confirmation of the pitch contrasts of instances such as (la-c). Simi-
larly, the instrumental measurements add no major difficulties to the analysis
of pitch sequences in the downstep series--nor are they able at the moment to
help very much (because of limits to current acoustic techniques) on the important
problem of stress.
6.2.6. Overlap of Conditioned Allotones
In the Igede, of §6.2.1, levels were clean-cut, and fairly uniform in their
respective heights throughout the phrase.
In downstep systems, however, conditioned variation of tone levels leads to
extensive problems. This statement, however, is historically stated backwards:
A system of conditioned variants of tone has led to the particular tone feature
known as phonemic downstep. The sources may be the' conditioned lowering of a high
by A preceding low ( §6.2.3); by conditioned lowering of high in its morphophonemic
source in a low ( §6.2.4); by a lowering of high after stress ( §6.2.5). Or an extra
high may take its source from a lost high ( §6.2.2) which was raised before a low.
Many morphophonemic changes ( §5.2', ,and Etung Appendix data )..reveal conditioned-
replacement of tones without observable current allophonic traces.
The extensive variety of such effects implies some general dynamic character

LI (see also §6.1) for these systems: A general tendency for the pitch level of a
6
phrase to 'drift' downward, rather than stay on an even key, in steady pitch
"bands".
El

77-111111101.1P-
 -140-

In principle, if the drift were rapid enough, a three-toned system could

have the following pattern:
H H
M H . M H
L M H L M H
L M# L M #

L L

in which each syllable can carry any one of three tone phonemes (H or M or L),
but the next syllable in the series (with the same set of possible contrasts,
no more and no less) would by conditioned variation (in the down-drifting con-
tour), have each tone phoneme suppressed in phonetic (not phonemic) height one
notch. The result would be : H + H would not be level (the second would be
lower); M + H would be level with each other (but in phonemic contrast, relative
to other potentially substituted elements at that point in that frame); L + M
7
would also be on a single level, etc.
In Degema, Thomas shows some of these phenomena. Between words (but not
within them, under penalty of collapsing some of the system) a word-final low
depresses a following word-initial low to its own low phonetic pitch, but a
following word-initial low to lower than the first low. Note:

(1) /mtmon ikpe meki/ 'I-saw goats in-market'

L H H L HL
but
. , .
(2) /mtmon spa kLkre/ 'I-saw frogs all'

L H HL L L
I 4\ ,.. 4..

In (1), note that L and ', in the. middle of ikpe meki are on the same level;
and that -ki is lower than the first syllable of mt,mon.
In (2), on the other hand, note that L plus L between tpa and kAre drops
sharply; the L of I- contrasts with (is lower than) the me- of meki in (1).
The phenomena, furthermore, are in addition to those for phonemic downsteps
!
/ / discussed for the same language in §6.2.4; the /1/ adds a nonpredictable con-
trastive drop outside the major system of two levels or the special high contrast-
ive level ( §6.2.2).
 -141-

In Izi, data from Paul Meier illustrate one of the most pervasive z-ld, at
first, disturbing--of these situations: A sequence of L H H has the low and
the downstepped high on the same pitch:
! . t ,
/ma re ji/ 'I-present sell yams' (with e and 'ji affirmed
to be on the same level; a statement clearly
confirmed, like many other crucial parts of
the tone phonetics of Meier's material, by
laboratory tests by Peck and Brend)
AA NI Ay/ ,v,
/muturogbonosi/ 'I threw a stick at Ogbona' (with -ro- and nosi
. .

. . . " . .v.
on same pitch; from *[mu turu ogbona osi])
!
But there is a contrast of LL with the HH, in the same environment:
. . . .
/muturogboniphe/
. .
'I threw something at Ogbona' (with -ni- lower
than ro--and - no--; and with -phe still
lower, since prefinal low-low drifts down
a bit; from *[mu turn ogbona iphe] where
the nouns belong to different morp5Ophonemic
classes)
It is the presence of contrasts in such comparable environments which justifies
the contrastive phonemic status of the elements; and it is the lack of more
than the topologically-equatable members of the sets which forces the analysis
of like-le 7els as phonetically-overlapping but phonemically contrasting tones
in the different environments. (Emphasis on different environments is main-
tained, lest that kind of intersection of phonemes be postulated which is un-
acceptable within postulates of classical phonemics.)
Many other intricate problems of contrast arise in the Izi data, but they
cannot be illustrated here in detail.
6.2.7. Towards Internal Reconstruction of Tone Chains
In scattered instances in preceding sections there have been given starred
forms, implying some degree of internal reconstruction. Can such tentative, in-
ternal reconstruction be carried substantially further? If so, could it illumi-
nate some of the deepest puzzles of Niger-Congo languages in which--for example--
the negation of clause involves pitch changes at discontinuous places? And
could it, furthermore, suggest sources for some of the most pervasive of tone
rules--such as a "flip" of high-low forms to low-high, etc.?
 ---

-142-

To begin with, such an attempt must be viewed as representing constructs

designed only to show correspondences between current spoken forms (or as a
generative base for those forms). My hope, however, would be that it would
stimulate longer range study into the history of changes in the respective
language families. For illustration of this approach--but with no guarantee
of accuracy--I choose a fragment of Etung data (from Thomas Edmondson, who is
studying the problem further).
I begin with a pair of illustrations plus their associated starred forms,
which I shall then attempt to justify:
(1) i) /. Ic% 1:,/ 'He (did) not come'
ii) *(t + a + kat + ba.)

(2) i) /. Ici 1:,/ 'They (did) not come'

ii) *(t + a.t + kat + ba.)
. .
The setting up of a difference between a 'lie' and a 'they' is based on
a large amount of data in non-negative forms, such as (3):
A I
(3) i) a ki gure 'He is selling'
I A I
ii) a ki gure 'They are selling'
The difference Jetween ka '11 ka is attributed to a raising influence assoc-
,
iated
fated with at 'they' but not a 'he':
(4) t + > I

It is important to note that this raising influence is not

--. a- phoneme
-.....- like the
lowering influence /!/, since it merely causes the replacement of /'/ by /' + !/

(already in the system phonemically) and neither adds a new level, nor a condi-
tioned variant of a level. It is a morphophonemic symbol, not a phonemic one.
(In some languages of the area there is the suspicion that a rising influence
/i/, analogous to the lowering /!/, des achieve phonemic status. Meier is
chocking on this possibility in Izi; so far, however, I have not seen evidence
which for me completely eliminates the nonphonemic interpretation. In Etung,

the kind of data giving rise to this possibility have not been found.)
The + before ba leaves it unaffected, since:
(5) t + >
I
!
(whereas the 1:, after of (2) develops an audible downstep of the 'ba complex).
Therefore morphonemic rising influence residing in kat is not manifested--not
detectable--in (2). Note, however:
. .
(6) i) a k ruse gile (did) not go'
.. .. I
ii) * (t + a + kat + rue)
(7) i) a ic. rile 'They (did) not go'
ii) *(t + afi + kat + rile)
 -143-

In (7) one finds the Basic form of rue,

rue, beginning with a low tone. When, there-
fore, the low of ru is replaced by high in (6), we deduce the inaudible presence
or the morphophonemic + in the preceding ka. But when the low of ru is replaced
by high, the + leaves its impact on ru, raising it to high. At the same time,
\
however, the e of rue is now heard in (6) as lower than the former-low-now-raised-
to-high preceding it. All of these phonetic characteristics are represented
. Ar. !
phonemically by ...ka ruse, within the regular interpretation of /'/ as causing
the high after it to be a bit lower than the one before it.
Note, therefore, that at this point there arises some hope of eventually
I \..

explaining reversals such as to etc., in morphophonemic terms whereby

,

a chain of elements affect the ones directly next to them. This would save
some of the problems of explanation of tone replacements (which appear to be
largely composed of assimilation phenomena) without requiring action at a dis-
tance. Explanation would then be in terms of the kinds of assimilative changes
actually observable, currently, in fast versus slow forms, or across closely-
related dialects.
Two further problems, however, would remain: (a) The origin of + would
itself have to be explained. This would now seem to comprise a reasonable re-
search task. Can it, for example, by comparative techniques be traced to a
lost high tone? Or does it suggest an easier way of searching for (or testing
present hypothesis of) a phoneme /i/ of rising influence analogous to /!/?
(b) The second fr'ature to be noted carefully is that such influences perhaps
do not alter the relevant domains of level pitches in those languages in which
several syllables in a row hang together (all up, or all down) in rules; and
several syllables of a domain operate in the rules like a single-syllable domain
in the same rules.
Returning now to (1) and (2) we note that + at the beginning of the formulas
has not been justified. The added basic assumption here is that the transformation
of a positive clause to a negative requires changes at more than one place in the
clause. This, however, is amply demonstrated in clause materials (see Chapter 1,
§§1.1.2). And if this is granted, the assignment of a second morpheme (or part
of a morpheme, in an alternate analysis) to the clause complex (+ before the
pronoun, in addition to the ka) seems no longer surprising. Yet it is precisely
this (or other?) morpheme with zero phoneineb but with significant morphophonemic
8
impact which in part makes the negative appear so mysterious. Here, however, the
4' plus third person a gives the a of (1).
Note, howeVer, that we would have expected + + to give AI, as in (4). How
can this absence be explained? Note that the downstepping influence would then
have preceded the low of ka, which it cannot affect since:
 -144--
(8) ! + >

and therefore, for (1) (and including rule 5) Li

(9) *(t + a + kat + ba)
.!
> *(a + kat + ba)
*(a + kat + ba)
. . .
> (a ka ba)

We would like, however, to find an instance where the implied

(10) *(t + + ') > *()
with actualized downstep. This situation, however, is present in (6), where Oat
raises the first syllable of rile to rd '6 with the expected resultant down-
step of the last syllable, which is high.
I have not developed these rules to cover much of the Etung data (see
Appendix, for material; Edmondson is working on a more extended analysis). I

do hope, however, that this kind of approach will open the door to a new type
of fruitful work in this area.
6.2.8. Theory of Phoneme Types as Item, Process, and Relation.
I now wish to suggest that these studies of tone (differentiating tone
phonemes-of-level, from tone phoneme-of-downward-influence, from morphophonemic
non-phonemic symbols of raising influence) can be fitted into a deeper theoret-
ical perspective and at the same time help to solve an old problem.
The general thesis: That _phonemes may be of three types--item, process,
and relation; that these in some sense are related, in turn, to tagmemic per-
spectives of particle, wave, and field. The specific claim: That "segmental"
elements such as /o, s, t/ are particle phonemes, as items; "suprasegmental"
elements of tone level, and of length, are field phonemes, as purely relational
elements; that the /!/ of Africa is neither a segment nor a relational pole of
a sector of a field, but a wave (or process) phoneme.
Note that, in general, the particles--once the system is known--can in
general be recognized in isolation, as 'things' in themselves. The field
phonemes are recognizable by paradigmatic contrast in a frame; the frame shows
contrastive relations within the subset of that field. The process phonemes
are recognizable only in a sequence--in a phonological construction, or wave- -
in reference to what has happened to an item previously identified in the
paradigmatic field.
The development of further implications of this view--and problems with
the well-known relational characteristics of segmental phonemes in a phonological
matrix--I must leave to another publication (now in process). This much, how-
ever, should indicate to the reader the deep theoretical significance of a
special phoneme such as the downward influence /!/.
 -145-

6.3. Segmental Phonemes

There is not space in this report to discuss at length the phonetic mat-
ters faced in the workshops in Ghana and Nigeria. Nor is it necessary to
sketch the general picture for West Africa, which has been done so well, and
9
so recently, by Peter Ladefoged. Here, therefore, I shall mention only a
few special problems.
6.3.1. On the Phonetics of Vowel Harmony
10
Working closely with Professor John M. Stewart on Twi, I was interested
in attempting to suggest classroom drills which would help students learn to
hear and produce the difference between "close" versus "open" varieties of
vowels. These sets are involved in vowel harmony, which is very extensive
0
El throughout the area of West Africa.
Stewart and I both operate on the assumption that it is the relatively
forward versus back position of the root of the tongue -no` height of the
tongue dorsum--which is crucial to the contrast between, say, [i] versus [0,
[u] versus [u]. Stewart's paper emphasizes the need for an added dimension--
tongue -root fronting--as making the [i], [u] set marked; versus the other set
unmarked. Linguistic morphophonemic problems of assimilation, he points out,
could not be neatly handled otherwise. He also discusses at length the relevant
literature.

My material, on the other hand, will--in a separate paper, in preparation-

show the relation of exercises for vowel production to the exercises formerly
11
published in my Phonemics (1947). Spectrograms by Peck and Brend, analyzing
both Ashante Twi data (Mr. Denteh, speaker) and samples of my artificial drill
material, show some of the kind of modification of height of the first formant
which might be expected by the postulated changes in the throat cavity.
6.3.2. General Phonetics
Various specific problems of detail were met, during the project, which
need further study in relation to the dynamics of larger phonological units.
u
The relation of the phoneme to the syllable: In Bette [kuk o] /kuk o/
'road' appeared dissyllabic in contrast to the trisyllablic [uku.o] 'bowl'.
Here problems of kwV versus kVV--long known--continue to require careful
attention in analysis.

Another problem related to vowel and syllable is apparently the contrast

between a downglide, fast, on a single vowel versus the same kind of glide,
slow, spread over two vowels: /g/ versus /aa/. Perhaps Abua furnished samples
 -146-

with obiim 'frog' versus oyil 'priest'. Such elements, it should be clear,
imply the need of careful handling of rhythm statements where--as in Abua --
some degree of evenness in length is found for feet with differing number of
syllables, or for syllables (in relation to feet) with differing number of
rowels (see §6.1).
A problem of certain consonantal contrasts warrants continuing study:
In Abua, the Gardners provide samples where I seem to hear phonemic contrasts
between flapped versus nonflapped [1]. The flapped one appears to be shorter,
weak, possibly shortening the preceding syllable; the nonflapped one appears
to be longer, perhaps fortis, and affecting (lengthening), in its turn, the
syllable before it. Note (with [1 as the longer, fortis, nonflap types,
[1] as flapped):
[O1.] 'bangle', [ola] 'intestines'.
For Agbo, Dr. Bendor-Samuel has in preparation, with Klaus Spreda, an
article treating in more detail the problem of fortis consonants there.
6.3.3. On Data Collection Preparatory to Workshop Analysis
For any report such as that of the preceding section, the experienced
analyst may retain a certain amount of reserve. Sample, isolated illustrations-
no matter how contrastive and relevant they may appear--sometimes disintegrate
under closer scrutiny. A supposed minimal pair, for example, may in fact con-
tain a factor unnoticed, which makes the pair nonminimal and cuts all support
from the analysis.
The reader may very well wonder, therefore, what kind of basic, solid
tests are made in workshops such as those reported here, to minimize such
sources of error in the phonemic work of some of the analysts who are in the
early stages of investigation. And how, furthermore, can the data be checked
rapidly--since time is limited--and problems isolated which warrant study in
depth.
The approach--which I first applied extensively, in this particular form,
in workshops in Peru in 1955-56--calls for the lining up of data in a specific
format: Sounds which are similar, and subject to confusion in analysis (giving
'suspect pairs' of sounds) are in each instance, for example, lined up with
hypotheses as to their contrast or as to their complementation, and accompanied
by contrastive word lists--and preferably with tape recordings. Segment se-
quences (like ts) which in general experience are known to be unit phonemes in
some languages, but sequences of phonemes in others, are required to be listed
 -147-

with arguments pro and con, supported in turn by data. Distributional patterns
and lists--necessary for decision as to consistency or proof of statements of
complementation or contrast are added.
Charts of phones and of phonemes show the patterned development of the
analysis. Frames, for tone contrast, also lead to illustrated statements of
morphophonemic rules and patterns of tone behavior.
12
With data before him in this form, the consultant is able to check and
utilize the underlying data rapidly--far more rapidly than he can do so from
the standard professional article on the same data--and suggest further work.
The body of this report does not allow a full illustration of this type
of data collection. But, since it is crucial to understand the technology of
13
the workshop, I have included in an Appendix a section on Agbo, by Klaus
ci
and Janice Spreda.
6.4. Intersection of Voice Quality with Gesture
Miss P. M. Revill attempted to go beyond routine linguistic analysis
in Ilbembe, by studying features of voice quality, speed, and rhythm with the
expression of emotion.
To her surprise, however, it became evident that the contrast between cer-
tain strong emotional states was in part carried by verbal cues and in part by
nonverbal ones. It was only by the intersection of vocal and gestural systems
thatsome of the emotional states could be recognized and differentiated.
Specifically, for example, both anger and surprise were breathy, fast,
and with low pitch on the final phrase. Yet portrayal of anger included fast
nods of the head, with eyes rolling and flashing--characteristics not present
with surprise. On the other hand, a surprised exclamation was accompanied
by a quick movement of the hands, with palms upwards, whereas anger was accom-
14
panied (contrastively) by fast pointing gestures.
For the details of this report--still in its preliminary stages--see the
Appendix.
 -148--

FOOTNOTES
1Kay Williamson, in A Grammar of Tjo Cambridge, 1965, p. 26, suggests, for
the Kwa language, a similar unit of timing. Within these 'tone groups' morpho-
,

phonemic changes of tone may occur (p.7), determined by the syntax. Tone groups
become part of tone phrases, and of breath groups (pp. 7, 11-12).
2"Tonemics, Morphotonemics, and Tonal Morphemes" General Linguistics, 4,
1959, pp. 1-9.
3
In A Grammar of the Asante and Fante Languages Called Tshi [Chwee, Twi], U
Basel, 1875 (republished in 1964 by Gregg Press Incorporated, Ridgewood, New
Jersey).
4
E. g., in A. N. Tucker and J. Tompo Ole Mpaayei, A Maasai Grammar, London,
Longman's Green and Co., 1955, p. 172.
5
In J. M. Stewart, "The Typology of the Twi Tone System," with comments by
P. Schacter and W. E. Welmers, preprint from the Bulletin of the Institute
of African Studies I, Institute of African Studies, University of Ghana, n.d.
See also,--amongst others--D. W. Arnott "Downstep in the Tiv Verbal System,"
African Language Studies, 5. 34-51, 1964, and F. Winston, "The 'Mid Tone of Efih,"
African Language Studies, 1. 185-92, 1960. Winston's phonemic treatment of the .

downstepping influence is of special interest.

6
Welmers has emphatically pointed this out--as in Stewart (reference in fn.
5), p. 54.
7
For a comparable situation of overlap in classical phonemics, see
before n But before r
li
e

r
a

in Danish in Andre Martinet, 'Ou en est la

phonologie?" .1111E2 1, 1948, p. 43.
For explicit discussion of this kind of problem in the analysis of tone,
see my Tone Languages, Ann Arbor: University of Michigan Press 1948, especially
p. 57 (for graph) and pp. 61 and 56 (for insistence on adequate procedure to
keep from being confused by change of general height (vs. phoneme height) from
one syllable to another. Welmers, in discussing rather rigid tone 'bands'
(see the Stewart article referred to above, in fn. 5), seems to have overlooked
these sections of my material.
 AP.

-149--

For further discussion of the more general problem in terms of typological

comparability of subsystems of phonemes, see my Language, Glendale [now Santa Ana],
California: Summer Institute of Linguistics, Part II, 1955, §§8.33 (7), 8.34,
8.441, 8.442.
8
See, for my earlier struggle with comparable problems, the 'zero word' of
Mixtec, in my Tone Languages, p. 82.
9
In A Phonetic Study of West African Languages, an Auditory-Instrumental
Survey, Cambridge: University Press, in association with the West African Lan-
guages Survey, 1964; West African Language Monographs 1.
10
Whose paper on Tongue Root Position in Akan Vowel Harmony was read at
the Sixth West African Languages Congress, Yaounde, Cameroun, March, 1966.
11
They cite H. K. Dunn, in "The Calculation of Vowel Resonances, and an
Electrical Vocal Tract" The Journal of the Acoustical Society of America, 22,
.0.1111

1950, pp. 740-53.

12
Or a later scholar or pedagogue who wished to work with informants in
the same language.
13
Some otners are available in field reports of the Summer Institute of
Linguistics. See, for example: John Callow, Collected Field Reports on The
Phonology of Kasem, Accra, The Institute of African Studies, University of
Ghana, 1965.
14
One of the astonishing features of the whole workshop was the sight of
a very quiet, shy, English girl transformed by the adoption of rolling eyes,
and 'voice flashing fire'--in African "translation". No more striking illu-
stration of the effective - -and culturally determined--role of such cues has
ever come to my attention.
 CHAPTER VII: IMPLICATIONS OF MATRIX APPROACH IN
PEDAGOGICAL PLANNING
How can a person effectively learn a language, preparing his own textbook,
when he has neither teacher nor analysis available? This question was relevant
to some persons in attendance as guests at part of the workshop. Their inter-
ests lay in techniques of language assimilation rather than in those of analysis.
In order to lead them to see the kind of problem they must solve, I in-
quired: What don't you know? When, of course, they could not answer this ques-
tion (which might have appeared unreasonable to them), I then asked: If you
don't know what it is that you don't know, how can you build lesson plans to
teach yourself those points?
The implication: (1) There is needed some way of guessing at the pre-
sence of areas of analytical ignorance, so that (a) research can be undertaken
in these areas and (b) drills could be set up for assimilating this data. (2)

In some sense this must be a "bootstrap" operation--i.e., the construction of

U
drills for assimilation cannot wait for a full analysis; analysis and assimila-
tion need to proceed together. The problem is to demonstrate how this can be 8.
done at all, and--preferably--how it can be done smoothly.
The start: Memorization and simultaneous actin& out of dialogue in cultural
context. Language behavior and nonlanguage behavior must be learned as a unit
(see paralinguistic tie in §6.4). A universal genetically determined seems to
be the fact that human 'computers' are designed to learn language as part of
other behavior--perceptual as well as social, etc.--and function very inefficien-
tly when language is abstracted from life. My belief: Language neither begins
nor ends with or within a context of sentences, but within a setting much larger
than language itself, with perceptual contrast and linguistic-structure contrasts U
learned together. But the 'learning-together' is not the learning-at-the-same-
time of disparate, disconnected, logically-separate events, patterns, rules, or
habits. Rather, the togetherness is a logical dependency, the mutual entwining
of axioms, definitions, and patterns: A thing cow as different from the thing
B
horse may be learned at the same time as the difference between the words "cow"
and "horse"; neither concept nor word can be experientially, logically, or
fl
linguistically completely disentangled the one from the other. (Note: I am not
saying that language determines behavior, nor that behavior determines language,
but that nonverbal and verbal behavior are in some sense a single package, with
partially independent and partially dependent relationships.)

-150-
 -151-

Such a behavior-integrated language view requires the discourse to be

treated as more basic than any isolated sentence, since it is through the
larger context that nonverbal and verbal behavior are entwined. Communication
(as in the paralinguistic data of §6.4) involves both. The question: Does
one know how to act in typical situations?--typical here referring to genetic-
, -:
expectable behavioral universals of eating, socializing,
etc.--can therefore substitute in a larger sense for the question: What don't
you know?
The transfer to linguistic search is immediate: Have I found linguistic
structures (which must be there, for reasons just given) which allow me to
discuss the needs or culturally-prescribed techniques of eating, or of dis-
course itself (with the presence of language as metalanguage being part of
the genetically-prescribed and expected language potential)? Or: Have I
found the means for expressing relations between the dramatis personae of a
'plot' (i.e., of a situation) in linguistic terms? Or: What are the lin-
guistic correlates of actor, action, benefactee, etc.?
Since the presentation of these dramatis personae can be equated with
components of linguistic structures, and more especially with tagmeme units
(though also through lexical terms--as we have just eivloyed the term "bene-
factee" for example), it follows that early search for crucial areas of be-
havior-ignorance can be achieved, heuristically, by an attempt to match guessed-
at plot-universals with known linguistic devices.
When a gap appears--say the inability to express the instrumental in the
target language--the analyst searches (through elicitation by translation, or
through building tales, or by hunting through a recorded corpus) for the ex-
pected presence of the component in various linguistic levels, areas, structures.
The first guess, for the place of expression of some components, will de-
pend in part on the cultural accident of the analyst's background. Both his
native language and his linguistic sophistication and biases will color his
early guesses. For the linguist with tagmemic tools available, however, effect-
ive guessing emphasizes two parts:
(1) He will attempt to get basic clause structures, in the forms of a
syntactic paradigm. (See §1, and note the close resemblance in content to
the base component which might be sought for by a transformationalist under
1
comparable circumstances--a base component which since 1965 includes
some explicit reference to syntactic features which begin to correlate with
 i

11.0

0
-152-

the semantic component of a tagmeme, and with the vectors of a clause matrix
of the tagmemicist.) Some plot relations will be found expressed as simple
expansions of (optional tagmeme additions to) the basic clause patterns. Others
(like the instrumental in Twi and Vagala, §1.1.1 and §2.2.4) may only be ex-
pressed by moving to structures on a different level.
B
(2) The learner will also attempt to set up matrices of clause types
(and of other constructional levels). Irregularities here--holes in the
pattern, or dangling, unintegrated bits--suggest search for the possibility of
regularity--that is, for example, if a declarative statement is found both
independent and dependent, but interrogative is listed only as independent, the
analyst might well search for a possible dependent interrogative.
Lesson plans, then, can be built from the beginning of a knowledge of how
to play-act and word-act. Discourse-in-play-act-context can be memorized (in
standard fashion) before analysis is finished. Drills (also in standard fashion)
can be set up using sample clauses, phrases, sentences from the discourse as
frames for substitution drills. Expansion drills can exploit the addition of
optional tagmemes of a construction. Discourse drills can be patterned on
minimum speaker exchange (e.g., standard question-answer techniques). Deriva-
tion of complex from simple matrices, when known, show transfer of dramatis
2
personae plot relations from one structure or level to another (or, in other
theoretical terms, transformation drills can be set up).
The advantage in pedagogical presentation, here, is the potential for a
linguistic bookkeeping which allows one more easily to see just how much has
been presented, relative to the partially-known system in each lesson, and in
the sequence of lessons. If one keeps in an appendix a set of syntactic para-
digms plus a set of matrices, each paragraph of the lesson can be keyed to
the paradigms and matrices, and--in turn--these latter may have for each tagmeme
and for each matrix cell a section number showing where in the text each item
has been presented, drilled, expanded, and related to other structures from dis-
course to word.

