Bantu expansion shows that habitat alters the routeand pace of human dispersalsRebecca Grollemunda,1, Simon Branforda, Koen Bostoenb, Andrew Meadea, Chris Vendittia, and Mark Pagela,c,1

aEvolutionary Biology Group, School of Biological Sciences, University of Reading, Reading RG6 6BX, England; bKongoKing Research Group, Department ofLanguages and Cultures, Ghent University, 9000 Ghent, Belgium; and cThe Santa Fe Institute, Santa Fe, NM 87501

Edited by Peter S. Bellwood, Australian National University, Canberra, ACT, Australia, and accepted by the Editorial Board August 10, 2015 (received forreview February 25, 2015)

Unlike most other biological species, humans can use cultural inno-vations to occupy a range of environments, raising the intriguingquestion of whether human migrations move relatively indepen-dently of habitat or show preferences for familiar ones. The Bantuexpansion that swept out of West Central Africa beginning∼5,000 yago is one of the most influential cultural events of its kind, even-tually spreading over a vast geographical area a new way of life inwhich farming played an increasingly important role. We use a newdated phylogeny of ∼400 Bantu languages to show that migratingBantu-speaking populations did not expand from their ancestralhomeland in a “random walk” but, rather, followed emerging sa-vannah corridors, with rainforest habitats repeatedly imposingtemporal barriers to movement. When populations did movefrom savannah into rainforest, rates of migration were slowed, de-laying the occupation of the rainforest by on average 300 y, com-paredwith similar migratorymovements exclusively within savannahor within rainforest by established rainforest populations. Despiteunmatched abilities to produce innovations culturally, unfamiliar hab-itats significantly alter the route and pace of human dispersals.

human dispersal | phylogeography | phylogenetics | languages | Bantu

Most biological species are confined to areas of the world forwhich their genes have adapted them, but humans, relying

on cultural innovations passed down for generations, have beenable to inhabit nearly every environment on Earth (1). Even so,from our earliest migrations as a species, there is reason to believethat modern humans, despite all of their cultural evolutionarypotential, might have preferred to follow habitats that did notrequire them to master new environments. The so-called “beach-comber” or “coastal routes” hypothesis proposes that the firstmigrations out of Africa might have followed a coastal route viaIndia to the Far East and eventually to Australia (2). Much morerecently, there was a suggestion that during the occupation of thePacific by Austronesian people ∼3,500 y ago (3), there were severalperiods during which the migration paused while people acquiredthe sailing technology to attempt further voyages (4). This tech-nology, in the form of boat designs, might also have been understrong natural selection (5), showing that cultural innovations arenot just a matter of whimsy. East–west migrations might in generalbe more common than north–south movements because the formerare less likely to encounter variation in climate and habitat (6).Bantu migrations swept out of West Central Africa beginning

∼5,000 y ago (B.P.) and eventually moved all the way down to thesouthern tip of the African continent. It was one of the most in-fluential cultural events of its kind, spreading over a vast geo-graphical area a new, more sedentary way of life that wasfundamentally different from that of indigenous forest foragers—ancestral Bantu speakers had mixed-subsistence economies, inwhich farming gradually gained in importance (7–9).Two major events in the recent paleoenvironmental history of

Central Africa might have influenced the route of the Bantu ex-pansion (10–18). The first was a contraction at ∼4,000 B.P. of theCongo rainforest at its periphery, for instance along the coasts ofSouth Cameroon, Gabon, and Congo (11, 16, 19). A second event

at ∼2,500 B.P. affected amongst others the western part of theCongo Basin, creating patches of more or less open forests andwooded or grassland savannahs (14, 15). These areas eventuallymerged into a corridor known as the “Sangha River Interval” thatrepeatedly facilitated the north–south spread of certain typicalsavannah plant and animal species (17, 20–22).The Sangha River Interval may also have been a crucial pas-

sageway for the initial north–south migration of Bantu speechcommunities across the Equator. The archaeological evidence isnot yet detailed enough on its own to test this idea (17). How-ever, the geographical expansion of the Bantu linguistic family,coupled with phylogenetic trees that make use of archaeologicalevidence, provides an opportunity to reconstruct how and whenthis cultural expansion moved through the varying habitats ofWest Central Africa.Here we use a new time-calibrated phylogenetic tree describing

the patterns of descent of ∼400 Bantu languages to study the routeand pace of Bantu speakers as they migrated from their ancestralhomelands. Our data include a dense sampling of languages thatdescend from the early phases of the Bantu expansion, along withfive now-extinct northern Bantu languages and several Bantulanguages spoken in the northeastern Democratic Republic of theCongo (DRC). In combination with information on present-daygeographical positions of the Bantu languages, the phylogenetic

