Work Papers of the Summer

Institute of Linguistics, University

of North Dakota Session

Volume 24 Article 1

1980

Notes on African linguistics

Terri R. Scruggs
SIL-UND

How does access to this work benefit you? Let us know!

Follow this and additional works at: https://commons.und.edu/sil-work-papers

Part of the Linguistics Commons

Recommended Citation
Scruggs, Terri R. (1980) "Notes on African linguistics," Work Papers of the Summer Institute of Linguistics,
University of North Dakota Session: Vol. 24, Article 1.
DOI: 10.31356/silwp.vol24.01
Available at: https://commons.und.edu/sil-work-papers/vol24/iss1/1

This Article is brought to you for free and open access by UND Scholarly Commons. It has been accepted for
inclusion in Work Papers of the Summer Institute of Linguistics, University of North Dakota Session by an
authorized editor of UND Scholarly Commons. For more information, please contact
[email protected].
 NOTES ON AFRICAN LINGUISTICS
Terri Scruggs

0. Introduction
1. Consonants
2. Vowels
2.1 Inventories
2.2 Conditioning Factors
2.3 Length
2.4 Nasalization
2.5 Epenthetic Vowels
2.6 Vowel Harmony
2.7 Elision
2.8 Miscellaneous
3. Tone
4. Noun Classes
4.1 Traditional System for Bantu
4.2 Vestigial Systems
4.3 Other Systems
5. Adjectives
5.1 Traditional Approach
5.2 Example from Xhosa
5. 3 ~lelmers I Non-Criteria
5.4 Structural Criteria
6. Multiple Verb Constructions
6. l Consecutivized Structures
6.2 Serialization

O. Introduction
This report is an overview of some of the common features of
African languages. It is difficult to present a concise and compre-
hensive summary of the features that are likely to be found in a group
of languages as large and totally unrelated as are the (probably) one
to two thousand found on the continent of Africa. Therefore I have
chosen representative languages or language families from around the
continent and will give a brief description of these. The map will
show their approximate locations. Appendix 1 gives a diagram of the
different language families.

SIL-UND Workpapers 1980

 143

1. Consonants
In African languages there are phonemes from all the basic types
of consonants that can be found -- plosives, affricates, fricatives,
nasals, laterals, vibrants, and semivowels. In addition, the distinc-
tive "clicks" of South Africa are quite interesting.
Dan (Bearth and Zemp, 1967 (B&Z)) is a language of the western
part of Ivory Coast spreading over the border into Liberia. Interesting
aspects of the consonant-phonology include the following points: in
addition to regular plosives, there are implosives 6 and d' as well as
labiovelars kp and gb. In situations of nasal context, the following
changes O£Cur: ~~+,or m' (syllabic or glottalized m), Nd'+ Q, kpV
+ kmV, gbV + gmV. Labiodental fricatives are formed by friction of
the upper teeth on the back of the protruded lower lip. The lateral
/1/ has two flapped allophones, an alveolar[~] and an alveolar vibrant
[r]. The archiphoneme N can occur presyllabically as NCVC or as a
syllable coda CVN. In neither position is it analyzed as the same as
/n/ or /m/, both of \'fhich are phonemes in the language, although it may
be pronounced similarly to either of these. Sequences of the form Cw
and Cy occur and have both been analyzed as a modification of the con-
sonant, rather than as consonant clusters. The sequence c1c2 does
occur in which c2 is /1/. In certain environments "an optional non-
phonemic transition vowel with the quality of the following main
vowel occurs between the onset and a prenuclear /l/ ... 11 (B&Z 1967:15)
Moving on into Nigeria, three different languages were looked .at.
In Higi of northeast Nigeria (Mohrlang 1972), there are three imploded
consonants - 6, cf, and~' and no labiovelars. Lateral fricatives exist
and can be voiced or voiceless; the fricative has a plosive offglide
in palatalized syllables).
(1) 1Y~a/
. [~dya'J "jealousy"
Vibrants have flapped and trilled variants. Palatalization, labiali-
zation, and prenasalization all occur as well. Voiced and voiceless
affricates also occur.
Angas is found in the central part of Nigeria (Burquest 1971).
The plosives are aspirated or unexploded according to the position in
the syllable. There is a voiced but no voiceless velar fricative.
Nasals have voiceless allophones. Laterals and vibrants have voiced
and voiceless variants. Labialization, palatalization, and prenasali-
zation all occur snparately and in various combinations, yielding
such sequences as rw. The three implosives&, cf, and 1 are mani-
fested as well.
Nupe (Smith 1967), in contrast to the other languages cited from
Nigeria, has no implosives, but there are labiovelars. There are
affricates and syllabic nasals. /y/ has an allophone [r] before nasal
vowels. Nupe is located near Bida.
1

SIL-UND Workpapers 1980

 144

While implosives and labiovelars are common in African languages

there are those which have neither, in spite of their proximity to
those which have one or the other or both. One such language is
Longuda (Westermann and Bryan 1952 (W&B)) found in Nigeria in the
provinces of Adamawa and Bauchi.
The following summary of data from Cameroon is based on Westennann
and Bryan and their linguisitc notes on class and nonclass languages
and Chadic languages. Some languages have labiovelars, implosives, and
imploded labiovelars. Glottal stop is often a phoneme as well (Bami-
leke). Syllabic nasals occur in two different positions NCV{C) and
CVN. Consonant clusters of the form Cr occur but are rare. Semi-
vowels wand y as well as 4 occur. Palatalization, labialization, and
prenasalization and various combinations of all three are manifested
by the semivowels. There are some instances of syllabic w (Mbembe).
The Chadic languages typically have lateral fricatives, q and G, and
no labiovelars.
In southwestern Africa the Wambo group of languages are found
(Baucom 1972); they are Bantu. There are no labiovelars, the affri-
cates only have voiceless representation, and nasals may be voiced or
voiceless. Prenasalization and palatalization are found but both are
limited in their distribution.
The Khoisan languages of the Bushmen and Hottentots in South
Africa have an unusual set of consonants - the clicks (Welmers 1973:50).
Clicks may occur with four different points of articulation - bilabial,
dental, palatal, and lateral. Each of these may be voiced, voiceless,
or nasalized and have an aspirated, unaspirated, or glottalized
release.
On the east side of Africa, the Orusyan language is found in
eastern Uganda {Huntingford 1965). The sequence bg is noted but it is
unclear if this is a normal labiovelar (gb) or merely a sequence with
a syllable break in between. Other consonant clusters do occur, notably
rmw, rny, and nyw. For these last it is again not clear if the cluster
is all within one syllable. Prenasalization and labialization both
occur; and they co-occur with /k/ and /s/. There are four nasal
contrasts /m,n,p,Q/.
Mor6, a language of the Nile-Saharan family, is spoken in the
southern part of Sudan (Cowan 1965). It has labiovelars as well as
imploded b and a. A voiceless alveopalatal is the only affricate
mentioned. Four contrastive nasals are found before vowels, /m,n,p,Q/.
Prenasalization, labialization, and a combination of the two occur.
The sequence Cr also occurs.
In the northeast section of the continent, it is said that nasal-
stop clusters (prenasalized stops) do not occur in Nilotic languages.
This feature. which is so common to African languages as a whole has been
found though in at least one language, Luo (Gre~ersen 1972). There are
examples cited of alternations between m and mb, n and nd, p and pj,
and Qand Qg. Luo also has 1 and nd alternations.