Such an index meets the requirement raised at the first of this section.
3
It allows the beginning analyst-learner to know, in principle, what he knows;
to guess at gaps; and to keep track of his assimilation responsibility toward
what he knows analytically. In addition, it would vastly simplify the task of
a textbook writer of a 'second-level' text by letting him see, directly, some-
thing of the structural rationale and developmental sequence of the data inherited
from the writer of the first level.
o -153-

Q FOOTNOTES
1
Noam Chomsky, Aspects of the Theory of Syntax, Cambridge: The M. I. T.

o Press, 1965.
2
See Kenneth L. Pike, "Discourse Analysis and Tagmeme Matrices," Oceanic
Linguistics, 3, 1964, pp. 5-25.
D
3
Herbert Stahlke, of the Evangelical Lutheran Mission, attempted to apply
these principles in carrying on his learning of Yachi. He made available to
D
me his report of the experiment, which confirms the usefulness of the approach
under field conditions.
D

a
D

0
a
 CHAPTER VIII: SUMMARY
I
Chapter 1; Studies are reported for certain Niger-Congo West African languages:
Kasem, Vagala, Sisala, Dagaari, Bimoba, Basare, Twi, Bariba, Degema, Engenni,
Igede, Izi, Abua, Mbembe, Agbo, Bette, Etung, Yachi. (Secondary sources are
used, also, for Bobangi [Bantu] and Hausa [Chad, Afroasiatic],).
Clauses may be basic, or derived, described in reference to tagmemic con-
trasts, variation, and distribution.
Chapter 2: Clauses differ according to the place they occur in the sentence.
In addition, clauses enter into specialized, characteristic clusters (serial
clauses), within these sentence parts. Clauses in clusters exhibit special
variants, loss of tagmemes, and co-occurrent restrictions. These processes
lead to semantic specialization of subclusters; development of verb phrases;
and classes of verbs seen as in a state of change. Some tagmemes and some
clause variants expected by the English speaker to occur within the separate
clause are expressed only within the clause cluster or subcluster. (See,

also, §1.)
Chapter 3: Sentence clusters (= paragraphs) show, on a higher level, structural
restrictions and relations analogous to those on lower level clusters. Nuclei
of paragraphs (topic sentences) are structurally analogous to the independent
clause of a sentence. Detailed study of discourse will apparently show similar
kinds of structure, of which only the first bits are reported here.
Further kinds of discourse-tie involve the intricate relation of prescrip-
tions for the choice of direct versus indirect quotation according--for example- -
to "on-stage" versus "off-stage" focus, status of speaker (chief versus commoner),
involvement of dramatis personae (as speaker versus addressee, for example).
Chapter 4: Noun phrases, like clauses, can be described and compared in type
by techniques of presentation in syntactic paradigm. Some noun phrases, like
clauses, can be seen in process of change to a lower-level construction.
Matrix treatment of nouns allows irregularity to be seen as distortion of
a simple field structure, and lays the ground for comparison across languages.
Concord of noun to other tagmemes of a clause, seen via matrix, sometimes
shows ordered regularities of a ranking type.
Chapter 5: Matrix presentation of verb structures allows the clear exhibition
of highly complex rules of fusion by ordered change from simple matrix to de-
rived matrix.

-154-

ri
B -155--

n Chapter 6: Phonological study, to show the deeper underlying characteristics

U
of these systems, must include relation to feet, in dynamic sequence (often
in isochronic units with a varying number of syllables).
The tone systems of the area are observably in a state of change. In

many of them a new, extra-high tone can be seen in a limited set of special
contexts (e,g., the negative). Otherwise, basic, contrastive tone heights
vary in number from two to four.
A special phoneme of process--a lowering (or "dowlstepping") influence- -
is characteristic of the region, and leads to a modification of phoneme theory
to accommodate phonemes of particle (or item), of wave (or process) and of
field (or relation). Segments like s, o, 1 would be phonemes of a particle
type; high versus low tone would be relational elements of field; and the /!/
lowering influence would be a phoneme of process (distinct in phonological
result from a hidden morphophonemic influence). Combinations of the three
types, plus morphophonemic symbols, suggest special results for internal re-
construction of tone chains.
Segmental phonemes involve various problems--including the role of the
throat in vowel harmony.
Phonological data can be presented, for testing a preliminary analysis,
in a convenient summarizing format.
Vocal quality comprises one component of certain cues to emotional
states. These ambiguities can be resolved only by the intersection of
these cues with gestural ones.
Chapter 7: For the learning of a language before the analysis of that lan-
guage is well along, tagmeme and matrix techniques allow for a bootstrap
operation in which learning and analysis proceed simultaneously. Language
is treated as but one component of communicative behavior. Discourse oper-
ates in a setting of action, and standard drill types exploit data from dia-
logue. But preliminary syntax paradigms allow the early use of substitution
drills, while suggesting search-leads for expansion of patterns of constructions.
Preliminary matrices of basic and derived constructions, on the other hand,
allow search for analogous constructions implied by the presence of holes in
the pattern. In addition, the combination of the two allows the construction
of an index matrix which permits the textbook writer to keep track of the
areas taught and those still to be taught--whether in a first text or on an
advanced level. Thereby the learner can answer the question 'What don't you
know?'--and set the stage to learn; while the textbook writer can discuss, in
reference to his own efforts or those of a predecessor, 'What remains to be
added before my task, as I define it, is done?'.
 or
Appendix I

NOTES ON THE MBEMBE CLAUSE SYSTEM -

A PRELIMINARY ANALYSIS

Kathleen Barnwell

Basic Theory:
(1) Any unit of purposive human behavior is well-defined if and only if one
describes it in reference to
a) contrast (and resulting identification)
b) range of variation (with its essential physical manifestation)
c) distribution (in class, in hierarchical sequence and in systemic
matrix) "

K. L. Pike 'On Systems of Grammatical Structure'

This study of Mbembe clauses, therefore, aims to describe the clause unit
with reference to contrasts of internal structure (the presence or absence of major
places and differences in the classes which fill them) and of transformations,
to variations in form (alloclauses) which occur in certain environments or which
are controlled by factors of emphasis or style, and to distributir within the
clause system and within the grammatical hierarchy, primarily within the next highest
level, the sentence.

(2) Halliday observed in 1961 that the transformational approach is a 'valuable

supplement' to, though not a replacement for other approaches to description.
(M. A. K. Halliday 'Categories of the Theory of Grammar?).
Longacre points out, in a recent article which demonstrates how tagmemics
c:in gain valuable insights from transformational grammar while defending the use-
fulness of function-set notation and of the matrix approach, that "Transformational
relations are not the only parameters which relate constructions. Rather transfor-
mations belong to a system of relations which includes non-transformational parameters
as well." (R. E. Longacre 'Transformational Parameters in Tagmemic Field Structure').
On this principle Mbembe clauses are described both by paradigms showing the
major places and fillers of the basic clause types, and by rules for the transfor-
mations of these basic clause types, the transformations being presented in matrix
perspective where this is useful, thus cutting redundancy in the description.

(3) Longacre's 'dual criteria'principle has been applied for the separation of
clause types. Clause types are separated if they contrast in two respects either

-156-
 -157-

in internal structure, or in potential transforms, or in distribution, or by one

such difference substantiated by the pressure of the system.

B Note: Thit description is limited to major clause types, which comprise about
95 per cent of the text examined, so far.

General Outline:

I Structural contrasts II Chart of primary III Chart of secondary

of basic clause types [widely applicable] [less widely appli-
transformations + cable] transforma-
rules tions + rules

Citation paradigm Citation Citation

of basic clause types paradigm paradigm

Detailed comment on Examples from Examples from

separate clause types text material text material
and examples from
text material.

IV Distribution of clause V Variants of VI Variants of clause

types in the grammatical clause types types determined by
hierarchy determined by emphasis or style
environment

{Serial constructions
VII Close-knit clause
Result-clause composite
composites
Introductory. clause composite

VIII Outline of sentence

structure

IX Illustrations of negative
formations

Bibliography

I STRUCTURAL CONTRASTS OF THE BASIC CLAUSE TYPES

The basic clause types are classified on the basis of distribution in the sen-
tence as Independent, those which may function as simple nucleus of a sentence,
and Dependent, which may notso function.

I
 -158-

Nine Independent and two Dependent clause types are set up. These show
structural contrasts in:
a) the class of verb-roots which may function in the verbal phrase filling U
the Predicate place
b) the relationship of the fillers of the Complementation places (charted
under the labels Object, Indirect Object, Complement and Goal) to
the other components of the clause
c) the restrictions on the classes of fillers which may occur in various
clause slots.
The Intransitive, Transitive, Ditransitive, Directive Motive, Stative and
Independent Introductory (1) clause types have an optional extension of the verbal
phrase which occurs after the Complementation places (or, if no Complementation
places occur, immediately following the Predicate) and which consists of a redup-
lication of the verb functioning as head of the verbal phrase. e.g.

bte 6ci eten Oci lit. 'father eats meat eats'

Marginal places All basic clause types may be expanded by optional
marginal places which may be diagrammed:

- Ex. 3:temporal/ + Nucleus Ex. 1:prep.p. - Ex. 2:adverb/

time phrase/ emphatic pronoun/
particle/ temporal/
modifier/
Expansions 1 and 2 do not occur in Dependent clauses.
Fillers of places in the Clause
Since this description is limited to the clause level a brief summary is
given at this point of the nature of the fillers of these places. In a description
of the whole grammatical hierarchy reference would be given back to phrase or word
level.
Fillers of the Subject and Complementation places comprise:
nominal phrase
The most common types of nominal phrase are:
+ noun - (+ demonstrative particle) e.g. eten 'Ado sa
'meat that there' tiV
+ noun - possessive eten 'cC
'meat his!
+ noun
+ +
- (-
.
min:a + numeral) an50 min:a afa
'people about two'
+ noun qualifier 51150 kpengokpen
'person every'
+ numeral Afa
'two'
 -159-

CHART I shows the structural contrasts of the Nucleus of the Basic Indicative
clause types.
Complementation
U
SUBJECT I PREDICATE I.O. ORJrCT comPumm COAL

+
- n.p. Intransitive v.p. - - -
I n.p.co
pro.

N
+
n.p.(p) Transitive v.p. - n.p. - -
n.p.co(p) n.p.co
pro. pro.

+ + +
E - n.p.(p) Ditransitive v.p. - n.p.(p) - n.p. - -
n.p.co(p) pro. n.p.co
pro. n.p.co(p) (pro.)
P

+
- n.p. Copulative v.p. - n.p.(p) - + n.p. -
E n.p.co n.p.co(p) n.p.co
pro. pro.

N
+ +
- n.p.(p) Directive v.p. - - - - n.p.
n.p.co(p) n.p.c.
D pro.

E - n.p.(p) Motive v.p. - + n.p.(p) - - n.p.

n.p.co(p) n.p.co(p) n.p.co
pro. pro.
N
4-
- n.p. Stative v.p. - - - -
n.p.co
T pro.

+ +
- n.p. Initiative v.p. - Complementation + verbal noun
n.p.co M. /
pro.

- n.p.(p)
+
Introductory (1) - n.p.(p) + n.p.(c) - -
n.p.co(p) n.p.co(p)
pro. pro.

D +
- n.p.(p) + +
E Introductory (1) - n.p.(p) - n.p.(c)
P n.p.co(p) n.p.co(p) - bt + speech/
E pro. pro. clause
N
D +
- n.p.(p)
+
Introductory (2) - n.p.(p)
N n.p.co(p) n.p.co(p) -
T pro. pro.

n.p.(p) and n.p.co(p) indicates that the head of the construction filling this
place must be a personal noun. However in fable style a non-personal noun may
also function here.

n.p.(c) Indicates that only a very limited group of nouns, which are semantically
cognate with the preceding verb, nay fill this place. e.g. gbA:ga 6d1 'speak word'
 -160--

nominal phrase composite

The most common nominal phrase composites are
n.p. + n.p. in appositional relationship
n.p. + link + n.p. in co-ordinative relationship
n.p. + relative clause/relative phrase in subordinative relationship
pro. + n.p. or n.p. + pro. in appositional relationship t
pronoun
The filler of the Predicate place is a verbal phrase with a verb of the
appropriate class as its head. The verbal phrase may be diagrammed:
+ r
- tense prefix aux 2 aux 1 + item + verb....- verb
e.g. Mbira nd nsi 'I also ? do'
aux2 auxl verb
'I will do again'
Expansion 3 place may be filled by:
time phrase In structure a time phrase is similar to a nominal phrase but a time
word (e.g. ewa 'day') always functions as its head. It may be expanded
n
by a relative clause.
e.g. awa '636 c'imakwd ma
'day that which they-came here'
temporal e.g. mada 'today' wendo 'long ago'
particle e.g. ma 'here, now'
modifier e.g. wura 'implies past' mina 'implies about to happen'
Expansion 1 place may be filled by:
U
prepositional phrase which may be diagrammed:
+ prep. - n.p./ (+ prep - n.p./)
n.p. co/ n.p.co/
pronoun/
In the full expansion the second place is very commonly filled by a n.p. or n.p.co B
with a personal noun as its head, or by a pronoun.
e.g. k ekw5:r k 5s3:m
'to tortoise to house'
'to tortoise's house'
Expansion 2 place may be filled by:
adverb e.g. matqhig 'plenty' mapyir 'completely' Cbargfiar 'quickly'

emphatic pronoun e.g. kalkam 'I alone' (reduplication of pronoun)

temporal
 -161-

Chart gives a Citation paradigm of the basic clause types.

Minimum forms of the clauses are given with the fewest possible changes of the
lexical items in order that the contrasts between the clause types should be
highlighted.

A full description would include notes on each of the clause types giving
further details of the classes of the fillers and any restrictions in their occur-
rence not shown on Chart I, and transformations or details of distribution relevant
to the separation of clause types, notes on frequency, and examples from text ma-
terial to illustrate the clause types and also to prove their validity from unelic-
ited forms. In order to restrict the length of this paper only the Transitive and
Directive clause types are here so described.*

CHART II Citation paradigm showing contrasts in the nuclear places of the basic
clause types
Clause type Subject Predicate Complementation
Intransitive Cite 6460"
father lingers
.t
Transitive Cite oti eten
eats meat
Ditransitive Cite Ago twa eten
gives child meat
Copulative Cite 6he 644:n50
is chief
.1.
Directive 'ate °yin eten
father goes meat (to fetch meat)
.1
or Cite °yin gPYg
goes market
t,
Motive ate 5to:ma jwa eten (to fetch meat)
sends child meat
Stative Cite 6311, 63111
is good rich
Initiative Cite 45m eten Oc=ci (- indicates open
begins meat to-eat transition)
Ind. Introductory 1
Cite 5ba:ga jwa Odig
talks child word
says
Dep. Introductory 1
.
ote ogba:ga 6we (±be)
II
says child that "
Dep. Introductory 2

4
Cite 6n
ob.! jwa
II
says child that "

*Illustrations of the negative are given in Section IX.

t.
 -162-

Detailed comment on se arate clause es with examples from texts specimen

Transitive clause type
1) About 60 per cent of verb-roots are classified as transitive.
ii) The Object place is very commonly filled. Where it is not filled the Object
referent is nearly always either explicit in the linguistic context (e.g. in a
serial construction the object referent may be explicit in a preceding clause)
or implicit in the non-linguistic context. In Conversation style discourse, where
the linguistic and non linguistic contexts are more closely interrelated than in
Narration style, the object is quite commonly implicit in the non-linguistic context.
iii) Transitive clauses may be transformed, without any apparent change of meaning,
by the use of the verb t5ga 'take' into a clause serial construction and quite
frequently occur in this form.
e.g, bre oci 6ten may be transfoilled into the serial construction
(clause group) (see section VII)
Ote 5t5ga eten Oci
father takes meat eats

This potential transformation further distinguishes the Transitive clause

type from the Directive which may not be transformed into a serial construction.
iv) The Object and Expansion 1 places are not very commonly filled in the same
clause, especially where one is filled by an expanded phrase composite, although
examples do occur in text. This is avoided by the use of the same serial trans-
formation.
e.g. oyiga ijb:g na 6kpa may be transformed into the serial
he-put snake construction
in bag
0 Ex.l
bt6ga ijb:g Oyigg na 6kpa
he-took snake put in bag
v) The transitive clause may function as the first element of a Result clause
composite (see section VII). This further distinguishes it from the Directive
clause which may not so function.
Examples from text
oyena eten kW (20.2)
he has meat that alone
S P 0 Ex.2
pro tr n.p. emp.pro.

all b:aba. Oyer' eilaiia

All Baba has a horse
S P 0
tr n.p.

El
 -163-

m Eta / bgb (20.1)

he-will him he-kill
-shoot
P 0 / P where the object referent of the second
tr pro tr clause is explicit in the linguistic context.

n Oddoa 6ten101,5 '6w6 fia 'Wg / 5t5ga / 61/44sa 64,6 kWe

he-picked bone- that on ground he-took he-put in bag his
Li -up
P 0 Ex.l. / P / P Ex.l.
tr n.p. prep.p. tr tr prep,p.
where the object referent of the second and
third clause is explicit in the, linguistic
context. The last two clauses illustrate a
serial construction transformation of a
transitive clause.
Directive clause type
i) About 2 per cent of verb-roots are classified as directive, the most common
being yin 'come, go, fetch' and tá 'go'.
ii) The Directive clause is further distinguished from the Transitive clause by
the fact that the potential transformations described for the Transitive clause
(above, iii and iv) may not be applied, and that it may not function as the first
element of a Result clause composite.
iii) Where a place name (e.g. Enugu) or a noun indicating a place e.g. gpya 'market'
fills the Goal place, a preposition may be inserted before the noun thus transform-
ing it into a prepositional phrase.
e.g. 3Va may be transformed to bca sa gpyg
gpyg
he-goes market he-goes to market
P Goal P Ex.l
dir n.p. dir prep.p

iv) Either the Goal place or Ex.l. place is almost always (perhaps) filled or is
explicit in the linguistic context. An exception to this rule is the common use
of the verb yin to mean 'to go to farm'. In this use the Goal place is not filled.
Examples from text
oyin Ar5b Irts5:0 (1.4)
she-goes mushrooms in evening (she goes mushrooming...)
P Goal Ex.l
dir n.p. prep.p.
..6yin 1357'd (19.1)
he-goes market
P Goal
dir n.p.
egbaji (18.3)
they-go hunting
P Goal
dir n.p.

MT'
 -164 -

II PRIMARY TRANSFORMATIONS OF THE BASIC CLAUSE TYPES

The basic clause types may be transformed by three dimensions of contrastive 8
coordinates:
i) Mood ii) Aspect iii) Neg /Pos

Indicative (basic) imperfective (basic) positive (basic)

Imperative perfective negative
Subjunctive
Interrogative
Subordinate
Relative
Of these, Mood transformations are considered primary because they are relevant
to all basic clause types and because they are relevant to the statement of restric-
tions on the distribution of clauses, which the other dimensions of coordinates
are not.

CHART III shows the possible transformations of the basic clause types into
contrastive moods

Indic. Im er. Sub . Inter. Subord. Relative

Intrans. X X X f X X X

Trans. X X X X X X

Ditrans. X X (X) X X X

Copul. X (X) X X X X

Direct. X X (X) X X (X)

Motive X (X) (X) (X) (X) (X)

Stative X (X) (X) (X) X X

Initiative X X (X) X (X) X

Ind.
Intro.l X X X X X (X)

Dep.
Intro.l X X (X) X X X

Dep.
Intro.2 X X (X) X X (X)

. -

(X) indicates that this form has not yet been found in text material, only
in elicited forms.
 -165-

Rules for the transformation of basic clause types into different moods
The points in which the Indicative mood contrasts with other moods are summarized.
Transformations of other moods are described in relation to contrast with the
B
indicative.
Indicative i) Has the order Ex.3 S P Complementation Ex.l Ex.2
ii) Obligatory presence of order 1 verb-prefixes (person)
iii) Optional occurrence of order 2 verb-prefixes (tense) and of
verb-auxiliaries within the verbal phrase
iv) Occurrence of indicative tone patterns
Imperative i) Obligatory absence of Subject place
ii) Obligatory absence of person and tense verb- prefixes
iii) Occurrence of imperative verb-prefixes zero 'sing'
ma- 'plu.'
iv) Occurrence of imperative tone patterns
Subjunctive i) Optional presence of particles c' or bg preceding the Subject
or Predicate place
ii) Occurrence of distinctive tone patterns
Interrogative i) Occurrence of an interrogative class word in the Subject place,
or in any one of the Complementation places, or in Ex.l place
Subordinate i) Obligatory occurrence of clause conjunction Examples of clause
conjunctions are Ode 'if' gra 'before' mink kw' 'like'
ii) Distinctive tone patterns (= same as subjunctive)
Relative i) Obligatory presence of a relative particle (in concord with the
noun which the clause qualifies)
Interrogative, Subordinate and Relative clauses are further distinguished by
differences of distribution. These are described in section IV.

CHART IV Citation Paradigm for the transitive vector of Mood transformations

[= second row of Chart III]
Indicative 'ate eten
oci
father eats meat
Imperative ci eten
eat meat!
Subjunctive c' (Ste OCi eten
let father eat meat
Interrogative am Oci eten or Ote OCi Mb60
who eats meat? father eats what?
Subordinate 6de Ote Oli eten (m5d5:b)
if father eats meat (he-will-be-satisfied)
Relative 51150 kw' OCi eten (m5d6:b)
someone who eats meat he-will-be-satisfied
 PRECEDING PAGE BLANK- NOI FILMED -167-

Examples from text

Imperative t5ga aka kw5 / time / itáng (68.3)
take mother your take go
B P 0
/ P
tr n.p. tr tr
mege M 5t5g (11.4)
give me thing
P I0 0

ditr pro n.p.

ben e (be) (11.3)
tell him that
P I0

intro 2
Pronoun
1

Subjunctive c' 6kp5ga min:A (12.2)

let it-end so

P Ex.2

intr adv.
Interrogative c' 51'150 kw' ape 6Bekwe 6niog men -o (126.5)
grave of person who died newly it-is where ?
- - S. - - - - P Ex.1
n.p.co intr interrogative
...5gben5ke' kwg 6kpeng yen (128.3)
brother his is-called what?
S P Comp
iQ
n.p. cop interrogative
6w6 ma 6kel, áîíì -o (85.1)
this here meets whom ?

S P 0

n.p. tr interrogative
Subordinate 6de aben bg/ Ayin 5t5g ( 4 iken yin (2.1)
if you-want that you-go thing early go
conj P P Goal Ex. P
intro2 dir n.p. prep.p. dir
isa 6yog Okwil / 5ta e / 545 ) (20.2)
if friend comes he-will he-kill
-shoot him
conj S P P 0 P
n.p. dir tr pro tr
 -168-

Relative 5n50 3d5g5d5g kw' 56wa gad sa / n Okwu ... (21.4)

person other who enters bush there he come ...
rel P Goal Ex.l
.dir n.p. prep
S:n.p. co
wa okpicen50 wane onio -e kw'cla opydm 6t5g... (13.3)

once blindman one was who begged thing

Ex.3 S P rel P 0

temporal n.p. intr tr n.p.

extension of S
6gban6kg' kwg kw' maidrg kastim 6nao guage... (117.2)

brother his whom they-call K. gave her guage

rel P Comp
cop

S: n.p. co P IO 0

III SECONDARY TRANSFORMATIONS OF THE BASIC CLAUSE TYPES

The clauses of the Indicative, Interrogative, Subordinate and Relative
vectors of Chart III may be further transformed into the perfective aspect, in
contrast to the imperfective aspect, the basic form.
The clauses of the Imperative, Subjunctive, Interrogative, Subordinate and
Relative vectors of Chart III may be further transformed into the negative, in
contrast to the positive, the basic form. Indicative imperfective and Indicative
perfective clauses may both be transformed into the negative.

CHART V showing secondary transformations of clause types

positive Negative x indicates

clauses reached
by primary trans-
imper ective x a
formations only.
INDICATIVE
perfe tive b c letters (a,b, etc)
aim to coordinate
IMPE NIVE x d this chr-t with
the rules for
SUBJU ICTIVE x e secondary- trans-
formations which
imper ective x are given in
INTERROGATIVE f paradigm format
perfe tive g in Chart VI.

imper ective x
SUBORDINATE h
perfe tive j

imper 'ective x
RELATIVE k
perfe :tive m
 -169-

CHART VI corresponds to Chart V. Rules for secondary transformations are given

in paradioi form

Positive Negative

Indic. imperf. x a Predicate occurs clause finally.

Negative verb-prefix m- occurs.
Distinctive tone pattern
Final vowel of group 1 on verb.

Indic. perf. b Predicate follows comp- c Predicate occurs clause finally

lementation, precedes margin. Negative verb-prefix k' occurs
clitic occurs before 1st Distinctive tone-pattern.
complementation place/or Final vowel of group 2 on verb.
before Predicate.
Distinctive tone pattern.
Verb-suffix -a.

Imperative x d Predicate occurs clause finally.

Negative imperative prefixes
kb.- and oka- occur.
Tone pattern as for Indic.
imperf. negative.

Subjunctive x e Transforms into a negative

imperative intro. composite
kacer 'do not agree' + indicative
clause.

Inter. imperf. x f Predicate occurs clause finally.

Negative verb-prefix ina-.
Tone pattern as for Indic. perf.
neg.
Occurrence of interrogative word
as filler of S. Comp. or Ex.l.
Inter. perf. g. Form as for Indic.
perf. pos. with substitution
of interrogative word as
filler of S. Comp. or Ex.l.

Subord. imperf. x h Predicate occurs clause finally.

regative verb-prefix ina-.
Final vowel of group 2 on verb.
Occurrence of conjunction.
Subord. perf. j Form as for Indic.
perf. pos.
Obligatory occurrence of conjunction.
Relative imperf. x k Predicate occurs clause finally.
Negative verb-prefix ina-.
Final vowel of group 2 on verb.
Occurrence of relative particle.
Relative perf. m Form as for Indic.
perf. pos.
Obligatory occurrnace of
relative particle.
 -170-

CHART VII Citation paradigm of the secondary transformations showing positive

and negative forms for the Indicative imperfective, Indicative perfective,
Imperative and Subjunctive

Positive Negative
Indic. imperf. Cite eten
aci a 5s5:m Cite eten sa 5s5:m mOci
tather eats meat in house father meat in house won't-eat
x a
Indic. perf. Cite Veten 5cg qa 5s5:m Cite eten sa 5s5:m koci
father meat ate in house father meats in house hasn't eaten
past b c

Imperative ci eten pa 5s5:m eten sa 5s5:m kaci

eat meat in house x meat in house do-not-eat d
Subjunctive Ote eiai eten pa 5s5:m kacer *ate OCi etan sa 5s5:m
let father eat meat in house do-not-let eat meat in house
x father

Examples of secondary transformationstaken from text

a) mina ramose (16.1)

us he-will-not-see
b) tata 61M. k' ij5:g 5d 5b50 sa q'gbg (14.3)

father that past snake caught in bush

k'gPa k k5' mfa (20.4)
I I

you-have-died past you I-have-killed

c) Okpicen50 two 565:g k gkpa k5A5:m (15.1)

blindman that hand in him in bag did-not-put

kam na:rg (10.5)
I

I English can't-hear

IV THE DISTRIBUTION OF CLAUSE TYPES IN THE GRAMMATICAL HIERARCHY

In the description of clauses so far, contrastive distribution has not been
considered. In this section the places in which clauses may function are listed.
Chart VIII shows what classes of clauses may occur in each place.
Clauses function primarily as elements of clause level composites and of
sentences. Clause constructs may also function as an element of the phrase.
Clauses may function in the following positions: (see sentence structure diagram
Section VIII)
i) as simple nucleus of a sentence
ii) as inner margin of a sentence
iii) as outer margin of a sentence
U -171-

n
iv) as 1st element in a result clause composite (see result clause composite
diagram section VII)

v) as second element in a result clause composite

IJ
vi) as primary clause in a clause serial construction (see clause serial
construction section VII)

vii) as secondary clause in a clause serial construction

viii) as first element of an Introductory composite (see Introductory composite
section VII)

ix) as an element of a nominal phrase composite

CHART VIII shows, in general, the distribution of clauses in these places

B
Indic. Imper. Subj. Inter. Subord. Relative Dep. Intro.

i) x x x x
U ii) x

iii) x
iv) x x
v) x
vi) x x x

vii) x x x

viii) x

ix)

V VARIANTS OF CLAUSE TYPES DETERMINED BY ENVIRONMENT

Independent and Dependent clauses have allovariants whose occurrence
is determined by their distribution in the environments summarized above.
In environment v: Clauses in this position cannot have the same subject refer-
ent as the preceding clause. The subject place must have the same referent as
some other place which is explicitly filled in the preceding clause. This must be
apparent either by an explicit filler of the Subject place or by the concord of the
verb prefix.
In environment vii: Clauses in this position have obligatory absence of
Subject place. Their subject referent must be the same as that of the primary
clause in the serial construction. This is indicated by concord of the verb-
prefix with the Subject of the primary clause.
In environment ix: Relative clauses functioning as elements of a noun phrase
composite may occur in one of two relationships to the nouns they qualify.
a) Objective: Only Ditransitive and Transitive clauses may occur in this
relationship. The Object place of a clause functioning in this way is
obligatorily absent, since the logical object is the noun which the

clause qualifies.
 -172-

e.g. may60a 1t5:0 k Ose kw' rita b50a ma

listen ears to story which I-will-tell you here
where the qualified noun ese is the logical object of the relative clause.
b) Subjective: The Subject place of a clause functioning in this relationship
is obligatorily absent since the logical subject is the noun which the
clause qualifies.
e.g. okpicen50 Opyem 5cat5g kw' 6d 6k1ken opyem
blind man begged food who after he-begged
Where the qualified noun okpicen50 is the logical subject of the relative
clause.