Significance

Humans are uniquely capable of using cultural innovations tooccupy a range of environments, raising the intriguing questionof whether historical human migrations have followed familiarhabitats or moved relatively independently of them. Beginning∼5,000 y ago, savannah-dwelling populations of Bantu-speakingpeoples swept out of West Central Africa, eventually occupyinga vast geographical area. We show that this expansion avoidedunfamiliar rainforest habitats by following savannah corridorsthat emerged from the Congo rainforest, probably from climatechange. When Bantu speakers did move into the rainforest,migration rates were delayed by on average 300 y comparedwith similar movements on the savannah. Despite unmatchedabilities to produce innovations culturally, unfamiliar habitatssignificantly alter the route and pace of human dispersals.

Author contributions: R.G., S.B., K.B., A.M., C.V., and M.P. designed research; R.G., S.B., K.B.,A.M., C.V., and M.P. performed research; R.G., S.B., K.B., A.M., C.V., and M.P. contributednew reagents/analytic tools; R.G., S.B., K.B., A.M., C.V., andM.P. analyzed data; and R.G., S.B.,K.B., A.M., C.V., and M.P. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. P.S.B. is a guest editor invited by the EditorialBoard.

Freely available online through the PNAS open access option.

Data deposition: The Bantu language data and the multistate encoding of the languagedata are available at www.evolution.reading.ac.uk/DataSets.html.1To whom correspondence may be addressed. Email: m.pagel@reading.ac.uk or r.b.grollemund@reading.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1503793112/-/DCSupplemental.

13296–13301 | PNAS | October 27, 2015 | vol. 112 | no. 43 www.pnas.org/cgi/doi/10.1073/pnas.1503793112

tree allows us to infer ancestral migration routes and then testamong proposed scenarios for how Bantu speakers moved throughthe savannah and rainforest habitats of Central, Eastern, andSouthern Africa.

ResultsDated Phylogenetic Tree of the Bantu. We derived a Bayesianposterior sample of n = 100 phylogenetic trees from linguisticdata on 424 Bantu and related languages (Materials and Methodsand SI Materials and Methods). The consensus phylogeny (Fig.1 and Fig. S1) depicts a progressive “backbone” or pectinateradiation from a common ancestor with the out-group Grass-fields languages. This radiating tree occurs in 100% of the treesin the posterior sample (SI Materials and Methods). The tree’sbroad outlines are similar to the tree that Currie et al. (23)report, but where those authors find paraphyletic groups forthe central-western and west-western Bantu, we reconstructmonophyletic groups.On the basis of four calibration ranges supported by archae-

ological studies (Materials and Methods and SI Materials andMethods), the root of the tree estimates a common ancestor withthe outgroup Grassfields speakers at ∼6,900 B.P. (node 0, Fig. 1;age = 6,929.7 ± 418.6 B.P.), a date considerably older than the5,000-B.P. younger limit suggested by our calibration range. Thetree then dates the remaining Bantu in-group (node b) to ∼4,800B.P. (4,846.5 ± 138.1), a time that is near to the older end ofdates suggested by archaeology (node b prior range = 4,000 B.P.to 5,000 B.P.; SI Materials and Methods).

The ∼4,800 B.P. date for node 1 can be compared with theresults from two recent genetic studies on the assumption that thein-group Bantu node coincides with the beginning of the Bantuexpansion. Gignoux et al. (24) report a population expansion of“sub-Saharan” people at ∼4,600 y ago, and Li et al. (25) find ev-idence for a Bantu population expansion at ∼5,600 y ago.