SIL-UND Workpapers 1980

 145

2. Vowels
Westennann and Ward claim that all African languages have an i
sound, au sound, and one or more a sounds and generally there is a
parallelism between front and bacl< vowels. Central and centralized
vowels are generally obscure and difficult to distinguish and usually
11 11

few in number. They further claim that back unrounded vowels had not
been found. Since then, however, they have been attested; a few examples
include Nsaw-Kom, Widekum, and Barnum - all of which occur within a small
geographical area in the western part of Cameroon (W&B 1952:124ff).
Rounded front vowels, central vowels, and back spread vowels are quite
rare but they do exist. Front rounded vowels have been noted in, for
example, Mambila in Nigeria (W&B 1952:143).
2.1 Inventories
According to Welmers (1973:20),. the vowel systems of most Bantu
languages are symmetrical, with either an even or odd number of phonemes.
(In the 11 odd 11 inventories, the odd number.is almost always a central
vowel.) From reading articles on various non-Bantu languages as well,
this seems to also be true for them.
Despite the fact that symmetry is claimed to be characteristic
of Bantu, languages have been found in which there is some assymmetry.
It seems that for Niger-Congo languages, there is however, "evidence of
at least an underlying symmetry" (Welmers 1973:21). In languages where
these exceptions appear, it seems that the assymmetrical phone only
contrasts in certain situations; otherwise it is an allophone of a
phoneme that fits the symmetry (for example, Efik).
For Moro {Cowan 1965), symmetry of the following system is achieved
by assigning the a to the back column. This however is not the way
11 11

the language actually works phonetically, according to the analyst.

(2) i a u
e
e 0
a

Fol lowing are some exampl,es nf types of S.Yf!lmetry:

smallest inventory found (_5) i u Wambo - Baucom 1972
E 0
a

largest inventory found (12) + u i u Dan - B&Z 1967

e e 0
E a 0
a a 'O

conunon example - odd number (7) u Efik - W&B 1952:134

e 0
e 0
a

SIL-UND Workpapers 1980

 146

convnon example - even number {6} u Ewe - Welmers 1973:21

e o
a :,

nonsymmetrical i u {some class languages of N. Nigeria}

L Irigwe - W&B 1952:108
e 0
e :, {some also have a, A, and u}
a

symmetrical i ..
a u Tiv 2 - W&B 1962:116
e a 0
a :,

symmetric a1 u i a {symmetry here is in three groups-

e 0 ; e T, u o a, and a aa ~}
aa a a Nupe - Smith 1967
Examples of languages with more than one central vowel include Dan,
Tiv, M;ro, Mambila, and Bamileke.
One language that has a very peculiar system is Higi of Nigeria.
Mohrlang gives this summary statement:
The vowel system of Higi exhibits a potential 4-way contrast in
11

word-final position and a 3-way contrast in word-medial position.

{These contrasts are i, e, e, a and~' e, a respectively.} This
already minimal number of contrasts is further reduced by (1)
neutralization of contrast medially in the pause group; {2} the
effect of prosodies; {3} grammatical conditioning; and {4} infre-
quent occurrence of one of the vowels. As a result, much of the
communication load in the language is carried by a simple 2-way
contrast. {p. 24}
11

The four word-final contrasts /i e ea/ are fairly static and ob-
viously always front vowels. Mohrlang does not attempt any explanation
of this. The three word-medial contrasts are fairly changeable:
(3) /+/ - [i L t U u]
/el - [ e o]
/a/ - [e A a:,]

depending on the presence or absence of the prosodies of labialization

and palatalization. {Mohrlang uses the Firthian approach to prosodies.}
2.2 Conditioning Factors
The main factors conditioning the pronunciation of the vowel pho-
nemes seem to be the following:
(a) placement in an open versus a closed syllable {that is, the presence
or absence of a syllable-final consonant)

SIL-UND Workpapers 1980

 147

(b) the preceding consonants or a combination of the syllable-initial

and syllable-final consonant
(4) Angas (Burquest 1971) /e/ [ea] / velars
/t/ [ tt>J /palatal~ r, velars
/ai/ [at] / in open syllables
(c) by the presence of labialization or palatalization (prosodies)
(5) Higi (Mohrlang 1972) /Ymt/ [mYi]

/Wwe/ [ wo]

(d) in Kpelle (Welmers 1973:23), a following vowel conditions a previous

one, with certain conditions on contiguous vowel sequences, presence
of certain consonants, and the features of the vowels in question.
(6) !oral [oara]
(e) tone rarely is a conditioning factor in vowel quality (Welmers
1973:23).
2.3 Length
Long vowels and diphthongs are another typical characteristic of
African languages, the former beinq much more prevalent usually than
the latter. Welmers says that phonemically long vocal segments can,in
11

every known case, be readily interpreted as double vowels" (p.24) and

if long vowels occur, clusters of heterogeneous vowels also are corrmon
(p.29). Welmers basically uses tonal evidence to support this claim.
Most analysts agree with this kind of analysis but in Dan, Bearth and
Zemp have analyzed length as a phoneme, a feature of the syllable.
Hence, in his analysis length is nGt writte~ as VV or V:.
In some languages, all the short phonemes have 11.mg counterparts;
in others only a few members of the inventory occur lengthened. In
many cases a vowel that appears long, V.V., is not basically V.: but
rather is a fonn derived from two diffete~t vowels assimilating one
to the other.
Languages with diphthongs usually have them begin and end on a
phone which is a member of the short vowel inventory and usually one
of the cluster is i or u. Occasional examplei of oa have been found.
No examples occur in the-literature of vowel clusters beginning or
ending on a central vowel other than a. However, in Yamba (Cameroon)
the following sequences do occur: ~i:- ae, and aa.

SIL-UND Workpapers 1980

 148

2.4 Nasalization
Nasalization is also a common phenomenon. Frequently it is pre-
dictable by its proximity to a nasal consonant but often it is not
predictable at all. It is also often the case that only a limited
number of the vowel inventory will occur nasalized.
Within vowel clust~rs, the distribution of nasality may ~e the_
same, that is, CVV or CVV or it may be different, that is, CVV or CVV.
These latter types are fairly rare.
2.5 Epenthetic Vowels
Vowels, mostly due to their sonorant qualities, seem to be gene-
rated often to fill in as a transition element. From the examples
seen, the vowel quality is either a duplicate of the main vowel of
the stem or a mid central phone a.
(7) Dan bt+ 3 "viper" [bt+ 3 J or [b+1t3] (B&Z p.15)
Bulu CVC#CVC (where# is a word boundary) is often pronounced
as CVCaCVC. There is still much discussion about the
true nature of this a. (Alexandre p.243)
2.6 Vowel Harmony
Vowel harmony is another phenomenon that occurs frequently. Niger-
Congo and Nilo-Saharan languages often have it, and non-Niger-Congo
languages often have a large vowel inventory which may have a partial
division according to some vowel harmony rules.
Vowel harmony is frequently analyzed as a feature (prosody) of
the syllable. Occasionally, however, it is predictable and therefore
not phonemic.
It typically occurs within a polysyllabic word or with pronominal
subjects harmonizing with the verb. It generally does not cross phrase
boundaries. In Yoruba (Bambgose 1967:268), there are also other elements·
called 11 verbal items 11 which will harmonize with the following vowel.
In Twi (Boadi 1963) the two sets of vowels are i u and 1. u
e o e ::,
a a
that is, a raised set and a lowered set. a is common to both sets.
Within a word, the main vowel is usually the second one so the first is
determined by it. In words with a as the main vowel, a factor of pala-
tality is relevant to choosing the vowel set. If the consonant preceding
the a is palatalized, the vowel of the first syllable (or prefix)
automatically has the corresponding vowel from the raised set.
The languages of the Wambo groups {Baucom 1972) are an example of a
system in which there is partial vowel harmony operating.

