Heine Bernd, Nurse Derek. A Linguistic Geography of Africa

Подождите немного. Документ загружается.

counterexample to the well-known generalization that the resumptive

pronoun strategy is very rarely used in prenominal relatives; in postnominal

relatives, African languages conform to a general tendency to use the gapping

strategy if the relativized role stands relatively high in the accessibility

hierarchy, the resumptive pronoun strategy being preferred if the relativized

role stands relatively low in this hierarchy.

Postnominal relatives are commonly (but not always) introduced by rela-

tivizers, and are often followed by determiners in languages that place deter-

miners at the right edge of noun phrases (which is a particularly common

phenomenon in West African languages). The following example from Kanuri

illustrates postnominal relatives immediately juxtaposed to their antecedent,

but followed by a deﬁnite marker. Note that the use of a resumptive pronoun in

the relativization of the dative, but not of the direct object, conforms to the

generalization mentioned above.

(46) a. ka

m [kasu

wu-lan ru

mma]-d@



person market-LOC see.TAM.S2SG-DEF

‘the person you saw at the market’

cf. shi



-ga kasu

wu-lan ru

mma ‘you saw him/her at the market’

b. ka

m [shi



-ro goro yi



na]-d@



person 3SG-DAT cola give.TAM.S1SG-DEF

‘the person to whom I gave cola’

cf. shi



-ro goro yi



na ‘I gave him/her cola’

Relativizers analyzable as pure subordination morphemes (which very often

are also used as complementizers, such as English that) are common in African

languages. True relative pronouns are very rare, if this notion is restricted to

words that stand at the edge of the relative clause but show variations in their form

and syntactic properties (e.g. combinability with adpositions), strongly support-

ing the hypothesis that they are pronouns fulﬁlling the relativized role but

occupying a non-canonical position. But a third type of relativizer is very com-

mon in Africa; relativizers of this type can conveniently be called relative linkers.

They are often confused with relative pronouns, because they show variations in

gender and number governed by the antecedent of the relative clause, and can

represent it; but nothing in their morphosyntactic properties can be viewed as

evidence that they are constituents of the relativized clause in a non-canonical

position, and very often (e.g. in Tswana, or in Arabic) they systematically

co-occur with resumptive pronouns. The following example is from Arabic:

(47) a. ar-raZulu llaDi: d

araba muammadan

DEF-man LINK:MSG hit.PERF:S3MSG Muhammad:ACC

‘the man who hit Muhammad’

Denis Creissels et al.142

b. ar-raZulu llaDi: qatalu:-hu

DEF-man LINK:MSG kill.PERF :S3MPL-O3MSG

‘the man they killed’ (lit. ‘the man that one they killed him’)

al-bintu

llati:Dahabtu ma¿a-ha:?ila:

DEF-girl LINK:FSG go.PERF:S1SG with-3FSG to

s-su:qi

DEF-market:GEN

‘the girl with whom I went to the market’ (lit. ‘the girl that one

I went with her to the market’)

Diachronically, there is often evidence that such relativize rs originate from

demonstratives in a sequence noun þ demonstrative þ relativized clause in

which the demonstrative has become obligatory and has changed into a mere

linker. In Tswana, the same linkers are obligatory with adjectival modiﬁers too

(see section 4.4.5).

4.14.4 Complementation

In a number of African languages, the clausal complement of verbs of saying,

perceiving, and thinking, and of verbs expressing desire, intention, command,

etc., is introduced by a complementizer that is quite transparently a gramma-

ticalized form of a verb ‘to say’ (or of a non-verbal predicator used in quotative

constructions), but this is a common typol ogical feature in other parts of the

world too.

There are languages with several types of complementizers. In Lele there is a

de dicto complementizer na, coding complements of verbs of saying and coding

indirect perception after verbs of perception, and a de re complementizer go,

coding direct perception after verbs of perception (Frajzyngier 2001).

In many Chadic languages, and also in Nigerian Arabic, the coding of the

subject of the embedded clause as object of the matrix clause has the function of

coding direct perception verbs, and realis wish with volitional verbs. This phe-

nomenon also occurs in English (Frajzyngier 1996; Frajzyngier & Shay 2003).