Historical Migration Route.We used information on the latitudinaland longitudinal positions of the languages to reconstruct theprobable ancestral geographical locations of each of the internalnodes of the trees in the posterior sample (Materials and Methodsand SI Materials and Methods). We then used these reconstruct-ions to record the routes of dispersal of Bantu speakers from theirhomeland, and we linked the reconstructed geographical positionat each node to its inferred time, as recorded on the tree, and toinformation from palynological and paleoenvironmental studies(13–15, 26) on the likely habitats at different times in the past.The reconstructions (Fig. 2 A and B) locate the ancestral

homeland of the common ancestors to the Bantu and outgroupGrassfields speakers (node 0, Fig. 1) in the savannah habitat ofNorthwestern Cameroon. The pectinate nature of the tree meansthat the Bantu language groups that descended from the Bantucommon ancestor (node 1, Fig. 1) would themselves become theancestors to the major radiation of the Bantu that eventually oc-cupied large parts of Central, Eastern, and Southern Africa.The principal dispersal route (Fig. 2A) first moves in a south-

easterly direction (approximately nodes 1–8), before traversing ina predominantly easterly direction along the southern boundary ofthe Congo rainforest [this is in contrast to Currie et al. (23), whosereconstructed route moves in alternating south and east steps,crossing the Congo rainforest]. We find no evidence for the sug-gestion (27, 28) that the main migration followed a coastal route(Fig. 2). A few early groups did explore coastal routes (Fig. 1), butthese groups moved in from the east after having branched off themain backbone migration, rather than being ancestral to it.At least three principal southern migrations branched off from

the backbone as it moved east along the southern boundary ofthe rainforest (Fig. 2A), the last of which were the ancestors tomodern-day South African Bantu speakers. This migration routeis consistent with proposals (29–35) that the ancestors of themodern-day Eastern Bantu groups diverged from the WesternBantu ∼2,000 y ago in the Congo region.However, our results reject the suggestion (36–38) that the

Eastern Bantu speakers in the Great Lakes region of East Africatrace their ancestry back to Bantu-speaking peoples who hadmigrated from the northern Congo. Instead, we find that theEastern Bantu are the descendants of people who moved northinto the Great Lakes region from the main backbone (brownlines, Fig. 2A). This result emerges despite the fact that our treeincludes five now-extinct Bantu languages, along with severalcontemporary Bantu languages, all spoken in the northeasternDRC and that have been proposed (39) to have shared a morerecent common ancestor with Eastern Bantu. Our findings arealso consistent with genetic studies (40) that have found a pos-itive correlation between genetic and linguistic distances, whichsuggests that a northern migration route was less probable.Our principal interest is in whether the early Bantu migration

(nodes 1–8 in Fig. 2A) took advantage of changes to the climateand habitat in the western Congo basin that created north–south“corridors” through the core of the Central African rainforest(dashed curve Fig. 2A). Before ∼4,000 B.P. (11, 16, 19), nearlythe entire light- and darker-shaded regions of Fig. 2A were cov-ered by rainforest (SI Materials and Methods and Fig. S2A). Then,palynological and geological data (11, 16, 19, 41) indicated that, byat least 4,000 y ago, climate changes had created encroachingsavannah habitats in the periphery of the rainforest (white andlight green shading, Fig. 2A)—for instance, along the coasts ofGabon and Congo. It is only toward 2,500 B.P. that climate change

JarawanA31-40-50-60

A10-20-30

A40-60-70

B20-A80-90

B20B10-30

C10-20-30-40

C60-70-80B80

B50-60-70

H10-30-42-B40

L50-60-K10-30-R10-20-30L20-30-40D10-20-30-43-55M40-50-60

P30-K21-S10-20-30-40-50-N20-30-40

JE10-20-30-40

JD50-60

M10-20-30F10-20-30

G50-60-N10-P10-20

G10-20-30

E74a-G20-G39

E46-50-60

G40-E70

CENTRAL-WESTERN

WEST-WESTERN

SOUTH-WESTERN

EASTERN

C50-D10

D20-30-C40

L10-20-H21a-H30-40

0100020003000400050006000

1

2

3

4

5

6

78

9

0

NORTH-WESTERNCameroon

NORTH-WESTERNGabon

a

b

c

d

Fig. 1. Consensus time tree of n = 424 Bantu languages, derived from n = 100trees drawn from the Bayesian posterior distribution. Triangles are pro-portional to the number of languages in the group, and the labels are thecodes used by Guthrie (65). Phylogenetic methods and full tree are reported inSI Materials and Methods. The four calibrations used are identified by redletters (a, 5,000 B.P. or older; b, 4,000–5,000 B.P.; c, 3,000–3,500 B.P.; and d,2,500 B.P.; SI Materials and Methods). (Inset) Map of Africa with colored dotsto represent the current location of the languages. Note: The age of the rooton the consensus tree differs from the average root in the posterior sample(text). This is because the ages of nodes on the consensus tree were recon-structed by fitting the phylogenetic model to the fixed consensus tree topol-ogy. All statistics reported in the text are based on the posterior sample, notthe consensus tree.