SIL-UND Workpapers 1980

 149

2. 7 Elision
Languages which have nouns that begin with a vowel often demonstrate
extensive elision. The fact of a vowel at the front of a noun usually
reflects the existence of a noun class system (which may or may not be
functional) and this sets the stage for elision particulary in two differ-
ent kinds of constructions: verb+ object, noun-noun associative phrases.
Elision can be merely an optional fast speech phenomenon or it can be an
obligatory process. Frequently it causes audible tone perturbations. And
it frequently co-occurs with consonant elision and causes segmental
changes as well as tonal ones. (The two environments mentioned are not
the only ones in which elision can occur but are merely exemplary.)
2.8 Miscellaneous
Westermann and Ward note that there is often an interchange between
front and back vowels within a language or between related ones. For
example, i might occur in one and u in the cognate in the other, ore and
o, ore and o. In Kpelle (Welmers1962), there are no contrasts between
/o/ and /we/-or between /o/ and /we/. In fact, Welmers feels that the
derounding and fronting diphthongs fit the pattern of the language better
if analyzed as /o/ and /o/.
Breathy vowels have been noted in several Nilotic lanquages.
In Beembe of the Congo (Jacquot 1962), the vowel inventory is as
follows:
i u t a i: u: t: a:
e 0 i:! cS e: o: i:!: cS:
a ii a: ~=
To eliminate some of the possible contrasts this can create, there are
rules of neutralization between a) oral and nasal vowels (both long and
short) if the vowel is high, b) oral and nasal vowels before a non-nasal
consonant, the oral is the representative, and c) long and short vowels
between consonants when a stem has more than four more.
3. · Tone
Until the last ten to fifteen years, it seems that tone was not con-
sidered important or very relevant to the study of African languages.
For example "tone has been noted in languages of Northern Nigera a
11 ,

conunent in Westermann and Bryan, p. 109. Those who did recognize tone
as pertinent, only understood it as far as it distinguished lexical items
or gramnatical functions, such as singular from plural on nouns.
The majority of tones noted are level - that is, spoken on one pitch
level, for example, high, mid, low, although contour tones, rising and
falling, are not uncommon. Most languages though are considered to be
level languages as opposed to "contour" languages. Within the desig-
11 11

nation of "level", languages labelled discrete have a fairly: absolute

11 11

pitch on which the different tones are uttered. Terraced level

11 11

languages function on a basis of relative pitch, each pitch relative to

the ones surrounding it.

SIL-UND Workpapers 1980

 150

Within both types of languages, discrete and terraced, languages have

been found with two, three, or four contrastive tones. Dan (B&Z) has
been analyzed as having five phonemic pitch levels, although this is very
uncommon in Africa and not accepted as a possibility by some analysts.
Some languages have a combination such as three level and two contour
tones, contours being tones of more than one pitch on one vowel without
additional length.
The terms high, mid, and low, are sometimes discarded in order that
a high may be called high even when it is not on the same absolute
pitch as a previous high. One phenomenon that can cause this change
in absolute pitch is called downdrift. For most languages it is purely
phonetic, the lowering of a nonlow by a preceding low. (This is typical
of Bantu languages.) Lows stay relatively static even in terraced
languages while it is the nonlows that really show the drifting or down-
stepping.
Downstep is another process whereby tones get lowered. This accounts
for a nonlow being lower than a preceding nonlow. (This definition
describes a two-contrast system but downstep can occur in.languages with
more contrasts than just low and nonlow.) Welmers defines downstep as
"a phoneme conditioning a lowering of the pitch of a high tone; it applies
to a sequence of syllables in one utterance." (p.89) A low tone can be
lowered by the feature downstep as well as a high. Downstep is generally
felt to be caused by a latent or lost tone between the two nonlows. Most
analysts feel this tone would have been low but Welmers feels it some-
times also may have been a nonlow (p.87). The reason this tone has been
lost could be due to segmental contraction or allomorphy in which certain
vowels are silent or 0. A phenomenon of tone raising in which a tone can
be raised above its normal pitch due to some conditioning factor can also
be a source of downstepping, the downstep showing up on the nonlow follow-
ing the raised tone. An apparent downstep may be conditioned by certain
consonants, but this can be explained better by phonological conditioning.
An opposite phenomenon of upstepping could be possible where a non-
low is higher than a previous nonlow which cannot be explained some other
way. This has not been too well documented.
Tone does not have to be thought of as a definite pitch that is
assigned to each and every vowel in a tonal language. It sometimes appears
as if the tone of certain vowels extends over more than one vowel of a
word or morpheme even when the vowels are not juxtaposed (as in a diph-
thong). Other times more than one tone can have dominion over one vowel.
Spears calls this the domain of a toneme. For example, a tonal unit may
be high and its domain one or more adjacent vowels. The same applies
for a tonal unit which is low. This may imply that some syllables are
toneless or neutral with relation to tone. If the tonal effects spread
across morphemes, this is now tonal extension. The domain of the last
tonal unit spreads to include the first (or only) vowel of the next mor-
pheme. Verb reduplications also may exhibit tonal extension. Bell used
the term "prosody" to label the morphemes in which a given tone covers
more than one vowel. Should these· morphemes be compounded, he has found
that the tones of the first element of the compound are completely
nullified.

SIL-UND Workpapers 1980

 151

There is also a phenomenon in which one tone affects the following

tone according to what follows that second tone. The middle tone of the
sequence is polarized that is, it becomes high before a low and low
11 11 ,

before a high or silence. This special effect is conditioned by the

presence of a polarizing tone, which is the first in the sequence. For
examples and further discussion, see Spears 1967 and 1968. A polarizing
tone may come from a polarized one if the operation has repeated appli-
cation in a long sequence.
As stated previously, there are two main functions of tone - gramma-
tical and lexical. For conditioning of both these types there is a
variety of factors involved:
(a) Phonological Conditions
One particular condition mentioned (Welmers 1973:94) is called de-
pressor onset. The features of the consonant or consonant sequence begin-
ning the syllable depress the tone and cause what looks like downstep.
In Ewe, voiced stops and fricatives condition low tone on noun stems.
Phonological conditioning should be the first possibility tried when
formulating tone rules.
{b) Lexical Conditions
In many languages a given string of segments may have different
meanings dependent entirely on tone. It is common that there will not
be a complete range of all possible contrasts, however, for even one
given segmental string. Minimal tone contrasts are not necessary to
call a language tonal. The distribution of tones on nouns and verbs
is often different within a language.
(c) Morphological Conditions
There are morphemes whose only realization is a tone, called a
floating tone
11 Because it does not have any segments to attach to. it
11 •

is absorbed in the preceding or following tone and very often causes some
tonal changes in that neighboring tone. One very corrmon instance of
this is associative noun phrases. in which there is a tone between the
elements of the phrase, usually reflective of some concord between them.
This tone is often very difficult to discover. There are also affixes
which consist of a tone that goes with the stem consonants and vowels
instead of the regular lexical tone of that stem. There are also affixes
which in addition to their own segments and tonal representations, have
a special tone for the stem segments.
(d) Syntactic Conditions
The grammatical relation that a word or morpheme bears in a sentence
may determine its tone. Examples of this include the following construc-
tions: certain relative clauses, imperative, negative imperative, loca-
tive phrases, verb tenses, various noun-noun constructions, or the relation
a given noun bears to the verb of a clause.

