Specific aims : Specific aims
Task 1 - Classification : Task 1: Towards a new classification of African languages
Task 2 - Reconstruction : Task 2: Historical inferences of linguistic reconstructions
Task 3 - Modelisation : Task 3: Languages, populations and environmental constraints
1. Technical and scientific description of the proposal
The place of the African continent as the home of Homo sapiens is not generally disputed any more
among paleontologists and geneticists. Human genetic diversity in Africa is much greater than
anywhere else and the oldest remains of H. sapiens have been found there. On that account, it seems
rather paradoxical that African linguistic diversity appears much lower than in say, New Guinea,
whose oldest settlement surely does not predate c. 60,000 BP , or the Americas (30,000 BP ?).
One should thus investigate the reasons for this apparent uniformity. A first answer which comes to
mind is that linguistic diversity on the African continent has been underestimated by existing
classifications. The four phyla identified by Joseph Greenberg (1963) and generally accepted by most
present-day African linguists (Afro-Asiatic, Khoisan, Niger-Congo and Nilo-Saharan) might include
languages actually unrelated to each other (since to a large extent Greenberg accepted wholesale
groups of languages established by previous linguists -missionaries, administrators, etc. - often on the
basis of very slender evidence).
It is thus quite likely that a number of languages have been misclassified, and indeed there is a
growing feeling among specialists of various linguistic areas in Africa that about 12 to 15 languages
should rather be classified as isolates (a list will be provided in 1.4.). Our own feeling on the matter is
that this figure might well be too low.
Whatever the case, and even with this proviso in mind, there is still much less heterogeneity in
Africa than in New Guinea (where no less than 60 different phyla have been identified - some of them
with an admittedly very small membership - for a total of 800-odd languages).
Another cause for this loss of linguistic diversity should probably be sought in the expansion of
food-producers (agriculturalists and/or pastoralists) into territories formerly inhabited exclusively by
hunter-gatherers. Such a process has been identified several times since neolithisation began (cf.
Diamond & Bellwood, 2003) and the expansion of Bantu languages into Central, Eastern and
Southern Africa constitutes a particularly telling example, and one which we intend to use as our
primary case study.
Processes of state-building and other centralized or semi-centralized polities certainly played a role
in more recent times, as exemplified by Vansina (1990) for the equatorial forest and Schoenbrun for
the Great Lakes area (1998), among others. It can be surmised that expansion of chiefly power into
ever-widening regions implied the concomitant growth of the language of the court at the expense of
the languages of subject peoples (the examples of the Kongo kingdom for West Central Africa or
Buganda in Eastern Africa immediately come to mind).
In order to unravel the complex interactions between peoples, languages and ways of life, linguists
have at their disposal the powerful tool of the historical-comparative method. Briefly, this method,
elaborated and refined since the mid-19th century, particularly by Indo-Europeanists, makes it
possible to reconstruct at least the broad outlines of an ancestral language, especially as concerns its
phonology, morphology and lexicon. By so doing, it is possible to identify contacts between daughter
languages and languages belonging to other families (as well as, with slightly more difficulty,
interaction among the daughter languages themselves), and to formulate hypotheses about the natural
environment of the societies in question, their mode of life (hunting and gathering, agriculture, animal
husbandry, etc.), their social systems and to some extent their ideologies as expressed in their
Since the earlier pioneers in comparative method, many new developments and refinements in
historical linguistics have taken place in recent years: new phylogenetic methods have been applied to
linguistic data, e.g. the reticulated network approach to the overall classification of Bantu languages,
conducted by Russell Gray (Auckland) in collaboration with our own laboratory (Dynamique du
Langage, DDL, Université de Lyon - CNRS), or the computational analyses of Gabon Bantu
languages, undertaken at the University of Groningen (Alewijnse et al.) on the basis of the ALGAB
(Linguistic Atlas of Gabon) collected and analysed by the Bantuists at DDL, to which one should add
the use of virtual reconstructions (Mouguiama-Daouda), further to be elaborated upon in 1.3.
We are fortunate, as Bantu-language specialists, in having at our disposal a very elaborate
comparative phonology and lexicon covering most of the Bantu area. It is thus relatively easy to point
out divergences and convergences among the various languages as well as the impact from
neighbouring language groups (hence peoples) on the Bantu languages themselves. Starting from
there, a fine-tuned examination of the cultural vocabularies mentioned above will lead us to well-
formulated hypotheses on the history of these languages (and peoples!), leading for instance to an
answer to the question : granted that the Bantu expansion did take place, as all evidence leads us to
believe, what were the paths followed by this expansion from Southern Nigeria-Cameroon, through
and/or around the rainforest, and into Central, Eastern and Southern Africa? Exciting new data from
other disciplines (archaeology, population ecology, demography...) will also be taken into
consideration to help us better achieve these aims.