Grollemund et al. PNAS | October 27, 2015 | vol. 112 | no. 43 | 13297

EVOLU

TION

also led to the development of savannah vegetation in centralparts of the Congo rainforest, yielding corridors such as the San-gha River Interval in the western part of the Congo Basin (SIMaterials and Methods and Fig. S2B), which connected northernand southern savannahs (14, 15, 17).To test the savannah-corridor hypothesis (that the backbone

Bantu migration followed savannah rather than rainforest habi-tats), we reconstructed the ancestral geographical positions ofnodes 0–8 (Fig. 1) for each of the trees in our posterior sample.Then, using dates from the trees along with the paleoclimaticdata (SI Materials and Methods and Fig. S2 A–C), we askedwhether at the time the Bantu speakers are inferred to have beenat those positions, the habitat had changed from rainforest toeither savannah or other nonrainforest habitat. The last of thesenodes (node 8) roughly corresponds to the point at which thesoutheasterly Bantu migration reaches the southern boundary ofthe rainforest, before turning east.We find that in all 100 trees in the posterior sample, the

backbone moves in a southeasterly direction toward the southernboundary of the rainforest (Fig. 2B). A small number of ancestralpositions are reconstructed in a “bulb” of rainforest habitat in thenorthwest, but the majority are not, suggesting that the main mi-gration moved around it (curved arrow). Thus, in n = 96 (96%) ofthe trees, the reconstructed positions of at least 7 of the 9 ancestralnodes miss the rainforest entirely (routes plotted in Fig. 2B): all9 nodes miss the forest in n = 73 of the trees, and at least 8 missthe forest in n = 87 trees, giving an average of 8.53 ± 0.96 of 9 ofthe backbone nodes falling in nonrainforest habitat.It is unlikely that the reconstructed migration route and fit to

the habitat could have arisen by chance: When we simulate mi-grations as random walks from the ancestral homeland, and byusing conservative criteria that favor the random-walks hypothesis(simulation details in SI Materials and Methods), we find that, atmost, 6.3–9.7% of the random-migration routes follow the sa-vannah corridor as closely as the real data (corresponding to 7, 8,or 9 nodes outside the forest; Fig. 2C). Only when we restrict thesimulations to move exclusively in a southeasterly direction do oursimulated routes coincide with the savannah corridor beyond a

negligible level (∼47% of routes; Materials and Methods and SIMaterials and Methods).An intriguing alternative to the proposal that the Bantu fol-

lowed emerging savannah habitats is that they created their mi-gration route by deforesting the Sangha River Interval region (42).However, we think this scenario is unlikely to have played a majorrole in determining the Bantu’s route. The thinning of the rain-forest occurred simultaneously over much of the region fromCameroon to the Congo (42), and it grew from southern as well asnorthern areas (Fig. 2A). This thinning has been linked to climaticchanges, but not to human deforestation (43, 44), suggesting thatif Bantu populations contributed to thinning, it was to a processthat was already underway.There is also no evidence to suggest that the predominantly

north–south movement of the Bantu through the savannah cor-ridor followed or was aided by rivers. Archaeological evidencefrom the Inner Congo Basin (45, 46) suggests movement of Bantucommunities along rivers mostly in a west to east direction andinvolving groups that are not part of the backbone or mainmigration lineage.

Migration Rates Within and Between Savannah and Rainforest Habitats.The tree, along with the dates and palynological and paleoenvir-onmental information, can be used to identify “habitat transitions,”defined as instances in which the geographical position and datereconstructed at the beginning of a branch on the tree implies adifferent habitat from the one implied by the geographical positionand date at the end of the branch.Across trees, we found an average of 52.7 ± 4.4 independent

habitat transitions, with 35.8 ± 3.4 corresponding to transitionsfrom savannah into forest and 16.9 ± 2.5 from forest back tosavannah: We say “back” to savannah because most rainforest-dwelling Bantu speech communities have an ancestral history ofresiding in savannah. The consensus tree records 48 transitionsbetween habitats, 31 corresponding to transitions from savannahto rainforest, and 17 from rainforest back to savannah (Fig. 3).The remaining branches record movement within the same habitat,either forest or savannah.