SIL-UND Workpapers 1980

 152

Tonal contrasts in any of these contexts or affected by any condi-

tioning factor need not be maintained in their entirety. That is, in
one situation, the language may exhibit a three-way contrast but in
another only a two-way contrast; two of the contrasts will have been
neutralized. (The tones manifested will be two of the three original
contrasts usually.)
Although early works on African linguistics make such comments as
"stress may be significant" or "in most Bantu languages stress is on
the penultimate syllable" (Westermann and Ward, p.114,115) or "stress is
of secondary importance" (W&B p.134) Welmers maintains that 11 no (African)
language has been reliably reported to have both tone and stress in the
phonemic system" (p.113). However, he does admit to a system of fntona-
tion covering a tonal system, but only a very limited set of intonation
patterns have been found to so exist (for example, in Hausa).
4. Noun Classes
4.1 Traditional System for Bantu
The expression "noun cl ass sy~tems 11 is traditionally used to describe
a system of affixes which appear in a language to classify nouns into
different categories or classes. All nouns in one particular construction,
for example, a. possessive phrase, will not have the same affix. The
affix on the noun stem may be different, the affix on the possessor may
be different, and they may all be different again depending on whether
the noun phrase is subject or object of the sentence or whether it is
singular or plural. Further differences may be phonological or morpho-
phonemic due to underlying forms of the root or stem of the nominal. Bantu
lanquaqes are typically associated with such a system and much study has
gone into the reconstruction of the proto-Bantu system of noun class
markers.
This system differs from the typical Inda-European system of genders
in several ways: there are many more classes than the typical two or
three genders, certain semantic distinction·s are relevant but generally
sex is not one of them, and number and gender have no correlation.
The classification into classes has nothing to do with anything in-
herent in the phonological shape of the stem but is entirely dependent
on the affixes the stem takes. Each noun belongs to a pair of classes -
the singular being one, the plural the other. Not all nouns that have,
for example, class 1 affixes in the singular, take class 2 in the plural.
Some may take class 10 or another class. Likewise, nouns that take, for
example class 10 affixes in the plural, will not all take class l affixes
in the singular. This system has been set up and used by scholars
for many decades now and the labels (numbers 1 through 23) are coreferen-
tial between languages even though the languages may not be related at
all. The numbering system is not merely a random assignment of a number
to the list of prefixes, one number for each prefix.
Following is a brief summary of the classes of proto-Bantu according
to semantic similarities:

SIL-UND Workpapers 1980

 153

1, 2a, 2b - kinship, personification, proper names, some animates,

rarely inanimates
3, 4 - trees, plants, inanimates
5, 6 - miscellaneous - augmentatives
6a - liquid masses {no plural)
7, 8 - miscellaneous - diminutives
9, 10 - animals, some inanimates
{in the above 10 classes, the odd numbers are singulars, the even plurals)
11 - long, thin objects, abstracts
12, 13 - diminutives
14 - abstracts,fermented beverages from grain or seeds
15 - verb infinitives
16, 17, 18 - locatives: near, explicit; remote, general; inside
{respectively)
19 - diminutives
20 - augmentatives
21 - augmentatives {pejorative)
22 - only found in one language so far, LuGanda
23 - locative
{classes 20 - 23 are all rare)
Occasionally a noun will have affixes of more than one class. The stem
may take the prefix of the class to which it belongs, as well as the pre-
fix of, for example, class 21, which adds a pejorative overtone to the
meaning.
In a few languages, class 5 has morphophonemic alternations in the
stem {b I J have alternants ts and c, and r g h have alternants sand sh
(Welmers 1973:168)). However, more general alternation occurs in prefixes
conditioned by the stem-initial vowel.
Not only are there affixes which appear on the nouns, there are
also elements of "concord" which show agreement between words in a con-
struction. For example, the concord system will relate a nominal to a
demonstrative, a possessive {personal pronoun may be distinguished
from nominal), an attributive, a numeral, an interrogative, a relative
pronoun, as well as concord for subject or object, and other categories
as well. It is sometimes the case that the noun class system is so com-
plex that the concord system is used to define the classes. This is
true for some languages in the Republic of Benin, for example. This
reduces the number of classes that a language will have.
4.2 Vestigial Systems
The Bantu system as described above is quite distinctive. The rest
of the Niger-Kordofanian family {see Appendix 1) with the possible excep-
tion of Mande acts somewhat differently with respect to noun classes.
Kordofanian languages mark classes with prefixes which are conson-
3
antal for the most part. They show similari ies with Bantu classes 3, 4,
5, and 6a. According to Greenberg's listing there are 25 classes.
There is also a system of concord.
These similarities between Bantu and Kordofanian suggest that there

SIL-UND Workpapers 1980

 154

must have been some sort of class system in proto-Niger-Kordofanian.

The other branches of Niger-Congo give some clues as to what this system
may have been.
Probably because these languages are non-Bantu, most analysts look
on the class systems of the following languages as embryonic; Welmers
takes the view that they are vestigial.
The Kwa Branch - singular-plural distinctions are marked by a prefix
which is a vowel or syllabic nasal. There are few traces of concord.
"These languages (Yoruba, Igbo, Efik, Akan cited as examples) have
no affixal pluralization of nouns, no concord, and very little else
immediately reminiscent of functional noun class systems. There
are significant features in the structure of these languages, how-
ever which are by all odds best explained in tenns of vestigial
noun class systems." (Welmers 1973:189)
Verb roots are typically monosyllabic: CV or eve. And nouns are
generally of the shape VCV or VCVC. Welmers feels that the basic dif-
ference between the shapes of these two groups of morphemes suggests
that the initial vowel of the nouns is a prefix. This is supported by
the fact that in related languages there often is a functioning system
of prefixes on nouns. Even within a language, there are forms which
suggest strongly that this is the case.
(8) Yoruba /ewe/ "leaf" /iwe/ "paper"
The Gur Branch - there is a functional noun class system, singular
and plural distinctions, some concord, but basically the system is
suffixal rather than prefixal. Generally there is no attributive concord.
Suppire, however, appears to have several types of concord, forms for
identifiers, subject and object, possessive, remote demonstratives, demon-
strative copulatives, attributive interrogatives, and independent or
nominal interrogat·ives. Gourma, another Gur language, has both prefixes
and suffixes, either identical or very similar to each other; the begin-
ning and ending of a nominal would thus be clearly defined.
Adamawa-Eastern - according to Greenberg there are suffixes to dis-
tinguish pairs of singular/plural noun classes. He feels that there are
many parallels between Bantu class prefixes in form and semantic cor-
relates and these suffixes. There is some concord, but not with plurals.
West-Atlantic Branch (Fula) - there may be up to 25 classes, marked
with suffixes. The choice of omorphy of the suffixes is lexically
conditioned. But there is also initial consonant alternation. It will
be a plosive if singular and a fricative if plural in certain classes.
This alternation of the initial consonants is very possiblv the remnaat
of a set of prefixes. Also, the initial consonant of a verb may be
determined by the class of a noun subJect - either a fricative or a nasal.
There is a full system of other concords as wall-