VI VARIANTS* OF CLAUSE TYPES DETERMINED BY EMPHASIS

Independent and Dependent clauses have allovariants whose occurrence is
determined by factors of emphasis and style.
Front shifting of the object
The Object of a clause may be moved to the initial position in the clause for
emphasis.
I I-
e.g. obo owo kw wura ote oben be ofo okpicenoo owo iva PE
death that which once father that blindman that children
said he-kill his
0
an50 afa bt n atera ape
people two they instead died
S P
0
tm50 6.0 mapyir 3d 5(5:n min:a (13).

song his all he-sang so

Ex.2

VII CLOSE KNIT CLAUSE COMPOSITES

Certain close-knit clause composites occur:
Serial clause construction
Introductory clause composite
Result clause composite
These only function in positions in which an Independent clause may also
occur. Most commonly they function as nucleus of a sentence. The internal
structure of such composites involves structural restrictions and sequence

* [In the report, changes caused by shifting tagmemes for emphasis are treated
as leading to sets of emphatic clauses, one set for each basic type. K. L. P.]
 -173-

restrictions of co-occurrence. The relationship within the composite is therefore

different from that between the nucleus of the sentence and the marginal places
which involves only obligatory versus optional occurrence.
In the following description these composites are treated as elements of the
sentence, on the same level as independent clauses.
Serial*_Clause Constructions
'Serial clause construction' = series of clauses
'Serial r1Pusel = one clause in series
Serial clause constructions consist of a sequence of clauses which:
(i) Share one Subject, the Subject being optionally explicit in the first
(primary) clause of the series and obligatorily absent in all secondary
clauses.
(ii) Pave verbal prefixes throughout the construction showing concord with
the Subject referent in person and number.
(iii) All occur in the Indicative mood, or all occur in the Imperative mood,
or all occur in the Indicative mood except the first or last clause which
is in the Interrogative mood.
Up to six clauses may occur in one serial. Two, three, or four is more common.
Introductory clause composites may also function as elements of the serial
construction with the restrictions described above.
Chart IX shows restrictions on the clause types which may occur in serial con-
structions and on their co-occurrence.
Relationships within the serial are described in terms of Wave (waves of refer-
ence) and Particle (clause, and clause group units) perspective.
Waves of Reference. Within the serial construction certain places, notably,
Subject, Object and Ex.l (Locational) may be shared by several clauses in implicit
reference, while being explicit in one clause only. The Wave of Reference of the
shared place extends over all those clauses which share the reverent, i.e., which
could be reconstructed as independent clauses in which the same referent is explicit.
The same phenomena could be viewed from a particle perspective in terms of deletion.
e.g., lkwan50 51(4:1,a gAi / owbna / ogwo / Onloa gwOr

woman she-fetches water she-pours she-drinks she-sits seat

S P:tr 0 P:tr P:tr P:tr 0

* [Called clause clusters in the report. K. L. P.

ci
 -174-

The construction could be reconstructed as a series of independent sentences.

0
S ikwan50 51ca:a 'woman fetches water'
ikwan50 OwOna 'woman pours water'

ikwan50 Ogw6 'woman drinks water'

ikwan50 onioa CwOr 'woman sits seat'
ikwan50 is the Subject referent of all four clauses
is the Object referent of the first three clauses
The Subject wave of reference, therefore, extends throughout the construction.
The Object wave of reference extends over the first three clauses. This may be
diagrammed:
lkwan50 5kg:IYa Ow60a Ogw6 Onfoa 6w5r

0-
Clause Groups reflect certain close-knit relationships within the serial con-
struction. They are set up on the basis of:
(a) frequency of occurrence of certain patterns
(b) restrictions of the fillers of certain Predicate places
(c) potentiality of transform
The Transitive group may be diagrammed:
+P:t6ga + 0 P:cga + P:tr - Ex.1
Lima tima

This may be transformed into a single clause P:tr + 0 - Ex.'. There is a tendency

to avoid the occurrence of two non-Predicate places explicit in one clause within
a serial construction.
The Ditransitive group may be diagrammed:
+P:t6ga + 0 + P:ditr +10

Vina
0-

This may be transformed into a single clause P:ditr + IO + 0

The Accompaniment group may be diagrammed:
+P:cfma +0 Plima + P:tr + 0 /
t6ga c6ga + P:ditr + 0
+ P:intr /

+ P:ditr + Goal /
0
 -175-

The Object referent of the two obligatory clauses must not be the same. This
group will not transform to a single clause.
Groups may be, but are not commonly, interrupted by other serial clauses
(occurring before the second predicate).
If the first clause of the group occurs within an Object wave of reference
the Object will not be explicit.
'Clause Auxiliary Verbs'

a Certain verbs deserve comment because of the frequency of their occurrence

in serial constructions. In order to show their place in the system they are
contrasted with two classes of Verb auxiliaries which function within the ver-
C
bal phrase.

- n Aux. 1 functions immediately preceding the verb in the verbal phrase.

u It comprises only one member -d' usually translated 'went did something'. It

never occurs in an unelided form but may be prefixed by the full range of verb
prefixes.

Aux. 2 functions immediately preceding Aux. 1. It comprises a small group

of members of which the most common are bira 'repeat' tuma 'yet' tera 'just'.
In structure Aux. 2 are identical with verbs. They may occur in an unelided
form and may be prefixed by the full range of verb prefixes.
Clause Auxiliary verbs may function as the head of the verbal phrase.
They may be modified by Aux. 1 and 2 and may be followed by the usual complemen-
tation and expansion places. They also function commonly in the serial con-
struction without any complementation or expansion. They are often closely
bound by a subordinating tone pattern to a following verb. In this position
they do not carry a heavy information load. They comprise the verbs
kwu 'come' to 'go'

fona 'go' bina 'get up'

Alternative treatments (1) Treat them as other verbs in the construction;
this would mean that factors of (a) frequency (b) restriction of internal
structure (c) tonal association (optional) with a following verb are not
accounted for in the description.
(2) Treat them as Clause Auxiliaries. In their unexpanded form they may
not function as the head of a clause but are treated as Auxiliaries to the
Predicate. However there are then certain border-line cases where it is not
clear whether a verb oelhis class is functioning as a Clause Auxiliary or as
a free form since other verbs may also occur without expansions.
However the second treatment is felt to be preferable since it accounts
for relevant facts and also reflects parallelism with Verb auxiliaries.

I
 -176-
fl

1;
E.g. oyog ofona Oyfn gpyg (19.1)

friend went came market

ri
S Paux P:mot Goal U

(6d od6ba ecirg:06 / Okwu 6k6 sd 5cat5g) tr. (19.4) fl

he-bought poison he-came put in food
P 0 / Paux P Ex.1
0
(S)

kg' ose oyog sg / obina Ota oyog / ayOg oto / ape (20.4)

he saw friend there he-got up shot friend friend fell dies

S P 0 Ex.2 / Paus P 0 / S P /P Li

Note:- Paux place would be filled by a verb phrase with one of these verbs as head.

Analysis of some sample Clause serial constructions:

/ indicates clause break
tr/ditr/acc
( ) indicates clause group. Clauses which interrupt the
group are marked //c1//
Ditransitive:
(5gban5ka' 614'6 5t5ga 5t5g iv6 / 5na0 ikw.an5o 6w6) (117.3)

brother that took thing that gave woman that

S P:toga 0 / P:ditr I0

0 >
S

Accompaniment:

(5t6ga gn5:ns6 nd6 / Otima / 61,áng k .61,6n) (103.3)

he-took bird's that took went to-village

head
P:toga 0 / P:tima / P:intr Ex.1
0 Inner Margin - -
(114.1)
(S)

all baba 5116:0E / abg sa' Etakpa min:a / ose b / c'fid aseig 610" iby6
7 1 7

Ali Baba moves- he-stands he-sees as they put things those

aside by rock so them
S P:intr P Ex.1 Ex.2 P:tr 0

S
 -177 -
E] Accompaniment; Tr-n.
El btg mgkpgr / (5t5ga OkorO / Ope oyog es6) / (50ga s / ole fia iqeg)
he-went at-once took knife cut friend head took him put in ground
P:dir Ex.2 P:toga 0 P:tr 0 P:toga 0 P:tr Ex.1
0
(S)
On5:0a
(7.5)
he-buried him
P:tr 0

Tran.
(obina 515ga gri5risti'AdO//Ottim epe mApyir // OyigA) (102.4)
he-got- took bird's that dug hole all put-in
up head
Paux P:toga 0 // P:tr 0 // P:tr

Tran.
(gOga aik5mgirido c' wura k' gkima / Agog sg) / (At5ga // gwOria //
you-take hot-water that which you put keep there take pour
P:toga 0 P:tr Ex.1 P:toga P:tr
0
(5)
Accompaniment
Arg:ita ab,A)
(56.1)
you-cook soup
P:tr 0

Introductory Clause Composite

+ Dependent cl bg + Speech/Clause
Where a speech functions as second element this
is direct speech
Where a clause functions as second element this
is indirect speech
A serial construction of which a Dependent clause is the last element may
function instead of a Dependent clause as the 1st element of an Introductory clause
composite.
e.g. obeli. ikwAn30 k* / bg 5kgo Okilg (133.4)
he-told wife his that she-fry yam
Dep.cl be clause (indirect speech)
 -178--

byOg bbC / m6fong (16)

friend said we-go where?
Dep.cl Speech (direct speech)
Result Clause Composite
+ Indep cl + Indep cl
Where the Subject of the second clause has the same referent as one of the
Post-predicate places in the first clause. This must be apparent through the
repetition of the Subject referent or through concord.
0
e.g. mabira Ak615' 661 tido matL'te ma / eci rid6 gbira &to
they-also cut tree that plenty tree that again fell
P:tr 0 Ex.2
bdal) fia 6nana matttt ma / td5:15 (116.2)
he-tied to horse plenty it-heavy
P Ex.I Ex.2
Where prefix of second predicate shows concord with hlaila

VIII STRUCTURE OF THE SENTENCE

A diagram of the structure of the sentence is given in order to show the
places in which clauses may function in the sentence.
Outer Margin Inner Margini + Nucleus + Inner Margin2
The Outer Margin may be filled by an Independent Indicative clause plus the
initial feature -o. It may also be filled by minor-clause and non-clause elements.
The Inner Margins may be filled by a Subordinate Clause.
The Nucleus may be filled by an Independent clause.
Close-knit clause composites may function in any of the positions in which an
Independent clause may function.

Negative formations in Mbembe

Present tense
Subject Predicate Object Location
1. Owe ode]) ewilr5 selpya
child buys dress in-market
2. 6b6r 6n6:ma s61s5:m
borrows shirt in house
3. brigo
gives
4. 6061)6
washes
5. 5f6:n6
irons
o -179 -

Subject Predicate Object Location

6. bk5:0
shows
7. 5sa:ma
forgets
8. oror
wants
9. Osog
keeps
10. Okdr6
looks at
11. okibg
spoils
12. saga
puts down I

13. Osood
knows/recognizes
14. 5targ
adds
All these sentences can be formed which can be made by interchanging substitution
items.

Structure of verb is verb-prefix (o/o 3rd person .7ing.; 3 before low vowels,
a o before high vowels.) + verb-root.
Subject and Location places are optional. Verb and Object obligatory.

Future Transformation
The above sentences can be turned into the future tense by changing the tone
on the verb-prefix to high, and further prefixing the future prefix m-.
In addition, in sentences 8-14 the first (high) tone of the Object noun is
slightly lower than the preceding high tone (downstepped). This is shown by ',
e.g. 1. 1,7A mode) dwilr5 snpyg 'the child will buy a dress in the market'
2. moror'ewiir5 snpyg 'she will want the dress in the market'

Perfective Negative Transformation:

Above sentences can be turned into the perfective negative by:
1) a change of word order from S V 0 Loc to S 0 Loc V7
2) the addition of the further prefix k- on the verb;
3) certain changes in the final vowel of the verb
verbs with final c/o in other tenses retain c/o in perf. neg. (4,5.10.11)
verbs with final a in other tenses lose this a in the perf. neg.
(6,712.13.14)
'verbs with no final vowel in other tenses add the final vowel c/o
in perf. neg.
c after a front vowel or a in the root
o after a back vowel
 -180-

In addition:
In sentences 1-7 the verb prefix o- becomes high tone with a downstep onto the
verb root. All syllables of the verb root are high tone.
In sentences 8-14 the verb prefix o- becomes high tone. The pattern on the
verb root is low tone followed by high tone. Where the verb root has two syllables,
the first will be low, the second high. Where the verb root has only one syllable
a low-high glide occurs on this syllable. (Glides are not normally a part of the
tone system--this represents a compression of a word tone pattern onto one syllable):
E.g. 1. Wa gwur3 sglpya ktideb6 'The child didn't buy the dress in the
market'
2. Wa 6115:ma k56am 'The child didn't forget the dress'
8. Wa gwiirb koror6 'The child didn't want the dress'
13. Wa gwiir5 k6s60 'The child didn't recognize the dress'

Non-perfective Negative Transformation

Above sentences can be turned into the non-perfective negative by: a
1) change of word order from S V 0 Loc to S 0 Loc V (as in perfective
negative):
2) the further addition of the verb prefix m, with the retenLion of low
tone on the prefix 0-;
3) high tone on all verb roots.
E.g. Wa ewiir5 model, 'The child won't buy the dress'
Wa 6wilr5 mogb61,6"The child won't wash the dress'
Wa 5116:mg s51s5:m mosog 'The child won't keep the dress in the house'
fjwa ewur3 mostiog 'The child won't recognize the dress'
Note, further:
1) Change in word order: S V 0 Loc in indicative;
S 0 Loc V in negative
2) Changes in final vowels of the verb root:
Verbs with 'basic' final 6/0 retain this in perf. neg.; Il

Verbs with 'basic' final a lose this a in the perf. neg.

Verbs with 'basic' no final vowel add final vowel c/o ;

in the perfective negative c follows front vowel or a in root,

-o follows back vowel in root.
3) Verb tones: Two classes of verbs must be set up. In the present and
in the non-perfective negative these are not distintive. In the future, Class 1
(1-7 'high' tone) verbs show no downstep after the verb onto the object noun, but
Class 2 (8-14 'low' tone) verbs have downstep onto the first syllable of the noun.
 -181-

Where the basic initial tone of the noun functioning as object is low it becomes
high after a Class 1 verb in the future tense, but remains low after a Class 2 verb,
but this is not illustrated here. In the perfective negative Verb-roots of Class 1
have high tone on all syllables, Verb-roots of Class 2 have the pattern low-high
spread over one or two syllables.
The same illustration can be amplified to include the plural distinction:
The verb-prefix for 3rd person plural is a- with same tones as in singular. In
addition the clitic m- is prefixed to the verb in positive tenses: but to the
object (in other clause types to the first complementation item- not illustrated
here) in negative tenses. In the future plural, where it precedes the consonant
m it carries low tone.
n Examples (plural prefix m- underlined):
Present plural
Iva madeb ewiirb 'The children are buying the dress'
iva makerE ewiirb 'The children are looking at the dress'
Future plural
iva ilmagdeb ewiirb 'The children will buy the dress'
Iva filmakerc'ewilrb 'The children will look at the dress'
Perfective negative plural
iva mewilrb kadebg 'The children didn't buy the dress'
iva mewilrb kaker6 'The children didn't look at the ''-ress'
Non-perfective negative plural
Iva mgwilrO madeb 'The children won't buy the dress'
lva mewilrb mAkerg 'The children won't look at the dress'

In all examples where two vowels occur together elision occurs; the first vowel
is lost, the second retained. In the examples given, elided vowels have no tone
mark on top of them, signalling that they do not in fact occur in normal speech,
but would occur in very slow speech.
 Appendix II
PRELIMINARY NOTES ON SERIAL
CONSTRUCTIONS IN KASEM
Kathleen Callow
CORPUS AND METHOD
The evidence for this preliminary survey is based on 100 serial construc-
tions taken from running text, and all percentages and frequency counts are
taken from this corpus. Conclusions were then checked against a larger body
of material, and a few examples of features which did not appear in the counted
corpus are taken from this extended corpus. The original 100 serial construc-
tions were taken from three different texts covering a total of about 400
clauses (150 sentences), and the further corpus used for checking was probably
slightly smaller than this.
REGULAR SERIAL CONSTRUCTIONS
Regular serial constructions are defined as those that consist of a pri-
mary clause followed by one or more secondary clauses in the same aspect. That
is to say, a regular serial construction is either entirely in the non-continu-
ous aspect (alias perfective, alias aorist), or entirely in the continuous
aspect (alias imperfective, alias stative). In the counted corpus, 79 percent
were non-continuous, 8 percent were continuous, making a total of 87 percent of
the serial constructions which were regular.
Certain statements can be made concerning the structure of the clauses in
a regular serial construction.
The primary clause may be any one of the neutral or non-terminal types de-
scribed in connection with the clause matrix, below.
The secondary clause subject. All secondary clauses must have a pronoun
subject agreeing in person number and class with the subject of the primary
clause. Its tone is mid in non-continuous clauses, low in continuous clauses.
The secondary clause verb mood. In a non-continuous series, the verb is
always in the consecutive mood; in a continuous series it is always in the non-
future mood.
The secondary clause Dreverbal tagmemes. No preverbal tagmeme ever occurs
other than the subject, which means that there is obligatory, absence of the in-
troductory and temporal tagmemes, and of all preverbal particles, including the
negative ones. Consequently, 50 percent of the counted secondary clauses occurred
in the minimum form, viz., S:Pn + Pred:V.

-182-
 VW'

-183-

The secondary clause postverbal tagmemes. Postverbal tagmemes occur, ap-

parently without restriction. 50 percent of the secondary clauses had a periph-
eral tagmeme. The object tagmeme was the most common, but location, agent, re-
lational, and adverb tagmemes also occurred. The occurrence of these tagmemes
li seemed to bear no relation to the position of the secondary clause in the series,
whether initial, medial, or final. While only one peripheral tagmeme normally

U occurs, it may be of any length (one object was 14 words long) and may include a
rankshifted clause.
All attempts have failed* to state restrictions in terms of verb occurring
in specified sequence. There is a tendency for the following verbs to occur in
fl
the final position in the series: dwey 'to exceed', gaali 'to exceed', gara
'to be better', gan 'to fail', pa (+ 0) 'benefactive', by 'to finish'. These
occur final in the series very regularly, but the non-continuous series consists
essentially of a sequence of successive actions, and it is only necessary to
add another action to reduce these verbs to medial position, e. g.,
awia duwri o daali "pilwri 3 yiy songo yiyga 3 dlari
Awia ran he exceed Apuwri he reach house first he leave him
'Awia ran faster than Apuwri and reached the house before him.'
It is probable, however, that these verbs would always be final in a continuous
construction, which describes simultaneous actions or states.

IRREGULAR (NON-FAVORITE) SERIAL CONSTRUCTIONS

Of the 100 counted series, 87 were regular, 10 involved a change of aspect,
and 3, all following auxiliary verbs, were not capable of proof. The 10 cases
involved an identical change, viz., a non-continuous series followed by one con-
tinuous clause in the final place. Seven of these 10 occurrences involved one
verb, wf 'say', which is inherently continuous in aspect, and always occurs at
the end of a series introducing speech. The other 3 described a state terminat-
ing a series of actions, e.g.,
omu wd zangi o vo o kasina gwaaru tulu na lana td o nagi 5
to sigi mankyaplina niy ni...
'She would get up, put on her best [beautiful] Kasena clothes,
go out, and remain standing by the compound entrance...'
Two points should be noted about the continuous clauses terminating a non-
continuous series:

*[But see §2.2.2. for more recent success. KLP]

3
 -184-

(a) they may have a particle between the subject and the predicate;
(b) there is an alternative construction in which the series is termi-
nated before the continuous clause and the conjunction yi 'and' inserted,
thus making a complex sentence. It is normal to form a complex sentence
using yi whenever a construction is required which would break the series,
by changing the aspect or by introducing a new subject in the same sentence.
(Query: Does this imply that yi introduces a change of subject which is
out of 'focus'?) E.g.,

zimbaaru wum mm yin yi o to keeyra

brother the refused and he still crying
'The (older) brother refused and kept on crying'

mu o ne tutwey di gy4 di soli gulf yi di nablyli sin zwin

El
focus he saw mouse it sitting it stirring porridge and its tail washing calabashes ij

'He saw a mouse sitting stirring porridge while its tai) washed calabashes'

STRUCTURE OF SERIAL CONSTRUCTIONS RELATED TO CLAUSE TYPES

Restrictions on the order of certain types
Non-terminal primary clauses must be followed by a series, and may be
followed by a series of any type; terminal clauses cannot be followed by a
series, and so far have been found following auxiliary primary clauses only.
These restrictions may be summarized by the following chart. (Numbers refer
to following illustrations.)
SECONDARY CLAUSE
Non-Term. Neut. Term.

PRIMARY Non-Term. x (1) x (2) x (3)

CLAUSE Neut. x (4) x (5)
Term. -
. v .. ...

1. a pLgi a yayni a vey da

I already I habit. I going there
'I have already been making it my practice to go there'
_ v - ..

2. o ma yayni o zangi o vuw o loowri...

she habit she rise she go she beg...
'As usual, she got up and went off begging...'
 -185-

.
3. o to kOwri o yi babia
he still contin. he is brave
'He was still as brave as before'
4. pe ma gya u 3 pa 3 diyga
chief took him he give him enter room
'The chief made him go into a room'
5. zimbaaru wpm ma zangi o swe o yira...
brother the - rose he wash his body...
'The older brother got up, took his bath...'

Restricted constructions with certain types

(a) Abilitative. This non-terminal type may not have a periphery of its
U
own (i.e., in the primary clause), but frequently 'adopts' the periphery of
the following secondary clause. The adopted tagmeme is usually the object
(either pronoun or noun phrase) but the location has also been found. E. g.,
-
a daa ba wani u a biy
I again will-not be-able her I raise
'I won't be able to bring her back to life again'
.
wuntu wu wani 6 biy u

that-one fut be-able he raise her

'he will be able to bring her back to life'
-
n we kaana dide n diy
you can wives many you eating
'you can marry many wives'
V
n we tiga ni n vara
you can ground on you hoeing
you can hoe on (level) ground'

(b) Causative. This is the only clause type which must be followed by a
different subject in the following secondary clause. Where this subject is a
pronoun there is no other contrast with a regular series, but the subject of
the secondary clause may also be a noun. This may point to the setting up of
a complex series to handle this.
 -186-

o pE a niingi

he gave I go-out
'he made me go out'

o pE kaani wum nungs

'he made the woman go out,'

Impersonal verbs
There are certain clauses with the subject ku 'it' which have so far been
handled as intransitive clauses, as they do not contrast with them in the prim-
ary clause place. However, they would'be better handled separately as they have
unusual functions elsewhere.
(a) They may occur as included (rankshifted) clauses, filling an object,
time, or adverb place without breaking the series. E.g.,
fl
...5 viiw 5 nowri mangi di siwla finto o gya ba Li

he go he beg it approx. with shillings 30 he take come

'he went out and obtained 30/- by begging and brought it back'
11
(b) When they function as primary clauses, an allo-construction permits a
following clause which is not possible in a regular series, as it changes the
aspect and the subject, and has a negative particle.
ku daani yi a wu nE m (regular form)
it delayed and I not saw you
410t
'it is a long time since I saw you'

ku clani a ura nE m (allo-form)

I

(c) There is a further construction, without any known allos, which also
permits a following clause which is not possible in a regular series, as it r

may change the aspect (first example) and the subject. E. g.,

ku da dibam sangi
it is-not us cooked food
'it wasn't us who cooked the food'

ku de n5onu didwa mu vey 8 lagi kaani

it is-not man one emph. goes he wants woman
'it isn't just one man who goes in pursuit of a wife'
 o
Appendix III

Q PRELIMINARY PARADIGM OF SOME DEGEMA

INDEPENDENT CLAUSES
D
Elaine Thomas

Q 1. Basic independent clause types

1. Transitive
o 2. Intransitive
3. Ditransitive

n 4. Stative
5. Equative
n
6. Benef active
tj
7. Directive

ri
LI
2. Citation paradigm
1. omo- yo m3- de isen
D 2. &16-37 rri-ue
1

3. omo-yo mo-kt.e 6371 isen

D 4. o'm 0 --/ y aib't.

5. 6m6-37 O-yrn O, MO-MOSI

u 6a. 6m6-37
, ,
m3-d? -ke Oyt isen
6b. omo-yo mo,a-k isen my 6n6m6 1-161)

6.c. 6m6-37 mo-d-1.c On6m6 nou

B
7. omo- yo me -tA Ci4en

0
U 3. List of lexical items
isen
sen fish -y emphatic
D (t6Ln wood mo- nonpast 3rd person
,!, singular, subject
omo child
. . o- 3rd person singular,
-most. male
Pi subject equative
. .
oyo him/her
-kc benefactive
6n6m6 mother
D .%
mu introductory particle
nou his for NP benefactive
t, V.
aido be there
D A buy
ki. give
LI to go
v(6 cry
-187-
o yin be

i S.'
 -188-

4. Translation of clauses
1. The child buys fish.
2. The child cries.
3. The child gives him fish.
4. The child is there.
5. The child is a male.
6.a. The child buys fish for him.
The child buys fish for his mother.
The child buys it for his mother.
7. The child goes for firewood.

Paradigm of clause nuclei in contrastive tagmemic form

Subject Predicate I. 0. D.O. Compl. Benefactee Goal

+ ± NP
Trans. NP + Trans VP
pro pro 1

Intrans ± NP +Intrans VP
pro
1 4.
Ditrans NP +Ditrans VP +pro - NP
pro

Stative
+
- NP +Stative VP
pro
+
Equative NP +Equative VP +NP
pro pro
+ + NP ± NP
Benefac. ± NP +Benefac. VP -pro
K.....,
pro o-- >Ben

Directive +
- NP +Direct. VP +
NP
pro

Note
1. If the I.O. is a personal noun a transformation into a serial clause is re-
quired.

Example 'The child gave his mother meat'

!i
6m5y5 51/611 risen 5v6 yl.ki4n 6116mb dip

child took fish took gave mother his.

 -189-

Marginal places
All basic clause types may be expanded by optional marginal places which may be
diagrammed as follows:
Ex.l: time phrase +Nucleus -Ex.2:manner phrase - Ex.3:loc. phrase
Ex.4: time phrase / temporal

Fillers of places
(a) Subject, Object and Complement places.
i. NP +Head ± Mod ± Num ± Poss. ± Dem.
So far only 2 attributes have been found in one phrase.
iI \ I
Examples sen ive noir 'his 2 fish'

isen vi nana 'these good fish'

ii. NP composite
Appositional 5m5y5 Amade 'the boy Amade'

Co-ordinative omoyo nv. onoma nov. 'the boy and his mother'
Relative omo nootdn mekiyo

child who went to market

iii. Independent Pronoun
1 1
fum.6 1I, eni we
! 1
IOW o 'you sing.' umal 'you pl.'

oyl 'he/she' I'lAt 'they'

Note
1. 1st and 2nd person singular. object = I.O. are expressed by a
verb suffix.
mo-mon-me 'he sees me'
'he buys (it) for me'
(b) Predicate place
The filler of the predicate place is a verbal phrase with a verb of the
appropriate class as its head. The VP may be diagrammed as follows:
± Aux 2 ± Aux 1 + Verb
Examples oda 'so) he still buys '
A.2. Al.vb
ti

3V3 ka d'E 'he takes also buys'

A.2 A.1. vb
 -190-

(c) Benefactee place

The filler is a NP of which the head must be a personal noun (see under

a). The NP is preceded by the particle me if the D.O. is expressed.