?

9

?

8

765

0020000222221

34

A B

C

Fig. 2. Ancestral migration route reconstruction. (A) Ancestral migration route reconstructed on consensus tree by using geographical locations of con-temporary languages and connecting ancestral locations by straight lines (true route will differ). Numbered positions correspond to nodes on the consensustree (Fig. 1). Curved dashed line indicates suggested migration route through savannah corridors (B). Lighter green shading corresponds to the delimitation ofthe rainforest at 5,000 B.P.; the darker green corresponds to the delimitation of the rainforest at 2,500 B.P. (text and SI Materials and Methods). (B) Mapshowing the ancestral locations of the backbone nodes (Fig. 1) for the 100 trees in the Bayesian posterior sample; curved arrow is suggested route for theearly migration based on a small number of reconstructed points that fall in rainforest. (C) Same as B but showing the ancestral locations of randommigrationroutes for nodes 0–8 (text and SI Materials and Methods).

13298 | www.pnas.org/cgi/doi/10.1073/pnas.1503793112 Grollemund et al.

On average, Bantu speaking groups that moved into the rain-forest (F) were significantly delayed, taking on average ∼300 ylonger than comparable transitions within savannah (S) habitats(Fig. 4). This significant delay is observed separately in at least 90%of the trees in the posterior sample and is not an artifact of S->Ftransitions covering a greater distance: Our analyses control for thedistance moved, implying that S->F transitions proceed at an ab-solutely slower pace. We think it is unlikely that the slower S->Ftransitions could arise from a higher extinction rate of groups thatattempted this transition: Even if there were higher extinction rates,because the analyses control for the distance moved, the finding ofa slower rate of movement of successful transitions stands.By comparison, transitions from the rainforest back to sa-

vannah take no longer on average than movements within eitherrainforest or savannah (not significant in any tree; Fig. 4). Thisfinding might suggest that the savannah is an easier habitat tooccupy or, more interestingly, that the rainforest-dwelling Bantucultures in our tree tend to descend from ancestrally savannah-dwelling cultures and retained some cultural knowledge of howto exploit the savannah environment.

DiscussionTogether, our results show that the Bantu expansion was char-acterized by a measureable preference for following familiar sa-vannah habitats as it moved from present-day northwest Cameroon

in a southeasterly direction, taking advantage of a savannah cor-ridor that began to appear by ∼4,000 y ago. This route avoidedrainforest habitats and spawned numerous migratory branches thatled to the occupation of nearly all of southern Africa, along withseveral independent movements north into the Great Lakes regionof East Africa.When savannah-dwelling Bantu-speaking groups did move into

the rainforest, their rate of migration was significantly slowed. Onits own, this result might not be surprising—the rainforest iscovered with dense vegetation that might have made subsistence(and especially farming) more difficult, and rainforest habitatsmight harbor more predators and organisms causing infectiousdisease. What is surprising, however, and relevant to the questionof human cultural innovation, is the extent to which the rainforestslowed human movement. Vansina (47) has written that “[Bantu]Farmers took some 2000 y to settle the rainforests of equatorialAfrica, and then, about another half millennium to absorb newtechnologies and to become finely attuned to all of the potential oftheir habitats.” Our phylogenetic reconstructions, showing thattransitions into the rainforest were delayed by ∼300 y comparedwith movements of a similar distance within savannah habitats, arein good agreement with Vansina’s observations and correspond toa 50% reduction in the pace of human expansion.Could transitions into the rainforest really delay movements

by hundreds of years? Our results curiously seem to fit withmodern studies that suggest that human innovation has less to dowith thinking hard until the right solution comes to mind (thelightbulb switching on in our minds), than with the slow accu-mulation of knowledge and technology principally resulting from“trial and error.” Thus, Basalla (48) and Arthur (49) both em-phasize the cumulative nature of human innovations, downplayingthe role of “genius” innovators. For instance, Henry Ford’s fa-mous assembly line production drew on earlier experiments withstreamlining assembly lines, and Watt’s steam engine was less ofan “out of thin air” invention than a development of Newcombe’searlier engine. Thomas Edison is often credited with “inventingthe light bulb,” but records show that his patent was for a betterfilament to a lightbulb, and his notebooks reveal that he triedthousands of filament materials before alighting by chance on hisfavored material. The typically low population densities of sub-sistence peoples such as early Bantu speakers would only haveexaggerated the difficulties of accumulating new technologies (50).Our approach shows that evidence bearing on subtle questions

of human history can be investigated by using phylogenies de-rived from languages, combined with relevant information oncontemporary cultures and appropriate statistical modeling. In-deed, there is reason to believe that language phylogenies mighteven be preferable to gene-based trees in this regard (51).Languages typically evolve at a higher rate than genes, meaningthat they can resolve shorter time scales, but languages mighthave an even more fundamental role. Languages track the in-heritance of culture, and it is this inheritance that is normallypertinent to questions of human cultural evolution. Genes, bycomparison, can readily move among cultures, without neces-sarily taking their cultures with them.