Among the SOV languages of Africa, clausal complements preceding the

verb (i.e. occupying the same position as nominal complements fulﬁlling a

similar function) are found for example in Kanuri, Amharic, Somali, and in most

Central Khoisan languages (although frequently in a reduced form), but clausal

complements may follow the verb even in languages that otherwise behave like

strict SOV languages, as illustrated by the following example from Ijo:

(48) a. eri



knI d



UU gbaamI

he one tale told

‘He told a tale’

Africa as a morphosyntactic area 143

b. er



ı gba a

m



 er



ı bo˛gimi

he said that he will.come

‘He said that he would come’

More generally, discrepancies between the position of nominal and clausal

complements of the same verbs are not uncommon. In some VS languages of

the Cha dic family (e.g. Hdi), the clausal complement of verbs of saying pre-

cedes the matrix clause, but the nominal complement of verbs of saying fol-

lows the matrix clause, whereas the clausal complement of verbs of perception

follows the matrix clause (Frajzyngier & Shay 2002).

However, this is not always the case: an interesting particularity of Somali is

to fully assimilate complement clauses in subject or object function to NPs: no

complementizer is present, and the last word of the complement clause (i.e. the

verb) is case-marked exactly in the same way as the last word of an NP

fulﬁlling the same function.

4.14.5 Switch-reference in complex constructions

and logophoricity

Comrie (1983) cites Gokana (an Ogoni language spoken in Niger ia) as an

exceptional case of an African language in which the morphology of dependent

verb forms marks the presence or absence of coreferentiality between the

subject of the dependent verb and the subject of the main clause. However,

Amha (2001) indicates that switch-reference marking on dependent verb forms

is widely attested in Omotic and in Cushitic languages. Mina (Central Chadic)

has a third-person pronoun whose function is to code contrastive focus on the

third person; another function of this pronoun is to code switch-reference, more

precisely, coding that the subject of the clause is different from the immedi-

ately preceding subject.

Another type of reference tracking, namely, “logophoricity,” is much more

common in African languages, in an area ranging from Senegal in the west to

Ethiopia in the east, and cutting across genetic boundaries. Logophoricity and

switch-reference are functionally similar, but logop horicity involves special

third-person pronouns occurring in dependent clauses only and expressing

coreferentiality with the subject of the main clause.

In the Chadic language Mupun, there are two sets of logophoric pronouns.

As illustrated by the following example, one set has the subject of the verb of

saying as its antecedent (sentences a and b with subject pronouns, and sen-

tences c and d with object pronouns), and the other set has the addressee of the

verb of saying as its antecedent (sentences e-f). This is perhaps the richest

logophoric system attested (Frajzyngier 1993).

Denis Creissels et al.144

(49) a. wu/wa/mo sat n@ wu/wa/mo ta ee n-jos

he/she/they say CMPL he/she/they stop stay PRE-Jos

‘He

/she

/they

said that he

/she

/they

stopped over in Jos’

b. wu/wa/mo sat n@ ta I/e/u ee n-jos

he/she/they say CMPL stop he/she/they stay PRE-Jos

‘He

/she

/they

said that he

/she

/they

stopped over in Jos’

c. wu/wa sat n@ n-nas wur/war

he/she say CMPL 1SG-beat him/her

/she

said that I beat him

/her

’

d. wu/wa sat n@ n-nas in/e

he/she say CMPL 1SG-beat him/her

‘He

/she

said that I beat him

/her

’

e. n-sat n-wur n@ wur ji

1SG-say PRE-3SG CMPL 3SG come

‘I told him

that he

should come’

f. n-sat n-wur n@gwar ji

1SG-say PRE-3SG CMPL 3SG come

‘I told him

that he

should come’

See Gu

ldemann and von Roncador (2002) for a recent survey.