(d) Goal place
The filler is a NP (see under a)
(e) Expansion places
Ex. 1 and 4 may be filled by
A A F
(i) time phrase oho awLyeyo
time morning
i ApA
(ii) temporal nina today ude yesterday/tomorrow
Ex. 2 is filled by a manner phrase
&to dier 'badly' (lit. hand bad)

aku vi 'well' (lit. method good)

Ex. 3 is filled by a locative phrase consisting of a NP preceded by the
particle mU
A A
mu, eki 'at /in the market'

7. Verb chart

Prefixes Root Suffixes

tonal tense/aspect derivational inflectional

neg.1 past -ne 'of own accord' -te completive

3
subj. nonpast -se 'causative' {-Vn} perfect
2 3
-kine 'reciprocal' -te habitual
{ -viii} 'repetition' -mun not again
-ke 'benefactive'

Notes on the verb chart

1. Negative is shown by an extra high tone on the tense/person prefix.
2. Subjunctive is shown by a high tone on the tense/person prefix.
3. Matrix of tense/person prefixes
0 -191-

a
Singular Plural
0 2 3 1 2 3

Past
mi- s u- o- e- a- e-
Q
Nonpast me- mu- mo- me- me- me-

o
4. There is vowel harmony throughout the verb word between the followiilig 2
sets of vowels; the vowel of the root determines the set used.
11
tense set i e e o u
lax set L
e a 0 I)

Examples to illustrate the verb chart

me-E 'I will buy/buy'
!

m E --of en 'I bought'

F. r: ;.
woe
-ole-vLL 'you bought often'
+
C o-cf& 'he does not / did not buy'
--tdZ 'we should buy'
+
a -a -miin 'you did not buy again'
e -de -te 'they bought already'

The extra high tone of the negative is shown thus +.

D
 Appendix IV

GRAMMATICAL PROSODIES ??
John T. Bendor-Samuel

Can any features be found within a grammatical system which correspond

to the prosodies found in a Firthian type of analysis of a phonological sys-
1
tem . To answer this question we must first ask what phonological prosodies
2
are . What features of the phonological structure are handled as prosodic?
Prosodies in Firthian analysis are phonological features which either
extend over more than one segment of a structure or have implications over
more than one segment. Either they cannot be adequately assigned to a sin-
gle segment or alternatively if occurring as an isolatable single segment
they have an implication for the structure as a whole, e. g., marking the
boundary of a phonological word. This second type of prosody resembles
Trubetzkoy's 'Grenzsignale'.
Are there grammatical features which parallel this? First, are there
grammatical features which can be said to extend over more than one segment
in the grammatical structure. Such a question presupposes that it is possible
to set up grammatical segments which correspond to the phoneme segments of
phonology. Any given grammatical structure can be viewed as a series of units
in certain relationships to each other. Each unit would correspond to the
phoneme segments in phonology. Thus a clause can be viewed as consisting of
phrases and a phrase as consisting of words. The segments of the clause would
then be the phrases and the segments of the phrase the words.
The question would then be, are there grammatical features which extend
over more than one phrase in a clause or over more than one word in a phrase?
As a rather simple and limited example, the singular/plural contrast extend-
ing over both the nominal and the verbal phrase in an English clause might be
adduced, as in
The boy is eating
The boys are eating
The contrast singular/plural can be seen to extend beyond the one phrase and
in fact marks both the phrases which can be said to make up such a clause
The clause can then be said to be marked as singular/plural and a grammatical
prosody of singular/plural could be set up.
The usefulness of such an analysis in English is rather reduced by the
fact that the object nominal phrase is also marked by the singular/plural con-
trast as in
The boys are eating the melon
The boys are eating the melons
-192-
 -193-

So there would still have to be a category singular/plural at the phrase level

as well as the clause level.
A much more extended and corn"- ...g instance of this occurs in languages
3
with a 'concord system'. In Etung phrases such as the following occur. (Hy-

phens mark morpheme boundaries and tone is omitted):-

e-yu ji t bi - yu bi - bae
yam one yams two
e-yu an ji bi yu am bi
yam that yams those
e-yu ej - e bi yu eb - e
yam his yams his
e-yu n - ji bi yu m bi
yam this yams these
n-kae yo t o fak bi t

wife one broom one

n-kae ow e o - fak ob e

wife his broom her

n-kae n- n_ o o- fak m- bi
wife this broom this

Each phrase consists of two words and each word includes a prefix or suffix.
The type of analysis proposed here would state that such phrases consist of two
main elements (the two words), together with one of a series of concord prosodies.
These concord prosodies are realized by the affixes. The selection of one out of
the set of concord prosodies is determined by the class of the noun head of the
nominal phrase.
Concord in Etung extends beyond the single phrase to include nominal and
verbal phrases. Clauses such as the following occur:
e yu. ji - t - man
yam one is finished
bi yu bi bae bi - man
yams two are finished
n - kop yi - t a man
box one is finished
o fak bi - t o man
broom one is finished
 -194-

The concord prosodies extend over the subject nominal phrase and verbal
phrase. Thus the clause as a whole is marked by the concord not just the
phrase. In a full treatment of Etung it would be necessary to state these con-
cord features as operating as a feature of certain clauses and phrases.
4
Another type of extension occurs in clauses in Bimoba . A conditional
clause is identical in structure to an independent clause except that it is
marked by the particle i and/or tii which occurs after the nominal phrase sub-
ject and before the verbal phrase, together with a low tone on the final syllable
of the clause. Sometimes the particle i occurs more than once in the clause,
occurring as the final syllable of some of the verbal particles. The combina-
tion of particle(s) and low tone must be regarded as the exponent of the sin-
gle grammatic category of "conditional". Instead of considering such a parti-
cle as a unit of the structure of the clause, which would raise problems since
it is clearly not a phrase, it is simpler to regard it as a part of a type of
grammatical prosody. The prosody would include the repeated occurrences of
the i particle and the final low tone.
Examples:'
kojok i dii napayon gbanl not saa mok n bik na
cock if cat stone one fidish he future marry my daughter this
"If the cock finishes eating one stone, he will marry my-daughter."

m i bo i biak i saa saa Naliatuk le bo siina

I if immediate again future go Nalerigu it immediate better
"If I were to go to Nalerigu again it would be better."

What of the other type of prosody - that which though not extending over
more than one segment has implications for the structure as a whole.
In "maay West African languages there are particles which can be regarded
as playing just such a role. To take another Bimoba example, the general sub-
ordinative clause is marked by the particle a (after pronouns) or mba (after
nouns). It is difficult to describe these particles as being either themselves
phrases or as being a constituent part of any other phrase in the clause. To

assign them to the nominal phrase or verbal phrase seems quite arbitrary and
their function is to mark the clause as a whole. This point is further illus-
trated since when a series of verbal phrases combine to make a serial clause
the particle only occurs once. These particles may then be:said to have an
 -195 -

implication of subordination for the structure in which they occur. This is

in fact the only way in which many subordinative clauses are distinguished from
other types of clause. Instead then of giving them the status of grammatical
unit it seems simpler to handle them as grammatical prosodies.

Examples
o a won wuur gben ti won joob ti sanjok
he yesterday washed finished we yesterday took our road
"Yesterday as he finished washing, we started on our way."

banton mba din poon o na o din wpb no pontr

crocodile remote made him he remote put on with frog
past past back him
"As a crocodile make him cross the river, he put a frog on his back."

In many languages a problem often arises as to where segmentation into

grammatical units should be carried out. Take for instance, the English phrase
"cats and dogs". Clearly the two main units in the construction are the two
nouns. What of "and"? Is it a third unit? If so, what is the status of 'and'
in the phrase "men, women and children"? To treat it as merely the third unit
in a string of four equal segments seems inadequate since though its physical
realization may be after the second noun its function is to link the first noun
to the second just as much as the second noun to the third. In fact it signals
5
the co-ordinate relationship of all three nouns . Furthermore what is the sta-
tus of "and" when linking two clauses. It can hardly be regarded as belonging
to either of the two clauses. How much simpler to regard "and" as a feature
of the construction as a whole marking the co-ordinate relationship of the main
elements concerned.
The term "grammatical prosody" may be ill-chosen. Prosodists may disown
it. "Syntagmatic features" might be a less provocative term. Whatever the
label, however, it seems there are grammatical features which are not well
handled by the usual segmentation into constituent parts of the construction.
Let them be handled as features of the total structure and not forced into an
arbitrary segmentation.
 -196-

FOOTNOTES

II
1
This note arises in part from the two workshops conducted by DI. K. L. Pike
under contract 5-14-065 of the U. S. Office of Education, through the Center
for Research on Language and Language Behavior, of the University of Michigan.
The underlying research on West African language has been carried out as a
member of the Summer Institute of Linguistics in association with the Univer-
sities of Ghana and Nigeria.
2
For an introduction to the theory of prosodic analysis see J. R. Firth,
"Sounds and Prosodies", Transactions of the Philological Society, 1948, pp.
127-152; and R. H. Robins, "Aspects of Prosodic Analysis", Proceedings of the
University of Durham Philosophical Society, Volume I, Series B, No. 1 (1957).
For an example of this type of analysis applied to a language see my "Some
problems of segmentation in the phonological analysis of Tereno", Word, Vol.
16, No. 3, pp. 348-355 (1960).
3
I am indebted to my S.I.L. colleagues, Mr. and Mrs. T. Edmondson, for
the Etung material used in this illustration.
4
For further details of this and the other Bimoba examples quoted in
this note, see my 'Problems in the analysis of sentences and Clauses in Bimoba'
to be published shortly in Word. I am indebted to my colleagues, Miss G. Jacobs,
for the Bimoba material.
5
Compare R. S. Pittman's treatment of conjunctions in "A grammar of
Tetelcingo (Morelos) Nahuatl", Language Dissertation No. 50, 1954. In setting
up the category of 'valence', in contrast to morpheme, he regards conjunctions
as 'valence-carrying' morphemes. Valence is used particularly for the relation-
ship between two immediate constituents, whether overtly or covertly expressed.
Conjunctions would exemplify overt valences carried by morphemes.
 Appendix V

NOMINAL AND VERBAL GROUP MATRICES FOR KASEM

J. C, Callow
NOMINAL GROUP MATRICES
There are four major nominal groups (i.e., pairings of singular-plural
classes) in Kasem, but only C and D are demonstrated here.
GROUP C
For this group, the singular suffix is a central vowel, the plural suffix
a front vowel, the phonemic quality being determined entirely by the stem phono-
logy.

Consonant-final stems

a Stems with a final /d/ are taken for illustrations in the singular below.
On the left is a matrix of all such stems, and on the right is the matrix of
the ci-14;ular forms, arrived at by adding the central vowel suffix to the stem.

d- final stems d-final stems, singular form

Upper
1 2
Cid- Cud- Cida Cuda
Harmony
3 4
0 Cad- COd- 0 Cada Coda

+1ICen. V
sg.
5 6
Lower C id- Cod - Cida Coda
Harmony 7 8
0 Cad- Cod- 0 Cada Coda

The Osignifies systemic non-occurrence, i.e., /e/ and let cannot occur in
stems of this general shape; the numbers refer to the examples below.
1. bide 'wall' 5. fida 'cane'

2. kukuda 'dog' 6. zoda 'poor (person)'

3. bakada 'boy' 7. sada 'mat'

4. koda 'voice' 8. moda 'pus'

Now note: (1) The topology (overall pattern) of the harmonies is the same,

1
and this is true throughout the language. So, replace i and I by the cover
symbol I; a and a by A; o and o by 0; u ando by U. (2) To repeat the above
matrix formula for each of the 5 final consonants (d, 1, n, 0, g) would be
redundant and inelegant, so reduce to one matrix formula by having one final
consonant per row, one stem vowel per column.

-197-
 -198-

Singular matrix for Nominal Group C, consonant-final stems (columns arranged

by vowel quality)
Consonant-final stems Consonant-final stems, singular form
CId- CAd- COd- CUd- CIdA CAdA COdA CUdA
CA1- C01- CUI- CI1A CA1A COLA CUlA
CIn- CAn- COn- CUn- + cen. CInA CAriA COnA CUnA
CID- CAD- C00- CUD- CIOA CADA CODA CUDA
CIg- CAg- COg- CUg- CIgA CAgA COgA CUgA

Having established this general pattern of matrix, the plural, which is more
complicated, can be exhibited in the same way.

Plural matrix for Nominal Group C, consonant-final stems

did- CAd- COd- CUd- CIdI CAdI COdI CUdI

CI1- CA1- C01- CU1- CI1I CA1I CObI CUlI

CIn- CAn- COn- CUn- + Fr. V, CInI CAnI COnI CUnI
CID- CAD- C00- CUD-
CwI /I CE CwE CwI/1
CIg- CAg- COg- CUg-
CI CE CwE CwI

Consonant-final stems, sg. form (N.G. The examples will follow the columns
of the matrix, and then the rows)
bide 'wall bakada 'boy' koda 'voice' boda 'fishing-net'
kikila 'round' fele 'white man' gola 'fish hook' fula 'granary'

mipola 'an ant' cana 'month' bona 'mosquito' Dona 'rope'

iLoa 'hand,arm' baDa'beam' cope 'path' zoDa 'calabash'

digs 'room' yaga 'market' koga 'back' bugs 'river'

Consonant-final stems, plural form (meanings as above)

bidi bakadi kodi bodt.

kikili fali goli tuli

mt.mt.mt cant bon t Dont
jt./j1 be cwe

di YE jwc bwi
 -199-

Singular and Plural matrices for vowel-final stems in Nominal Group C

The vowel-final stems can be handled in a similar manner, except that
there is only one row, and there are five columns, as CE- occurs as a vowel-
final stem.
B cen.V-, CIA CIA 1 CAA CUA CUA

CI- CE- CA- CO- CU-

'1 II
fr. V-4 CI., CE CEO CwE I CwI
1

Vowel-final stems, Nominal Group C, sg. and _pl. forms

sg. via 'poison' pia 'sheep' daa 'stick' yua 'hair' noa 'finger'
pl. vl pe de ywe nwt.

Group D

to For this group, the singular suffix is a back vowel, and the plural suffix
one of the three alveolar medial consonants (d, 1, n) followed by a close back
vowel. As before, the phonemic shape of the suffixes is determined by the
phonological shape of the nominal stem.
Singular and plural matrix for Nominal Group D, consonant-final stems

[Stem] [Singular] [Plural]

CUd- COd- CUdU COdO CUddU i CwA:dU

CU1- C01- CU1U C010 CU11U I CwA:1U
r-
CUn- COn- + bk CUnU COnO +TU CUnnU CwA:nU
CUD- C00- CUgU C000 CUnnU CwA:nU
CUg-- COg- CUgU COgO CUdU i CwAdU/COdU

Consonant-final stems, Nominal Group D, singular and plural

sg. sg. pl. pl.

modo 'thousand' kodo 'calabash for milk' moddo kwaadu

culu 'taboo' bolo 'valley' cullu bwaalu
fofono 'speckled' tong 'skin, paper' fofonno twaano
yituou 'stool' n000 'lion' yitunnu ilwaanu

kogo 'animal hair' dogo 'clay' kodo dwad / dodo

 -200-

Vowel-final stems, Nominal Group D

(For symbols in matrix representing analysis of this group, see final com-
posite chart of the nouns) of
sg. (-U) pl, (-TU)

piu piidu 'rock'

sto seedo 'shea -nut'

maao maad 'catapult'

V33 voodo 'leaf'

tuu tuudu 'elephant'

NOMINAL GROUP Bl
Stems, Nominal Group Bl
(For matrix see following composite chart; group B2, not illustrated here,
includes just a few irregular forms)
sg. (-I) pl. (-A)
tidi tide 'forehead'
wad'. wada 'mud brick'
pori pore 'okro' (sometimes spelled 'okra')
kuri kure 'stone for cooking on'

it
nabili nabila 'tail'
Fa

yali yala 'tooth'

goli gola 'club (for driving animals)'
golf gola 'millet porridge'

bLni. bLna 'year'

ban'. bana 'anger'

ton!, tona 'beard'

kuni kung 'knot'

p L pLa 'yam'

de da 'day'

gwe gwa 'cola nut'

nakwA nakwa 'elder'
 -201-

COMPOSITE NOMINAL MATRIX

Having produced a considerable variety of smaller matrices, the question
naturally arises--is it possible to devise some form of matrix which would dis-
play the whole noun class system at once? This will not be attempted here,
but the following matrix does cover nominal groups Bl, C, and D; A has been
omitted for the time being as it is much more complicated and irregular (also,
in A, nouns derived from verbs have been omitted as they also introduce stem
patterns not otherwise found in the nouns).

VERBAL GROUP MATRICES

There are four verb groups (i.e., pairings of Perfect-Imperfect classes)
in Kasem, labelled' A to D in the cells of the matrix defined* as follows:
0 Impf.-Absolute**
-A
-A
Perf. -Abs.
-I/O

Further, all the morphological forms (which function verbally, i.e., not
nominalizers) exhibited by any verb can be exhibited on a matrix of the follow-
ing pattern:
MOOD
Past Future

Abs.
Perf.
NAbs.
ASPECT
Abs.

Imperf.
NAbs.

Each of the verb groups, then, will be displayed in this matrix to show
the considerable variety of pattern that results, restricting the material,
however, to consonant-final stems.

*Definition, here, is accomplished by giving the minimal criteria needed

to characterize a class. This can be done by noting just two appropriately-
selected words of the paradigm, in respect to the vowel of the suffix. The
suffix vowel is indicated by a cover-symbol A (not to be confused with the
class label A inside the matrix) or I outside the matrix, representing upper
or lower harmonies. By turning to the last (composite) chart of verb groups,
the reader can see that first and fourth rows contain perfect past absolute
and imperfect absolute respectively. In the rows, within the columns, under
the appropriate class symbols, the two defining forms are found. No further
 -202-

NOMINAL GROUFS
STEM
SHAPES Bl C D

-I -A -A -I -U -TU
sg. pl. sg. pl. sg. pl.

CId- CIdI CIdA CIdA CIdI 0 0

CAd- CAdI CAdA C4dA CAdI 0 0
COd- COdI COdA COdA COdI COdO CwA:dU
CUd- CUdI CUdA CUdA CUdI CHU CUddU

CI1- CI1I CI1A CT1A CI1I 0 0

CA1- CA1I CALA CA1A CA1I 0 0

C01- CO1I COLA COLA CO1I COLO CwA:1U

CU1- CUlI CU1A CU1A CUlI CU1U CUIIU

CIn- CInI CInA CInA CInI 0 0

CAn- CAnI CAnA CAnA CAnI 0 0

COn- COnI COnA COnA COnI COnO Cwa:nU

CUn- CUnI CUnA CUnA CUnI CUnU CUnnU

C10- 0 0 CI0A Cl/T 0 0

CAD- 0 0 CADA CE 0 0

COO- 0 0 CODA CwE CODO CwA:nU

CUD- 0 0 CU DA CwI/F CUOU CUnnU

CIg- 0 0 CIgA CI 0 0
CAg- 0 0 CAgA CE 0 0
COg- 0 0 COgA CwE COgO CwAdU/COdU
CUg- 0 0 CUgA CwI CUgU CUdU

CI- CI CIA CIA CI CIU CI:dU

CE- CE CA CIA CE CIO/U CE:dU
CA- 0 0 CAA CE CA:U CA:dU
CO- CwE CwA CUA CwE COO CO:dU
CU- CwI CwA CUA CwI CUU CU:dU

forms are needed to separate the words into classes--but many further
forms are needed to describe concomitant variants--KLP
**The absolute form usually precedes zero or some variety of juncture;
it is not followed by forms in closest-knit relation to it.
 -203-

VERB GROUP A
Verb Group A is divided into four subgroups according to the pairing of
the stem-final consonants in the Perfect and Imperfect as follows:
g-d, g-n, g-zero, n-n.

Verb Subgroup Al
Past Fut. East Fut.

Abs. CVgA CV vaga va

Pf.
NA.,. CVgI CV vag t va
e.g.,
Abs. CVdA CVdA vada vada 'to farm, hoe'
Impf.

NAbs. CVdI CVdI vadt, vadt

Verb Subf,roup A2

Past Fut. Past Fut.

Abs. CVgA CV 05ga 05

Pf.
NAbs. CVgI CV e.g., 06gt 05 'to steal'

Abs. CVnA CVnA Ona Ona

Impf.
NAbs. CVnI CVnI 06nt orit

Verb Subu2uE_A3

Past Fut. Past Fut.

Abs. CVgA CV 15ga 15

Pf.
NAbs. CVgI CV 16gt 15
e.g
AbF. CVA CVA lba laa 'to build'
Impf.
NAbs, CV CV 15 15

Yf7-EIatgLTIPA
Past Fut. Past Fut.

Abs. CVnA CV/V dana dO/O

Pf.

NAbs. CVnI CV/17 demt di/6 'to chew'

e.g.,
Abs. CVnA CVnA dOna dOna
Impf.
NAbs. CVnI CVnI d6nt
 r.

-204-

VERB GROUP B
Past Fut. Past Fut.

Abs. CVgA CV tags to

Pf.
NAbs. CVgI CV tggI to
'to shoot'
e.g.,
Abs. CVI CVI tg tg
Impf.
NAbs. CV CV to to

VERB GROUP C
Past Fut. Past Fut.

Abs. CVn/ I CVn/ I bogi bOoi

o o
Pf.

NAbs. CVn/ I CVn/ I-

o o e.g., 'to call'
Abs. CVI CVI bwe bwe
Impf.
NAbs. CV CV bo bO

fl
VERB GROUP D [The most commonly-occurring class--though with fewest contrasts
within the system]

Past Fut. Past Fut.

Abs. I CVCI CVCI t5g1, t5g1,

Pf.

NAbs. CVCI CVCI t5g1, t5g1,

e.g., 'to follow'
Abs. CVCA CVCA t5ga t5ga
Impf.

NAbs. CVCI CVCI t5g1, t5g1,

OF
COMPOSITE VERBAL MATRIX
Having displayed the different verb groups separately, it is now interest-
ing to ask whether the whole system can be displayed on a single matrix, so
as to make comparison between the groups easier, and perhaps comparisons with
other languages. This is attempted in the following matrix.

no
 -205-

Matrix for all verb groups, consonant-final stems only

VERB GROUPS

All l A2 j
A3 A4 B C D
Past Abs. CVgA CVgA CVgA CVnA CVgA CVn/oI CVCI

PERFECT NAbs. CVgI CVgI CVgI CVnI CVgI CVn/oI CVCI

Future CV CV CV CV /V CV CVn/ I CVCI

o
Abs. CVdA CVnA CVA CVnA CVI CVI CVCA
IMPERFECT
NAbs. CVdI CVnI CV CVnI CV CV CVCI

No. of Contrasts 5 5 4._ 3 4 3 2

Notice two interesting features demonstrated in this matrix: (1) With

one exception, there is a decreasing number of contrasts in the successive

columns, or, put in another way, there is increasing neutralization. Even
so, the contrast between Perfect and Imperfect is maintained in every case,
B when Absolute, so this leads to the assumption that this is the fundamental
contrast in the verb system. (Cf. singular and plural in the nouns.)
(2) The 8-hole matrices used above have been condensed to a 5-row column
B
vector, as there are never more than 5 contrasts for any verb. Notice how
unevenly distributed the contrasts are. Thus, there is no past-future dis-
tinction in the Imperfect; and there is no Absolute-Non-Absolute contrast
for the Future.
If the auestion is asked--where does the Imperative fit in? The answer
would be that so far as the Imperfect is concerned there would be no change,
but in the Perfect it would form a three-way contrast with the Past and
Future, and, like the Future, would have no Absolute-Non Absolute contrast.
 Appendix VI
NOUNS OF ETUNG* CLASSIFIED BY THEIR SINGULAR-PLURAL PREFIX PAIRS
Eileen Edmondson

[The stems are first subdivided according to the prefixes which mark their singular
versus plural forms. Within this grouping, stems are subdivided according to the
tone patterns which cover the word as a whole. The tone pattern is given only once,
for the class, (determined by the stems), since the tone of the word in singular
form is the same as in the plural form (if /6/ is low, /bi/ is low: if /g/ is high,
/bi/ is high).
Loans--0 to /ba/--are not included.]

*Dialect of Bendeghe Ayuk in Ogoja Province, Nigeria. Informant: Moses

Abang E'ta.
 -207-

e 'Mass' Nouns (57 items)

... .... ,..

or or
1
eban ' dislike. elAt privacy.

ebio ' call. eCene ' break in rainy season.

eboge ' time. eCit ' ooze.

ebok ' willful accusation. efep ' air.

ejiri ' stupidity. ejiri ' darkness.

ediri ' tramp. ejil ' stealing.
efit ' ill temper. eka ' semen, stale food.
efi ' kind of plant. ekok ' harmattan.
egome ' snore. erik ' time.
f
ege frequent cry. eyo sleep.
1B
'

egibe ' reference, mention.

ejen I
travelling.
v or

ejok noise. ebigim ' satisfying.