The starting point of our hypothesis is to be found in the path-breaking work of Cavalli-Sforza and
colleagues (1994) which first established, on the basis of classical genetic markers (blood groups) that
the phylogenetic diversity of African populations by far exceeded that found in the rest of the world.
More recently, work on DNA itself (mitochondrial DNA and Y-chromosome) have amply supported
the original findings, confirming that the origin of H. sapiens is indeed to be found in Africa, as
already mentioned under 1.1.
One great breakthrough brought about by molecular studies is the possibility to date, albeit
approximately, by using the "molecular clock", the most remote common ancestor of present-day
populations. A very noteworthy recent contribution has been published in the Proceedings of the
National Academy of Science. Resulting from a joint study undertaken by linguists from our
laboratory (under the direction of Professor Lolke van der Veen), anthropologists from the Museum
National d’Histoire Naturelle (Serge Bahuchet) and geneticists from the Institut Pasteur (Quintana-
Murci and collaborators, 2008), it was shown that the most recent common ancestor of Bantu-
speaking and Pygmy populations probably went back approximately 70,000 years back. Furthermore
it was established that Western (Gabon) and Eastern (Ituri) Pygmies shared practically no common
genetic markers, showing that the unity of so-called "Pygmies" is in fact spurious and various
independent groups of hunter-gatherers existed in Central Africa and elsewhere long before the
present phyla began their expansion (thus making it highly likely that extinct language groups,
unrelated to any attested phylum, were also present).
A second source of inspiration for our project is to be found in some recent developments generally
grouped under the heading of "language ecology" as developed inter alia by Nettle, Nichols and
Blench among others. Basically, language ecology aims at analysing language distribution and
diversity in relationship to the environment, both natural and cultural, in which the speakers of the
language are immersed. For instance Nettle (1999) has highlighted the correlation between climate,
food production and the minimum size of linguistic communities, showing that in wet tropical areas
where sustainable agriculture can be effective, smaller, hence more diverse, discrete communities are
to be found, whereas in areas more marginal to agriculture, with a longer dry season, for instance,
linguistic communities tend to be larger, due to the necessity of establishing wide-reaching exchange
networks to face the vagaries of the environment and the risks of crop failure and starvation.
On the other hand, hunting-gathering groups, living in the same marginal environment, will
probably face the problem of diminishing resources by moving somewhere else, their mobility being
much greater than that of land-tillers, on account of their lacking material possessions in great
number. Thus, it is to be expected that they will exhibit more linguistic diversity than agriculturalists,
although living in the same marginal environment. As far as pastoralists are concerned, although they
are undoubtedly quite mobile, they also have very extensive social networks of partners, with whom
they leave part of their livestock to minimize risks and thus find themselves rather in the same
category as agriculturalists in these semi-arid environments, i.e. one of reduced linguistic diversity.
A very interesting contribution to the question of language diversity in Africa in relation to
geographical factors is to be found in a paper by Bernard Comrie and Michael Cysouw of the Max
Planck Institute for Evolutionary Anthropology in Leipzig. Basing themselves on the WALS (World
Atlas of Linguistic Structures) the authors show that typological (thus not necessarily genealogical)
similarity in Africa is distributed along an East-West axis north of the Equatorial forest, that is in the
area of maximum language diversity, which is also likely to include the centers of agricultural
innovation; likewise, in this area - at the forest savanna-ecotone - are to be found a variety of
subsistence resources at the disposal of hunting-gathering groups (and even agriculturalists); on the
other hand in Eastern Africa, typological similarity is oriented along a North-South axis, clearly
illustrating the expansion of food-producers (Bantu, Cushitic and Nilotic speakers) expanding into
hunter-gatherer territory and absorbing the latter's languages.
The overall hypothesis of our project is thus that considerably more linguistic diversity existed in
sub-Saharan Africa before the expansion of food-producing populations and that part of this diversity
can still be recovered by: 1) examining the remnant language isolates (Task 1); 2) analysing in-depth
the vocabularies of contemporary populations (food-producers as well as hunter-gatherers) in order to
discover traces of former languages at present submerged (Task 2); 3) correlating paleo-demographic
and paleoclimatic data to putative population expansion and replacement within and around the
equatorial rain forest (Task 3).