A system of anti-logophoricity marking has been reported by Kouadio

(1996: 384–9) for the Kwa language Attie, and by Andersen (1999) for the

Western Nilotic language Mabaan. In these languages, in indirect speech,

third-person pronominal forms identical to those used in independent clauses

indicate coreferentiality with the subj ect of the main clause, whereas other

third-person referents are represented by special pronouns (“fourth-person

forms”), as illustrated by the following example from Mabaan:

(50) a. ?



k



 g



Oke

ge¯ ?



k



 ka

ﬁJ





3SG say:AP:3 INIT.3SG swim:FUT:INDIR:3SG

‘He

says that he

will swim’

b. ?



k



 g



Oke

ge¯ ?

kta

ﬁJ





3SG say:AP:3 INIT.4SG swim:FUT:INDIR:4

‘He

says that he

will swim’

4.15 The grammatical coding of spatial relations

Most languages use locative case markers or adpositions to encode basic

spatial relations (in/on/at N), but in many Niger-Congo languages, among

those that have the noun class system characteristic of this phylum in its most

typical form, this function is fulﬁlled by noun class markers: class afﬁxes

inherently associated with the noun ‘place’ may also be added to the inherent

Africa as a morphosyntactic area 145

preﬁx of other nouns; nouns combined with locative class markers fulﬁll

the function of locative argument or adjunct without necessitating the

addition of any case marker or adposit ion, but, at l east in the languages

that have this mechanism in its most typical form, thelocativemor-

pheme clearly remains a noun class marker, since non-locative nouns

combined with it lose their inherent concordial properties and behave in

concord rules like the noun ‘place.’ This is consonant with the fact that

in many Bantu languages, locative constructions may function as syn-

tactic subjects or objects of verbal predications involving motion verbs,

postural verbs, or verbs of existence and availability. This strategy as

well as other strategies whereby locative constructions are treated as core,

rather than peripheral, constituents, are further discussed in Dimmendaal

(2003).

Another remarkable feature of the grammatical coding of spatial relations

in Niger-Congo languages is that, as a rule, the locative adpositions or afﬁxes

do not code the distinction between loca tion at/movement towards/movement

from, which implies the existence of two distinct classes of motion verbs

assigning the role of goal or source to a locative complement. In addition to

that, the set of motion verbs that can assign the role of source or direction

to their complement is often very limited, hence the frequent use of verb

sequences in order to code the starting point and the direction of a motion

(for example, ‘run from A to B’ is often rendered literally as leave A run reach

B), even in languages that otherwise show no marked tendency towards

serialization.

Similarly, in some Chadic languages, directionality away from a source or

towards a goal is coded through serial verb constructions, where each verb

codes just one parameter, such as ‘go,’ ‘leave,’ ‘arrive,’ ‘pass’; in some other

languages, the verbs that once were part of a serial verb construction have

become extensions to the main verb.

In languages that have an applicative derivation, a possible use of the

applicative (observed in particular in Bantu languages) is to derive verbs

assigning the role of goal to a locative complement from verbs that, in their

non-derived form, either don’t have any locative argument, or have a locative

argument to which they assign the role of sourc e, as in the following example

from Tswana:

(51) a. k



I-t¸a

-hu

v-a





S1SG-FUT-move-FUT PRE Kanye

‘I am going to move from Kanye’

b. k



I-t¸a

-hu



la



Uro



S1SG-FUT-move-APPL-FUT PRE Gaborone

‘I am going to move to Gaborone’

Denis Creissels et al.146

c. k



I-t¸a

-hu

v-a



 k



I-hu

v-e



I ko

S1SG-FUT-move-FUT PRE Kanye S1SG-move-

APPL-SEQ

PRE



Uro



Gaborone

‘I am going to move from Kanye to Gaborone’

The chadic language Mina has a “locative predicator” used with locative

predications when the predicate is inherently non-locative. A locative

preposition is used only when the com plement is inherently non-locative. If

both the predicate and the complement are inherently locative, neither the

predicator nor the preposition is used. In the following example, the locative

predicator a

must be used in sentence (a) because the main verb ‘called’ is not

inherently locative; in sentence (b), tsu

‘went’ is inhe rently locative, and the

locative predication does not require the locative predicator:

(52) a. nd-a

ngu

lng@



go-GO call husband 3SG LOCPRED room

‘And [she] called her husband into the room’

b. se

ym@



ngu

lti



y-u

so REL husband see 3SG see-3SG wife

tsu



went EE bush

‘So the husband saw that the wife went into the bush’

The tendency to code the distinction location at / movement towards /

movement from exclusively on verbs is typical of languages spoken in a large

area including the whole of Niger-Congo languages, but not of African lan-

guages in general. In particular, if more or less grammaticalized uses of a verb

‘leave’ are not taken into account, ablative case afﬁxes or adpositions seem

to be totally absent from Niger-Congo languages; by contrast, ablative

case afﬁxes or adpositions are common in several branches of Afroas iatic and

Nilo-Saharan.

