International Symposium on Linguistic Diversity and Language

Theories
Boulder (Colorado) May 14-17, 2003

Language processes, theory and description of language change,

and building on the past: lessons from Songhay

Robert Nicolaï
Abstract
This paper is a continuation of earlier work on the controversial genetic classification
of Songhay as Nilo-Saharan (Nicolaï 2003), where I show that the results thus far (cf.
Bender 1995, Ehret 2001) are unsatisfactory. This is attributable to the authors'
models and methods, the assumptions these rest upon, the nature of the data, and a
priori factors in research.
I discuss theoretical hypotheses regarding the ways in which languages change, the
methods and procedures used by scholars, and the ways in which empirical data are
used. This leads me to ask the following questions at the outset:
1) Are we entitled to make explicit use of virtual or real
multilingual/multidialectal factors as normal theoretical parameters in
building models of language change and proposing long-distance
relationships? (This would imply changes to the classical tree
representations and facilitate the incorporation of areal and contact
phenomena.)
2) Would it be helpful to postulate an anthropologically defined setting for the
interaction of structural linguistic forms, basic cognitive processes, and
punctual input from historical contingencies, resulting in rearrangements of
norms of use and formal structures? (This question is vital to the study of
areal phenomena and situations lying beyond the theoretical constraints of
the standard genetic model.)
These questions are in line with others raised during the Symposium with regard to
the relations between theories of language structure and theories of cognition, the
degree of conscious motivation for processes of language change and
grammaticalization, and adabtability in language change and functionalization.
In answering them, I make a number of suggestions based on my overall views on the
practice of comparative linguistics, the construction of theoretical frameworks, and
my recent empirical results in Nilo-Saharan and Afroasiatic.

1
Il faut résister à un positivisme de premier
examen. Si l’on manque à cette prudence, on
risque de prendre une dégénérescence pour une
essence (Bachelard 1934:159).
Prima facie positivism must be avoided.
Carelessness here could lead us to take a
degenerative form for an essential one.

Lessons from Songhay

It is always instructive to see methods of proven efficiency encounter
serious difficulties in a particular instance. This awakens us to the
possibility of widening our viewpoint, rearranging our theoretical and
methodological apparatus to deal with the rebellious data, and defining
previously unrecognized problems. A striking example of this in the
field of problematic genetic classification has recently been provided
by the Songhay group in Africa, composed of languages, most of
which are trade languages with no written tradition, spoken mainly in
the region of the Niger bend and displaying little internal
differentiation beyond a division into northern and southern subgroups.
Let us examine this case more closely.
I will divide my discussion into three parts. The first is empirical, the
second looks to the future, and the third draws conclusions. To provide
a concrete illustration, I will begin by summarizing the latest ideas on
the genetic classification of Songhay which will show both the
inadequacy of tree diagrams for representing language change and the
fundamental importance of language contact in the origin and
development of this particular language (for further details, see Nicolaï
2003). I will also account for the participation of Songhay in what
seems to be an area of convergence with the Mande languages.
In view of these results, I will suggest changes to our analytic al
framework which will give a central role to language contact
phenomena and sociolinguistic hypotheses while in no way
diminishing the importance of the more traditional approach. I will try
to show that the resulting change in factor ranking will be crucial to the
understanding of processes of language change.
Finally, I will make a few remarks on the possible impact of
modelization in the field of language change.
Apparent genetic relationship.
The inclusion of Songhay in the Nilo-Saharan family, first suggested
by Greenberg (1964), was subsequently impugned (Lacroix 1969,
Nicolaï 1990), and then ostensibly confirmed by studies of Nilo-
Saharan as a whole (Bender 1995, Ehret 2001), though these two
authors disagree regarding its position on the family tree. The latest

2
study1 (Nicolaï 2003) provides a detailed critique of the works which
attach Songhay to Nilo-Saharan, and attempts to show that the models
(tree model of language diversification) and methods (reconstructions,
phonetic correspondences; isoglosses) they use, their basic theoretical
assumptions (linear development), the nature of the data (no written
tradition), and a priori factors in research2 all contribute to a mistaken
conclusion in this sense. Consequently, it is no longer possible to
maintain that Songhay belongs to Nilo-Saharan, if ‘belonging’ is taken
in its usual genetic sense and ‘Nilo-Saharan’ is meant to be a family
(or a family- like phylum) of genetically related languages capable of
being represented by a tree diagram.
My own analysis, founded on the entirety of the available Songhay
dialectological data, has led me to seek hypotheses which might
explain the absence of clear morphological correspondences with other
Nilo-Saharan languages and the high proportion of likely Afroasiatic
lexical items 3 which are neither obvious recent loans from Arabic (e.g.,
àlbésèl bs‚l onion, àlkámà qmh‚ wheat, etc. or the religious vocabulary
of Islam) nor the result of other bounded contacts (e.g., àddà machete
from Hausa, tukamaaren cheese from Tuareg, etc.) but part of basic
vocabulary; and not isolated units but sets covering complex lexical
domains down to their fine structure (cf. Table 1). Songhay is thereby
reoriented towards Afroasiatic, which is not to say that it stands in any
genetic relationship with this family.
Table 1: Lexical designations 4 of body parts.
head bòÑ [Kbl5: abbaà head; goiter bókò, soft spot below the lower jaw
cranium] bokolo [Kbl: ffeqlej be flabby, fat, soft]
palate dáanà (daÃna) [Kbl: aneà / gums díinì [Hgr: ta–yne gums; Tmz:
ineÃ; Amh: sŒnag, tŒnag, lanqa taniwt gums ]
palate]
hair/feather hámní; himbiri [Ar: h‚abl nerve, tendon linji Ar: ‘irq root; Tmz:
string; a‘bal thick, tightly-woven rope; lÆer nerve, tendon, vein, artery]
Hgr: éhafiilen long body hair; téhafilt
short body hair; Wlm: abŒndal hairy

1
Bibliographical recapitulation: 1) Songhay an isolated unit (Westermann 1927),
typologically close to Mande (Delafosse), 2) Songhay a member of Nilo-Saharan
(Greenberg 1964), 3) hypothesis impugned by Lacroix (1969), 4) Songhay a Tuareg-
Mande creole (Nicolaï 1990), 5) return to Nilo-Saharan (Bender 1995, Ehret 2001),
6) Songhay derived from a Afroasiatic lingua franca (Nicolaï 2003).
2
I distinguish ‘assumptions’ relating to the model (thus, the tree model of language
diversification assumes a linear development) from ‘a priori factors in research’
relating to the sociological and at times emblematic features of scholarly activity
(thus, choosing Bachelard rather than Feyerabend or Lakatos as one's epistemological
reference necessarily categorizes the proponent).
3
Not previously identified owing precisely to a priori factors in earlier studies. There
is no space here to present the data and analyses which support this affirmation.
Nicolaï (2003) provides a detailed discussion. See Nicolaï (in press a) for a summary
version. A few examples are nevertheless appended.
4
These are only a few examples. Nicolaï (2003:296-306 and 384-497) contains an
extensive list of useful items. Even that list, however, is neither exhaustive nor
definitive.
5
Kbl: Kabyl; Amh: Amharic; Hgr: Tahaggart; Tmz: Tamazight; Ar: Arabic; Wlm:
tawellemmet; Gz: Gueze; Tms: Tamajaq.

