The second task aims at determining the specific modes of contact between food-producers and hunter-gatherers throughout Central, Eastern and Southern Africa.
Since, as mentioned in 1.1, it appears likely that the expansion of food-producing peoples was conducive to a reduction of linguistic diversity, there remains the apparent paradox that Central African hunter-gatherers all speak languages identical or at least closely related to those of their agricultural neighbours, whereas this is not the case in Southern Africa where "San" hunter-gatherers have maintained a considerable amount of linguistic diversity (all the more so, since "Khoisan" as we saw in 1.4.a is probably not a valid genealogical grouping). Nettle's explanation as we saw in 1.2. has to do with environment, where hunter-gatherers in arid environment rely on mobility to escape stress and so are better able to retain their own linguistic and cultural identity. One should however notice that in no less arid areas of East Africa, linguistic diversity among hunter-gatherers is less, although still fairly significant - in this case, the time factor is doubtless decisive, interaction between East African hunter-gatherers and food-producers predating by at least 1,000 years - probably more - similar contacts in Southern Africa.
Second, the concrete trajectories followed by the expanding food-producing populations need to be precisely determined. To take one very significant example, it is not known - and has been in dispute over many years - whether Bantu-speaking food-producers traversed the entire breadth of the rain forest before emerging in East Africa, or followed a path north of it, or again crossed it on a roughly north-western / south-eastern orientation to emerge in the savannas south of the forest. We favour the second interpretation, and will thus attempt to support it by selecting specific languages situated at various strategic points on the northern rim of the rain forest (e.g. the Grassfields languages, languages of the Congo Interior basin and languages of the Great Lakes area in East Africa.
In order to achieve this aim, we will have to set up data bases for significant domains of specialised vocabularies in the languages under study: these would concern wild fauna and flora, hunting and fishing techniques, agriculture, animal husbandry, pot-making, iron-working, etc.
A not insignificant amount of cultural vocabulary has already been identified for Proto-Bantu. Some crop names can be confidently reconstructed: cowpea (Vigna unguiculata and / or V. sinensis), yams (Dioscorea spp.); Bambara groundnut (Vigna (= Voandzeia) subterranea); and some species of pumpkin (Cucurbita spp.). It is noteworthy that no cereal names are reconstructible at this level, and indeed expanding Bantu populations do not seem to have cultivated grain crops until they reached East Africa (Ehret 1998). The case of bananas is more complex: they provide the staple crop of most groups in the equatorial forest as well as a substantial number in the moister areas of East Africa, but it is very difficult to reach a consensus on the date of their introduction; the traditional view is that they came late but some archaeological data point to their presence in Cameroon c. 500 BC, so that the actual role they played in Bantu expansion is by no means settled.
We also have good surveys of the vocabulary of pot-making (Bostoen), fauna (Mouguiama-Daouda) etc. These mostly concern Bantu languages however. Much work remains to be done on other groups, although a fair amount of cultural vocabulary (mostly animal husbandry) has been reconstructed for Eastern and Southern Nilotic (Vossen and Rottland respectively).
We will obviously have to complement the available data on specialised lexicon for carefully selected languages (in particular Tsogo, Mongo, Ganda and Kikuyu). This will be achieved by field trips as required.
By applying the method of virtual reconstructions, one can propose putative ancestral forms for vocabulary items attested in present-day languages, which allows a stratification of these items in terms of their relative antiquity. As indicated in 1.3. establishing virtual reconstructions means the application of regular correspondences to items found in present-day languages so as to mimic the shape of the putative ancestral items. In some cases, these reconstructions would coincide with roots already identified (for instance in Guthrie's or Meeussen's lists), in other cases, they will reconstruct to regular proto-roots which would then have to be reconstructed to a regional proto-language (daughter language to Proto-bantu) and in others yet they will point out irregular correspondences, and probably the result of loans or other lateral influences. We intend to develop a program for the automatic generation of these virtual reconstructions, which should be fully operational during the third year of the project, the data bases themselves having been completed by the end of the second year.