B
'

ekin ' alligator used in ebogim ' adding.

grinding snuff. ebii ' intercourse.
eket ' destitution.

ile in fable.
v
ebat ' light. ecabe ' disgrace.
ebibe ' dust. egbut ' swelling disease on the hand.
ebimim hiding oneself. ekap mark of ring worm or leprosy.
U ' '

ebumum ' being drunk. eklt ' kind of soft cocoyam.

eComim ' being insane. ekp40 ' dish made from cocoyam.
efet ' chance. ere ' out-side.
efuo ' kind of rope.
...
ejim ' onion. or

ekip ' srew-pine. efoge ' that which prevents.

ekio native guitar.
' ..
or

esebe sand.
 -208-

e 'Mass' Nouns Con't

or Others.
ekpin life. ^"ebek sweet alligator.
1
epit superciliousness. eRagasi 1
that which is quiet, gentle
1
eriri deceitfulness. and polite person.
1
eyuk cold. eYCitiikibi
1
that which is cut into particles.
eirie e grass land.

e- to ni- (48 items)

or or
ebam bag. ebit soft cocoyam.
efare trip, giddiness. eb in farm.
egan bundle. ecue heap, mass.
ekam company, age group. efe mirror, globe.
ekpa mat. ef4ge fan (alt. plural in bi-).
ekpak kind. efut feather.
etam hat. ekim deep pool.
ekup parcel.
"1/4

or
esome comb.
ebam' compound.
etim hut, shelter.
ebaoe iron-thing.
eyom hoe (alt. plural in bi-).
e6oe load.
AA
efuk axe. or
ejio pile, heap (alternate ebuge tray for loads.
plural in bi-). ebune mud-bed.
er4m pod. ekoo spine (fruit edible).
eroo line. ekp4me bottle.
es4n trunk, stalk. etage flat basket for drying pepper.
esum clearing.
etek or
village, region.
etim ebln dance (alt. plural in bi-).
bush.
eb i palmnut.
or ebum heap (alt. plural in a-).
eCan electric fish. efit definite length of cloth.
efo cloth. ekim half: part which is short.
ekun spiny fish. ekon song.
ekpao hamper. etum work.
eya year.
Others
esome worm.
 -209-

e- to bi- (140 items)

Tone class Tone class
or or
eba ' fetter. eban ' iron rod.
ebage ' kind of monkey. ebat ' pancreas.
ebe ' belch. ebuya ' white edible substance.
ebi ' fox (alt. plural in a-). ejigi ' sill.
eboe ' trace. eCome ' appetizer with food.
ebuk ' gong. eCobe ' eraser.
ebume ' pile. eCo ' something pounded.
e6ak ' hair at the end of tail. efu ' cut, wound.
efe ' sloth. efime ' wicked spirit, witchery.
efun ' foreign place. efooe ' similarity.
egae ' trial, beach. efini ' that which locks.
egak ' cactus. egbu ' kind of food from corn.
ego ' forked-stick. egbuoe ' momentary closure of eyes.
egba ' box (on ear) fist. ejio ' pile, heap (alt. plural in n-).
egbaoe ' strength, power. ekae ' crab.
egbe ' menstruation. ekit ' widowhood.
egboge ' creeping. ekii ' madness.
ejak ' pride, pomposity. ekum ' kite.
ej4m ' kind of tongs. ekpe ' brush.
ejigi ' pond, stagnant water. ekpuo ' calabash with neck.
ekara ' iron implement held by priest. erim ' bitter yam.
ekioi ' stammerer: small wasp. esare ' latrine (place).
ekube ' conversation. eti ' advice.
ekume ' function, office. epi ' cow's horn (drinking vessel).
eme ' neck.
enae ' rainy season.
enare ' kind of spider.
enok ' fight.
enuo ' bed.

eyu yam.
DIALECT OF BENDEGHE AYUK, or /Bindege/,
erio sign.
in Ogoja Province, Eastern Region,
esore clay plate.
Nigeria.
esup court.
INFORMANT: MOSES ABANG ETA
ewao implement for plucking cocoa.
eyak rubber.
 -210 -
e- to bi- con't.
Tone class Tone class
44. ,.
or or
ebe escape. ebaoe squirrel.
eblk path; something partly filled. eblge flying squirrel.
ebln dance. ebit ceremony after death.
ebop parcel. eboge second occasion.
V.
eciri sneeze. ebuoe stack of yams.
ef une red fly. efup fear.
efiffam kind of tree. efloe key.
eflO stomach. efige fan (alt. plural in n-).
efap floor, brooklet. efiri screw.
eflp polygyny. egome plantain.
egbuk ravine. egan story, parable.
ejimi custom. ejlp lake.
ejo little brush. ejo civet cat.
ekon song. ekebe small sickle.
ekop ladle. ekime small basket.
ekii cry, death. ekun adultery.
eribe tongue. . emaoe sickness.
eru manner, behavir:-.7. emege ankle, wrist.
esikb0 ' mud or wooden pipe. etan ant-hill.
eyat chin. etiin hair left after barbering.
eyem wound. eyom hoe.
eyilk wooden drum. eyu ray.
eyum voice.
Tone class
or
Tone class
ebore yellow scum. or
**.

egome ' insanity. egom ' jaundice.

edida round basket. egak ' kind of game.
ego ' hawk.
egbao ' bell. Tone class
or
A* egbogbok something old.
ebip ' guinea pig.
A egbiigi someone moving sluggishly.
ekpiri ' small wooden drum.
A ejojop stupid person.
erop ' spear.
A ejume ' low bred fellow.
esebe ' sand.
ejiigi vapor, steam.
esikok ' hiccough.
ekit ' kind of soft coco yam.

* A's indicate that I think they are the Etung way of handling Adjectives; they
need a following noun to make sense, but they act like nouns themselves in that they
perturb the following noun according to normal N + N rules.
 -211-

e- to bi- con't.
Tone class
.V
or
edimba narrow necked pot.
eginlm something or (someone) bent over.

Other tone patterns.

ebare smoothing instrument (alt. plural in a-).
ece pepper pestle.
A?
/A ebime greyness.
efomnte puff adder.
/ ej agba big needle for sewing sacks.
A? ejiji stupid person.
A? ejogejoge rude person.
A? ejuoejuoe that which is high and unsteady.
el4mamgba intimacy.
A? ejuoguruo that which is tall.
ekogarim rainy morning.
ekumorok kind of fish.
AA
ekli owl.

ebigibigi shout for gladness.

D eresi rice.

eruoa water-yam.
 -212- a
e- to a- (85 items)
or
n. n
or B
, , ,
ebi fox (alt. pl. in bi-). ebe breast.
, , ,
edil maggot. eb un heap (alt. pl. in N-). I]
, ,
efabe wing. A? ' eCak something torn.
, ,
a
,
efu marble. efao shoulder.
, ,
eglre ,
ring, girdle. ekim half, part which is short.
egut ,
parrot (alt. pl. in bi). (alt. plural in n-).
egbaogbao ,
big vessel. zinc. ,
eko ,
stem (of leaf).
n
, , ,
ej ik kind of maggot. ekup tick.

ejooe ,
thorn. ,
ekiik
,
cooking-stone. o
, ,
ejti
,
native pear. emao palm-leaf.
, , ,
ekek molar, black tooth. enyin dot.
, ,
ebare ,
finger-nail. eoan bat.
, , ,
A? etlm old thing. erok sore.
, ,
eyioi ,
root. esikoo wooden or mud pipe.
, ,
eyem wound (alt. plural in bi-).
.. A , ,
or eyio tooth (plural am*0).
, , ,
A? ebage something divided, e.g. yam. eyit eye (plural amit).
, ,
A? ebare ,
something growing wild. eyoo drop (e.g. of water) eczema.
,
ebi pot: placenta.
, V
ekae basket. or
, , ,
ekat leg, foot. ebat side, part.
,
ekm ,
stump of tree. ebip weal.
, ,
A? ekpage ,
empty thing. efut feather (alt. pl. in m-).
, ,
ekpage ,
skin (of tree). egbem hollow place (not filled up).
, ,
emage ,
reed. ej are foot-print, foot step.
; , ,
ej i egg.
,
emap ,
dint, dent. B
, ,
emage ankle, wrist (alt. pl. in bi-).
, ,
enyot hair.

El

,..
 -213--

e- to a- con't.
IA
or or
ebare hollow rock. enyio ' guinea fowl.
A? ebare white thing.
...v
ebare clay pot. or
A? ebi red thing. egbo ' kind of pin used in decorating
ebiri hole, pit. gourds and calabashes.
eftoi broken crockery.
efare shoulder blade. ' Others.
"
A? efebe that which is light. ' A? ebege ' something broken in
efoge empty space. pieces.
efu branch of plant with efae wooden spatula (for fufu).
1 .......
leaves. efue ' yellow fly (not harmful).
......
efi ' boil. I
egbeme ' hollow place (not filled ill
A? egbogbok ' something old e.g. I
up) term of an abuse. i

An..
machine, (alt. p1. in bi-). ' ejagba ' big needle for sewing
I
A? egbligi someone moving sluggishly ' sacks (alt. pl. in bi-).
as a leper with no toes ' A? "" ejoge ' that which is broken
(alt. pl. in bi-). I
into pieces.
.......
ejak I
a laughing bird with ' A? enyage ' black thing.
......
long beak. ,
etugi ' piece of burning wood.
ekap ' mark made by ringworm. I

eklp ' bone. I

A? ekimi ' that which is cut. I

ekpak ' grasshopper. I

eEigi ' crooked (alt. pl. in b). I

eslin ' smelling insect, family I

of ant. I

etambuo mat (for drying beans).

eti wood, tree.
eyege palm-rod.

A? efoge ' that which prevents.

 -214-

N(i)- (32 items) 'mass' nouns

or .V
or
nabe hopping.
mbume rafters.
ndlm beans game.
ngun world.
nduge beni-seed.
nnyilo vegetables.
mf an number.
nsi mud.
mf em flower.
ngom fine.
or
mgbeyuk severe fever.
mbim ' fruit with black paint.
nkoge a tree fallen across road.
nnycre shea-nut.
Others
nntik porridge water. .0 1
mfiri ground-nut.
nnyage black juice leaf.

or

nciri gunpowder.
mf a green leaves.
mfune shredded leaves.
njue sun.

mkpaoe cover for water pot.

ntao refusal, objection.
nnyub e sweat.

or
fl
robot pulp of fruit. LI

mbuta rain.

ncen noon.
mfam kind of juju.
I

mfep wind.
4

mfili scum.

ngum places.
nnti hair.

nson shame, shyness.

a
 -215-

N(1) -to N(1)-

(8 items)

.14. 414.

or
nkoge f
tree fallen cross road.
nnyore f
shea-nut.
ntip ' sticks used for mud houses.

414. do., 0. .0 .
i
or
nkap ' kind of weaving.
nkat ' kind of fruit (edible).
nkik4t ' small mortar.

or
nsa ' draughts.

Others
mbanamba""' f
swallow (bird).
 fl
-216-
N (i) to o-(66 items)
*a. S.
or or
U
mbae two spotted palm civet. mbue goat.

mbi name. njo dog.

f
ngon horn-bill. nju house.
f
ngu skin. nka load-pad.
njm back. nkim f
circumcision.
nj ok elephant. nkiin-
f
nkaoe bamboo, rib. dak peacock.
f
nkon bee. mtAk glass, tumbler.
f
nnyagakokon big millipede. nnyo snake.
nnyam meat, animal. nnyop ' porcupine.
nsum drill.

nwet book, school. or

nyip alligator. mbar) ' kernel.
mbi ' red-feather bird.
or mbUi ' louse.
mba ' road (could also have a ndae ' hernia (pl. orae).
concord). ndem ' dream (pl. orem).
1
mfuo ' buffalo. ngume bush-pig.
f f
njio fly. ngao chest.
f f
nke white ant. njae hunger.
f
nkem mamba, (snake). nkok ' fowl.
nkm ' python. nkam ' drum, (tied with skin).
nko ' large snail. nkop ' box.
nkoolm ' bush cow. ntbe f
okro.

nku ' rufous-bellied rat. ntine ' penis.

f
mkpo ' shoe. ntuo throat.

nsim ' genet. nyok ' chimpanzee.

1
nsi ' fish. nsun red-deer.
ntUt ' female private part.
ntam ' trap.
 -217-

(i)
N - to 0- con't

or
T
mbim mosquito.
mboe T
heart.
T
robot sandfly.
T
nce small snail.
T
ncit garden-egg.
T
ndoo jigger (pl. oroo).
mfem T
cockroach.
7
ntiir) navel.

.40 " .,....

or
ngare ' pepper.

Others
I
mbiantuo-"" T
spitting cobra.
I
mfiri "" /
ground-nut.
nkupeyit-"" T
eyebrow.
 -218-

N(i)- to a- (22 items)

.. ...
or i

njom ' juju, sacrifice. T

nkap ' fruit. i

mkpoe ' corner. i

mkpUtli ' sludge (waste water from i

palm-nuts). V

nsom ' paddle, oar. i

nsure ' blister. i

n6ok ' kind of grass. I

or V

mkpaoe cover for water pot.

ntae stone. V

ntare main beam. V

ntme ' pole (used in boat) .

or
ndap ' branch. 1

mfut ' yellow fly. I

mkpun ' club, baton. I

nsan ' thunder. I

nsum ' finger (of plantain). I

nt4m ' end, time. I

I
.V ...
or
mbue cover, lid.
ncot trouble.
nsure ripe plantain.

or
nje ' leaf.

Others
V
.^..
ndumoto ' kind of gourds
(pl. arumoto) .

.,../1116,,
 -219-

bi- (23 items) 'mass' nouns

or

bia'ao visit.
biji body.
bika0 wild nut with strong
smell.

birimtim violent.
bisi0 wrestling.
bitop muddy lane.

or

bibi small pot.

bifame scolding.
b imo filthiness.
biruk wickedness.
bisek joke.
bisere after.dinner drink.
bita war.

or
1
biC'oe bad luck.
I
bikep spy.
1
binot evening visit.
0 bisuge 1
kind of fruit with
U
sticky juice (edible)

or

bifigi musical instrument.

b iko plant with strong odor
(edible).
bimeoe senses, cunning.
bisu alligator-pepper.

...v
or

bidlme small mortar.

bilam sluggishness.
 -220-

a
bi- to N (35 items)

... .0...
or or
I I
bisoo ' hip. biCage toe-sore.
i
bisu ' dove. t
biome I
calf of leg.
t
bit4me ' knife, machete. bifage ' piece of wood with fire.
t
bifure ' cane rope.
...
or bifio ' space between toes and
t
D
bibore ' kind of bird. finger.
t I
bifam ' wedge (for breaking bikebe rod for fishing.
wood). t
bikon ' wasp, or mason fly.
D
I t
bifige vampire-bat. bikpa ' whip.
bikae I
pangolin. I
bisuge I
weaver bird. o
t
bikoe ' curve. bisao ' frog with long legs which
biru ' defective pupil of the eye. ' jumps very far.
t
bise ' plate, basin.
i
bisooe ' spoon, fork. or
1 I
bitin ' calabash. biCit sunbird.
t t
bikpae ' brown bird with yellow bifet mongoose.
t
feet.

bine ' image. ' Others

t n. I %. a. a. t
binon ' bird. bikonakom a singing bird.
bikumnsi ' Otter shrew.
...
or
t
biki ' tail.
,
bikli ' water beef.
0.
t
bise ' duiker.
,
biya ' calabash with handle.
t

...V
or
biter) ' plait.
biton ' hair on woman's pelvis.
 -221-

bi- to a- (4 items)
or

bisao ' Sticks for propping yams.

or

bitare ' A tree with small leaves.

or
bibar) ' horn.

vein.

(o)
N to a- (20 items)

or or

nguoe young-man. mgban servant.

lazy person. nse ' father.

nne person.
nsop girl, lady.
ntem- friend.
I
nCok fool.

0 ....
or

nal dumb person.

G njene stranger.
njop fool.
(]
C] nkae wife.

mkpak poor person.

nstio slave.
0
or

ndum husband.
mf4p co-wife.
nje witch.

*. V
or

mfone owner.
mgban powerful man.
nnyen mother.
 -222-

0- (no plural)

or or

oboma plant with paint. oda ' kind of snuff.

ofi pawning.
ojo narrow base water pot. Others
oka skill. . oso " ' sky.

ono good, good thing.

otasi creeping plant with
bitter leaves.
otem friend-ship.

II
or
obi evil.

obone sleepy'dust'in eyes.

oCu bitter leaf.
J
V
ofu day.
V

oke marriage.
osam diarrhoea.
osi chicken-pox, measles, fish.

6. A
or
V
obibe tear (from the eyes).
ob im group hunters.
V
ofi selfishness.
V
otue chieftaincy.

.11
or

obp mushroom.
ogm ' backbiting.

III
or

omao fruit used in killing

fish, treating measles.
otae kind of game.
 -223-

o- to N (1) - (3 items)
S. A
or
okun ' fire-wood.
osio ' mango tree.

Others.

oki tailor ant.

o- to a- (44 items)
. v
or or
obukpoO trumpet. okpi ' boat, canoe.
ofun leprosy. osep soup.
ogm market, week. osim web, (spider's).
1
ogbe special dance.
or
ojU native pear tree.
okoro
okut wooden barrel.
down-wine.
okpa first born.
okpere gourd used as plate.
9 okpuga six pence, (pl.- a-
osere down, under.
which is money). osogo stack (of yams).
omone monitor, (kind of lizard).
A%
or
oraga bridge.
orer
oda ' drying platform.
sponge.
otere vulture. A^
or "'
otun funnel.
ogbiri chair.
otU jar.
oya stomach, pregnancy. Others.
o54m grave. okuOa chameleon.
oyUk arrow. okpiton umbrella.
A
or

okm cotton tree.

orikao matches.

ft.e %A%
or or
obi palm-tree. okik cheek.
obo arm. okoe cough.

9 on ' face. okpooe quarrel.

ofak ' broom. ooa cat.
ogbap ' armpit. orao fufu.
ogbu ' shelf. ota thigh.
okak ' special type of wood. otuo ear.

.111,4
 -224-

.(a)
- to o- (3 items)

or

mouth.

or
mbuk ' forehead.
mmi nose.

N(a)- to a- (9 items)

or
ndini ' boundary.

or
mbum ' new palm-leaf. v

nd4m ' fairy. 1

ndigi ' rope. v

mfuk ' hole. v

or
t
n6ok pestle. 1

mf en ' shaft v

1
ngun gun, fire.
njan ' medicine. v

V ..
or

z-77"W77-7'77=n,.-, . '
 -225--

a- 'mass' nouns (62 items)

%A
or or
ajobe miscellaneous things. abio excrement.
abakpa Hausa people. acik carbon, (from cooking place)
abin pus. afige urine.
abUrU dirt in the water. afom fat.

ac am inferior cloth. akaO salt.

ace game, playing. akin bahama grass.

agore apology. as an syphilis.
agura ring of irc the legs atoo ash.

of circumc.....sd woman. awao smell from fresh fish.

agara mucus stuck in the throat,
or Cont.
agbaoe side pain, lungs.
akpabe threat.
ajlo lever.
ami MUCUS.
akpao hero, brave person.
anabe partiality.
akpara prostitute.
anik epilepsy.
or anwam seizure.
abu dew. ap a the least piece of money.
adun kind of corn. ariri temporary decorated pillar.
afuO juju horn. asigi tickling.
agba verandah. ason saliva.
akaba potash. asuge inciting to anger.
akpa- atao Indian bamboo.
small pox.
nyao atem beads.
akpe palm-tree. atigi doing something without
ar4m debt. consultation.
atim lungs. atop swamp, muddy ground.
awowo little bells. ayebe imitation.
ayuo blood. ayigi removal.

%,, %%,
or Others
I ,%A I
aam palm-wine. abii race.
I AA I
adagat kind of yam. ace cutting sticks.
I %%A I
agun rust. aceme one who tramps heavily.
I %%A I
ajiri round beans. ac ±be twister.
I
aron brain. afere dust of roasted yam.
%%A I
asabe ' itching leaf with dangerous anabe repairer.
hairy fruit.
ayu willful accusation.
i
1

-226-

a- to a- (14 items)
or '"
y
aga needle.
y
akaCak rattle.
atma y
old farm ground.
y
atore mahogany.
y
ad" night.
y
aya river.
y
ayip water.

or
adiidii ' long haired goat.
afat ' set of twins.
akpakak ' chair.

.... I...
or

afek ' hair at the temples.

afifoo ' handkerchief.

agira ' wooden instrument.

agbara ' soft red wire.

ba- 'mass' or 'abstract' nouns (6 items)

,..... . .... ....

or
Banao ' deceit.
Batm ' beating.

Bafao untruthfulness.

II III
or
Bakpii ' wickedness.
 Appendix VII

PRELIMINARY DESCRIPTION OF
SOME VERB STRUCTURES IN ETUNG1
Tom Edmondson

0
Introduction
The focus of this paper is upon some segmental and tonal structures of
2
verbs in Etung , a Cross-River Bantu language of Southeastern Nigeria.

The verb structures handled here are those found in the predicates of
independent and dependent clauses (i.e. not subordinate clauses) with the
exception of (independent) imperative clauses. These clauses are so decribed
because of their distribution:in the sentence.
Verb structures are discussed here under two heads: basic forms (section
1.) and amplifications of basic forms (section 2.). On the basis of structural
differences two sub-groups of basic forms are set up: (a) indicative-prefixed
forms (section 1.1) and (b) minimal and suffixed forms (section 1.2).
Features relevant to each section, though not necessarily treated in this
order are: (1). Principal classes of verb stems; (ii). the segmental and CV
structures of basic forms of verb structures and their amplifications; (iii).
The relations of these structures to corresponding tonal structures, and (iv).
the varying nature of these tonal structures in terms of predictability and
unpredictability.
1. Basic Forms of Verb Structures.
1.1. Indicative-prefixed Forms.
1.1.1. The structure of any indicative-prefixed form consists of an obliga-
1

tory pronominal prefix, followed by an obligatory indicative prefix, followed

by an obligatory verb stem. Two classes of verb stems are set up on the
basis of tonal differences and since these differences most clearly appear in
indicative-prefixed forms, it is more appropriate to begin with these forms
than with the minimal forms (see section 1.2). The two classes of verb stems
are designated A and B respectively, and some of their tonal differences are
set out in Table II. But first, the segmental and CV structures of the in-
dicative-prefixed forms are shown in Table I.

-227-
 -228-

TABLE I. Segmental and CV Structures of Indicative-Prefixed Forms.

Pronominal Indicative Verb Stem of
D
Prefix Prefix Class A or B

C CV
CV V CV CVC
Structures CV CVV
CVCV

1.1.2. Given the indicative-prefix, the tonal structure of the indicative-

prefixed form is pi dictable and can be stated as follows: -
G32
The following tone patterns were earlier found:-

Tone 3 2 1
Patterns Syllable Syllable Syllable
1 LLL LL L
2a LHH LH LH
I I

2b 'HH 'H

3 HLL HL HL
4 HHH HH H
5 LHL La -

6 LLH LLH -

7 HHL HHL -
t fil
8 HIM H't H H:H
r\
9 HLH HLH -
1 'TN i

10 LH'H LH'H 1 -
1
0 i eTh
11 ' HH'H 1-1Hi li

Another tone pattern, H', could now be added.

3
a
The tone patterns of the verb stem are determined by:-

a. the tone pattern of the indicative-

prefix, and
b. the class to which the verb stem be-
longs, whether A or B. B
The tone pattern of the pronominal prefix is determined by the tone
pattern of the indicative prefix in all persons except 3rd person plural where
the tone pattern is invariably High. In Table II, which illustrates these

wr-
 -229-

patterns of co-occurrence, it is to be noted therefore that the tone patterns

of the pronominal prefixes in the left-hand column represent all persons ex-
cept 3rd person plural.

TABLE II. The Relations of Segmental Structures to

Tonal Structures in Indicative-Prefixed Forms

Tone Patterns of 1

1
Tone Patterns of ! Tone Patterns of Tone Patterns
Pronominal Prefixes ! Indicative Prefixes 1 Verb Stems of 1 of Verb stems
Class A of Class B
t

L H (kf m6 kA) LH '1111

t t

H .1-i (16) LH .11H

1

H L (kA) H.H HH
L HL (kk.) HL HL
i

The meanings of the Indicative prefixes are as follows:-

ki 'present continuative'
4 , - .
mo mo 'future negative'
. t ,
mo occurs with 3rd person pl; 'mo elsewhere
ka kA 'simple past negative'
kA occurs with 3rd per. pl; ka elsewhere
. - .
ro do 'future negative repetitive'
dO occurs with first per. sg;rO elsewhere
ka 'negative optative' [or negative cf obligation]
The pronominal prefix segements are as follows:-
I sg. N-
II sg. 0-

III sg. a-
1 pl. e-

II pl. o-
III pl. a-
Third person singular and plural indicative-prefixed forms are nearly
always distinguished by tonal differences over the pronominal prefix segment
or over the indicative prefix segment where the latter requires a special
tone in the 3rd person plural.
 -230-

Second person singular and plural indicative-prefixed forms are distin-

guished by the plural verb suffix -wfin -fin. The first of these allomorphs
occurs following a vowel, the second following a consonant.

1.1.3. Examples for

TABLE II: Class A Verbs; 3rd pers. sg.
'he-is-selling'
.
a- mo-gure 'he-won't-sell'
6-ka-gu re 'he-did not-sell'
.!
6-rii-gu re 'he-will not again-sell'
a-ka-eire 'he-should not-sell'

Supplementary Examples:-
'he-is-washing'
. .!
a-ka-su
'he-did not-wash'
'he-should not-wash'

Examples for
TABLE II: Class B Verbs.
!
(6) a -ki bome 'he-is-putting on D
(7) a- mo- bome 'he-won't-put on'
.
(8) a -ka-bome 'he-has not-put on'
(9) 'he-will not again-put on
(10) a -ka 'he-should not-put on'
Supplementary Examples:-
.
a -ki ka 'he-is-staying'
-ka -k 'he-did not-stay'
a -ka -ka 'he-should not-stay'

The 3rd person singular forms cited above represent tones of all other
persons except 3rd person plural. Under section 1.1.2. above, it was observed
that 3rd person singular and plural are distinguished nearly always by tone,
either over the pronominal prefix segment (compare examples 1 and 5 above with
examples 11 and 15 respectively below: similarly, compare examples 6 and 10
above with examples 16 and 20 below.) or over the indicative prefix segment
(compare examples 2 and 3 above with examples 12 and 13 below: similarly, ex-
amples 7 and 8 above with examples 17 and 18 below.). Examples 4, 9, 14 and 19
show the exception to the general tonal differentiation of 3rd person singular
 -231-

and plural, it will be noticed that while in 3rd person plural forms,
the relations of pronominal prefix and indicative prefix tone patterns do
not generally correspond to Table II, then relations between the indicative
prefix tone patterns and verb stem tone patterns are perfectly regular.

TABLE II: Class A Verbs . 3rd pers. pl.

(11) 'they-are-selling'
(12) a-mo-gure 'they-won't-sell'
(13) a-k6-eire 'they-have not-sold'
. . .1
(14) a -ro-gu're 'they-will not again-sell'
(15) a.-ka-g6re 'they-should not-sell'

TABLE II: Class B Verbs.

(16) a -ki - bome 'they-are-putting on'
.
(17) a -mo bome 'they- won't -put on'
!
(18) a -ka bome 'they-have not-put on'
I
(19) a -ro -borne 'they-will not again-put on'
(20) i-ki-b6me 'they-should not-put on'

1.2. Minimal and Suffixed Forms,

1.2.1. The minimal form of the verb consists of an obligatory pronominal prefix
and an obligatory verb stem. This form may be suffixed, in the case of verbs
having CV stems, by a consonant or vowel suffix, and in the cases of verbs hav-
ing CVC, CVV or CVCV stems, by a vowel suffix only Di-syllabic verb stems
exhibit mono-syllabic stem allomorphs before suffixation. CVV verb stems have
the stem allomorph CV and CVCV verb stems have the stem allomorph CVC, e.g.
a-yue 'he killed'
(verb stems, their allomorphs and their meanings are underlined), 'he-
kills-habitually'; a-b6me 'he-put on', 6-b6m-a. 'he-puts on-habitually'.

1.2.2. In the phonological analysis of Etung, pitch has been treated at the
level of the phonological word in terms of tone patterns, so that each phono-
logical word is marked by a tone pattern. The phonological word is not, how-
ever, necessarily congruent with the grammatical word and in fact the constitu-
ent parts of the verb are themselves regarded as phonological words. For example,
the pronominal prefix is regarded phonologically as a word and therefore carries
a tone pattern. The same is true of other verb prefixes and of the verb stem.
 r
-232-

Upon the same principle the suffix too is a phonological word except when it
ii
is not a vowel. Li
The suffixed verb stem might then be treated as a complex unit bearing
two tone patterns, i.e., the tone pattern of the verb stem plus the tone
pattern of the suffix. But it facilitates description to regard the verb
plus suffix as a complex unit, both phonologically and grammatically. As a
complex phonological wor.! comprising two such words, it is regarded as being
marked by one tone- pattern. As a grammatically complex unit, here termed a
verb core, its tonal structure generally is predictable, being determined by
the tone pattern of the suffix morpheme and the class to which the verb stem
belongs. Table III illustrates the predictability of tone patterns over verb CJ

cores.

TABLE III. The determination of tone patterns over Verb Cores.

ITone patterns of Resultant Tone Patterns Over Verb

suffix morphemes Cores shown at Interstices
Class A Verbs Class B Verbs.
H L H H H
, ,

'H H', H ICH

-
ICH L 11.11

The core whose tone pattern is LH!H is confined to Class A verbs. Some a
exceptions to these rules are stated below under Suffix Morphemes.