SIL-UND Workpapers 1980

 155

Tivis a non-Bantu language of the Niger-Congo branch. There are

11 classes: l has no affix, 5 have prefixes only, 2 have suffixes only,
and 3 have both prefixes and suffixes. To identify each class, a con-
cordial morpheme is used rather than the nominal affix.
On the basis of the evidence that suffixes seem to play an important
role in noun-class systems, as well as do prefixes, Welmers comes to the
conclusion that at some stage of pre-Bantu, nouns in all classes had both
prefixes and suffixes. Over time, some languages have been left with
only prefixes, others with only suffixes, and others with both which
gradually lost some of one or the other at different times in their
development. He makes this comment:
11 It appears that prefixes alone, suffixes alone, or both prefixes
and suffixes are no strangers to Niger-Kordofanian noun-class
systems. Although systems with prefixes only or with suffixes only
are the most common, there is evidence of· both prefixes and
suffixes in every branch of Niger-Congo which has noun classes
at all.
11 (p.204)
4.3 Other Systems
Mande has a distinction in nouns between "relational" and free 11 11 :

free being those whose stem can constitute a whole noun phrase, relational
being those that need an explicit possessor. A possessed free noun will
have special markings. There is also a secondary distinction - personal
versus nonpersonal. A distinction between singular and plural is not as
relevant as one between generic/general and individual/specific. The
distinction of individual nonpersonal nouns and general personal nouns
is reflected in other branches of Niger-Congo as well.
Afro-Asiatic languages have another different system. They have
two genders, reflected in the forms of nouns and pronouns. Masculine
nouns are male persons and animals and various inanimates; feminine
nouns are female persons and animals and other miscellaneous. In the
Berber branch there are gender and pluralization differences. In addition,
kinship terms act differently from all others.
Cushitic languages also generally have a two gender system. In Saho,
masculine nouns have stress and feminine nouns are those-without stress.
Intersecting with these two genders there are three categories: l - mass
nouns, 2 - generic nouns (unspecified quantity), 3 - nouns with singular
and plural.
In many languages, in addition to having gender and number distinc-
tions, for any noun there are two forms which are called the absolute and
11 11

and "construct". The construct form is used when a noun is used in a

particular grammatical construction, for example, the second noun of a
noun-noun phrase, after numerals, after prepositions, or for noun subject
when it follows the verb (the normal position). The absolute will be

SIL-UND Workpapers 1980

 156

used elsewhere. The rules governing the choice will be language-

specific.
In Nile-Saharan languages, the most complex and irregular group of
nominal variations is found, particularly involving plural formations.
This part of the summary (section 4) is based mainly on chapters
6 - 8 of Welmers, African Language Structures.
5. Adjectives
5.1 Traditional Approach
In many languages it seems that the postulation of a class called
adjectives 11 is based on a semantic relationship that one word holds to
11

another, comparable to a similar relationship in Inda-European languages

which is traditionally called "adjectival". That is, if an utterance is
translated, for example the good boy", there must be a word within the
11

phrase which means 11 good 11 and bears an adjectival relationship to the

noun 11 boy 11 • The following are given as examples of adjectival construc-
tions. (Data from Westermann and Bryan)
(9) Tiv ub6 mbA ukasev "ugly women 11 {p.118)
ugly genitive female
Songhai bolo bi II black man 11 {p.47)
bolo-bi-ai 11 black men 11
Mande pe: le: kwele "white house" {p.45)
pe: le: kweler:ia "white houses''
be: le: kwelai "the white house"
be: le: kwelel)ai 11 the white houses 11
This then could be considered as a word class approach to the analysis
of adjectives, in which the definition of an adjective is dependent on
its semantic function of qualifying/modifying a noun.
5.2 Example from Xhosa
The Nguni languages of South Africa (Jordan 1967) have traditionally
been analyzed this way.
(10) um-ntwana omhle "beautiful child"
class prefix-child beautiful
omhle is analyzed as a- "qualificative formative"
-um- "noun class prefix"
-hle "beautiful"
When the formative and noun prefix are juxtaposed, certain predictable
vowel changes occur. a + um = om

SIL-UND Workpapers 1980

 157

Jordan has found however that, while this analysis appears tenable
in the affirmative, in the negative there are problems. In the first
place he notes that only predicates can be negated in these languages.
Xhosa then should not allow that adjectives can be negated. Furthermore
the negative morpheme -nge- splits the coalesced form of "formative +
noun class prefix" and he feels that this nullifies that analysis of om.
(11) um-ntwana o-nge-mhle "a not-beautiful child"
Other morphemes can also be placed in this position.
(12) um-ntwana o-se-mhle "a still-beautiful child 11

If then omhle is actually a predicate, because it can be negated, and

not an adjective modifying a noun, because the obligatory prefixes can
be split, a different analysis for the om of omhle must be found. What
Jordan proposes is that omhle is actually a relative clause.
(13) um-ntwana o mhle "a child who is beautiful"
Theo becomes the relative marker and mis the noun prefix. (The u of
um- is omitted as Jordan regards it as-a definitive article, not as part
of the prefix.) Note the following examples.
(14) um-ntwana o li-layo "the child who is crying"
um-thi o khu-layo 11 the tree that is growing"
And note the following pair:
(15) um-ntwana o mhle "the child who is beautiful" in Xhosa
um-ntfwana lo mhle "the child who is beautiful" in Bhaca
(].Q_ is analyzed as a relative clause marker for Bhaca.)
(16) um-ntwana o ngemhle "the child who is not beautiful"
um-ntwana o semhle "the child who is still beautiful"
Jordan is saying that there is not a class of words called "adjec-
tives" but rather that the concept of qualifying a noun is expressed by
the use of a type of relative clause.
5.3 Welmers' Non-Criteria
For the purpose of trying to resolve the problem of how to define
an adjective, Welmers tries to set up criteria on which to analyze the
concept of modifiers. He devotes a chapter to the notion of "adjectives"
especially as they relate to Niger-Congo languages. The following criteria
are discussed as being inadequate bases for the establishment of a class
called adjectives.