3
man]
fontanelle lòÑgò [Hgr: élenõeou large mouth mê [Hgr: émi mouth, Tmz, Wlm:
nape of neck (derisive); Wlm: tallŒka imi mouth, entry, orifice], eye mòy,mò
sinciput // fontanelle] [mghb: m-mm-w iris of eye, Hgr:
emmah pupil of eye; Kbl, Tmz: mummu
pupil, iris of eye]
face mòydúmà [Hgr: ôudem face; cheekbone/smile múmúsú [Tmz:
Kbl, Tmz: udem face] smummey smile, pout; Wlm: ƒŒmmŒƒmŒƒ
smile]
sweat súÑgéy [Hgr: enõi trickle; Kbl: sneeze tísôw [Hgr: tôusou cough
ssengi cause to flow; Gz: ’ngy, sngy, regularly; Kbl: tusut whooping cough;
sng, sgd(d ), gy melt, flow, sweat] ent‚ez sneeze; Gz: ‘at‚asa sneeze]
have diarrhea sóorú [Ar: isha¯l tear(s) múndì [Ar: dama‘a tear(s); Hgr:
diarrhea; ‘as‚ara press; Kbl: esrem a–mit tear(s); Kbl: imet‚t‚i tear(s); Tmz:
cause diarrhea; Tmz: nmarsi diarrhea; amet‚t tear(s)]
Tms: zarrat diarrhea; Gz: ‘as‚ara
press out, press, squeeze, wring out]
urinate tòosì [Hgr: a–se´„as bladder; spit túfà [Ar: taffa (tff) spit; Hgr: soutef
Tms: tasŒyast bladder] spit; Kbl: t‚te‚ ft‚ef foam with wrath; Gz:
taf’a spit, spit out]
defecate wá [Gz: ‘Œba¯ dung] drool yólló [Hgr: a–lidda drool (n); Kbl:
aledda drool (n); arch: rayyal drool,
foam, salivate]
vulva bùtè [Ar: bud‚‘ vulva] chin danka [Ar: d„aqan chin; d„aqn beard,
whiskers]
breast fòfè [Ar: ‘ubb breast, gusset; arm/hand kàbè [Ar: k‘b ankle, heel;
Hgr: éfef breast, teat; Kbl: iff teat] kaff, kaffah palm of hand]
lung kùfú [Hgr: ekef (be) inflated; liver tásà [Hgr: te´s„ a belly (of person or
Kbl: ik‚uftan foam] animal); Kbl, Tmz: tasa liver]
Etc.

Stratification of structural isomorphism and isoglosses.

At the same time, the typological structure of Songhay shows marked
morphosyntactic affinity to that of the Northern Mande languages. The
examples in Table 2 below (taken from Mandinga and Zarma,
important representatives of West Mande and Songhay, respectively)
show how extensive the isomorphism is.
Table 2: Examples of Mande-Songhay structural isomorphisms.
Derivation Word formation:
abstract quality: -ya -ta˜ra•y derivation and
mo˜ko >> mo˜koya• bo˜ro• >> bo˜rta˜ra•y 'huma compounding:
'humanism' nism' Overall, both groups use

4
Compounding comparable derivational
jo˜li 'blood', si•la 'road' ku•ri• 'blood', foóndo˜ 'ro processes and have highly
jo˜lisi•la 'vein' ad' productive compounding
ku•ri•fo•ndo˜ 'vein' according to identical
Reduplication formal patterns;
hu˜la• 'two', si˜di 'tie' i˜hi•nka• 'two', ha•w 'tie' reduplication is also very
hu˜lahulasi•di 'tie two by t ha•w i˜hi•nkahi•nka• 'tie t productive.
wo' wo by two'
Genitive construction7 Noun modification:
(mdk) de˜ndikoo jifoo ba˜nka˜araóa zi•iba˜a Noun-modification
the garment + the pocket the garment + the pocke structures are often
'the pocket of the garmen t parallel, including the use
t' 'the pocket of the gar of two reversed orders:
ment' {Modifying Noun +
Adjectival modifier Modified Noun},
mo•ngo•n kE•rEn nu˜ '(the) gre ma•ngu ˜ bo•ogo•o 'the g {Modified Noun +
en mangos' reen mango' ; ma•ngu Modifying Adjective}.
bo•ogu 'a green mang
o'
Transitive proposition Predicative propositions:
se•ku˜ di• mi˜si• sa˜n 'Seku bo da•wda˜ na˜ ha•wo• da•y 'D There are clear similarities
ught the cow' awda bought the cow' in the structure of
predicative propositions {S
se•ku ma•n mi˜si• sa˜n 'Seku da•wda˜ ma˜n ha•wo• da•y Aux O V Cpl} and
did not buy the cow' 'Dawda did not buy th numerous affinities in the
e cow' TAM system and the
negative conjugation.
Grammaticalization of lexical items:
In both Mande and Songhay, the heads of adjunct phrases are noun suffixes. Some of
these derive from still extant lexical items (this phenomenon is of course far more
widespread).
Semantic structure:
Subject to further information, these affinities in the structure of semantic fields and
categorization would seem to be shared with languages across all of West Africa,
rather than being characteristic of the Songhay-Mande alone
Phonological system9 : West Mande Songhay
The table opposite lists primacy of disyllabic lexemes
a few indicative seven-vowel systems (5 in 5- or 7-vowel systems
similarities, to which western Mandinga and
may be added structural Soninke, 6 in Kita
features such as the Maninka)

7
Juxtaposition expresses "inalienable possession" in Mandinga. This is the structure
chosen here. It corresponds exactly to Songhay; the presence of the connective in
"inalienable possession" does not affect the parallel in constituent order. The four
examples marked "mdk" are from Mandinka, cf. Creissels (2001). The choice of
Mandinka rather than Malinke does not mean that the structures illustrated do not
exist in Malinke; rather that the data available to me for the latter do not contain a
suitable example.
9
This subtable summarizes Vydrine's (2000) conclusions, which stresses the
difference between West Mande and Mani-Bandama.

5
absence of /p/ and the relatively large medial full consonant inventory in
absence of /r/ in initial consonant inventories, medial position
position. particularly in the north
length contrast in non-final length contrast in all
position in many languages positions
(in all positions in
Mandinka)
2 tones; transition to accent 2 level tones + rising and
systems in Mandinka, falling contours analyzable as
Kagoro, and some Jallonke a succession of level tones;
dialects; 3-tone systems transition to accent systems
reducing to 2 in Kpelle (northern Songhay); loss of
prosodic contrasts en eastern
and western Songhay

But once overall isomorphism is recognized, its linguistic and

geographical stratification must then be examined in full detail. Such
examination has led us to observe that the structural processes
operating in the subsystems of the language have been differently
conditioned by the historical factors in play:
- phonological change seems more sensitive to the effects of contact
than morphosyntactic change,
- morphosyntactic change seems more sensitive to functional
phenomena of simplification deriving from use as a trade language.
At the same time, the contexts of language use (which may change
over time) give rise to processes which can modify prior
configurations. Thus, we find continual Songhay-Mande contact
bringing about an intensification of apparent convergence on some
points through the creation of a Sprachbund. This is evident from a
detailed study of the geographical and linguistic stratification of the
shared features (e.g., the geographically bounded merger of /s/ and /z/
in western Songhay in direct contact with Mande languages).
Elsewhere, however, we find that convergence has been obscured by
subsequent change resulting from the use of the languages in contexts
requiring simplification (e.g., the loss of SOV word order in western
and northern Songhay). For all these reasons, the phenomena of
linguistic stratification themselves require particular attention;
otherwise, should they be ignored, the resulting input will necessarily
lead to mistaken interpretations (see Nicolaï, in press b, on these
points).