Suffix-Morphemes.
These suffix morphemes and the tenses they indicate are now listed to-
4
gether with the tense abbreviations employed in Table IV below.
a
-á Past habitual (PH) [also -k for certain verbs with CV stems]
f,
-a -'a Present habitual (PrH) [also -k for certain verbs
with CV stems.]
_!A
occurs with third person plural
-a occurs elsewhere
(in class B verbs, the suffix -a produces tone
pattern LH over the verb core, an exception to
the rules above.)
D[
 -233-

x ti. -x -n Pluperfect (PP)

-n occurs with all persons except 3rd person plural
of class A verbs.
-IA occurs elsewhere

-x Pluperfect variant (PPv)

The symbol x represents:-
(i) reduplication of high front vowels where
these occur in the verb stem,
(ii) the vowel -e elsewhere.
(in the third person plural of class A
verbs the suffix -A produces the tone
pattern HH over the verb core, an ex-
ception to the rule above)
1.2.3. But while some degree of predictability exists in the formation of
the tonal structure of the verb core, the co-occurrences of pronominal prefix
tone patterns with the verb stem and verb core tone patterns, has not been
found to be predictable. [ "Table IV suggests that these relations are arbitrary'!]
For example, in the first person singular of the Past Habitual tense (PH), the
tone pattern of the verb core derived from a Class A verb stem is LH; so too is
the tone pattern of the verb core in the first person singular of the Present
Habitual tense (PrH); yet in the first case, the tone pattern of the pronominal
prefix is L while in the second case the tone pattern of the pronominal prefix
is H. But in the third person plural of the Past Habitual tense (PH),while
the tone pattern over the verb core is still LH, the tone pattern of the pro-
nominal prefix is now HL. Equally, difficulties of this kind arise with verbs
with Class B stems. In the first person singular of the Past Habitual (PH),
the tone pattern of the verb core is HH; so too is the tone pattern of the verb
stem in the first person singular of the optative (OP). Yet in the first case
the tone pattern of the pronominal is L, while in the second case the tone
pattern of the pronominal prefix is H. However in the third person plural of
the Past Habitual tense,while the tone pattern over the verb core is still
HH, the tone pattern of the pronominal prefix is now HL.
rR

1.2.4. In Table IV irregularities in the over-all pattern of verb stem and

verb core tone patterns are marked in two ways. First, dotted rings indicate
interstices where the expected tone patterns do not appear. For example, in
 -234-

o
vectors IP, the dotted ring shows that in the third person singular of the
immediate past /future tense (with both class A and class B verb stems), the
, G
expected tone pattern LL does not occur over the verb stem segment. Similarly,
in vectors 0 (with both classes of verb stems) it might be expected that
third person plural would exhibit tone pattern HL over the verb stem following
o
tone pattern L over the pronominal prefix. But this is not the case and the
dotted rings show this. Second, complete rings at interstices indicate either, o
(i) that the expected tone pattern over the verb stem or core occurs in an un-
expected position, e.g., verb stem tone pattern in third person plural of vector
0 (both classes of verbs), or (ii) that an unexpected tone pattern appears in
an unexpected position, e.g., in vector OP (class A verb stems) the verb stem
fl
U
tone pattern in third person singular is LH following pronominal prefix tone
pattern L. Similarly in vector PPv (class A verb stems) the verb stem tone
D
pattern in third person plural is HH following pronominal prefix tone pattern H.
It will be observed that all these irregularities only affect third persons
singular and plural which are thus always distinguished tonally. There are three a
types of distinctions:
(i) distinctions dependent upon dissimilar tone patterns n
over both pronominal segments and verb stems or cores
(see Table IV: Class A verbs, vectors IP, OP, PPv,
II
and PP; Class B verbs, vector IP);
(ii) distinctions dependent upon dissimilar tone patterns
of the pronominal prefixes only, while tune patterns
a
over verb stems and verb cores remain identical (see Table
IV: Class A verb stems, vectors PH and 0; Class Bverb stems,
0
vectors OP, PH, PPv, 0 and PP), and
(iii) distinctions dependent upon dissimilar tone patterns a
over the verb cores only (see Table IV: Class A verbs,
vector PrH; Class B verbs, vector PrH). These differences
of differentiation between third persons singular and plural
are supporting criteria for distinguishing between the two
classes of verb stems.
It has been noted above under 1.1.2. that second persons singular and
plural are distinguished by the plural verb-suffix -wlin rl, -tin. This suffix
is secondary in position to the tense suffixes when they occur.
 -235-

TABLE IV:-
FORMATION OF MINIMAL AND SUFFIXED FORMS OF THE ETUNG.VERB

Pronominal Tone Patterns of Verb Stems and Cores Indicated at Intersects

Prefixes
Per- Seg- Tone T E N S E S: Class A Verb TENSES: ClassBVerb Stems
sons ments patt- stems
erns .12_ .ag, PH PPv PrH 4 PP _a OP PH PPv PrH A PP

I sg L - - LH LH - HL EETTIT - HH HH - HL H'1 H

1H
H LL H' LH - LL HH - - LH - -

IIsg o- L - LH LH - HL LH .H HH HH - HL H't il
...,
t
H LL H. H - LH - - LL HH - - LH -
n -- 1

U 1-
I pl e- L - - LH LH - HL LOH - HH HH - HL H' H
1

1
H LL H.H - LH - - LL HH - - LH - -
I

4- /-
IIpl o- L - 1- - LH LH - HL LH.H
t

%....
- HH HH - IHL H'H
t

H LL HIH - LH - L HH - - LH - -

IIIsg a- L E) (D LH LH - L LHIH
..-
40 Ci HH HH 1 - HL H'I H

H (2eN -1 - LH - CI, :) LH
.._/ k.....1

, 4 te- \ .--- .,- ...

r
IIIpl a-

H
L -

LL
-

H1H
1/4-.i

LL HH
t_)
...)

-
C._
...,

HH
_
rt---
.1 `J
100 EHL e
_

el
_ _ _ _
3 HL r"--- - _
I

Notes: (1) In vector PP (Class A verbs) the ligature linking H with 'H
indicates that these two tones occur over the same syllable.
(2) Tenses constituted by minimal forms [by change of stem tone]
are indicated at the heads of relevant columns by abbreviations
in italics.

io
 -236-

Partial Citation Paradigm

In this partial citation paradigm from Table IV, examples are restricted
to first person singular (which is representative of first person plural and
second persons singular and plural as far as tonal structure is concerned)
El
and to third persons singular and plural. Citations are made from both classes
of verbs, A. and B.
Class A verbs: first person singular.
Reading from left to right across the first row:
. v
PH n-su-k 'I have been washing'
. . A
PPv n-su-e 'I had washed'

0 n-su 'I washed'

% . A A
PP n-su-e 'I had washed'
Reading from left to right across the second row:
. .
IP n-su 'I have just washed/I am just about to wash'
A AI
OP n-su- 'I should wash'
A 11
PrH n-su-k 'I wash'

Supplementary Examples:- with two-syllable verb stems

. .. A
n-gur-a 'I have been selling'
. .. ..

n-gur-e 'I had sold'

. A
n-gur-e 'I sold'
. , ..
n-gur- e 'I had sold'
A . .
n-gur-e 'I have just sold'
t

n-gu' r-e 'I should sell'

A . A
n-gur-a 'I sell'

Class A verbs: third person singular:

Reading from left to right across the ninth row:
IP a -su 'He has just washed/is just about to wash'
. V
OP a -su 'He should wash'
PH
. v
a -su -k 'He has been washing'
u
. ... A
PPv a -su -e 'He had washed'
. n
0 a -su 'He washed'
. I'
PP a-su-e 'He had washed'
Reading from left to right across the tenth row:
A V
PrH a-su-k 'He washes'

u
 -237-

Class A verbs: third person plural:

Reading from left to right across the twelfth row: (It will be noticed
that the eleventh row is not occupied)
IP a-su 'They have just washed/ are just about to wash'
/ A /
OP a-su' 'They should wash'
/ A A
PPv a-su-e 'They had washed'
t
p t . ,
PrH a-su -K 'They wash'

O a-su 'They washed'

/ A
PP a-su- ?ee 'They had washed'
Reading from left to right across the thirteenth row:
PH a-su-k 'They have been washing'
Class B verbs: first person singular.
Reading from left to right across the first row:
PH n-ka-k 'I have been staying'
. ,
PPv n-ka-e 'I had stayed'

0 ft-ka 'I stayed'

, t.
PP n-ka-.e 'I had stayed'
Reading from left to right across the second row:
A
IP n-ka 'I have just stayed/ am just about to stay'
A A`

OP n-ka 'I should stay'

A V
PrH n-ka-k 'I stay'

Supplementary Example:
mI I
-bo m-a 'I put on'

Class B verbs: third person singular

Reading across from left to right the ninth row:
,
IP a-ka 'He has just stayed/ or is just about to stay
, .
OP a-ka 'He should stay'
... -
PH a-ka-k 'He has been staying'
. .. ..

PPv a-ka-e 'He had stayed'

. ,
0 a-ka 'He stayed'

PP ,-...-re 'He had stayed'

Supplementary Example:
\ A
a -bom -a 'He has been putting on'
Reading from left to right across the tenth row:
A V
PrH a-ka-k 'He stays'

Supplementary Example:
U
a -bom -a 'He puts on'
 -238- 0

Class B verbs: third person plural.

Reading from left to right across the twelfth row:
IP -ka 'They have just stayed/or are just about to stay'
OP a -ka 'They should stay'
(Class B verbs: third person plural. cont'd)
PPv a-ka-e 'They had stayed'
PrH A-kAl'-k 'They stay'
, ^
0 a-ka 'They stayed'
, , !,
PP aka- e 'They had stayed'
Supplementary Example:
, ,
a-bom-' a
t

'They put on' 0

U
Reading from left to right across the thirteenth row:
^ ,
PH a ka k 'They have been staying'
Supplementary Example:
a-bom-a 'They have been putting on'

2. Amplifications of Basic Forms.

2.1. Both indicative-prefixed forms and minimal and suffixed forms may be
amplified by optional prefixes. A distinction is made between the repetitive
optional prefix, which is at once affected tonally by the tense it amplifies
and yet also determines the tone pattern of the verb stem, and initial optional
prefixes which have no effect upon the tonal structure of the verb, other than
that of adding an extra tone pattern to it.

2.2. The repetitive optional prefix kpo occurs invariably

before the verb stem. Verb cores are not found following this prefix and
those tenses having verb cores are reduced by de-suffixation following this
prefix. Table V displays the range of amplification of basic forms by kpo.
It should be noted that all indicative-prefixed forms are included in the ex-
treme right column.
TABLE V. Repetitive Amplifications of Basic Tenses.

Basic Forms
IP OP PH PPv PrH f) PP Indic-Prefix
Repetitive
Optional x x - x - x x i x
Prefix
kpo

Note:- For abbreviations, see Footnote 4.

 -239-
B

LI The behaviour of the repetitive optional prefix is interesting in that (i)

its tone pattery, are determined as though it were a class B verb stem,
and (ii) it has its own characteristic effect upon the tone pattern of the
following verb stem. It should also be noted that when following an in-
dicative-prefix, although its tone pattern is determined by the indicative-
fl prefix tone pattern in the same manner as a class B verb stem, yet it is

now the optional-prefix, not the indicative-prefix that determines the tone
pattern of the verb stem. The relations involved here are stated in Table
VI.

TABLE VI. Verb Stem Tone Pattern: Following Optional-Prefix kpo.

Tone Tone Patterns of Verb Stems

Pattern Class A Verb Stems Class B Verb Stems
of kpo

H LH 'H

'H LH 'H

L LH H

HL LH H

Under 2.1 it was noted that the repetitive optional prefix is tonally affected
by the tense whose form it amplifies. In minimal and suffixed forms this means
that the tone pattern normally found over a Class B verb stem or verb core in
the tense to be amplified is now found over the repetitive-optional prefix ex-
cept in the amplifications of pluperfect variant (PPv) and Pluperfect (PP),
where the tone pattern of the optional prefix becomes that of Zero tense (0).
For example, in the optative tense (OP), whether the form to be amplified in-
volves a Class A or B verb stem, the repetitive optional-prefix tone pattern
is HH, while the verb stem tone pattern is LH in the case of a Class A verb
stem or 'HH in the case of a Class B verb stem.
In the partial citation paradigm which follows, both basic and amplified
forms are given to facilitate comparison. The paradigm is confined to first
person singular and to verbs with Class A stems. Also included are the tone
Patterns of Class B verb stems and cores in basic tenses.
 -240-

Partial Citation Paradigm

Reading from left to right in Table V.

Unamplified Class B Verb Amplified

Stem/Core Tone
Patterns in
Basic Tenses
IP 'I have washed' L 'I have again washed'
OP 'I should wash' H i n-kpo-su 'I should wash again'
PPv n-sue 'I had washed' H n-kpo-su 'I had washed again '

n-su 'I washed' HL 'I washed again'

PP n-sue.' 'I had washed' H'H ii-kp6-sii 'I had washed again'
Indicative-prefixed forms.
ki n-ki-su 'I am washing' 'H ' 'I am washing again'
! ! v
mo m- mo-su 'I won't wash' 'H m-i mo- kpo-su 'I won't wash again'
. . t.
ka' n-ka-su- 'I have-not was hed' H 'I have-not washed again'
. .
ro n-do-sub' 'I won't wash a gain H 11-(18-kp6-sil (more emphatic)
ka 'I shouldn't wash' HL 'I should not wash again'

(Partial Citation Paradigm cont'd)

Supplementary Examples: Verbs with Class B stems.
IP m-kpo-bome 'I have again put on'
!
OP m-kpo- bome 'I should again put on'
^
PPv m-kpo-bome 'I had again put on'
m-kpo-bome 'I again put on'
PP m-kpo-bome 'I had again put on'
. ,. ! ,. ,.

ki n-ki-'t kpo- bome 'I am again putting on'

1 1 1
t -
'mo m- .m6 -.46 -.b6me 'I won't again put on'
, , ,. ! ... ...

ka n-ka-kpo- bome 'I did not again put on'

,. .t . ,.

ro n -do -kpo bome 'I won't again put on'

ki n. -ka-kpo-bome 'I shouldn't again put on'

2.3 Initial optional prefixes occur before the pronominal prefix. Not only
do they have little effect upon the tonal structure of basic forms, but with
the exception of kln they are severely restricted in distribution. Table
VII displays their distribution.
 -241-

TABLE VII. Amplifications of Basic Forms by Initial Optional Prefixes

Initial Basic Forms

Transform
Prefixes IP OP PH PPv PrH 0 PP Ind-Prefix
kin x x x x xx x x
- kin + tik x
+ - v.
- kin + car) x

The meaning of -1(11 is generally what might be described

as past optative,
fcr example its amplification of Zero tense (0) produces
the meaning 'should
have done', e.g., a-ji 'he went' when amplified by ki.n produces ki.n-a-ji the
should have gone'.
But its transform of the optative tense (OP) and the future
tense (tik) (see second row of Table VII) yields the meaning
of uncertainty,
e.g: g.-riPe 'they should go' when amplified by k4n produces kin-a-rte1.
'Should
they go?' (-I'm not sure'). Similarly with the future 5
I tense (tik) e.g. tik-a-
1300 ojo 'he will pluck tomorrow' when amplified by 14n
" I It
produces ktn-tik-a-
1300 ojo' 'he should pluck tomorrow -(but I'm not sure if he will)':
The future
initial optional prefix is self-explanatory and the
negative initial optional
prefix can is restricted in distribution
to the transform of the Past Habitual
tense (PH).

In the partial citation paradigm which follows, the

amplifications of Table
VII are set out in first person singular and restricted
to verbs with Class A
stems so that comparison may be easily made with the amplified
paradigm under
2.2 at the conclusion of that section.

Partial Citation Paradigm.

Reading from left to right of the first row of Table VII:
1 I \
IP ktn-n-su 'I should have just washed'
I /
OP (me A-kerA me se) kin-n-sub'
'I think I should have washed-but
I'm not sure'
I \
PH kin-n-suk 'I should have been washing'
PPv kin-n-sue 'I should have washed'
I e
PrH kin-n-suk 'I should have been going to wash'
3.
(but plans have been changed)
0 kin-n-su 'I should have washed'
*.
PP kin-n-sue* 'I should have washed'
 -242-

Cl
(Partial Citation Paradigm, cont'd)
Indicative-Prefixed Forms. 0
. ... . V
ki k4n-n-ki-su 'I should be about to wash'
!
. . .... ,
mo kin-m-!mo-su 'I would not have washed' (meaning difficult
I A
. . .1.. to ascertain.)
ka Icri-n-ka-su. 'I should not have washed'
! ' . . . .1.
ro kin-n-do-su. 'I would not have washed again'
(if it hadn't been necessary)
ka kin-n-ka-su 'Perhaps I should not wash'
(I'm not certain).
Reading along the second row of Table VII.
tik-n-suk 'I will be washing'
tik-n-su 'I will wash'
. v
kin-tik-n-suk 'Perhaps I will be washing-I have actually
planned to do so'
'Perhaps I will wash (if plans work out)'
Reading along the third row of Table VII.
v. v
cap -n-suk 'I don't wash (never do)'
.
kin-card -n-suk 'I should not have been washing'

2.4. Double amplifications are possible through the multiplication of Table V

by Initial optional prefixes. These amplifications are set out in Table VIII.

TABLE VIII. Double Amplifications Through the Multiplications of

a
Table V by Initial Optional Prefixes.

Initial Optional Repetitive Amplification of Basic Forms (Table V)

Prefixes
IP OP PH PPv PrH 0 PP ; Indic-Pref. 4
i

k4n x x x x x I x
+ , i

- k±n + tik x
+ 4, VP
- k±n + cap x
I

1
I

Note:- crosses at intersects indicate what transformations occur.

The double amplifications are completely predictable in that while the
repetitive optional prefix affects the verb stem tones and is affected by the
tense of the form being amplified in the usual way, the initial optional pre-
fixes simply add their morphemes to the morphological structure of the verb.
 -243-

In the partial citation paradigm which follows, these double amplifications

are set out in first person singular and restricted to verbs with Class A
stems so that comparison may be easily made with the partial citation para-
digm under 2.3.
Partial Citation Paradigm.
Reading across the first row of Table VIII: from left to right.
. . v
IP Icn-n-kpo-su 'I should have washed again'
. . v
OP Icn-n-kpo-su 'I should have washed again?-I'm not sure'

PPv k4n-a-kp6-sil 'I. should have washed again'

0 k4n-a-kpb-sii 'I should have washed again'

v
PP ktn-n-kpo-su 'I should have washed again'

Indicative-prefixed forms:
k4n-a-ki-I V
kI l(po-su 'I should have been about to wash again'

mo k4n-i-!mO-!kpO-sii 'I would not have washed again'

v
ka k4n-ii-ka-kpo-su 'I should not have washed again'

ro k4n-n-dO-kpO-sii 'I would not have washed again (if it

hadn't been necessary)'

ka 'Perhaps Ishould not wash again'

The second and third rows of Table VIII are omitted from the Iaradigm.

2.5. Finally, it should be noted that amplifications of basic tenses are rare
in text data, those found being the repetitive optional prefix, and the initial
optional prefixes tik and CaO. The repetitive optional prefix is commonly used
in speech with tenses Immediate Past/Future (IP) and Optative (OP), while_ the
initial optional prefix tik is usually used in speech with the Zero (0) tense.
The other initial optional prefix,C.a.0 is commonly employed in speech with the
Past Habitual (PH) tense. Other amplifications than these are nearly all
6
elicited forms.
 -244-

FOOTNOTES
1
This paper presents a preliminary analysis only The author is currently
preparing a further paper in which it is intended to state a more comprehensive
analysis in terms of rules.
2This study is based upon the dialect of Bendeghe Ayuk in Ikom division
of Ogoja Province. This dialect is almost identical with those of Etome and
Abia, the nearest Etung villages to Bendeghe Ayuk and constituting with
Bendeghe Ayuk the Mbuma Clan. Some differences in the verb structures have
been noted between the dialect of Bendeghe Ayuk and other Etung dialects (than
those mentioned above) in Ikom Division, These differences involved tone, pre-
fixation and suffixation.
3
In Etung, each phonological word is marked by a tone pattern "consisting
of a sequence of three level tones. With three syllable words, these patterns
are realized as a series of three level tones spread out over the three syllables.
With two syllable words, which are marked by tone patterns involving dissimilar
tones for second and third tones, the pattern is realized with the final two
tones compressed as a glide. With one syllable words, when the patterns involve
any dissimilar tones, the tone pattern is realized as a glide." This statement
is taken from T. Edmondson and J. T. Bendor-Samuel "The Tone Patterns of Etung",
To be published in JAL vol. iv, 1966. In that article, it is stated t'iat there
are 12 tone patterns.
4
A list of the tenses (and abbreviations used in this paper) constituted
by minimal forms and suffixed forms is given here. Minimal Forms in Immediate
past/future (IP), Optative (OP) and Zero (0). Suffixed-Forms:- Past Habitual
(PH), Pluperfect-variant (PPv), Present Habitual (PrH) and Pluperfect (PP).
5
The initial transform prefix tik has two allomorphs: -
tik employed with 3rd person plural
tik elsewhere
6However since this writing the following example of a double amplification
has been found in text data:
kin n kl ikpO

should I - am again -returning (back)

ie. 'I ought to be going back again'

 Appendix VIII

PRELIMINARY REPORT ON PARA-LINGUISTICS IN MBEMBE (E. NIGERIA).

P. M. Revill

Aims 1. To report on preliminary work on the linguistic and non-linguistic

signals of emotional overtones in Mbembe.
2. To indicate the direction of subsequent work.

1.0. Report on preliminary work.

1.1. Methods of Approach.
Initially, in order to avoid a biased interpretation of emotional over-
tones in elicited material, a recorded text was taken as a basis. An informant
was questioned on his interpretation of the speaker's feelings at certain
points. Specific emotional overtones emerged in certain utterances and the
informant was able to point out some of their characteristics and contrast
them with the same utterances in normal speech patterns.
Secondly, the informant was asked to use utterances from text material
and elicited utterances to demonstrate the effect of certain emotions on
pitch, voice quality and other linguistic features. He also demonstrated
certain non-linguistic features, such as gesture or facial expression which
might accompany these utterances. In each case when an utterance was elic-
ited a situation was described which might be expected to give rise to emo-
tional overtones in speech. The situations used are noted below.
Subsequently in this report observations arising directly from unelic-
ited text material (i.e. the first method described above) will be marked
"A", those arising from the second method will be marked "B".
1.2. The initial findings have been recorded below under
general headings of the emotions displayed.
1.21 FEAR Situation. Nervousness at speaking into a recorder
fll.
"He is afraid, he has never seen such a thing before.
Linguistic signals. Tense voice quality and slow
speech. Short forms will be used (e.g. fit5ma 'I begin?'
A
v. fit5m6 O'I begin' question marker).

Non-linguistic signals. "You will hear it in his throat",

i.e. gasp for breath before speaking. "He will tight
himself." i.e. sit hunched up without moving.

-245-
 -246-

1.22 ANGER Situation 1. A command had been given and disobeyed,

the speaker is angry and wishes to enforce the command.
Linguistic signals. Voice quality will be breathy ei-
ther throughout the utterance or on the final syllables
Faster and more intense speech than normal. "You will
speak fast", "you will speak with power". The pitch
will be lowered on the final syllables. (Phonemic tone
contrasts are preserved within this.)
Non-linguistic signals. Almost invariably signalled by

B fast gestures. e.g. In the command kwu 'Come!' there

will be a swift beckoning movement with one hand. In
a longer command there will be a movement of the whole
arm with the finger pointing downwards, "to show that
you must do it". In certain contexts the linguistic
features of anger may not be distinguishable from those
of general excitement but, "they will know by your eyes 13

whether you are angry or surprised". i.e. The eyes

open wide and the eyeballs roll. "You may nod your
' head." i.e. fast.

Example showing pitch contrast. a) Normal speech.

4'
Elided form with Okwcitilm5vg:n50114Oci
intonation and a
phonemic tones.

.
Unelided form Okwii &gm 5vg:n5Q pira o. ci II

Meaning. He-comes he-should-dash chief before he-eats.

1
b) Angry speech.
Okwiittim5g:n5" plrOci
1
Breathiness

13

13

U
 -247-

/-\ Situation 2. (Volunteered by informant) The speaker has

been taken to the edge of a deep pit and has been told
to jump down into it.
Linguistic signals. In a situation such as this which
causes distaste and horror, in addition to the phenomena
of voice quality, pitch and speed noted above, the ex-
clamation cg will be used. This morpheme has no trans-
latable meaning but it is used frequently in the language
B
to signal distaste or merely surprise. When it signals
distaste it may include an extra syllable, ci:g. The
exclamation w6 may be used for something surprising but
it will never be used to signal distaste. (See 1.25)
Non-linguistic signals. With the exception of gestures
specifically used in command the non-linguistic signals
mentioned in situation 1. will occur. In addition the
v speaker will shrug his shoulders when using an exclamation.

1.23 RESPECT Situation. The informant was asked to make the words of
the example in 1.22 a request to a chief or older man.
Linguistic signals. The respectful utterance is always
spoken slowly and quietly. There is a tendency to
glide the phonemic tones giving an undulating effect
over the whole utterance. The final syllables will
bear low pitch and may be preceded by juncture.
B
Non-linguistic signals. A slow nod of the head and a
movement of both hands away from the body, the wrists
turning and finishing with the palms of the hands
facing upwards. The beckoning gesture with kwe:
2
'Come!' will be slight and slow.

Example. a) Normal speech.

Elided form with
intonation and 6kwcittim6vg:n50 oci

phonemic tones.
I I / I I
Unelided. form okwu otum ova:noo pira oci

Meaning He-comes he-should-dash chief before he-eats.

b) Respectful speech.
2
Elided form showing Okwaumova:noo pira ocie
showing junctures.
 -248-

1.24 SORROW Situation 1. A friend has promised to come on a visit

but he has not done so. Situation which might cause
disappointment.
Linguistic signals. Voice quality will be breathy on
the final syllables of an utterance. These final syl-
lables will be spoken quietly, will bear extra low
B pitch and may be preceded by juncture. There may be
strong accent on the syllable preceding the juncture.
The exclamation 6: may be used in such a situation as
an expression of sympathy. The slow form ci:a may be
fl
used to express disappointment and annoyance. Li
Non-linguistic signals. You will hear how you breathe."
i.e. There will be a sighing intake of breath before
speech.

Example.
Elided form with
intonation and wurobenbokwu kokwu cya
+ +
phonemic tones.

Unelided form wura oben be okwu kokwu cya

Meaning. he-said that he-come Neg-he-come cya!

(he did not come)

A Situation 2. A situation in which distress rather than

U
disappointment would be called for, such as the death
of a friend.
Linguistic signals. The same linguistic signals will
be present as in situation 1, except that the exclamation
nso: will be used to express sympathy with the relatives
B
and a high, tense exclamation wo or kwa will be used in
hearing the news.
Non-linguistic signals. The speaker will beat the fist
of one hand into the palm of the other.
 -249-

1.25 EXCITEMENT" Situation 1. The climax of a story.

Linguistic signals. Wide relative pitch intervals,
with extra high pitch and marked crescendo on one
word or syllable in the utterance. Voice quality
A will vary between breathiness on the low soft por-
,
tions and clear intensity on the high crescendo
portions. Pauses between intonation groups will
decrease in length as the high climax approaches.

Example. (from text) Spacing between groups indicates length of pause.

Elided form with
intonation and owaolwa it bi re min: a W1.1

phonemic tones.