As regards the grammatical coding of spatial r elations, another very

common phenomenon in African languages (not restricted to particular

families or areas) is that place names in locative argument or adjunct function

need not combine with the locative case afﬁxes or adpositions that are

obligatory for common nouns fulﬁlling the same function. The case of the

Chadic language Mina has been mentioned above. In the Chadic language

Hdi, the locative preposition has a different tone when occurring with

a toponym than when occurring with an inherently non-locative noun

(Frajzyngier & Shay 2002).

Africa as a morphosyntactic area 147

Directional morphemes (deictic or non-deictic) attached to the verb are not

very common in African languages in general, but they are common in Nilo-

Saharan languages (in particular venitive and andative) and are sporadically

attested elsewhere (for example, Wolof has venitive and andative verb forms).

An interesting particularity concerning the grammatical coding of spatial

relations in t he Omotic as well as the Cushitic branch of Afroasiatic, is

that locati ve complements of verbs of motion are frequently treated on a par

with com plements of prototypical transitive verbs such as ‘eat’ or ‘beat,’ i.e.

as direct objects. Thus in Maale (a language with a case contrast of the

“marked-nominative” type), the complement of the verb ‘go’ (accordingly

best translated as ‘reach (somewhere)’) takes absolutive case:

(53) a. ?i



hell-e

-ne

3MSG:SUBJ reach-PER F-POS:DECL

‘He arrived’

b. ?Ia

zIa

gurd-o

hell-e

-ne

3MSG:SUBJ village-ABS reach-PERF-POS:DECL

‘He arrived in the village/He reached the village’

This phenomenon is spora dically found in other languages too, for example

in the Southern Bantu language Tswana.

4.16 Conclusion

As already observed in the introduction, our conclusions are limited by the fact

that they are not backed by appropriate quantitative data, for example, in the

form of a survey that would ensure a ﬁne-grained picture of the linguistic

diversity of the African continent. Moreover, in order to establish in an

uncontroversial way to what extent Africa can be considered as a morpho-

syntactic area really different from the rest of the world, similar data would be

needed for the other parts of the world too.

As the preceding sections have shown, Africa’s internal coherence cannot be

established on the basis of features common to all African languages: the only

morphosyntactic features found in all African languages are presumably uni-

versals, or quasi-universals (see chapter 2). It is however possible to list a set of

features that certainly would deserve thorough investigation in further studies

of the morphosyntactic diversity of African languages com pared to the

diversity observed in other parts of the world. What we propose below is a

tentative list of morphosyntactic features that seem to concern a proportion of

African languages sign iﬁcantly different (higher or lower) from that observed

at world level.

Note however that, as mentioned in the relevant sections, most of these

features are not evenly distributed on the African continent. In some cases,

Denis Creissels et al.148

their distribution may cast doubt on their relevance from the point of view of

areal typology, and suggests that the genetic factor may have been crucial, and

that contact may have played only a limited role in their diffusion. In particular,

several features listed below (for example, the relative rarity of case-marked

subjects and objects, or the frequency of double object const ructions), do not

really characterize Africa as a whole, but rather an area including mostly

Niger-Congo languages, plus possibly languages belonging to groups in

relatively close contact with Niger-Congo languages (i.e. Songhay, Chadic,

and Central Sudanic).

With all these caveats, leaving aside features that concern the major ity of

African languages but seem to be equally well attested in most other parts of

the world, or features that concern a minority of African languages but seem to

be equally rare in most other parts of the world, we would like to propose the

following as the features most likely to be relevant, at the present state of

documentation, for a characterization of the African continent as a morpho-

syntactic area:

(a) The ergative type of core syntactic role coding is exceptional among

African languages (section 4.2).

(b) Case-marked subjects or objects are less common among African lan-

guages than at world level (section 4.2.2).

jects and objects is exceptional in other parts of the world but very

common among African languages that have a case contrast between

subjects and objects (section 4.2.2; see also chapter 8).

(d) Obligatory agreement of transitive verbs with their object does not seem

to be attested among African languages (section 4.2.3.2).