In conclusion, we may assume that, barring future evidence to the

contrary, Songhay arose in a complex way from a lingua franca, an
ancient sort of trade language, whose precise nature (Berber, earlier or
later form of Semitic, Ethiosemitic, or other) remains to be established.
This language, which was not necessarily homogeneous and probably

6
of a simplified nature, took on stable form when it was appropriated by
populations which originally spoke neither Semitic nor Berber
languages. Two hypotheses in this respect are possible a priori, neither
of which can be excluded at this stage:
- a no longer existent lingua franca strongly impacted on and
widely relexified another regional language, giving rise to
Songhay, or
- that lingua franca "was" (in a sense yet to be defined) what has
now become Songhay; in such case, Songhay was simply the
result of the nativization11 of this language.

Inferences
The language formation hypothesis illustrated by the Songhay data
is concordant with the generally recognized criteria for the
development of stabilized pidgin languages, and furthermore accounts
for the impossibility (or perhaps merely the difficulty) of establishing
strict phonetic correspondences despite the kinship in basic
vocabulary.
At the same time, Songhay's continued status as a trade language,
the anthropological diversity of its speakers, and the correspondence of
the current situation to what we know of the medieval African world
are all historical features which fit well with this hypothesis. It thereby
becomes easier to understand the following three things:
1. The diversity of the Afroasiatic sources to which the Songhay
lexicon can be related.
We would expect that a putative lingua franca with an Afroasiatic base
spreading over white and black Africa would have taken over the
lexical material required for its use from a number of languages, from
Cushitic to Egyptian and Arabic to Berber. What we know of the
Mediterranean lingua franca suggests variation of lexical sources
(Venitian, Genovan, Provençal, etc.) over time showing that change in
a comparable situation can be fairly fast.
2. The extent of lexical diffusion of the Afroasiatic vocabulary
shared by Songhay with neighboring African languages.
If there really was such a lingua franca as I am suggesting, it is to be
expected that many of its lexical items would have been incorporated
into the neighboring Northwestern Mande languages, Wolof, the
Saharan languages, and many others including the "truly" (?) Nilo-
Saharan ones. A process of this kind would explain the amount of
shared "ancient" lexical items revealed by a study of lexical diffusion
over the entire West African region. It would also account for the
similarities with Mande or Chadic that have been noticed (cf. Creissels
1981, Mukarovsky 1989, Nicolaï 1977, 1984, Zima 1988).

11
Let us speak of ‘nativization’ of a language as we might speak of ‘ethnicization’ of
a culture. Ethnic groups, like languages, are not necessarily formed by genetic
descent; they may quite well come into being without deriving from some other one,
and then only later provide themselves with a history.

7
3. The morphologization of the pidginized language into the
prevalent typological framework of the Mande area vs. the
relexification of Mande12 .
This is the phenomenon reflected in Songhay-Mande
isomorphism, the extent of which is apparent from illustrations given
above. Indeed, Songhay morphosyntax is relatively simple
(economical in Houis's 1971 terminology), typologically similar to
Mande, and classifiable as type B2 (as defined by Heine 1975) with
respect to syntactic ordering.

Genealogical hypotheses and apparent areal convergence.

Clearly, once Songhay is assumed to be a Nilo-Saharan
language, it follows that Mande and Songhay belong to two distinct
genealogical units. Thereupon, any isomorphism can also be logically
interpreted as the result of convergence between two groups of
languages in close contact. But if the Nilo-Saharan affiliation of
Songhay is incorrect, then the present form of the Songhay language is
conceivably (or better, in all likekihood) not an outcome of a process
of linguistic convergence; though long-term contact may reinforce the
apparent convergence and end up superimposing a Sprachbund
situation on the original one. It is impossible, on the basis of either of
the two hypothetical modes of the creation of Songhay suggested
above, to interpret this overall isomorphism with Mande as simply a
phenomenon of language convergence in the sense of the archetypal
phenomena observed, for example, in the Balkans. This impossibility,
as implied by my modified conclusions, is methodologically
instructive insofar as it illustrates the degree of interdependence among
rival explanations and the effect any ill- founded empirical hypotheses
will have on the construction and defence of an overall explanatory
system; cf. Nicolai (in press b) for further discussion of the
implications of Songhay-Mande isomorphism.
In sum, we are led to the conclusion that convergence is only apparent
in the Songhay-Mande case. Rather than highly unlikely ‘generalized
convergence’, we may assume that a ‘new variety’ of language
displaying the major typological features of one preexistent group of
languages and containing much of the lexical stock of another must
have appeared in a specific sociological contact situation (or sequence
of situations). It can be shown that only later did some ‘classical’
convergence phenomena, particularly of a phonological nature, affect
the resulting languages on a limited scale.
The lesson to be learned is that not all observed isomorphisms can be
attributed to what is generally known as a Sprachbund or area of
convergence. Indeed,
- the identification and interpretation of what might seem a
priori to be an area of convergence is intrinsically linked to the
cultural and anthropological setting in which the languages
involved arise. Comparable systemic features and a common

12
Cf. Creissels (1981), Lacroix (1969), Nicolaï (1977, 1984, 1990) for remarks on
Songhay grammatical morphemes akin to Mande. For a broader view, see Nicolaï (in
press b).

8
geographical location are thus insufficient evidence on which to
base a conclusion regarding the kind of process whose outcome
is the current situation, particularly when the historical facts are
poorly documented;
- isomorphisms which cannot be explained by genetic
relationship may just as well result from processes of language
creation as from processes of modification of preexistent
languages.
An inventory of the kinds of contact situation which give rise to
isomorphisms should therefore be established. This involves taking
into consideration not just the analytical operations conducted by
speakers, but also those of the descriptive linguist himself. While it
may be absurd or simply wishful thinking to try to establish a one-to-
one relationship between types of language development and types of
contact situation or anthropological setting, such an attempt provides
the groundwork for shaping hypotheses and in any case is essential to
the descriptive process.
The consideration of these facts leads to two more conclusions of a
more general nature:
- Whenever a model is improperly imposed on recalcitrant data,
there is a danger that conceptual arm-twisting will give rise to
fallacious representations (see Nicolaï 2003). Forcing Songhay
into the Nilo-Saharan framework provides a good example of
this.
- Whenever the shape of a phenomenon (e.g., an area of
convergence as defined by the set of isomorphisms found there)
is established by linking concurrent factors and underlying
processes without regard for any theoretical framework or prior
analysis, there is a danger of a semantic cover-up, as when
Songhay-Mande isomorphism is characterized as a
convergence phenomenon (see Nicolaï, in press b).
In the light of this case study, I should like to consider the frameworks
which can be helpful in apprehending phenomena of language change.
Their use as points of reference and the way they are linked to wider
typological questions should, I believe, be seen in the light of two other
issues, the role of multiple codes 13 and the anthropological context of
their use, though I am ha rdly able to provide definitive judgments on
either of these points. My questions are:
- Should real or virtual multilingualism/multidialectalism not be
one of the normal parameters of theoretical models of language
change and long-distance relationship? If so, the shape of
traditional tree diagrams is liable to change, and it becomes
easier to integrate the effects of language contact and areal
phenomena.
- Should an "area" not be defined on the basis of anthropological
criteria within which language structures, elementary cognitive
processes, and matters of historical contingency all come into

13
I use the term "code" here in the widest sense to refer to any formal feature shared
under an explicit or implicit convention which allows a meaningful distinction to be
made in a communicational exchange.