SIL-UND Workpapers 1980

 158

a) Semantics - having a qualificative or attributive meaning is not

sufficient evidence in itself. A phrase, noun +adjective as attributive,
may not be what it seems. Consider the following data.
(17) Suppire N;.re-qe
r,.~' ' ' !<Ere b6 1 -a "farm big farm"
(p.263)
'
kak5-0 kak5' b6 1 -a 11 1izard big lizard"
The first noun belongs to one class and the second to a different one,
as shown by the fact that they have different class markers as suffixes
in isolation. However, when the concept of 11 big 11 is added to the noun,
the marker of the second element of the phrase remains the same, showing
that the relevance of the class of the first element has been negated or
eliminated. The class marker is determined by the qualificative rather
than the nominal. Welmers feels that the data suggests that the modifiers
are also nominals and that they combine with the head noun to form a type
of compound noun. This may partially explain why the first noun is not
marked for its own class. If this analysis is correct, there is no class
of 11 adjectives 11 in Suppire despite the semantics of qualification. To
distinguish between noun +adjective phrases and noun +noun phrases, there
would have to· be significant differences especially with regard to tone.
b) Being a quantitative, numeral, or demonstrative - these categories
also are rejected as adjectives as there are often distributional consid-
erations which set them apart from other qualificatives. For example, in
Igbo there is a limited set of adjectives from which numerals and words
such as 11 some 11 and 11 any 11 are excluded because they function as nouns in
places similar to the phrases analyzed as noun-noun for Suppire (above).
c) Verbal morphology - a situation of a verb acting as an adjective or an
adjective as a verb is not acceptable analysis: a verb is a verb.
5.4 Structural Criteria
Welmers seems to indicate that the definition of an adjective must
be structural, that is, based on distinctive morphological or distribu-
tional characteristics. There are several ways these changes can be marked.
a) Segmental changes - a prefix or a suffix can be added to distinguish
adjectives from verbs or nouns. Verb roots may undergo a process of
reduplication or affixation.
(18) Root Adjective 4
Reference
se le I) ':i. le rJ-o 11 hang 11 /"hanging 11 Kpelle W:251
j~
• I
J 1-J~
•
11 eat 11 / 11 eating 11 Yoruba W:257
qa qa-to 11 illness 11 /"ill 11 Mal inka W&B:43

SIL-UND Workpapers 1980

 159

b} Tonal changes - a verb stem plus a tone which is not the regular verb
stem tone is another type of distinctive morphological change. Redup-
licated fonns also may have distinctive tone patterns which mark them as
adjectives.
(19) Root Adjectives Reference
waa ~a 11 wash 11 /"clean" Kpelle W:251
hwa h6-hwa 11 carve" /"pointed 11 Jukun W:254

worn weswom 11 be dry"/"dry" Jukun W:254

c) Distributional restrictions - in Fante an adjective can be followed

by an adverb but a nominal cannot.
d) Welmers also discusses a fourth way of expressing qualificative con-
cepts for some languages - by using ideophones. Although (or perhaps
because) the definition of "ideophone' varies considerably among linguists
Welmers defines some reduplicated forms as ideophones even though some of
them may have root forms with similar meanings. The reduplicated forms
function attributively following nouns. Discussion of this is not ex-
tensive (see section 15.71n Welmers 1973).
For many languages a variety of these criteria together decide
whether a word can be termed an adjective or not. In Igbo there is
a limited set of adjectives (with eight members) from which numerals
are excluded on the basis of semantics and other (typically adjectival)
words are excluded on the basis of tone changes. Of the 8 that are
accepted as adjectives, tonal behavior is the criterion for 3 while the
other 5 would probably be excluded on the basis of tone alone as there
are no changes. These other 5 are adjectives based on semantics - an
adjective makes a noun become a member of a category of things which the
adjective describes, for example, "new" or 11 large 11, whereas a simple attri-
butive would behandled by a relative clause with the nonadjectival form
of the attributive in question.
\ , , • ., I .,
(20) (p.259) uwe OJ I I "dark clothing" for a particular function
, eg. a uniform
\
uwe ojf "dark clothing" in general
( !' 1d ! i S the verb "be described as")
onye 1uwe oj f If "person with dark clothing,
eg. a policeman"
Some languages have no distinct class of qualificative adjectives at
all. Others have a limited set. Igbo as mentioned previously has a clear
set of 8. Swahili has a list of 50, distinguished from noun stems in that
they are not restricted in class membership as are nouns and they take the
concord of the noun referred to.

SIL-UND Workpapers 1980

 160

Bariba {Welmers 1973:268) has 3 types: invariable, class-bound, and

class-inflected. The first 2 types may possibly be types of nouns as the
concord of the modified noun does not affect the "adjective" but the
third type is definitely adjectival as inflections and concord are direct-
ly related to the class of the noun.
In Gbeya {Samarin p.80) there are reportedly 3 types also: preposed
class A, preposed class B, and postposed. The postposed class is comprised
of only 2 members which both .act as a kind of demonstrative which probably
would not be accepted as adjectives under Welmers··· conditions. The pre-
posed class A group has only 4 members which also have restricted environ-
ments and rules of operation. These also may not be actual adjectives
according to Welmers. But the preposed class B group seems to fit, as the
forms have derivational affixes and do differ from noun-noun associative
constructions in some respects.
Returning finally to the examples given in data {9) from Tiv,
Songhai, and Mande, Songhai and Mande show that a plural marker or a
definitive marker are added to the "adjective" of the noun-adjective
phrase instead of to the "noun". {Hyphens in the data were given.} From
the structural point of view this may indicate an analysis similar to that
proposed by Welmers for Suppire (section 5.3). The nominal + adjective
is actually a compound noun.
Tivis slightly different. The "genitive marker" in the following
examples is a type of concord from the noun class system. Westermann
and Bryan posit that there are 3 types of adjectives: those that maintain
all class affixes, those that maintain only the suffix, and those that
drop all affixes. Examples of each type are given below.
{21) {a) U - b6 rhba U - kcise - V
class 11 - ugly class 11 class 11 - female - class 11
prefix genitive prefix suffix
ub6 rhba ukasev "ugly women" {lit. "ugly ones of
female")
{b) kase - v mba b6 - V
women - class 11 cl ass 11 bad - class 11
suffix genitive suffix
ka sev mbA b6v 11 bad women"
{c) kwas~ u b6
woman class 1 bad
genitive
kwas~ u b6 "bad woman 11
Because both 11 bad/ugly11 and 11 women 11 have identical shapes in their
"nominal and 11 adjectival II forms (including the tone) (compare the (a) and
II

{b) examples, nominal forms always precede, adjectives follow), Welmers

SIL-UND Workpapers 1980

 161

would insist that they should not be disti·nguished and that these are
all actually noun-noun constructions.
There are 11 real 11 adjectives in African languages but until this area
is understood more fully the use of that term will always open the door
for much debate.
6. Multiple Verb Constructions
In many West African languages., a sentence may consist of several
verbs strung together. It is generally felt that these strings function
differently than the classic coordinate or subordinate relationships in
other languages. There are two types of structures in which these strings
occur: a consecutivized structure and a serialized structure.
The key question in the analysis of these verbs is: where do they
come from? What is the underlying structure? Hyman suggests that they
probably come from a structure that looks like the following:
(22) So
~2
Many authors have proposed various analyses but no one has yet been able
to prove his theory better than all the others.
6.1 Consecutivized Structures
According to Hyman ( 1971 :31), a 11 consecuti ve structure" contains (at
least) two verbs of a sentential conjunction, the second verb of which
represents an action subsequent in time to the first verb and is done for
the purpose of the first verb. Mainly with reference to Fe Fe (Bamileke) 1 1

he lists 4 different types. The first type he calls coordinate and

subordinate conjunction.
(23) (a) coordinate (i) ~ ka. s6.? n ZO: wu ZO:
he PAST come &eat food
11 he came and ate 11

