Cultural phylogeography of the Bantu Languages of sub-Saharan Africa

Abstract

There is disagreement about the routes taken by populations speaking Bantu languages as they expanded to cover much of sub-Saharan Africa. Here, we build phylogenetic trees of Bantu languages and map them onto geographical space in order to assess the likely pathway of expansion and test between dispersal scenarios. The results clearly support a scenario in which groups first moved south through the rainforest from a homeland somewhere near the Nigeria-Cameroon border. Emerging on the south side of the rainforest, one branch moved south and west. Another branch moved towards the Great Lakes, eventually giving rise to the monophyletic clade of East Bantu languages that inhabit East and Southeastern Africa. These phylogenies also reveal information about more general processes involved in the diversification of human populations into distinct ethnolinguistic groups. Our study reveals that Bantu languages show a latitudinal gradient in covering greater areas with increasing distance from the equator. Analyses suggest that this pattern reflects a true ecological relationship rather than merely being an artefact of shared history. The study shows how a phylogeographic approach can address questions relating to the specific histories of certain groups, as well as general cultural evolutionary processes.

Keywords: cultural diversity; language diversity; language evolution; phylogenetic comparative methods; phylogenetic inference; phyloplasty.

Figure 1.

Alternative Bantu dispersal hypotheses and the different phylogenetic tree structures they imply. Alphabetically labelled regions refer to the Guthrie zones commonly used in Bantu studies. Asterisks (*) represent the approximate location of the ancestral Bantu society. Under the ‘deep-split’ scenario, there was an early split between East (EB, yellow) and West (WB, green) Bantu clades as one group went along the northern edge of the rainforest (approx. regions A–D). In contrast, the ‘pathway through the rainforest’ scenario implies a more chained topology as groups spread and diversified through the rainforest, with the later formation of a monophyletic East Bantu clade. Guthrie zone map is adapted from map created by I. Edricson (http://en.wikipedia.org/wiki/File:Bantu_zones.png).

Figure 2.

Phylogenetic tree structure and explicit phylogeographic mapping support the ‘pathway through the rainforest’ dispersal scenario. (a) Simplified consensus phylogenetic tree of 542 Bantu languages constructed from posterior sample of 500 trees. Triangle size is proportional to number of languages. Letters refer to Guthrie zones (figure 1). (b) Ancestral locations of numbered nodes in the tree and pathway of expansion are shown in relation to extant languages. (c) Distribution of inferred ancestral location of some nodes across PP trees shows uncertainty about precise locations but the overall pathway implied by these analyses remains the same. The dashed-line ellipse refers to the location of dispersal for languages leaving the rainforest as hypothesized by Rexova et al. [22] and the star represents the location of dispersal for East Bantu languages under the simplified rainforest-first dispersal scenario of de Filippo et al. [8]. These locations are not supported by the dispersal pathway inferred from our analyses.

Figure 3.

Language area increases with increasing latitude. The solid line is the ordinary least-squares (OLS) regression line (R² = 0.21), while the dotted line shows the regression line based on parameter values of the PGLS analysis, which controls for phylogenetic relatedness (R² = 0.07).