9
play in the recomposition of norms and codes? This would be
vital to the study of areas of convergence and situations which
do not fit well within the standard genetic model.
Any attempt to answer these questions should lead to a clearer view of
the entire problem of language change in situations where the
appearance of languages is inherently linked to contact phenomena and
an anthropological dimension. I shall thus try to establish a connection
between two deceptively complex intuitive notions which can be
subsumed under the terms ‘contact’ and ‘genetic origin’. In trying to
do this, I shall keep in view what seems to me to be the essential nature
of language itself: its social dimension and the inherent heterogeneity
in the way humans exercise the cognitive capacity to restructure and
rationalize what they construct as a language.

The invariables of language change

Linguistic situations of the Songhay type have shown how great the
need is for analytical models of language change which give suitable
priority to considerations regarding language contact and social
behavioral contexts in processes of linguistic communication. Such
considerations have, of course, never been totally ignored; yet they
have often been treated as epiphenomenal on the basis of ordinary
models of individual language structure within a given theoretical
framework. Consequently, contact phenomena have been described as
(unnecessarily) complicating a simple situation, rather than as part of a
basically complex initial state of affairs, the very frame of study. Must
we accept that linguistic processes can be correctly apprehended only
on the basis of the theoretical a priori hypothesis that the right way to
start is by postulating a homogeneous structural system? Or would we
not be wiser to avoid such reductionism and try to set out from an
initial postulate of complexity?
This choice of initial postulate is fundamental insofar as it must affect
the framework for the explanation of the observed phenomena, as we
shall see below from the interconnection of the four themes which I
shall develop within the framework of an avowedly multilingual
approach. It must nevertheless be said explicitly that this approach is
tentative and does not aspire to be strictly theoretical. A priori
strictness is precisely one of the defects I wish to reject, along with the
absence of any theoretical framework whatsoever.

The multilingual approach

1) Multilingualism and/or multidialectalism 14 are both fundamental to,
and commonplace in language in general: this is the canonical
situation for the description of linguistic communication and the
analysis of the processes it involves. The need to deal with more than
one linguistic and/or other code is evident even in such extreme
situations as monolingual groups which reject everything extraneous

14
Note that multidia lectalism should not be treated as a first stage to multilingualism,
even though there are models which consider it to be historically prior. Rather,
multilingualism and multidialectalism appear concomitantly.

10
and condemn all departures from the norm (e.g., the Bororo Fulani,
adolescent groups, etc.). This ineluctable diversity of codes is one of
the necessary conditions of symbolic behavior in general and language
behavior (the interaction of language use and language structure) in
particular. Rhyming slangs, pig Latin, and other language games can
be interpreted as proof of this necessity. The social functions of these
practices as identity-building and exclusivist are founded on the
functional dynamics of multiplicity. This is why it must be brought to
the fore in all analytical discussions.
The apparently simpler option of taking monolingualism as the normal
state of affairs is the result of a rationalization which cannot account
for the commonest situations of communication and hence fails to
provide the means for describing them properly. Indeed, the analytical
process is thereby blinded insofar as the dynamics of multilingualism
cannot be (re)constructed by the mere induction of complexity from a
set of juxtaposed monolingual situations. This, of course, has nothing
to do with any holistic philosophical assumption; in this connection,
we might recall Bachelard's notion of "generalization by negation":
"Generalization by negation must appropriate the negated term. All
the last century's advances in scientific thought can be set down to
such dialectical generalizations which appropriate what they negate,
as non-Euclidean geometry appropriates Euclidean" (Bachelard
1940:137). The inversion of the canonical situation as suggested here
allows the monolingual situation to be appropriated as simply a
particular case of the multilingual one.
Hence, the first component of any explanatory approach must
obviously be this requirement that more than one code be available to
speakers. Whether such codes are actually different languages is less
important than the recognition of their availability and possibilities of
development. The consequence, though trivial, should be made
explicit: this fact that codes can be altered within a specific
anthropological setting which guarantees their meaningfulness
provides the basis for the processes of emergence and material
transformation of languages.
2) The communities within which language processes take place are
likewise not homogeneous. They must therefore be considered to be
areas of contact by definition (this again is not simply a factual
observation but also a theoretical postulate), whose features
predetermine the processes of communication. It is perhaps preferable
to speak of social fabrics (with emphasis on texture or structure) rather
than communities (with emphasis on partition and borders), since the
participants have a concrete apprehension (whether conscious or not)
of the nature and rules of the communicational structure within which
they interact and the reasons why they are engaged in it (what good it
is to them). On the other hand, they do not generally find it helpful to

11
have a precise idea of the limits of their community or a symbolic
representation of it. Hence, the first practical object of study is not
linguistic structure, which is a construct, but exchange and the contact
of languages and speech varieties within social fabrics through the
interplay of the available ranges of codes.
A language community can be variably defined according to the
boundaries or set of boundaries recognized by the participants in a
given act of exchange or how they categorize it. Language community
is thus a derived notion. Consequently, a general condition of
heterogeneity must be regarded as an elementary principle of language
behavior. The norm for the linguist should thus be that any linguistic
exchange in a given functional setting must be stably defined as a
potentially multilingual or multidialectal situation.
There are two corollaries to this: first, contact situations are inherent in
the constitution of any language community whatsoever. This means
that, even in an ideal case where there is null internal (lectal and/or
social) differentiation, some such differentiation would ultimately
emerge and become established de facto. Secondly, linguistic
exchanges necessarily transcend the limits of any ostensibly
homogeneous community (cf. in particular Nicolaï 2001). An offshoot
of this is that the boundaries of any language, dialect, or other lect, so
often viewed as essential, are a social construct which can be
manipulated and reshaped according to the strategic needs of the
moment. See Canut (1998:163-4) and Juillard (2001) for an approach
to heterogeneity and the construction of boundaries.
Let us nevertheless not forget that, for obvious reasons, this does not
entail that any empirical process in a multid ialectal context will have
the same outcome and be directly comparable to one in a multilingual
context (we need only recall here the case of koines). It simply implies
that the two contexts will be subject to the same heterogeneity
condition which governs any exchange.
3) The ‘finely layered range’ of codes (rather than the ‘languages’)
potentially available to the individual and/or the community constitutes
a continuum for linguistic rearrangement. This continuum is not a
finite space: it can always be structurally replicated merely through the
use made of it. I relate fine layering to all speech activity in the course
of which norms and expressive traditions are created, new ways of
speaking come into existence, and customary uses of language (often
reified by linguists as registers or genres) are utilized. Such activity,
definable rather in terms of communicational situations than with
reference to any specific language, may give rise equally well to stable
and lasting speech forms as to ephemeral phenomena. For example,
today, in France, the highly symbolic speech forms which have arisen
among the young, particularly in the housing estates of the poorer
suburbs of large cities, can thus be apprehended as an illustration of

12
fine layering. Any differential speech activity whatsoever could,
however, be interpreted in this way, provided it carries with it the
production of a new norm or expressive tradition.
As a process, fine layering operates by putting out and picking up
those phonetic, prosodic, morphological, lexical, syntactic, discursive
and conversational features, selected from an available repertory,
which have as one of their functions the ability to serve as contextual
cues in discourse (cf. J. Gumperz 1982). The use and reuse of such
features is closely monitored by legitimate participants in the given
type of speech activity and treated as indicators, in conjunction with
other symbolic and behavioral markers, of the continual formation and
dissolution of transient human groupings. The reuse of any linguistic
feature more for its contextual significance than for its referential value
is technically one of the procedures involved in fine layering. I shall
speak in this case of the anaphorization of past uses. This is ordinary
behavior and no language can fail to show evidence of it.