(ii) c!i k6. s6.? o zci wuza., -

he PAST come you eat food
11he came and you ate 11
(b) subordinate - marking purpose or intent
( i) ~ k6. so.? (6.) zu wu zu
he PAST come CONJ eat food
11 he came to eat 11

{ii) ! k6. so.? 6. o zu wuzu

he PAST come CONJ you eat food
11 he came in order for you to eat 11

SIL-UND Workpapers 1980

 162

In (23aii) there are no overt markings while in (23ai) there is deletion

of the second subject because it is coreferential with the first and also
N-insertion, a marking of consecutivization on the second verb. In the
subordinate relationships expressed in· (23b), there are no verb markings
but there is a relative clause marker inserted (optionally if the second
subject is deleted).
In addition ba "to be" can be inserted between the two verbs meaning
"continually or simultaneously". (ba becomes mba becomes ma.)
(24) a k6. su? ma n 20. -
WU' 20.
he PAST come &be &eat food
"he came eating"
a k6. su? (a) ba za , -
wuza.
he PAST come CONJ be eat food
"he came (only) in order to be eating"
The second type Hyman di scu·sses is consecutives with the verb "to
take" lah. This verb in many languages has come to mean "accompaniment",
"instrument", or even manner
11 Giv6n's article discusses verbs with
11 •

similar meaning and he sets forth these questions: (a) are they syn-
chronically verbs or prepositions? (b) if verbs, are they synchronically
coordinate or subordinate structures? (c) diachronically, does serializa-
tion (or consecutivization) arise from conjunction or subordination?
He does not attempt to arrive at any indisputable answers, but tries to
show that for cases of serialization, what has happened is that one verb
in a string of multiple verbs has become "grammaticalized". By that he
means there have been three types of changes: (a) semantic - depletion
of meaning from the grammaticalized form, (b) morphological - loss of
ability to take verb affixes (agreement, etc.), and (c) syntactic -
maintains position of verb but acts like a conjunction. These shifts
occur gradually and various stages of each type of change can be seen
in languages that are grammaticalizing some of their verbs.
In Hyman 1971, he presents both ma (from ba "to be") and rah (from
lah "to take") as grammaticalized forms (he calls them grammatical mor-
phemes). In their "grammatical II form they occur as the second in a con-
secutive series.
(25) a k6. su? m6. c6.k
he PAST come &be pot
"he came with the pot"
a k6. t hi p le: nah ncwee mb6.a.
he PAST forge knife &take &cut meat
"he forged a knife and cut the meat with it"
"Take", the verb, can also be consecutivized with "take", the grarruna-
tical form, and another verb.

SIL-UND Workpapers 1980

 163

(26) a k6. la h c6.k nah ns6.?

he PAST take pot &take &come
11 he took the pot and brought it 11

a k6. la h c6.k (6.) la h s6.?

he PAST take pot CONJ take come
11 he took the pot in order to bring it 11
Perhaps this type would better be called grammaticalized consecutives and
should probably include the last data in the discussion of the first type
as wel 1.
The third type of consecutivization that Hyman discusses is multiple
consecutives, when there are more than two (main) verbs. Whether or
not a linker (like nah) occurs, you can have an infinite number of con-
secutives each subsequent to the one before. This is also true if the
conjunction nf 11 and then 11 occurs between each clause or only between the
last two clauses. That is, the following four diagrams are all equivalant.
(27)

~
s1 s2 _ ~
S3 s4
=
~~
s1s s s
nah 2 nah 3 nah 4
=

s
/1\,~
s
1 nis s2 ni 3 ni 4
= . ,,,....
__

s. 1 s2 s3
-r~ "
. ......
. ···-.......................
ni:::...~s4
(with or without nan) (with or without nah)
However a conjunction, like nf, between any two other than the last
two changes the meaning. ~

(28) or

but s1 ni s2 s3 can only be so

/ )fa
S1 / \
S2 s3
More explicitly, the sentence below can have three structures:

SIL-UND Workpapers 1980

 164

11 He forged a knifek, carved a spoons with itk, and ate with itk."
s1 , nan S2, nan s3 .
Sp nah S2, rii" nan s3 .
s1 , n, nah s2 , n1 n~h S3.
But s1 , n, nah s2, nah s3 can only mean one thing: "he forged a knifek,
carved a spoons with itk, and ate with its." The n'i seems to disasso-
ciate clauses.
The sentence "he went to market, bought yams" implies that he bought
yams at the market. However, he went to market n'f bought yams" does not
11

imply either that he bought them when he went to market or that he bought
them there at the market.
The fourth type of consecutive mentioned in Hyman is consecutives
within auxiliaries. These act somewhat similarly to grarmnatical forms
but are not well understood. Examples of such auxiliaries from Fe'Fe'
include the fol lowing: :&!.. "to stay behind", pat to answer", vah to 11 11

pass the day", ta'lsf "to embracell, kwee( ns)) "to join". One thing
though is fairly certain: the following sentence is not to be taken
literally.
(29} po ka. pat nta?sT mfet?
they PAST answer &embrace &work
It does not have the structure s
/10~
S1 S2 S3
but rather it means "they worked again together again" and can be
diagrammed
So
Alx \
I /a\
s1 A~X s3
S2
Welmers has a little broader definition of consecutive He says that
11 11 •

the definition depends to some extent on the language. Hyman seems to say
that consecutivization occurs before serialization synchronically, the
latter being derived from the former, and that at a given point in time,
a language will have one or the other. But Welmers says that there may be

SIL-UND Workpapers 1980

 165

languages in which no distinction is made at all between the two types

or those for which the 11 consecutive 11 refers to the verb in a sentence,
not a relationship to time, or languages which have both consecutives
and serialized forms, being differentiated by syntactic structures. He
sets out a definition though of a "construction used to refer to actions
after the first in a sequence .11 {p.364).
As examples he includes the following language-specific information.
In Kpelle the consecutive is used for showing purpose or for simultane-
ous action; two different constructions handle this. Swahili has a
definite marker for a consecutive construction and both of these languages
necessarily repeat the subject pronoun with each verb. In Efi"k, consecu-
tives may express simultaneity but the verb semantics are not as closely
linked as they are with serializations.
6,2 Serialization
Hyman defines this as "cases where two verbs occur within one sentence
but do not enter into (that is, are not marked for) any of the coordinate
or subordinate relationships defined elsewhere in the language" (1971:30).
As mentioned above, serialization evolves from consecutivization. And in
one language (Nupe) there are examples of three structures which all mean
the same thing.
(30) "he brought the pot" iT 1a duka iT Cl t>e'
he take pot he and come
iT 1a duka' c'f be'
he take pot and come
(serialized form) iT 1a duka' b~
he take pot come
Giv6n's article mainly discusses the further evolution of serializa-
tion to grammaticalization and how this has affected word order in lan-
guages. He states that grammaticalization comes from a specific type of
verb serialization. Hence you would have verb strings (-+ consecutivized
forms)-+ serialization-+ grammaticalization. This seems to contradict
Hyman's view of grammaticalized verbs as a type of consecutivization and
shows how much disagreement there is in this area of verb analysis. In
fact Hyman feels that Giv6n says that serialization means that two verbs
are one event or action, that is, that one verb is usually grarrmaticalized.
Welmers' definition of 11 seria l ization 11 is based 1argely on the fact
that new information cannot be introduced in the second or sequential 11
11

verb but the verbal semantics must be quite closely related. For example,
the two ideas "he is going to market" and "he is carrying a headload"
have to be associated in time and person if they are to be used in a
serialized construction. In Efik it appears that both serialization and
consecutivization may occur but in different situations.