(e) Second-position clitics are relatively common in the languages of the

world, but exceptional among African languages (section 4.2.3.3).

(f) In a relatively high proportion of African languages, the construction of

verbs with an argumen t frame of the type giver – given – recipient tends

to assimilate the recipient (rather than the thing given) to the patient of

prototypical transitive verbs, and double object constructions are par-

ticularly frequent (section 4.2.5).

(g) Focus strategies implying morphosyntactic alterations, and in particular

focus marking by means of verbal inﬂection, are particularly common in

Africa (sections 4.3.2 and 4.13).

(h) The use of special verb forms in seque ntial constructions is particularly

widespread among African languages (sections 4.3.4 and 4.14.2).

(i) Applicatives are particularly common in Africa, and a relatively high

proportion of African languages make a wide use of obligatory applica-

tives and of various types of non-canonical applicatives (section 4.3.5).

Africa as a morphosyntactic area 149

(j) Classiﬁer systems are exceptional among African languages (section

4.4.1).

(k) Relatively few African languages are devoid of morphological plural, or

have a morphological plural restricted to a subset of nouns occupying a

high position in the animacy hierarchy (section 4.4.3).

(l) African languages that do not use the same morpheme as a noun phrase

coordinator and as a comitative adposition are relatively rare (section

4.4.8).

(m) The proportion of languages with a syntactically ﬂexible constituent

order is much lower among African languages than at world level

(section 4.8).

(n) The constituent order SOVX, relatively rare at world level, is relatively

frequent among African languages (section 4.8).

(o) Clause-ﬁnal negative particles occur amo ng African languages much

more frequently than in other parts of the world (section 4.12.1).

(p) Changes in the constituent order triggered by negation are particularly

common among African languages (section 4.12.1).

(q) True relative pronouns are particularly rare in African languages (most

words currently identiﬁed as relative pronouns are relative linkers), and

the use of dependent verb forms in postnominal relatives, relatively rare

in the languages of the wor ld, is common among African languages

(section 4.14.3).

(r) Logophoricity is particularly widespread among African languages

(section 4.14.5).

(s) Systems of coding of spatial relations in which the distinction location

at / movement toward / movement from manifests itself exclusively on

verbs are more frequent in Africa than in most other parts of the world

(section 4.15).

Denis Creissels et al.150

The Macro-Sudan belt: towards identifying a

linguistic area in northern sub-Saharan Africa

Tom Gu

ldemann

5.1 Introduction

It has been recognized for a long time that languages across a broad

sub-Saharan belt from the western end of the continent to the escarpment of the

Ethiopian Plateau in the east display certain linguistic afﬁnities. At the same

time, it has been difﬁcult to identify precisely the nature and range of these

afﬁnities and to provide a plausible explanation for them.

I propose in Gu

ldemann (2003a) that the distribution of logophoric marking

in Africa follows an areal pattern in that it is regularly found in languages of the

sub-Saharan belt referred to above, but is virtually absent from the rest of the

continent. This ﬁnding is the starting point for a more systematic investigation

of the following questions:

(a) Does this geographical region share other linguistic traits?

(b) If so, do these deﬁne some sort of linguistic area?

I present in section 5.2 several linguistic features which appear to share a

roughly similar distribution across the African continent, as well as additional

candidate features which may support the evidence provided in this chapter.

Section 5.3 brieﬂy surveys previous approaches to the general observation of

linguistic commonalities across the sub-Saharan belt, among which there is the

proposal that most of the languages involved belong to a genealogical lineage

comprising Greenberg’s super-groups Niger-Kordofanian and Nilo-Saharan.

On the basis of the data give n here, section 5.4 argues against such an

explanation and, in so doing, tries to make a case for an areal language group

which I propose to call the ‘‘Macro-Sudan belt.’’ Section 5.5 discusses a few

preliminary points on the possible historical emergence of this linguistic area

in the center of Africa. The ﬁnal section 5.6 addresses the potential relevance

of the hypothesis for areal-typological research in Africa as a whole.

To begin with, I will characterize the area at issue in more detail and outline

the language groups involved. As remarked already, the area is a broad belt south

of the Sahara and can be fairly well deﬁned in purely geographical terms: it is, so

151