Let us look at two correlative aspects of fine layering: layering

as activity and layering as outcome.
As activity, fine layering can be seen as the generating force for
paradigm building17 , given that creation, retention, or rejection of this
or that expression brings about change, whether simplification or
complication, in the entities contained in the non- finite space of the
repertory18 within which the outcome is being built up. Correlatively, it
should be obvious that such alteration / retention / suppression of
linguistic features by virtue of this eminently sociolinguistic process
has the effect of constraining the linguistic results which can be
attributed to the “structural mechanics” alone of a given language; for
this is the true locus of transformation and reorganization of linguistic
structures through the processes manifested in the contingent and
continuously interpreted historical unfolding of verbal activity.
As outcome, fine layering can be apprehended in the concrete
stratification of the repertory available to the speaker: it involves on
the one hand the specific forms and units which speakers choose
(consciously or unconsciously) because they value them highly, and
use strategically in communication, and on the other hand, the use of
specific discursive or conversational sequences, which are no less
identifiable for being procedures rather than components. Differently
stated, this aspect involves both the functionally essential components
of a linguistic structure (e.g., the alteration, retention, or suppression of

17
By this expression I designate the operation, usually enriching the repertory, which
consists of assimilating alternative, discursively developed forms and manners of
speaking for subsequent use. These constitute a paradigm insofar as, despite having
identical reference, they have different meaning (cf. Frege 1892). Note however that
this operation does not invariably involve enrichment; in a suitable anthropological
setting, the result can be the impoverishment of the range of paradigmatic choices.
18
In another area of paradigm building, we might also see a fine layering effect in the
stratification of the utterances which are potentially entailed by the paraphrasablity of
utterances produced in any verbal exchange; indeed, there is rarely only one way of
saying what one wants to say, with regard either to intended reference or to implicit
correlates and any symbolic intentions.

13
morphosyntactic constructions) and the functionalized positive signs
(e.g., the choice of lexical items or phonetic or prosodic features) used
in a signalling system lending itself either to further development or to
rejection. It is further determined by that cognitive operation which
organizes linguistic processes according to its own inherent structural
principles and whose apprehension is the ultimate object of typological
description.

In sum, we may say that there is a fine internal layering of the

repertory which is essential both to the language faculty itself (as
activity: it provides the generating force) and to that of the
functionalization of individual languages (as outcome: new layers can
always be added), and that this factor cannot be ignored in the
description of language change. I therefore use the expression fine
layering to refer to the capability of any language repertory to behave
as a source for the reworking and, on occasion, even the splitting off of
lects and verbal practices created by the refunctionalization of features,
linguistic forms, and materially available discursive and attitudinal
fragments (Nicolaï, 2001).

In even more synthetic terms, let us say that the notion of fine
layering rests on the following hypotheses:
- within the range of layers delimited by each de facto
exchange, whatever their number (drawing from the lattice of
languages, lects, habits, forms, standards, interpretations, etc.), a
restructuring of the given set is always possible without the
intervention of external factors, and
- a new layer can always be added or an existing one eliminated
by the simple self-referential process of fixing upon a linguistic feature
of some previous discourse, whether this be anaphorization, cross-
discursiveness, or something else again.
It thus becomes possible to grasp the overlay, the interweaving, and the
multiplicity of the variants and practices in the repertory without the a
priori assumption of their structural homogeneity.
Fine layering is thus both the outcome and the substance of a
continuous stratification giving no guarantee of any inherent regularity
which would allow us to anticipate the development of the layers or the
shape they will ultimately take on.
Finally, to the extent that we concern ourselves with speakers’
repertories rather than with the languages they know, we can see that
all the speech norms, both conscious and “infraconscious” (whether
subject to negotiation or not), the lects, and the manners of speaking
which mutually define, contrast with, and condition one another are
being constantly impacted upon by a variety of factors including
continual splitting and regrouping, which may bring into play the
individual’s consciousness of his own identity; and this is precisely the
process by which language undergoes change and acquires stability.

14
4) The anthropological setting: All of the above suggests that
the processes of language change with all their relevant parameters
occur within a conventional, contingent, historical setting. Hence, if
acquired and transmitted linguistic forms (these of course include
norms, regular variations, particular lexical and phonetic features,
syntactic constructions, and so forth, all of which are potentially
vectors of identity and sociocultural classification) other than structural
organizations in the structuralist sense and the elementary cognitive
schemata which underlie various contemporary theoretical approaches
are to play a role in the explanatio n of language change, we must
apprehend these forms as anthropological constructs arising from a
particular anthropological setting, which can be defined as the
substrate for their appearance.

The intervening space:

Lastly, the theoretical and methodological clarity of language
description will be enhanced (even in the absence of any empirical
necessity) by the definition of the framework within which these
anthropological constructs and the processes they involve are to be
situated. I call this framework the intervening space.
The intervening space is thus a theoretical construct deriving
from the requirement that any description be articulated according to
the full set of factors which allow us to apprehend the dynamics of
language change as set forth above. This space is the locus of neither
the ‘subject’ of the psychologist, nor the ‘speaker’ of the linguist, nor
the ‘group’, the ‘network’, or the ‘community’ of the sociolinguist, but
rather of another ‘agent’ which I shall call the homo loquens and which
I shall define as an active entity whose form of activity remains to be
described. For our purposes, we need simply assume that this homo
loquens is none other than the agent, theoretically constructed,
cognitively and historically specified, but not linguistically determined,
required by the anthropological constructs which take form, then
structure and define one another over and again according to the
necessities of the moment within a communicational space where the
fine layering of the always non-finite repertory has been identified in
its full linguistic specificity.
Specific processes of the kind observable in Songhay, no less than the
phenomena of areal convergence, analyzable in terms of metatypical
processes (cf. Ross 1997), which depend on other determining factors,
provide good examples of change which can be set down to processes
directly determined at the level of the collective representations
generated in the intervening space. But the reference to this
coordinating framework is even more us eful when one stands on the
limits of language apprehension in situations of need and structural
crisis of the kind envisaged by Manessy (1995) in his study of creole
languages, which led him to develop the notion of semantax. All of
these cases involve processes which cannot be accounted for by mere
structural and cognitive contrasts, lie beyond the domain of the
language as unit of reference, and assume a higher degree of
complexity.

15
My answer to my two original questions is thus strongly affirmative.
The multilingual/multidialectal dimension must indeed be made an
explicit component of models of linguistic processes, and a relevant
anthropological setting extending beyond the individual language must
be defined if the processes the latter undergoes are to be properly
accounted for. We may conclude with another relevant remark by
Bachelard (1934:142), who observed that "the Cartesian method is
reductive rather than inductive, and reductive in a way that distorts
analysis and hinders the extensive development of objective thought.
[...T]he Cartesian method, so successful in explaining the world, is
incapable of complicating experience, as all objective research
should".