SIL-UND Workpapers 1980

 166

As can be seen there are inconsistencies in the definitions and

environments in which these two types of verbal strings occur according
to the person who is analyzing the data at hand. It is fairly certain
that this structure has its distinctive features, and is a good topic
for more in depth research.

FOOTNOTES
1 This report was written for a reading course in African linguistics
at SIL, North Dakota, 1979, based on materials available in the SIL-UND
library.
2 Welmers 1973:21 cites Tiv as having the same system as Ewe listed
above. It is not known if the difference is a change of analysis or
merely different dialects.
3 All references to Greenberg's work in this report come from Welmers'
African Language Structures.
4 Wis Welmers 1973 and W&B is Westermann and Bryan 1952.

SIL-UND Workpapers 1980

 167

APPENDIX 1
Afro-Asiatic NilAan Khoisan Niger-Kordofanian

Semi(;J Chari-Nile others

~
Kordofanian Niger-Congo

A /'\\
Higi Hausa E.S~tic

I
Dho-Luo
I Hatsa
·. Tumtum 4 others

Niger-Congo
~-- ·---=----
Gur~ /
West-Atlantic ""' ~wa - ---::_---
Adamawa-East~e--rn-:-----.B~e=n--ue--......eongo

A
Dan Kpel le
\.
Tufo
A··,.,_
~- / I
K/kan ~:~ant J
/~
I/"' .~
Benue-Cross Bantu

Gourma Fula Twi Ewe Gbeya Tiv Bamileke Jukun

Suppire Efik . I
Senari Igbo Bulu Fe Fe 1 1

Nupe Dual a
Yoruba Swahili
Xhosa
Zulu
LuGanda

Approximate Locations of Certain Groups

== Afro-Asiatic
"' Ni lo-Saharan
··~ Khoisan
~ Niger-Kordofanian
~ Bantu

SIL-UND Workpapers 1980

 168

REFERENCES
Alexandre, P. 1962. "Sur la Voyelle Suffixielle du Bulu". Journal of
African Languages. 1:243-252.
Baucom, K.L. 1972. "The Wambo Languages of South-West Africa and
Angola". Journal of African Languages. 11:44-73.
Bearth, Thomas and Hugo Zemp. 1967. "The Phonology of Dan (Santa)".
Journal of African Languages. 6:9-29.
Bel 1, Herman. 1968. "The Tone System of Mahas Nubian". Journal of
African Languages. 7:26-32.
Boadi, L. 1963. "Palatal ity as a Factor in Twi Vowel Harmony".
Journal of African Languages. 2:133-138.
Burquest, D.A. 1971. Studies in Nigerian Languages, No. 1, A Pre-
liminary Study of Angas Phonology. Zaria:Insti"tute of linguis-
tics and Kano, Nigeria:Centre for the Study of Nigerian Languages,
Abdullahi Bayero College, Ahmadu B"ello University.
Cowan, W. 1965. "A Note on the Phonemes of /M~ro/". Journal of
African Languages. 4:114-117.
Dunstan, E. 1971. "Noun Class Systems in Mbam-Nkam". Journal of
African Languages. 10:15-28.
Eastlack, Charles L. 1968. "The Noun Classes of Ngemba". Journal of
African Languages. 7:33-40.
Given, T. 1975. "Serial Verbs and Syntactic Change: Niger-Congo".
In C. N. Li, ed. 1975. Word Order and Word Order Change.
Austin:University of Texas Press.
Gregersen, E.A. 1972. Nilotic Nasal-stop Clusters".
11 Journal of
African Languages. 11:79-81.
Heine, B. 1968a. "A Preliminary Survey of the Noun Classes of Basila".
Journal of African Languages. 7:1-13.
--------- 1968b. "The Allocation of Loan-Words Within the Nominal
Class System of Some Togo Remnant Languages". Journal of African
Languages. 7:130-139.
Huntingford, G.W.B. 1965. "The Orusyan Language of Uganda". Journal
of African Languages. 4:145-169.
Hyman, L.M. 1971. Consecutivization in Fe Fe
11 1 11 1 • Journal of African
Languages. 10:29-43.
---------- 1975. "On the Change from SOV to SVO: Evidence from Niger-
Congo". In C. N. Li, ed. 1975. Word Order and Word Order Change.
Austin:University of Texas Press.

SIL-UND Workpapers 1980

 169

Jacquot, A. 1962. Notes Sur la Phonologie du Beembe (Congo)

11 11 •

Journal of African Languages. 1:232-242.

Jordan, A.C. 1967. The Adjectives in the Nguni Languages of Southern
11

Africa Journal of African Languages. 6:132-145.

11 •

Mohrlang, R. 1972. Studies in Ni~erian Languages, No. 2, Higi Phonology.

Zaria:Institute of Linguistics, and Kano, Nigeria:Centre for the
Study of Nigerian Languages, Abdullahi Bayero College, Admadu
Bello University.
Ne\tAllan, R.M. 1971 11 Downstep in Ga'anda 11 • Journal of African Languages.
10:15-27.
Sarnarin, W.J. 1966. The Gbeya Language, Grammar, Texts, and Vocabularies.
University of California Publications in Linguistics Volume 44~
Berkeley and Los Angeles:University of California Press.
Smith, N.V. 1967. 11 The Phonology of Nupe 11 • Journal of African Languages.
6:153-167.
Spears, R.A. 1967. 11 Tone in Menda 11 • Journal of African Languages.
6:231-244.
----------- 1968. 11 Tonal Dissimilation in Maninka 11 • Journal of
African Languages. 7:88-100.
Vail, L. 1972. "The Noun Classes of Ndali". Journal of African
Languages. 11:21-47.
Voorhoeve, J. 1971. Tonology of the Bamileke Noun".
11 Journal of
African Languages. 10:1-12.
Welmers, W.E. 1962. "The Phonology of Kpelle 11 • Journal of African
Languages, 1:69-93.
------------ 1973. African Language Structure. Berkeley and Los
Angeles:University of California Press.
Westermann, Dietrich and M.A. Bryan. 1952. Languages of West Africa,
Handbook of African Languages Part II. published for the Inter-
national African Institute by the· 8xford University Press.
-·------------------- and I.C. Ward. 1933. Practical Phonetics for
Students of African Languages. published for the International
African Institute by the Oxford University Press.

SIL-UND Workpapers 1980