Unelided form. owaowa birobire min:a okwu

Meaning child
r small thus he-came
.f.
m
makwube
obir;;;LT

obira okwu
ma wube
makwube matete ma
,,
maitet_ma piIan conina
oo

pira nnabene be
I/
conina
they-drove-him he-came- they-drove- plenty before they- let-him
(away) again him say-that stay
/
o ioa
0
o nioa

he-stayed.
Situation 2. The speaker expresses surprise at
someone's action.
Linguistic signals. There may be wide relative
pitch intervals, and it is usual for one word or
syllable to be extra high. This word may also be
lengthened and stressed. Certain lexical items
B
appear frequently in this climactic position.
e.g. mapyir 'all', matete ma 'plenty'. (When
mapyir is used indicating surprise it will be
lengthened on the first syllable - ma:pyir 'all',
when it is used in a story it may be lengthened
stylistically on the second syllable-mapyi:r).
 -250-

Also as in situation 1 there will be breathiness on those

parts of the utterance which bear a low pitch intonation
pattern. The exclamations wo or ca (short fast form) may
be used to express surprise (e.g. "if you hear that one
of your friends has been arrested.")
Non-linguistic signals. When the word ma:pyir 'all' is
used (i.e. as lengthened in conversation not in narrative)
the finger will be drawn across the mouth.

Example
Elided form with
intonation and
. . , ., .
ono° febeba
01
L
,F_1
ici.ma:pyir
....
(gesture)
phonemic tones.

Unelided form onon ofo ebebe oci mapyir

Meaning Man he-killed cutting-grass he-ate all!

" Situation 3. The police are looking for a man, causing

general excitement. The informant was asked to show how
in such circumstances he would tell someone in secret
where the man was. The situation therefore implies ex-
citement and secrecy.
B Linguistic signals. Throaty, intense voice quality.
Wide relative pitch intervals and accent on the climactic
word.

Non-linguistic signals. Pointing may be with the finger

or with the chin or both.

Example, a) normal speech.

Elided form with

intonation and osensi uois :ta:oa
.. I
phonemic tones.
a .
Unelided form. oseo sa iduoo se eta:oa

,eaning he-went there room his inside.

b) in excited speech.

Elided form osensi uo'se:ca:oa

sr
 -251-

1.26 Forms dictated by social situation.

1
Some forms have been found which cannot be described as
j
reflecting feelings on the part of the speaker, but they will
be used in certain social situations.
a) Final feature -e:/i
(i) This will be used in story-telling or in a village
announcement to give emphasis to certain statements.
A e.g. ova:nou onio e:

chief was
There was once a chief...

okwotumova:nool pira oci:

he-should-come dash chief before he-eats.

(i.e. general announcement to the village)
(ii) It is used in conversation to ensure better under-
standing on the part of the hearer. It is frequently
employed for this purpose in speaking to a child or to
a foreigner. If an utterance has been misheard it will
be repeated with the final feature attached, and the final
feature will be heavily stressed.
14.

B e.g. ino:ni
Bird (it is)
(iii) The use of this form for emphasis may be associated
with distance from the speaker to the hearer. e.g. The
form i: 'yes' will be used when speaker and hearer are
I close to one another. At a distance of about 12 yards
or more the form iye: will be used for 'yes'.
(iv) The final feature also has a polite connotation and
will be used for instance in summoning a chief with the
word kw4 'come!' (cf. kwu 'come!' familiar, non-emphatic
form)

(v) The use of the final feature may indicate cheerful-

ness or playfulness, in which case it will be lengthened
and spoken in a singing tone.
 -252-

p
ti

b) Use of unelided forms.

When speaking to a foreigner many speakers will use un-
elided forms.
e.g. oleo siduoo sc:ta:oa
He-went there-room his-inside
It is interesting to notice that the unelided form will
not be used in speaking to a child.

2.0 Direction of subsequent work.

The above findings show the result of very preliminary work using one in-
formant and recorded text material mainly of the narrative type. Further study
would need to include observations within the real cultural situation and the
use of recorded conversations. The present study takes no account of the degree
of emotion shown or of variation between individual temperaments.
The observation of section 1 may be helpfully summarized in Matrix form.
Such a display will indicate areas in which ambiguity might arise unless the
cells of all the matrices were examined. Three linked matrices are shown, the
vertical parameter of the linguistic and non-linguistic matrices shows the vari-
ous emotions, the horizontal parameter shows linguistic and non-linguistic emo-
tional signals respectively. Below these two matrices is a third showing the
effect of social situations superimposed on the emotional situations.
Subsequent work might be expected greatly to increase the complexity of
the matrices. The display on Page 8 shows how the area of linguistic ambiguity
between anger, distaste and excitement is cancelled by the application of the
non-linguistic matrix. The addition of further matrices and of details within
existing matrices would be expected to reduce further the possibility of ambi-
guity.

Key to Matrices on Page 8

Linguistic signals Social Situation
B Breathiness M Mime.

LF - Low on final syllables EH Exaggerated hand

movements
Q Quick intense speech
e:A Accent on e:
WI Wide relative intervals
LC - Length on climactic word
IT It
AC - Accent "
 D -253-

0
AJ - Accent preceding juncture
JF - Juncture before final low syllables
S Slow speech
G Gliding pitches
SS - Slow soft speech Area of linguistic
ambiguity
T Tense
L - Lengthening of all syllables
Non-linguistic signals
E Eyes open wide and rolling
QP Quick gestures, pointing
QN Quick nod
S Shrug shoulders with ci:a (Exclam.)
I - Intake or sigh before speech
B Beating fist of one hand into palm of other
H Gesture of both hands with palms turning upwards
SN Slow nod
IG Gasping intake of breath

E3

00 FOOTNOTES
1
Quotations are from informant's comments.
2
See section 1.26 a) iv

io
t
 LINGUISTIC SIGNALS EMOTION NON-LINGUISTIC SIGNALS

PHONOLOGICAL LEXICAL IGRAMMATICAL FACE HANDS HEAD BREATH BODY

Voice Pitch Length Junc- Speed Exclam-

quality ture ations
Accent Intens- etc.
ity

+ B + LF + Q ANGER + E +Qp + ON

+ B + LF + Q + ci:a DISTASTE +S

+ B + WI + LC + Q + w9 + mapyir EXCITE-
+ AC ca maaama MENT
+ B + LF + JF + S +-1
ci:a SORROW + I
+ B + LF + JF + S + fisb: + B
w6,
kwa DISTRESS
+ G + JF + SS RESPECT + H +SN
+ T + L + S shortened SOCIAL
forms. FEAR + IG
SITUATION
+ -e: + Eli Emphasis
Repetition
+ -e: Emphasis
Clarity
+ unelided + EH +M Foreigner
forms
le111
-e: + EH Child

See key on pages 8 and 9.

4
 Appendix IX
AN INTERIM WORKSHOP REPORT ON
THE PHONOLOGICAL DATA OF AGBO
Klaus and Janice Spreda

Table of Contents

1. Introduction.
2. Phonological Hierarchy.
3. Non-suspect CV Patterns.
4. Interpretation.
5. Phonetic Workchart.
6. Evidence for Uniting or Separating Phones.

7. Chart of Phonemes.
8. Formational Statement of Phonemes.
9. Distribution of Phonemes.
10. Suprasegmental Features.
11. Unsolved Problems.
12. Orthography.
13. Text.

14. Word Lists and Comparative Notes.

I. Introduction

Non-linguistic
The Legb6 language is spoken by the people of the Agb6 clan which inhabits
the banks of the Cross River. About two thirds are living on the western bank
in Afikpo Division, the rest live on the eastern bank in Obubra Division. Both

divisions are in Abakaliki Province.

Their usual settlement is a compact village of several hundred up to over
a thousand inhabitants. As these villages form townships they are often called

wards. A township is a cluster of two or three villages directly adjoining

as a centre and others loosely scattered around it in a couple of miles'
distance. The villages or wards are made up of compounds which can vary
greatly in size.

-255-

.kospofF-
 -256- 13
sp 2
The clan consists of four townships with roughly the following numbers
LI
of inhabitants:
Igbo Ekureku 12.000 +
Igbo Imabana 8.000 +
Adadama 6.000 +
Itigidi 2.000 +
The figures are approximations based on the unpublished census results
of 1964. Igbo Imabana is the part situated in Obubra Division.
ED
All townships are accessible by road all the year round and have a daily
lorry service. The markets are frequented by members of the surrounding clans
and Igbo traders. However the outside influence does not seem to affect the
n
clan too strongly.
a
Linguistic
No mention of the language in linguistic literature has been discovered. U
However observations made in passing seem to indicate a certain relationship
with the neighboring languages Yakur and Mbembe (in lexical items).
Efik used to be the trade language and the majority of men have some
knowledge of it, however the use of it is decreasing in favor of English.
U
The women are mostly monolingual.
Each of the townships speaks a different dialect but the dialects are
mutually intelligible. This statement presents the dialect of Adadama.
The language has been studied in Adadama during three periods totalling
1
thirteen months between August 1964 and January 1966 . The principal informant U
was Mr. John Ebong, but about a dozen further informants of both sexes have
been used to various degrees.
This preliminary statement follows the outline suggested in "Format for
a Routine Phonological Statement", West Africa Technical Helps No. 4 of the
Institute of Linguistics.
We are greatly indebted to Mr. R. Stanford and Dr. J. T. Bendor-Samuel
for their advice during the studies and in producing this statement.

Symbolization
Mid tone and stress are not marked.
N stands for syllabic nasal. Or
 -257-
II. Phonological Hierarchy

n The following levels are pertinent in the phonological hierarchy of

Legb6:
syllable level
word level
utterance level

The Syllable
The syllable is defined as the tone bearing unit. Its structure can be
summarized in the following formula:
+ Marg + Nuc + Marg
1 2
C N/V
Restriction: If the nucleus is manifested by N the margins are obliga-
torily absent.
Within the syllable the nucleus is considered the tone bearing unit.
For distribution of phonemes in these three slots see section 9.
The syllable functions as a unit on the next higher level, the word level.

The Word
The word can be defined as the stress bearing unit (stress group). As
to distribution of the stress within the word see section 10.
The second criterion for the word is potential pause on either side of
it.

The word consists of one to six syllables. For details about distribution
of syllables within the word see section 9. There,details about the three
E11
types of phonological words set up according to distributional criteria can be
U
found as well.
The word has a suprasegmental feature of tenseness which is discussed in
section 10. Another feature of the word is a limited vowel harmony, for details
see section 9.
The phonological word largely corresponds to the word unit in the gramma-
tical hierarchy. It functions as a unit on the next higher level in the
phonological hierarchy, the utterance.

The Utterance
The utterance is made up of phonological words. Its criteria are pause
before and aftel it and secondly a slightly raised pitch on the first syllable
 -258--

with a decrescendo of prominence. The final syllable has a rapid pitch

drop with a decrescendo in prominance plus a very slight nasalization. The

main body of the utterance is level.

The question utterance is a type or variant with the following extra fea-
tures: The contour as a whole is higher in pitch. It is more strongly modu-
lated and spoken louder. The last syllable is also raised.
Statement Question

High
Mid
Low SOS

["> atata'ma i'sol bafuu'gi] [" atata'ma i'sol bafuuTgi]

/ atatama is5 bafinigi./ / atatama is5 bafdugn/
'Adadama all theyare-coming.' Adadama all they-are-coming?'

A feature of the utterance is elision. For elision rules see section 11.
Tentatively one can say that the phonological utterance corresponds
roughly to the sentence unit in the grammatical hierarchy. It is however likely
that one or more phonological levels between the word and the utterance must
be constituted (see pause [I] in the examples above), but the necessary inves-
tigations have not yet been carried out.

III. Non-uspect CV Patterns

The non-suspect patterns of the syllable are:
V. [6.be.be] 'dust' [6.0] 'cloth'
[a.nan] 'oil'

CV. [11.mal] 'door' [a.so] 'soap'

[e.ta.pa] scar
'scar'

CVC. [an] eye

'eye' [lgv51] 'town'

[e.lo.lon] 'snail'

For further syllable patterns established through interpretation and

word patterns see section 9.

IV. Interpretation
It proved convenient to base all the interpretation on the non suspect
syllable patterns and also to lean heavily on tone as being a feature of the
syllable and marking the syllable nucleus.

A
 -259-

A. Single segments which may be consonants or vowels:

non-syllabic:
[y] word initially [y3o] 'weave' [yaa] 'scratch'
medially [EyAli] 'paddle' [aye] 'he'

[w] word initially [wAdum] 'man' [waled] 'brother'

medially [Ew50] 'storm' [Awo] 'you sg.

These are interpreted as C because of the predominant syllable pattern

CV/CVC. There is no tone on [y] or [w].

Syllabic

[i] word initially [iban] 'k.o. fish' 'hen'

medially [ebin] 'mortar' 'bush'

finally [eyali] 'paddle' 'stick'

[u] word initially does not occur

medially [gezuolj 'night' [mbekum] 'tortoise' rt.

finally [gitil] 'k.o. lizard' [eyu] wave

'wave'

These are interpreted as V according to the predominant syllable pattern

CV/CVC. Also the V in question bear their own tones.

B. Segments which may be clusters or single units:

VN or syllabic N:
'worry' [Mbeke] 'pawpaw'
1

'I made' [ndumi] 'sandfly'

1

'I hated' [Igw3gwo] 'promise'

1

These are interpreted as syllabic N. They bear their own tone.

Long V or VV
[dyaakpu] 1 cassava' [essii] 'he is doing'
[atgemi] 'farm' [doom] 'burn!'
[zee] 'see!' [ewitu] 'he has already grated'
[ea55gi] 'he did not
vomit'
These are interpreted as vowel clusters because each vowel bears its
own tone though they are very closely knit together and because there are
vowel clusters in the language according to interpretation in paragraph C.
 -260-

C. Sequences which may be clusters or units:

[kp] [kpakpa] 'groundnut' [akpao] 'plate'

[gb] [gbOogboo] 'tin' [lcgba] 'time'

[dz] [dzi] 'eat!' [adzgol] 'k.o. ant'

These are interpreted as units (C) because of the predominant syllable
pattern CV/CVC, while there are no non-suspect C clusters.
Besides the above which involve consonants only there is a series of
sequences consisting of a consonant plus either the vowel [i] or [u] or plus
labialization or palatalization:

[I]
22(palatalized C followed by or c followed \T-clus ter :

[ty] [tyotyo] 'small frog' [etyi.0] 'porcupine'

[tyaki] 'to be dried' [fityano] 'tin'

[tycgi] 'fell!' [etiltye] 'palmwine'

[tyame] 'divide!' [ttyt] 'imitation'
[dy] [dyagkpu] 'cassava' [adyl] 'Festival'
[dyai] 'shallow' [lcdyg] 'sacrifice'
[dyala] 'be astride!' [cdyog5n] 'evening'

[lidy6go] 'cluster'

w
C V (labialized C followed by V) or CuV (C followed by V-cluster):

[kw] [kwal] 'canoe' [ckwa'ali] 'he tried'

[kwe] 'shout!' [ikwo] 'a weekday'

[ekwe] 'he shouted'

[tkwen] 'fire'
1
[gw] [gwano] 'woman' [ligwa] 'leaf'

[gwo] 'flay!' [igwo] 'debt'

[gwe] 1 recultivate' [egwe] 'he recultivated'

[gwu] 'whistle!' [ligwuti] 'whistle'

[Owl [r)wa] 'go to bed' [Ewa] 'flood'

MouTLn't] 'book'

These are interpreted as palatalized and labialized consonants

1. Because of the predominant syllabel pattern CV/CVC.
2. Because the element in question, i.e. the [y] and the [w], does
not bear tone.
 -261
Contrasting both with the group above and with the non-syllabic group
0 under paragraph A is the following group:

CYV or CiV or CiyV

[pi] [gepion] 'afternoon'

[Epia] 'market'
0
Rpi50] 'hornbill'

Rplaml] 'k.o. grass'

[IA] [biala] 'singe!' [6bia] 'native doctor'
[bia] 'bad'
0 [bbib] 'black' [ffibiokao] 'charcoal'
[biei] 'suck!' [gabi.t] 'darkness'
[di] [didia] 'be sweet'
[vi] [viai] 'squeeze' 'driver ant'
[aviaogba] 'scissors'
[wi] [ewia] 'hips'

w
C V or CuV or CuwV.

[bu] [baa] 'follow' [gabua] 'water storage pot'

[tu] [tua] 'be strong!' [atagn] 'bush cow'
[etui] 'he danced'
[du] [dua] 'hide!' [idua] 'sleep'

[edui] 'he beat'

[ku] [kaa] 'open' [ektia] 'he opened'
[gu] [gua] 'pull out' [egua] 'he pulled out'
[vu] [vua] 'pile' [livaal] 'k.o. ant'
These are interpreted as consonant followed by two vowels (CiV and CuV)
because the [i] and [u] bear their own tones. Since by definition of the
syllable there is only one tone to the syllable this means that the two vowels
belong to different syllables. However, the two syllables are phonetically
closely knit together, as in normal speech both tones are heard only if they
are different. This distinguishes this group from the one under paragraph A,
where both tones are always clearly heard, and where there is an audible [y]
or [w] between the syllables.
The following group is similar to the one above but has the [i] as the
second member of the cluster. The parallel constellation with the back vowel
[u] does not exist, since [u] occurs only after [u], and this is treated under
section B.

-7777-77711WPRr7777-7=-.. r-
, .

.1
 -262-

CVi or CVyi
[ekpei] 'dove' [Ilya] 'wing'

[ewai] 'comb' [edal] 'pepper'

[enil] 'he has already given' [ettoi] 'he is crying'

[effoi] 'he is returning'

These are also interpreted as vowel clusters across syllable boundaries,

the syllables being closely knit together, because each vowel bears its own
tone, but there is no consonant audible between them.

V. Phonetic Workchart

labio- alveo- labio- labialized

Consonants: bilab. dent. alveol. alat. alat. velar velar velar
vl. p t ty kp k kw
Stops
vd. b d dy gb gw

vl. f
Fricatives
vd. v

vl. s
Sibilants
vd. z

Affricates vd. dz

Nasals vd. m ow

Laterals

Semivowels
vd.

vd. w
dillillMil
\
Y

Syllabics vd. \J ni
1111111111N
,_.

Vowels:
Front Central Back

close
High
open

Mid

Low

a
Features on higher level: fortis as feature of the word final nasalization as
feature of the utterance.
 I
-263-

VI. A. Evidence for Uniting Phones as Members of the Same Phoneme

[g] [y] = /g/ as fluctuating in certain environments but never in
contrast.

Word initially

[g] before central and back vowels

[y] [g] before front vowels

[gabi] 'entwine!'
[gali] 'lift'

[geti] [yeti] 'tree'

[genedze] [yefiedze] 'banana'

[gtt55] [ytt55] 'work!

[gita] [yin] 'lizard'

Generally women and softly spoken men prefer [y] while older men and force-
fully speaking people prefer [g].

Word medially

[g] always as C in nouns of the pattern N/V.CV always when gV is

reduplicated
[y] [g] elsewhere independent of surrounding vowels:

[eg5] 'cloth'

[ogg] 'bullet'

[alga] 'needle'

[eg6g6 mbala] 'sweet potato'

LiLi [tugi] - [tuyi] 'drive away'

Dc-Peg( _ [kpeye] 'announce'

0 [Wgg] - [Wyg] 'children'

[koogo] - [kooyo] 'show'

g[dogo] _ [doyo] 'sleep'

[kpaga] _ [ kpaga] 'lock'

Generally [y] is more frequent than [g].

- .
 -264-

[m] [m] = /m/ ; [n] [n] = /n/ Co] [0] = /0/.

as being in complementary distribution.

[N] carries its own tone and stands in the nucleus of the syllable
1

[N] (non-syllabic) has no tone and stands always in the margin

of the syllable.

1
[1E60] 'mosquito' [mama] catch!'
[dzellima] 'today' [fizam] 'back'

[lipanA] 'worry' [ltmA] 'door'

[fittabe ] 'grasshopper' [Arian] 'oil'

1

[fidena] 'there' [nerd] 1 own!'

1

[nnii] 'I gave' [num] 'take!'

1

[owaani] 'I cooked' [Anao] 'four'

1

[hke] into [eveoi] 'sandbank'

[Ogg] 'bullet'
1

[L] = /i/ and

[u] [U] = /u/ as freely fluctuating in certain environments but never
in contrast:
More specifically:
Older people will almost always prefer [L] or [u] in CVN syllables,
while young people tend to use [i] and [u] but sometimes use [L] and [U].

[amin] [amen]

[iffn] [ifLn] 'cricket'

[eyim] [eyLm] onion
'onion'
[(g)ezin] [(g)ezt.n] 'waist'

[sin] [sLn] 'hair'

[wadum] [wAdum] 'man'

[enilo] 'salt'

[fiztio] 'nose'

[num] 'take!'

Sometimes older people will give [L] and [u] word medially before a syl-
lable beginning with N or /1/, while young people almost never do this:
 265-
_
[nnina] ti [nnLna] 'I placed'
1 1

[tkpAminA] ti [tkpAmtnA] 'bed'

[lebilA] [16bL1A] 'bundle'

[viii] [vili] 'cut!'

[ndumi] ti [ndUmi] 'sandfly'

[gezil0A] ti [gez00a] 'sweat'

[gekull] ti [gekull] 'many'

VI. Evidence for Separating Phones

[p] [b] [kp] [gb]

epA 'hawk' cbg 'centipede' tkpa 'body' tgbA 'talisman'

pa 'pluck!' ba 'ask!' kpa 'drum!' gba 'join!'
pole 'excl. of be to 'then'
surprise'
pAla 'search!' bala 'resaine

apo 'you ab o 'you died' akpei 'cheek' agbei 'Agbo pl.'

belched'
60 'he will 6bo 'he will apb 'type of egbei 'Agbo sg.'
belch' die' tree'

ltbaal 'breast' legbAl 'time'

lebel 'beard' legbel 'flute'

Okpg 'spear' ogbA 'junction'

ikpayan 'button' igbAyan 'birdtrap'

[k] [g] [kp] [gb]

tkA 'mother' tgA 'k.o. rope' kpa 'body' tgbA 'talisman'

eko 'he hated' egb 'cloth' ckpo 'drawnet' cgbo 'he jumped'

aka 'mothers' AgA 'k.o. plant' AkpA 'tongs' agbA 'chin'

oka '..with' ogg 'nut' okpa 'spear' ogbA 'junction'

eko 'friendship' ekpo 'he stayed ggbo 'Agbo sg.'

[k] [g] [kw] [gw]

tkA 'mother' tgA 'k.o.rope' ckwa 'instr. for tgwA 'hunger'

tapping'
eko 'he hated' ce 'cloth' tkw5 'snail' cgwo 'she weaned'
Ala 'mothers' AgA Tk.o.palm' AkwA0 'nat.bed' .agwa 'you drank'

Ogg 'nut' ogwa 'I drank'

 -266--

[k] [g] [kw] [gw] (con't)

ekegn 'torch' tkwen 'fire'

ku 'stay!' kwu 'chase'

rjkali 'trick' okwali 'I am

trying'

kwe 'shout!' gwe 'recultivate'

[t] [ty] [dy] [d]

ady i 'a feast' gdi 'you will say'

lttál 'stone' ledygl 'sacrifice' ltdal 'tongue'

lidyal 'rattle' 'Ida]. 'story'

itye 'fear' idyi 'dew'

getye 'part of gedyege 'being tall'

roof'

6tio 'he will etyio 'porcupine' edyioa 'he shivered'

lay'

etyaki 'it was dry' tdyagan 'tree fork'

tte 'father' ttyt 'imitation' edg 'he greeted

tt6 'house' etyo 'he quarreled' Edo 'again'

gtt5 'work' gtd5 'throat'

etetela 'madman' etyetyena 'stranger'

[z] [dz] [dy] [d]

Azi 'blood' adzi 'you ate' Aayi 'a feast' gdi 'you will say'

lidzi 'food' idyi 'dew'

gedzet 'being wide' gedyege 'being tall'

lidzA0A 'quarreling' etidyAgan 'forked stick'

gtdza 'bladder' ledyal 'sacrifice'

ezge 'he edzg 'he finished' edg 'he greeted'

walked'
ezi 'bushpig' edzi 'sun'
137Pra
gezee 'thorn' gedze
zee 'see!' dze 'be wide'

edz6 the forgot' Edo 'again'

[s] [z]

tst 'leopard' ezge 'he walked'

gtsg 'tail' gtzte 'journey'

es1 'he made' ezi 'bushpig'

Asi 'water' azi 'blood'

su 'steal!' zu 'rot!'
 -267-

[f] [v] [w]

efa 'power' ova 'dog' ewa 'lightning'

gEtil 'he will gvene 'ladder'

shave'
efila 'it germi- evila 'it was distant'
nated'
efu 'he will evil 'monkey' ewu 'he will grate'
come'

efi 'boil' evi 'he went oue._

,
vaan 'wrestle!' waan 'cook!'

V33 'flow!' woo 'bail out!'

[w] [Ow] [gw]

we- 'child' owa 'go to bed!' gwa 'drink!'

Ewa 'lightning' E0wa 'flood' 'tgwe 'hunger'

ewu 'he will grate eowilla 'snuff'

liowe 'fingernail' ligwe 'leaf'

[1] [y]
. ..:.. .
aviya 'you wedged' aviiya 'brains'

[n] [n] [0] [ow]

eneen 'oil' Anao 'four'

akpaan 'servants' akpao 'plate'

Ncpan 'cobra' ncpeo 'cough'

emgani 'she delivered' EinagOi 'it is sour'

-:
ntitioa 'broom' liowe 'finger-
I
1 nail'
1
J
enii 'he gave' 'ea 'elephant' eveoi 'sandbank'

'tnee 'skin eruption' gfie 'age set' E0wa 'flood'

enaanaa 'sediment' efiana 'he eiaoa 'paagoli'

I scattered'
AkAfia 'bell'

efiele 'insult' eoelioa tk.o. tree'

 -268-

[i] [e] . [6]

gezi ropey gezee 'thorn'

lidzi 'food' ledzi 'day'

dzi 'eat!' dze 'be wide'

edi 'palm' ede 'he is'

vili cut!' 'cut!' veli 'lend!'

emin emir' ameen 'it has

germinated' 4

emmini 'he is emmCni 'he is emmeni 'it is

lying' lying' germinating'
zee 'see!' ZEE 'walk!'
fl
kpe 'teach!' kpe 'sell!'

geve 'fishpond' geve 'foot'

gtee 'it will rain' eta 'his father'

fini 'give!' fini 'give!' fene 'shave!'

efili 'he is efili 'he is efele 'it is a

cutting' cutting' light'
. .
hzioili egg nzioLli 'egg' nZgni 'I walked'

[a] [0]

l6val 'running' lev51 'town'

ege 'fishtrap' 'cloth'

ekpa 'body' ckpo 'net'

fa 'rub!' fo 'return'

[o] [o]

lik61 'sleeping mat' likoL 'bush'

ab6 'you slashed' Abo 'you pl.'

akp6 'cheek' Akp6 'rubber'

livol tk.o. fruit' livol 'abdomen'

to 'cry!' to 'falll'

[u] [U] [o]

eku 'white ant' ekb 'friendship'

ekp 'crowd' ekpo 't.o. tree'
eyu wave' 6576- 'friend'

lekul 'navel' lekbl 'neck'

evuli 'plus' eval 'plus' ev6li 'Ikwo man'

Azazao 'fly' Azaz00 'fly' ezozbo 'sr oke'

evam 'he will 6NTEI 'he will evom 'tsetse fly'

move' move'

bedum 'men' bedUm 'men' bgdbom 'they will burn'

 -269--

VII. Chart of Phonemes

bilabial labio- lalveol. alveo- palat. labio- velar labio-

dental ' palat. velar lized
Consonants velar

vl. p t ty kp k kw
Stops
vd. b d dy gb g gw

vl. f i

Fricatives
vd. v
4 r

vl. s
Sibilants $

vd. z
n
a Affricates vd. dz
.. _ _

Nasals vd. m ii 0

Laterals vd. I

Semivowels vd. w y
:

Vowels Front Central Back

High i u

Mid e o

Low 6 a o

VIII. Formational Statement of Phonemes

For further examples of the phonemes and their submembers see sections 6 A
and B.