Tools
At this point, a distinction must be made between the attempt at
theorization and the task of creating descriptive tools. I shall refer here
to a few tools which various scholars have used in recent years to
account for some of the processes involved in language contact. My
list will not, of course, be exhaustive; I shall limit myself to
mentioning some of the notions and/or images which can be used for
internal or external models.
I use the expression ‘internal models’ to refer to those which are
intended to represent shapes (see above) in terms of processes such as
pidginization, creolization, koinization, and so forth. These are notions
which bear the imprint of the historical context in which they appeared,
but which have evolved through efforts to conceptualize them in a
context- free manner. These efforts have made it possible to shift them
from reference to empirical observations towards reference to a
notional process 19 .
Terminological inflation is an obvious problem in this field: pidgin,
creole, vernacular language, trade language, continuum, prepidgin,
postcreole, semipidgin, semicreole, creolization, pidginization,
recreolization, decreolization, etc. These are all terms which are often
hard to correlate with at least identifiable if not stable contents, to
relate to their historical substrate, and to take over into a theoretically
stabilized explanatory structure. We might thus find ourselves
speaking of Western Songhay as partially pidginized, or of Dendi (the

19
This discussion forcefully recalls Bachelard's comments on the strength of
imagery. Discussing the importance of the image of the sponge in the 18th century to
explain a number of phenomena of physical extension, he cites Réaumur and remarks
(1947:75), "His entire thought derives from this image, and is incapable of
progressing beyond this primary intuition. When he tries to erase the image, its
function persists [...] He might ultimately agree to give up the sponge, but he wants
to hang on to sponginess. This is proof of an exclusively linguistic development
entailing the conviction that, by associating an abstract word with a concrete one, an
intellectual advance has been achieved. A coherent doctrine of abstraction would
require a greater degree of detachment from primary images". Bachelard is speaking
here of historical configurations, but the fact remains that vigilance is required
whenever we come across notions whose definition is overloaded by their reference
or their connotation. There are many of these.

16
Songhay dialect spoken in the Niger/Benin/Nigeria border region) as a
revernacularized trade language.

As an examp le, let us take the notion of ‘vernacularization’ used by

Manessy (1995:96), who defines it as "the effect produced on a variety
of languages by two complementary processes: the simplification of
grammatical structures and the compensatory development of other
means of expression. In a way, simplification is given at the outset [...]
It results [...] from the relaxation of a sociocultural tradition so as to
free a language from normative constraints. The common factor in all
these situations is that the use of a simplified variant is interpreted not
simply as a way to achieve common understanding but as an
expression of solidarity which transcends ethnic differences within a
given framework such as town, region [...] or nation state [...] This
solidarity is manifested through the communality of discourse
conventions". Clearly, the essential point from the linguistic standpoint
is the existence of phenomena of simplification in the form of
perceptible changes each time a given community uses one of the
languages in its range and restricts its use, whether deliberately or not,
to some part of the functions which it might otherwise have.
Psychosocially speaking, the development of a situation of complicity
is the factor which stabilizes a given linguistic form by creating
specific conventions of discourse.
Vernacularization is thus linked both with pidginization (insofar as
both are marked by the same type of simplification) and with
creolization (insofar as the latter implies the normative stabilization of
a simplified variety of language). This is the process which gives rise
to a representation of the speech practices of a community which do
not as yet show the increased complexity characteristic of creolization,
though they are its precondition. Vernacularization is thus the first
stage in the process of creolization, as Manessy stresses when he says
(1995:129), "We propose to use the term 'vernacularization' to
designate the set of linguistic processes which are set in motion within
a given language variety as it is appropriated". The process of
appropriation, which is psychosocial in nature, is thus the essential
factor in this development. The field of application of the notion has
been defined; it now remains to organize the theoretical field within
which it must apply and prove its validity.
The question may be stated as follows: once these notions have been
extracted from their historical context and defined, to what extent do
they capture the elementary linguistic processes involved and/or to
what extent are they the representation of still vaguely defined
operative phenomena which remain to be analyzed within an as yet
undeveloped theoretical framework? This question cannot necessarily
be answered here; it is even unclear whether it has an answer. At the

17
same time, there is no reason to think that the absence of an answer
constitutes a handicap for understanding the observed phenomena.
In any case, the processes (such as those which bring vernacularized
forms into being) are neither described nor explained simply because
they have been given a name. At best, the term ‘vernacularization’
delimits a category of empirical phenomena which are classified
together on the basis of supposedly shared features such as
simplification; not because they are less historically marked do they
become more precise. Processes are thus notions which require an
explanation and not tools for providing one.

Of more recent origin, external models make use of mathematical

representations or schemata of regular processes. In developing
metaphorical extensions of models applied in other domains to
apprehend complex phenomena which are beyond the reach of a
deterministic approach, they abandon explanatory aspirations and try
only to account for processes which can be characterized geometrically
or topologically. This is modest on the one hand, in that ineradicable
pretentions of explaining linguistic facts are set aside, and ambitious
on the other, in that a new pretention of shifting the domain of
relevance comes to the fore. Recourses to catastrophe theory (Thom,
1974), fractal theory (Mandelbrot, 1975), dissipative structure theory
(Prigogine & Stengers, 1979), and chaos theory (Ruelle, 1991) fit into
this category.
Among those who have taken an interest in the heuristic capabilities of
this kind of model for the study of language processes and language
change is Lass (1997). His perceptive study explores the explanatory
potentialities at the intersection of biological models and metaphorical
conceptualizations drawn from chaos theory (cf. point attractors, sinks,
limit cycles; cyclical attractors; arrows and cycles; flow; chreods; drift,
etc.) . He postulates contingent topologies (epigenic landscapes, etc.),
which are not far removed from the substrate spaces of catastrophe
theory, correlated with the identification of causal mechanisms and
evolutionary configurations (cf. stasis, punctuation) which relate
explicitly to the biological evolutionary models of Gould and Eldredge
(1977).
The question which arises, whenever a model is to be exported, is
exactly what is being modelled and what relationship exists between
the properties of the model and those of the objects to which it applies.
Do the former mask or illuminate the latter? Do they have an effect on
perspective? What is added by the newly transferred model to the
already available perceptions? Could one perhaps identify a specific
anthropological setting such as the one characteristic of the emergence
of the Songhay languages in terms of processes occurring in a
definable "epigenic landscape" which differs structurally from some
other type of anthropological setting, such as the one characteristic of
the Oceanic et Melanesian area, which has given rise to the notion of
metatypy (Ross 1997)? More importantly, is anything to be gained
from such a modification in terms of general linguistic theory and an
understanding of the phenomena of language change?

18
Moreover, there can be more than one level of models corresponding
to incommensurate scales in the apprehension of the phenomena. For
example, the tree model of language change operates on a level which
is totally unrelated to those on which morphological representations of
the Cusp of Whitney found in catastrophe theory (cf. Thom, 1974:157)
or the strange attractors of chaos theory (cf. Ruelle, 1980:131) might
conceivably apply. The fractal approach, which apparently has
applications in demography (cf. Le Bras, 2000), might provide a
broad-scale approximation to phenomena of linguistic diffusion. In
these cases, are the metaphors involved (for these are indeed still
metaphors at this stage 22 ) of heuristic interest or, on the contrary, do
they create opacity?

Bricks and metaphors: What are bricks for? For building houses,
obviously. Well, both internal and external models referred to above
provide nothing more than "conceptual bricks". There are other equally
well organized, factually oriented approaches which have made
contributions to the same edifice (cf. Thomason & Kaufmann 1988 and
their analysis of interference, Manessy 1995 and his development of
semantax, and Lass 1997 and his thoughts on the processes of change).
Each of these conceptual bricks bears its factory imprint (i.e.,
metaphorically speaking, is referenced). This is both enriching and
problematic. In the end, however, the important thing is what can be
built from the bricks and the architectural idea which brings the whole
together 23 , since that idea must be anchored in empirical reality and
lead beyond self- reference.
The aim is thus not simply to find / invent a local or overall model
which captures certain apparent phenomenal regularities. It is also to
evaluate the extent to which this translation can account for properties
whose relevance has previously been recognized / posited / envisaged
on a theoretical level, given a few a priori principles and an at least
momentary correspondence to a particular class of empirical
phenomena. The import of these apparent regularities and the way in
which the model helps to make them meaningful must also be
evaluated. In short, the thing apprehended must be conceptually
grasped in its initial coherence (which is thereby put to the test) and in
the characterization of its relevant functions; and the model must be
found capable of magnifying precisely the desired defining functions

22
It should not be forgotten that all these theories have mathematics as their domain,
and that the relationship between the formal properties of a model and the objects to
which it applies must be established or at least explicitly set out in order for the
exportation to go beyond the level of the approximate metaphor, however
enlightening this may be in itself.
23
Thus Lass (1997:293) finds his tools helpful but stresses that "what counts is the
image of an evolving system as a kind of ‘flow’ in some n-dimensional space, and the
existence of regions in that space towards which the flow tends to converge".

19
and relations by making them explicit and translating them. We
thereby return to the subject of metaphor.
When reflecting on the process of conceptual elaboration, Lass
(1997:42) remarks that "[a]mong the important constructivist devices
available to the historian is the creation of metaphors; metaphorical
images can define and create new natural or conceptual kinds, which
then become legitimate objects of exploration, and enrich the
discipline’s universe". Again, "we may notice (a) that [our own
metalanguage] is much more metaphorical than we think, and (b) how
important these metaphors are as devices for framing our thinking,
and how much of our theory they actually generate", and this is in a
way evident. To this, Bachelard's reply (1947:38) might have been, "A
science which accepts imagery is the most vulnerable to metaphors.
This is why the scientific method must constantly struggle against
imagery, analogy, and metaphor". Or again (1947:81), "The danger of
immediate metaphors for the development of scientific thought is that
they are not always transitory images; they instigate autonomous
thought and tend to expand and reach fullness in the domain of
imagery". The debate remains open.

Perspectives
The issues and the framework for the issues: The discussion hitherto
should suggest a framework whose objective would be to retain the
explanatory power of existing theories while placing them in context
(see the remarks on negative generalization above). An aim of this kind
helps to place the problems of language processes and description of
change in a new focus. It is in this respect that my concerns merge with
others expressed at this symposium trying "to identify hitherto unstated
or understated fundamental issues in linguistic theories taking into
account the rich variation of forms and functions observed in the
languages of the world". Some of these "unstated or understated
issues" are illustrated by the preceding debate, since there is no
assumption of the multicode reference framework I have been
suggesting in the stance that makes utterances such as "What should be
the proper object for theories of language structure?" or "What should
a theory of language structure explain?" comprehensible.
This is also so of questions such as "What are the motivations for
language change and grammaticalization?" or "Does human conscious
choice play a part in language change?" Taking a position from the
outset that recognizes multicodism as a defining feature of
communication will entail changes in the analyses. Allowing for
processes within what I have called the intervening space will reorient
analyses in directions which it is far too early to spell out 24 . At the

24
We are no longer faced with a simple process of change which can be analyzed
almost deterministically with perhaps some allowance for contextual effects (the
ordinary framework for the a posteriori explanation of a clearly identified and
localized change) but are not yet (far from it) in a situation governed by chance and
preconstrained by a specific substrate. The most accurate prediction in the treatment
of complex phenomena of any kind (in meteorology, economics, etc.) remains that of
the margin of error.

20
same time, the focus on the need for models suggests that we should be
far more demanding before we agree to use them.
Grammaticalization can thus be set down to a process governed by the
dual effects of wear and cognitively oriented reorganization of
language structures (dominant and recessive structural types, shift from
concrete to abstract, etc. 25 ). But are these two processes what they
seem or do they, like the notions mentioned above, rather stem simply
from giving a name to an outward appearance or an a priori? Are they
not modified by the postulate of multiple codes?
Is the suggested link between dominance and improved cognitive
adequacy anything other than a hypothesis based on the consideration
of a known contingent process of formal structuring? Is the expansion
of a given form / structure necessarily linked to its effectiveness within
a cognitive and neo-Darwinian conceptual background or are there
other perhaps less intuitive alternatives? How can this expansion be
justified without circularity? Is this anything other than the kind of
shaping which one of the external models which account for the
creation of order in accumulations might provide? Or could other
models originating from interactional sociology not be used via, say,
the interplay of agents involved in a system of interdependence with
the development of an emerging phenomenon? Even without lengthier
analysis, several possible approaches appear clearly, each relying on a
given level of explanation and system of relevance.
Again, is it true that the concrete precedes the abstract and what might
this mean? How could we imagine, even at the "beginning", a language
without the inherent correlative abstraction which a symbolic system
implies by and for its very existence? Perhaps in this case, another
process should be conceived which, on a given occasion, might look at
first sight like a passage from the concrete to the abstract of the kind
which is axiomatic in many works on natural semantics. Finally, it
might also be better here to apprehend this fundamental duality as a
complex phenomenon rather than trying a priori to assign a direction
from one facet to the other of a process of symbolization and find
proof / traces of this in language change.
Finally, what we find here is that no one model will suffice to account
for the phenomena under consideration. While it is clear that each
subdomain is apprehended through a system of relevance and that each
model gives preference to its own representation, the set of
subdomains is in constant interaction, and this in turn validates other
models and brings out new relevancies. There is thus nothing
surprising in the fact that questions originate in different temporalities
and refer back to different dimensions so as to support heterogeneous
modelizations.
In the same way (initially, there is no alternative), their apparent
simplicity disguises a formidable semantic plurality: to speak of
"human conscious choice" is to bring in the potential effect of all

25
All of these are notional tools from the conceptual arsenal whose history could
doubtless be followed back via a cross-disciplinary analysis of scientific discourse.
Their common use and status as elementary concepts do not make them accurate
descriptions nor a fortiori valid explanations.

21
manner of legislative activities; yet there is action without explicit
legislation and without reference to institutionalized normative
representations (cf. the development of discursive and/or linguistic
norms in many communities). But here again, on what scale is the
effect of choice being recorded (from the level of the temporary
interactive group to that of a splitting of the community; from the
interpretation of a process as a function of contemporary use to the
recording of change over several thousand years, etc.)?
Finally, if it is accepted that neither the individual nor the speakers
have any direct effect on their language, what inferences can be drawn
regarding the construction of a collective entity referring back to a
potentially active homo loquens, which is as much of a construct as the
‘speaker’: an entity with no consciousness though necessarily
possessed of memory and normative references defined at a level
independent of the one on which the structural processes of languages
are built; an entity which can be defined to the dimensions of an
anthropological space whose limits are not necessarily those of a given
language since it comes under the effects of linguistic, cognitive, and
cultural dimensions all at once?
I am fully conscious of going beyond the limits of what can be
expressed within the framework of a symposium. It is likely that such
issues could only be dealt with in a fully fledged program of new
research.

Towards a conclusion
In this paper, I have discussed the nature of language contact and
diverse aspects of language change in the light of a particularly
difficult case. I have also set forth a number of theoretical
considerations which will help to grasp the importance of certain
aspects of language change which I feel have too often been
neglected. Finally, I have briefly suggested ways in which notions of
model making might be used to describe the results of these processes.
No one of these three themes is directly dependent on any other, none
can be induced in any way from another. It is nevertheless obvious that
they require correlative consideration and development. The
implications of a number of simple ideas require closer examination.
Fuller appreciation of the complexity of the structures we deal with
must be sought, and descriptive models 26 need to be developed which
identify agents which are intrinsically active in their field of reference
(intervening space, linguistic space, etc.) and provide more than
unanalyzed representations from this standpoint (such as those which
are characterized as processes deriving directly from the consideration
of historical phenomena: "tendencies" such as convergence)27 .