Consonants
in /p/ [p] voiceless bilabial stop

[61)] /epa/ 'hawk'

[po] /po/ 'belch!

/t/ [t] voiceless alveolar stop

[ki] Ati/ 'stick'

[ta] /ta/ shoot!'

swag...mar
 -270

/ty/ [ty] voiceless palatalized alveolar stop

[tyo] /tyo/ 'quarrel!'
[itye] /itye/ 'fear'

/kp/ [kp] voiceless labio-velar stop

[bcpa] Akpa/ 'body'

[kpo] /kpo/ 'nail!'

/k/ [k] vciceless velar stop

[lik61] /hick/ 'sleeping mat'
[IQ] /k3/ 'hate!'

/kw/ [kwj voiceless labialized velar stop

[kwAl] /kwAl/ 'canoe'

[kw6] /kw6/ 'snail'

/b/ [b] voiced bilabial stop

[abo] Abo/ 'you pl.'

[ba] /ba/ 'ask!'

/d/ [d] voiced alveolar stop

[ed6] /ed6/ 'throat'

[di] /di/ 'say!'

/dy/ [dy] voiced palatalized alveolar stop

[ledy6] /ledyA/ 'sacrifice'
[dyirja] /dyirja/ 'shiver!'

/gb/ [gb] voiced labio-velar stop

[legbA] /legba/ 'time'

[gbo] /gbo/ 'jump!'

/g/ [g] voiced velar stop

[Y] voiced velar fricative

[gJ occurs word initially before non front vowels and
word medially in VCV nouns and in reduplicated syl-
lables.

[agA] /agA/ 'needle'

[eg6g6] /Ag6g6/ 'sweet potato'
[gabi] /gabi/ 'entwine!'

[g.] and [y] fluctuate word initially before front vowels

and, word medially between any vowels except in VCV nouns
and reduplicated syllables.
 -271-
[g] and [i] con't.
[tugi] [tuyi] /tugi/ 'drive away'

[-yt66] /et66/ 'work'

/gw/ [gw] voiced labialized velar stop

[gwA] agwA/ 'hunger'

[gwe] /gwe/ 1 recultivate!'

[f] voiceless labio-velar fricative

[Min] /ifiln/ 'cricket'

[fo] /fo/ 'return!'

/v/ [v] voiced labio-dental fricative

[vAAm] avAam/ 1
'cow'

[vina] /vina/ 'go home'

/s/ [s] voiceless alveolar grooved fricative

[sa] asa/ 'back yard'
[si] /si/ 'do!'

/z/ [z] voiced alveolar grooved fricative

[Azam]
1
/Lam/ 'back'

[zee] /zee/ 'see!'

/dz/ [dz] voiced alveolar affricate

[iedzi] /ledzi/ 'day'

[dze] /dze/ 'crocodile'

/m/ [m] voiced bilabial nasal

[m] syllabic voiced bilabial nasal
[m] occurs as the nucleus of a syllable, bearing tone
[dzeiima] /dzeMma/ 'today'

[Mbeke] /Mbeke/ 'paw paw'

[m] occurs as syllable margin without its own tone

1
[aamAn] /aaman/ 'our'

[mana] /mana/ 'hold!'

/n/ [n] voiced alveolar nasal

[li] syllabic voiced alveolar nasal
[n] occurs as the nucleus of a syllable, bearirg tone
[itdA] /AdA/ 'where'

[Azam] /AzAm/ 'back'

 -272 -

[n] occurs as syllable margin without tone of its own

[Lin] /Lin/ 'waist' a__

[num] /num/ 'take!'

/fi/ [H] voiced alveo-palatal nasal

[nfianol /n-nano/ 'hors'
[Flo] /no/ 'look after!'

/0/ [0] voiced velar nasal

[g] syllabic voiced velar nasal
[9] occurs as the nucleus of a syllable, bearing tone
[11(6] /61(6/ 'into'

[0gA] /oga/ 'bullet'

[9] occurs as syllable margin without its own tone

[Anao] /Anao/ 'four'

[lzaoa] /11zA0A/ 'river'

/ow/ [ow] labialized voiced velar nasal

[Owed] /60wgn0 'book'

[owa] /r)wa/ 'go to bed!'

/1/ [1] voiced frictionless clear palatal lateral

[ledzi] /ledzi/ 'day'

[ekalala] nkalala/ 'European'

/w/ [w] voiced bilabial semivowel

[wadum] /wadum/ 'man'

[Awo] law)/ you sg.

/y/ [y] voiced palatal semivowel

[iyAa] /iyAa/ 'grandmother'
[y6o] /y6o/ weave!'

Vowels
/i/ [i] voiced closed unrounded front vocoid
[0 voiced half closed unrounded front vowels
[i] and [L] fluctuate in closed syllables and word medially
before a syllable beginning with [1] or a nasal, but not
in other positions:
[Amin] -[AmLn] /Amin/
[lebilA] -[lebila] /lebila/ 'bundle'

-,I=N=7:13====2.0,21=r5eGir,
 -273--

but:

[Atibb6] /Atibb6/ 'branch'

1 .

[ibam] /ibam/ 'Itigidi'

[geti] /geti/ 'tree'

/u/ [u] voiced close rounded back vocoid

[U] voiced half close rounded back vocoid
[u] and [u] fluctuate in closed syllables and word medially
before a syllable beginning with [1] or a nasal, but not
in other positions.
[wadum] [wadum] /wadum/ 'man_'

[gektili]- [gek6li] /gektili/ many'

but:

/gitA/ 'lizard'

[ebtikr6] /ebtikp6/ 'board'

/e/ [e] voiced unrounded half close front vocoid

[ibeke] Alibeke/ 'paw paw'
[ebo] /ebo/ 'lizard'

/o/ [o] voiced rounded half close back vocoid

[isokolo] /isokolo/ 'orange'
[dzo] /dzo/ 'keep!'

/6/ [6] voiced unrounded half open front vocoid

[pc] /66/ 'moon'

[veeme] /vccmc/ 'go away!'

/0/ [o] voiced rounded half open back vocoid

[et66] /ggt66/ 'work'

[h6n] nn6n/ 'person'

/a/ [a] voiced unrounded open central vocoid

[0p6.0] /epAqA/ 'part'

[Aye] Aye/ 'he, she, it'

In addition to these descriptions of the vowel phonemes and their members the
following remarks about the scatter in the area of articulation, illustrated by
the phonetic work chart, may be useful:
The half close vocoids [e] and [o] tend more towards their close correspond-
ing neighbors. [o] has also a tendency to be very slightly centralized. Also
the half open vocoids [e] and [o] tend towards their half close neighbors. The
central vocoid [a] generally tends to be fairly far forward.
 -274-
IX. Distribution of Phonemes

The distribution of the phonemes in Agbo will be described in terms of the

syllable. The phoneme distribution within the syllable itself will be dealt with
first, then across syllable boundaries. Finally, the distribution of the syllable
within the word will be dealt with.

Distribution of Phonemes within the Syllable

The syllable patterns of Agbo are: V CV N
VC CVC

Any vowel may occur alone as a syllable. Front vowels are more frequent in
occurrence than back vowels. Examples are given below.
moon 'hen' AkpAikon 'shoe' ekpei 'dove'
'fish trap' ipftnAn 'vessel for g& 'a belch'
oil and pepper'
esi 'bag' gezee 'thorn'
Akpan 'plate' enaanaA 'sediment of oil' epia 'market'
iboOkolO 'dormouse' eyoo 'civet'
ittlutu 'baby' esuu 'he has
filled'
etopsi 'he threw' too 'throw!'

VC

The vowel of this syllable may be e e a o, and the consonant 1 n 0.

Examples of VC syllables:
etyeel he will resemble'
AnenH1 'type of caterpillar'
biaal 'chief'

el31-11 'hornbill'
gepite 'forenoon'

CV and CVC Syllables

All consonants occur in initial position in the syllable. Final consonants
are restricted to nasal and lateral continuants. There is no restriction on the
occurrence of vowels. All occur. CV combinations are complete with the exception
of combinations with labialized and palatalized consonants plus dz. These are
 -275-

the phonemes least frequently found in the language which no doubt accounts for the
gaps. Combinations of vowels with these phonemes are illustrated in the chart below:

ty 1

dy

d +1 + ! + + +
+ + 1 + + i +
Ow 1 .
+
kw 1 + +1 +
FY +_, + + I +

Examples of CV Syllables
bUge 'children' mMoo 'answer!'
dee 'buy' nana 'run away!'
dyaakpu 'cassava' nanaoili 'fowl with rough feathers'
dze 'alligator' esuaa 'he filled'
fa 'rub!' eowA 'flood'

egA 'cloth' 'moon'

4ba r abcess' esti]. 'he stole'

gwano 'woman' Atli 'wine'

kele 'ch-ose!' tyeeo 'groan!'

kpc 'sell!' LTAzi 'hoop f. climbing palms'

kwe 'shout!' wadum 'man'

]lbal 'beard' 1-1.Aa 'old woman'

rt ibozu 'calabash'
U
Examples of CVC Syllables
a num 'take!' din eye
'eye' 'jigger' lOcal 'fishtrap'

fl The Syllable N
N is a syllabic nasal m n o as illustrated iZ the following words.
mbu 'coco yam'
ndegba 'road junction'
agwati 'fence'
 -276--

Consonant and Vowel Distribution across Syllable Borders

Consonant clusters may occur across syllable borders, and consist of syllabic
nasals, homorganic with the consonant they precede. This may be within the morpheme.

E.g. lb60 'mosquito' A6t30 lashes' Lcali 'trick'

or across morpheme borders as in the verb where the first person singular, Is
indicated by a syllabic nasal.
E.g. mbaagi 'I tied' ndaka 'I dreamt' alsele 'I chose'

Vowel clusters occur across syllable borders both within the morpheme and
across morpheme borders. They may be of the following combinations.
1. Each vowel followed by i
2. Each vowel followed by a reduplication of itself.,
3. i followed by a o o
4. o or u followed by a

2nd. member
OOOOO
ieeaoou
i + + ++
e +
1st. member e + +
a +
+
o +
u +

In the following examples of vowel clusters '-' indicates a morpheme break.

I
aviiya 'brains' e-s1-1 the is making'
lvia 'driver ant'

eploo 'hornbill'
gepiOn 'forenoon'
ekpei 'dove'

gezee 'thorn'

liveil 'wing'

gezft 'journey'

AkpAikon 'shoe'

vaal 'chief'
 -277-

Distribution of Syllables within the Word

When the V syllable is a back vowel it never occurs word initially. V only
occurs word initially if it is a front or central vowel. The VC syllable always
occurs finally in a word. CVC syllables occur alone or finally in a word. N may
be an initial or medial syllable in a word. The maximum number of syllables per
word is six, of which no more than four may be CV syllables.

Vowel Harmony
Syllables are restricted in their distribution in the word to accord with a
limited system of vowel harmony.
If the vowel of the first syllable is a low front vowel, the vowel of the
second syllable is also low.
There is no vowel harthony after high or central vowels.
If the vowel of the first syllable is a front mid vowel, the vowel of the
second syllable is high or mid.
This harmony is restricted torthe first two syllables of a word except in
the verb where three harmonizing V syllables may precede the' verb root.
In the verb root only, the vowel of an initial CV syllable may be a back
vowel. The vowel of the second syllable may then be:
1. i

2. a repetition of the first vowel when this is o or o

3. when the vowel of the first syllable is u, the vowel of the second
may be a

2nd. syllable
Lee a o o

i + + + + + + +

B .e + + + +
1st. syllable E + + +
ai + + + + + + +
0 + +
;13 + o

us +
 -278-
The Noun
The syllable pattern of the noun is
1-4 + + 1-2
1. + (N/V/CV) + CV VC V
2. + CVC
3. + CV
The minimum forms are CV dze 'alligator'

CVC din eyef

The maximum form is CV.CV.V.CV.V: genAanA 'sediment of oil'

Out of 750 nouns the first syllable of 380 begins with a consonant, of these:
145 have an initial g
ft
132 " 1
tt ft
85 " N

Of the 370 beginning with a vowel

86 have initial i

110 have initial e

17
137

37 " a

All nouns with an initial 1 also have a final 1. Many nouns beginning with
ge- or ge- have alternate forms beginning with e- or e-.
Some of the nouns pith an initial N have an alternate form with an initial
e- or e-.

This would seem to give validity to the hypothesis that there was a noun class
system in Agbo. A further pointer in this direction seems to be that certain nouns,
especially names for relatives, change the e- or e- to a in the plural, while
most of the nouns do not have a sg. vs. pl. distinction. The related language,
MBEMBE, retained a noun class system.

The Verb
The syllable pattern of the verb is:
+ +
[- CV.CV (t CV N/V1-2)] + CV [ ±(V CV) + VC]
CVC
Prefix Root

The minimum form is CV or CVC and occurs in the imperative:

gwa 'drink!' num 'take!'
 -279-

The maximum form is CV.V.V.CV.V.CV: geeekoOgO 'he will not show'

The initial C of a verb root is never /0/.

In the CV.CV verb roots the C of the second syllable must be one of the

following:
b g k m n 0 1

Examples: wuki 'follow!' gabi 'entwine!'

baga 'tie!' dumi 'sew!'

fne 'shave!' dza0a 'quarrel!'

gali 'lift!'

Particles
Under this everything that is not noun or verb is collected. The syllable

pattern of the particle is:

1-3
CV + V + CVC
1-2 1-2
+ N /V
CVC

The minimum form is CVC di0 'hard'

The maximum form is N.N.CV.CV riamana 'here'

X. Suprasegmental Features

Tone
Legb6 has three tonemes as can be shown for nouns as follows:

Comparison 1 Comparison 2 Comparison 3 Results

Preceding Substitution List Following

Frames: Frames:
A B C A B C

s h h 1 akpa0 'plate' 1 1 s h H M

s h h 2 waka 'brother' 1 1 1 s H L

1 s h 3 ekpon 'hill' h s h h M H

1 s h 4 kpakpa 'gr.nut' s 1 s h M M

1 s h 5 ego 'cloth' 1 1 1 s M.L

1 1 s 6 leta 'stone' h s h h L H

1 1 s 7 wadum 'man' h 1 s h L M

1 1 s 8 eten 'animal' s 1 1 s L L
 -280- 0 4r
The frames are:
Preceding: Following:
A AdA 'where is?' = H A s6 'the' = H
B aye ezee 'he saw' = M B aaman 'our' = M
C aye Lee 'he will see' = L C w3ni 'one' = L

Comparisons: h = higher s = the same 1 = lower

1. substitution item in relation to preceding frame.
2. second syllable of substitution item in relation to first syllable
3. second syllable of the substitution item in relation to following
frame.

There are so far no disyllabic nouns of the tone pattern H H, although the tone
sequence H H does occur in nouns of more than two syllables. A strong preference
for the lower tones is evident in the language and can be shown in the following
statistical table on disyllabic nouns:

Total sample of disyllabic nouns 283:

H H: 0 M H: 16 L H: 33
H M: 4 M M: 27 L M: 68

H L: 12 M L: 26 L L: 97

16 69 198

No perturbation has been observed, however tone changes are widely used to
mark grammatical distinctions like tense/aspect in the verb on the word level,
relationships between nouns on the phrase level, and certain dependent clauses
on the clause level.

Stress
Stress is not phonemic. Two factors control the distribution of stress within
the word:
(1) Tone: the stress is born by the syllable with the highest tone.
(2) Linear position in the word: out of several syllables with the
same height of tone the one nearest to the end of the word bears
the stress.

Feature of the Word: Tenseness

Phonetically the feature is manifested by a combination of fortis and length-
ened pronunciation of the consonant on which it focuses. The following vowel is
 -281-

appreciably shortened to compensate for the length of the consonant. A long

vowel is reduced to normal length. It can be said that the fortis consonant
has a certain ambivalence since the vowel preceding it is also shortened.
There is a phonemic contrast between presence and absence of the feature
in corresponding words:
(The consonants on which the tenseness focuses are written double,
of digraphs only the first member is doubled.)
1,,
1. p eppoi 'he is belching' epci 'he belched'

2. t etti0i 'he is laying' etimi 'he laid'

3. kp ekkpei 'he is learning' ekpei 'he learned'

4. k ekkumi 'he is sewing' ekumi 'he sewed'

5. b bba 'block!' ba 'ask!'

6. d eddui 'he is beating' edui 'he beat'

7. gb eggboi 'he is jumping' egbo 'he jumped'

p. 9
8. g eggel 'he is listening' ege 'he listened'

9. kw ekkwali 'he is trying' ekwaali 'he tried'

10. gw eggwai 'he is drinking' egwa 'he drank'

11. ty ettyeoi 'he is groaning' etyeeoi 'he groaned'

12. dy

13. f effai 'he is rubbing' efa 'he rubbed'

14. v vvei 'pour out!' vei 'boil!' int.

/. 8
15. s essii 'he is doing' esi 'he did'

16. z

17. dz eddzii 'he is eating' edzi 'he ate'

18. m ekkAmmi 'he is big' ekkAmi 'he was big'

P
19. n ennumi 'he is taking' endmi 'he took'

20. il envoi 'he is looking after' efici 'he looked after'

21. 0 ese001. 'he is going' eseoi 'he went'

22. ow e00wSi 'he is going to bed' eowA 'he went to bed'

23. 1 elldi 'he is making dirty' eldi 'he made dirty'

24. w ewwui 'he is grating' ewui 'he grated'

25. y eyyali 'he is paddling' eyaali 'he paddled'

Out of these examples the Nos. 2-6, 14, 18, 19, 21, 23, 24 are minimal

a pairs, the others represent contrast in analogous environments. At present there

are no examples for z and dy. The reason for the missing /zz/ is that as a rule
a /z/ which becomes fortis also becomes voiceless. The /dy/ is a rare phoneme
and so far it is not found at the beginning of a monosyllabic verb root.
 -282-

Distribution of this feature:

fl
It occurs focussed on the last syllable's initial consonant of the verb
root. If the verb root takes a CV suffix, it focusses on the suffix:

Pattern Root Suffix Pres. Continuous Meaning

CV ba -mi ebammi 'ask'

CV.V du -i eddui 'beat'

CV.V kOo -i ekkoi 'grind'

dua -bi edubbi 'hide'

CVC kum -i ekkumi 'sew'

CV.VC taaO -i ettaoi 'sweep'

CV.CV nomo -i enommi 'stretch'

It seems preferable to set this up as a prosodic feature rather than analys-

ing it in terms of a series of fortis consonants; firstly because it seems to be
closer to the phonetic facts, secondly because it allows a more concise descrip-
tion, and thirdly because its principal use in the language is to denote one
particular tense/aspect of the verb, viz the present continuous. There are
a few occurrences in the basic verb root (see Nos. 5, 14, 18) and much fewer
nouns.

gitta 'life' ebbf 'goat'

ebbe 'steam' b136 'branch'

istiddu 'wine jug' etc.

It would be possible to analyse tenseness as a feature of the syllable

rather than of the word. However certain details point the other way:
(1) the focussing on a part of a word as defined above,
(2) the partial carrying over of the tenseness to the preceding
syllable (This is specially noticeable, when in a verb form the con-
sonant of the preceding syllable is originally a voiced fricative,
because this then turns mostly into the corresponding voiceless
fricative.),
zoo 'find!' esokki 'he is finding'

XI. Unsolved Problems

Elision
Elision can take place across a whole phonological utterance or only
between parts of it, depending on the rapidity of speech. This means that
 -283-

it takes place between various word classes. The elision rules for all these
combinations have not been fully investigated. However some work has been done
on the most frequent combination of verb followed by noun with the following
result:

Verb + Noun
(Nouns can begin only with a front or central vowel)
The verbs of the pattern CV elide differently from those of the pattern CVCV.

CVCV Verbs
These elide quite regularly: the final vowel of the verb assimilates
3 completely to the vowel of the noun.

CV Verbs
Verb final front or central vowels assimilate completely to the following
vowel.
Final back vowels remain unchanged except in the following cases:
+i = uu o + i = uu
+ e = oo u + i = uu
+ e = oo

Tone
It seems very difficult to determine what happens to the tone on the inter-
acting vowels as there seems to be much fluctuation according to speed of speech.
We are not certain that one of the two tones ever disappears, if so this happens
only at extreme rapidity, in which case the lower of the tones disappears.

Consonants
The final consonants 1 m n 0 are dropped when the following word begins
with a consonant. However here too the details have still to be investigated,
as there are instances, when the final nasal is not dropped:

eb660 'grasscutter' 6660 'the grasscutter'

vs. Anaan 'oil' Anaa s 'the oil'

The conditioning factor may be the vowel, the consonant, or the tone.
 -284-

XII, Orthography
The following orthographical symbols have been proposed:
Consonants
Phoneme Orthogr. Symbol Phoneme Orthogr. Symbol
p p Ow nw
t t 1 1

kp kp

k k y y

b b

d d Vowels
gb gb

g g u u

ty

dy
kw kw E er

gw gw 3 or

f f a a

dz dz

z z
m m
n n

ny

0 (n word initially before k, g, kw, kp, gb, nw)

(ng word initially before gw and w, word medially and
word finally) .

Vowel Clusters
Vowel clusters which involve the half open vowels er and or are writ-

ten with the "silent letter" r at the end of the cluster to avoid the appear-
ance of the r between vowels, where readers who are familiar with English
may pronounce it instead of recognizing it as a "silent letter" the function of
which is to mark the half-openess of the vowel:
/livei/ liveir 'wing'
 Fortisness
This is written by reduplicating the consonant on which it focuses. In

case of a digraph only the first member is reduplicated.

Tone
It will certainly be a help for reading if tone is marked in the ortho-
graphy. The suggested symbols are:
over the V or N for high tone
over the V or N for low tone
no mark for mid tone.

XIII. Text
The following few lines of text are meant as a sample of the language
written in the proposed orthography. The first line is a phonemic transcription,
the second is in the crthography, and the third is a word by word translation.
/ means short pause, // means a more definite pause with difinite drop of voice
before and definite rise of voice after it.

ledzil 146ni wAdum w6ni ekA / aama eevooni bg6g6

Ledzil worni wkdum worni ekA, aama eervUrni beerger
day one man one he-was, who he-begot children

Afoo // ftL ede flan dto gesuu / emeni v6m /

Aforng. Erteer ede ernOrn eertor gesuu, ernerni vOrm,

two. His-father was person who-had riches, he-owned something,

IndeAte ebbi / lidzil / bi gineneni isagisa //

intergiAter, ebbi, lidzil, bi ginernerni isagisa.

animal, goat. food, and possessions several.

ledzil woni awake / aana eede wake / etOliye 61ce /

Ledzil worni AwaAyer, aana eede wagger, etOliyer Aker,

Day one child-his, who was the youngest, he went to him,
B
edlye Abee / tete tyke gineneni so get6g3 afoo /

edlyer abeer, terter cee ginernerni sor gertorgor aforng,

he-said quote: My father divide possessions your parts two,

 -286-

num Atyeegi 16min / nim / ede aama AkA gittA /

num Aceegi lOrmin, nim, ede aama AkA gittA,

take portion my, give-me while you-are life,

ebi Ana geede / ede Abbo / bAkas6 be mi gaAate ...

ebi Ana geede, ede Abbo, bAkasOr ber mi galfter....
fl
if so not-it-is, when you-will-die, brothers-your they me
not-will-allow...

XIV. Word Lists and Comparative Notes

Phones peculiar to some of the dialects

There are three phones which are peculiar to the Ekureku and Imabana
dialects: [b] [h] [hw] for examples see word lists below

Evidence for uniting phones (Imabana dialect)

[h] [hw] = /h/

[hw] occurs only before [u] see 41, 64, 80, 91, 152
[h] elsewhere

Evidence for separating phones (Imabana dialect)

[h] is suspicious with [w] and [y]
[h] [w] = /h/ /w/

[wa] 'cook!' [ha] 'wash!'

[h] [y] = /h/ /y/

[ahe] 'you walked' [aye] 'he'
n
[aha] you wash' [ayA] 'that' U

[b] is suspicious with [b] [v] [w]

11
[b] [b] = /b/ /b/

[abe] 'you killed' [ibe] 'they'

[b] [v] = /b/ /v/

[03A] 'cow' [Eva] 'dog'

[b] [w] = /b/ /w/

[be] si ng.fT [we] 'swim!'

a
 B
-287-

4 The following comparative lists must be considered with a certain amount

of caution since they are very early notations which could not be sufficiently
checked, especially for tone:
Examples of [b] in Ekureku and Imabana with corresponding words in Adadama
and Itigidi:
9
No. Engl. Ekureku Imabana Adadama Itigidi -,
26 root ketibe getibg gatizal gativg
62 kill! be be Ve Ve
83 roast! ba bamili vami ?

117 wind keb4e keb4e gevevL gevbg

127 leg keb keb ggvg eve
138 grass libiyele liblyelg livbelg liviyfte
153 cut! ? bill vili vili
156 far : ebAla evila gevila
186 sing! biya be live viya
thief ? ebebe Aveve ?

fresh ? ebabA nvaavaa ?

cow ? 4A gvaAm ?

120 lake Icbakill kebe geve lenme

Examples of [h] and [hw]

9 all k6613 kholo abini Abini ebdni ebdni
1
20 bird iho iy6 lizol izol
30 blood ahi ahi a zi azi
33a eggs 01163ill 0h60ili ;21_0111 litziajni

34 horn leho leho lava liva

41 nose 15hwu 15hwu izdo izdo
I .
49 belly keh6 keh6 liv6 1V3

57 see! he he zee ze
64 fly hwuoa hwuoa w610 wuoa
65 walk he he zee ze

78 soil lehe lehe lezg leze

XV. Conclusion
More investigation needs to be done on the higher level phonological
phenomena, which will no doubt shed light on the grammatical structure of
the language.

7 .79qpiw
 -288 -

FOOTNOTE
1
The final write-up of this material was undertaken at a workshop led by
Dr. K. L. Pike and financed in part by a grant from the United States Office
of Education, under contract 5-14-065, through the Center for Research on Lan-
guage and Language Behavior, of the University of Michigan.
 Appendix X

Some Articles in Preparation

Brend, Ruth M. 'Acoustical Support for West African Field Work.'

Callow, John C. 'Patterns of Neutralization Across Levels in Kasem Narrative.'
Callow, Kathleen. 'Comparative Notes on Serial Constructions in Three Ghanaian
Languages.'
Jacobs, Jill and Pike, Kenneth L. 'Matrix Permutation as a Heuristic Device
in the Analysis of the Bimoba Verb.'
Meier, Inge and Bendor-Samuel, John T. 'Some Contrastive Features of the Izi
Verbal System.'
Meier, Paul. 'Izi Tone and Grammar.'
Pike, Kenneth L. 'Suprasegmentals in Reference to Phones of Item, of Process,
and of Relation.'
Pike, Kenneth L. 'Grammar as Wave.'
Soutar, Jean and Pike, Kenneth L. 'Matrix Patterns in the Use of Direct and
Indirect Quotations.'
Spreda, Klaus and Bendor-Samuel, John T. 'Fortis Consonants in Agbo.'
Stahlke, Herbert and Brend, Ruth. 'Use of Index Matrices in the Preparation
of Language Textbooks.'

-289-