26
These must be neither opacifying (hiding what they are intended to illuminate) nor
residual (oversimplifying what they are supposed to account for). In either of these
cases, the problem is avoided, and the sum of the knowledge acquired is either nul or
negative.
27
Like many others, I once found notions of processes such as pidginization,
creolization, etc. stimulating and perhaps heuristic. I do not now renounce them but,
after further reflection, I hope to have assigned them to their proper place. This is
proof of their usefulness and opens the way to moving beyond them.

22
In the intervening space, for example, to what extent would models in
terms of systems of interdependence with their emerging forms not be
better adapted to interpreting linguistic transformations so as to
provide feedback for refocussing certain strictly formal processes? The
appearance of normative representations and their structuring could
probably be interpreted, according to the context in which they are
realized, as emerging effects of a complex system in which functional
role and interdependence work together. And this is equally true of the
description of processes apprehended over their historical
development. But this is perhaps oversimplifying.
Finally, in order to stress the fact that my discussions are creating a
local norm, I shall conclude with a counterpoint in the initial key.

Science on Cartesian principles quite logically complicated the

simple, but contemporary scientific thought tries to read real
complexity beneath the simple appearances of adjusted
phenomena. It tries to find pluralism beneath identity, imagine
occasions when identity might be broken down beyond
immediate experience, itself too readily compacted into a broad
view. These occasions do not come of their own, they are not
found on the surface of being, in fashions, in the picturesque
aspects of shimmering, orderless nature. They must be picked
out of the heart of substance from within the contexture of its
attributes. (Bachelard 1934:143)

Références
Bachelard, Gaston. 1934 [1999]. Le nouvel esprit scientifique. Paris :
PUF.
- 1940 [2002]. La philosophie du non. Paris : PUF.
- 1947. La formation de l’esprit scientifique. Paris : Vrin.
Bender, Lionel M., 1995. The Nilo-Saharan languages, a comparative
essay. München: LINCOM Europa.
Canut, Cécile. 1998. « Perception des espaces plurilingues ou
polylectaux et activité épilinguistique ». Language and location in
space and time, ed. by Petr Zima & Vladimir Tax, 155-170, München :
LINCOM Europa.
Creissels, Denis. 1981. « De la possibilité de rapprochements entre le
songhay et les langues Niger-Congo (en particulier mandé) ». Nilo-
Saharan, ed. by Thilo C. Schadeberg & M. Lionel Bender, 185-199.
Cinnaminson: Foris Publications.
- 2001. « Catégorisation et grammaticalisation : la relation génitivale
en mandingue ». Leçons d’Afrique, Filiations, ruptures et
reconstitution de langues, ed. by Robert Nicolaï, 433-454. Louvain-
Paris : Peeters.
Dixon, Robert M. W. 1997. The rise and fall of languages. Cambridge:
Cambridge University Press.
Ehret, Christopher. 2001. A historical-comparative reconstruction of
Nilo-Saharan. Köln: Köppe.

23
Frege, Gottlob. [1892] 1971. Ecrits logiques et philosophiques. Paris:
Le Seuil.
Gould, Stephen Jay & Eldredge, Niles. 1977. “Punctuated equilibria:
the tempo and mode of evolution reconsidered”. Paleobiology 3. 115-
151.
Greenberg, Joseph H. 1963. The languages of Africa. Bloomington:
Indiana University Press.
Heine, Bernd. 1975. Language typology and convergence areas in
Africa. Linguistics 144. 27-47.
Houis, Maurice. 1971. Anthropologie linguistique de l’Afrique noire.
Paris : PUF.
Juillard, Caroline. 2001. « Une ou deux langues ? Des positions et des
faits ». La Linguistique 37 (2). 3-31.
Lacroix, Pierre-Francis. 1971. « L’ensemble songhay-jerma:
problèmes et thèmes de travail ». Actes du 8e Congrès SLAO, 1969,
Annales de l'Université d'Abidjan, Linguistique, Série H, Abidjan. 87-
99.
Lass, Roger. 1997. Historical linguistics and language change.
Cambridge : Cambridge Univ. Press.
Le Bras, Hervé. 2000. Essai de géométrie sociale. Paris : Odile Jacob.
Mandelbrot, Benoît. [1975] 1995. Les objets fractals. Paris :
Flammarion.
Manessy, Gabriel. 1995. Créoles, pidgins, variétés véhiculaires,
Procès et genèse. Paris : Editions du CNRS.
Mukarovsky, Hans. 1989. « Songhai – eine tschadische Sprache? ».
Frankfurter Afrikanistische Blätter 1. 16-29.
Nicolaï, Robert. 1977. Sur l'appartenance du songhay. Annales de
l'Université de Nice 28. 129-35.
- 1984. Préliminaire à une étude sur l’origine du songhay. Berlin :
Dietrich Reimer.
- 1990. Parentés linguistiques (à propos du songhay). Paris : Editions
du CNRS.
- 2000. La traversée de l’empirique. Essai d’épistémologie sur la
reconstruction des représentations de l’évolution des langues. Paris :
Ophrys.
- 2001. « La « construction de l’unitaire » et le « sentiment de l’unité »
dans la saisie du contact des langues ». Langues en contact et
incidences subjectives, 359-386. Montpellier : Traverses 2.
- 2003. La force des choses ou l’épreuve ‘nilo-saharienne’. Questions
sur les reconstructions archéologiques et l’évolution des langues.
Köln : Köppe.
- in press a. « Aux marges de l’espace chamito-sémitique: songhay et
diffusion aréale. Les langues chamito-sémitiques (afro-asiatiques) »,
ed. by Amina Mettouchi & Antoine Lonnet. Paris : Ophrys.
- in press b. « Mande-Songhay: Facts, questions, frames ». Africa as a
linguistic area / Areal typology and African languages, ed. by Bernd
Heine & Derek Nurse. Cambridge Univ. Press.
Prigogine, Ilya & Isabelle Stengers. 1979. La nouvelle Alliance.
Métamorphose de la science. Paris : Gallimard.
Ross, Malcom S. 1997. « Social networks and kinds of speech-
community event ». Archeology and Language 1, ed. by Roger Blench

24
& Matthews Spriggs, 207-261. London – New-York: Routledge.
- 2001. “Contact-Induced Change in Oceanic Languages in North-
West Melanesia”. Areal and Genetic Inheritance. Problems in
Comparative Linguistics, ed. by Alexandra Y. Aikhenvald & Robert
M.W. Dixon, 134-166. Oxford: Oxford University Press.
Ruelle, David. 1980. « Strange attractors ». The Mathematical
Intelligencer 2. 126-137.
- 1991. Hasard et chaos. Paris : Odile Jacob.
Thom, René. 1974. Modèles mathématiques de la morphogenèse.
Paris : Christian Bourgois.
Thomason, Sarah. 1996. Contact languages, a wider perspective.
Amsterdam - Philadelphie: John Benjamins.
Thomason, Sarah G. & Terrence Kaufman. 1988. Language contact,
creolization, and genetic linguistics. Berkeley: California Univ. Press.
Vydrine, Valentin. in press. “Areal and genetic features in West
Mande and Mani-Bandama (East Mande) phonology: In what sense
did Mande languages evolute?”. JAWL International Symposium
"Area Typology of West Africa" ed. by Bernard Comrie & Ekkehart
Wolff.
Westermann, Dietrich. 1927. Die Westlichen Sudansprachen und ihre
Beziehungen zum Bantu. MSOS Beiheft (Jagrg. 30). Berlin: De
Gruyter.
Zima, Petr. 1988. “Songhay, Hausa and Chadic (Preliminary notes on
lexical affinities)”. Progressive Traditions in African and Oriental
Studies: Asia, Africa, Latin America 21. 185-192.

25