Savannah -> RainforestRainforest -> Savannah

Fig. 3. Consensus time tree with panels that enlarge the clades that havesavannah to rainforest (n = 31 independent transitions) and rainforest tosavannah (n = 17 independent transitions). Numbers of each kind of tran-sition vary in the posterior sample (text). Both kinds of transition are widelydistributed among the clades near to the rainforest, and S->F transitions arealways ancestral to F->S transitions. Some lineages have experienced threetransitions in their history.

hannavaShannavaS tseroFtseroF 300 400 500 600 700 800300 400 500 600 700 800300 400 500 600 700 800

Fig. 4. Posterior distribution of times taken for four different habitat transitions, controlling for distance moved. Savannah to forest transitions are significantlyslowed (Tukey honest significant difference test; P < 0.05) compared with transitions within savannah in 90 of 100 trees in the posterior sample. Rainforest tosavannah transitions take no longer on average than movements within either rainforest or savannah (not significant in any tree). Mean in years ± SD: S->S =368.6 ± 13.9; S->F = 662.8 ± 78.7; F->S = 446.0 ± 64.7; F->F = 420.6 ± 24.2. All significance tests were performed on log-transformed data to normalize variances.

Grollemund et al. PNAS | October 27, 2015 | vol. 112 | no. 43 | 13299

EVOLU

TION

Materials and MethodsLinguistic Data. We collected lexical data from published sources and fromfieldwork for 409 Bantu and 15 Bantoid languages. Our wordlist is a modifiedversion of the Atlas Linguistique du GABon (52). This list comprises 159 wordsfrom which we have sampled 100 words that are the best documented forthe languages we studied (Materials and Methods and SI Materials andMethods). We then classified the words into cognate sets and built a binary-coded dataset (each column identifies a unique cognate class), yielding 3,859cognate classes for the 424 languages.

Phylogenetic-Statistical Methodology. We inferred a time-dated phylogenyfrom the lexical dataset using a variable-rates molecular clock model that allowsthe rate of evolution to vary among branches of the tree. The variable-rateclock ismodeled by applying a scalar multiplier to each branch of the tree thatalters the rates by some fixed amount (53). We assume these scalars are drawnfrom a log-normal prior distribution with μ = 1 and unknown σ2 that we es-timate from the data. Node ages were estimated by using a Yule process (54).

Trees were inferred using Markov chain Monte Carlo methods (55) thatimplemented Tuffley and Steel’s covarion model (56), which allows the rateof evolution to jump between an “on” and an “off” state throughout thetree (model testing and selection is detailed in SI Materials and Methods andTable S1). The covarion model is well suited to binary-coded cognate data,owing to the fact that each cognate class ideally arises just once on the tree. Thevariable rates and covarion models were implemented in our BayesPhylogeniessoftware (57). The ladderised or pectinate phylogeny of Fig. 1 is robust tosubsampling of the n = 100 words (SI Materials and Methods).

Chains were run for 3 × 108 iterations, with a sampling period of 10,000iterations. We used the Tiv and the Grassfields languages as out-groups toroot the tree.

Calibration Ranges.Weused archaeological data to propose date ranges, and inone case a fixed date, for four nodes of our tree (labeled a–d in Fig. 1). The fourcalibrations are as follows: (a) 5,000 B.P. or older for Bantoid, non-Bantu (58);(b) 4,000–5,000 B.P. for Narrow Bantu (13, 14, 16, 44, 59, 60); (c) 3,000–3,500 B.P.for the Mbam-Bubi ancestor (61); and (d) 2,500 B.P. for Eastern Bantu (62). Weused a uniform prior in our Bayesian tree inference for all calibration ranges.

Geographical Data. We recorded the latitude and longitude of the approxi-mate centroid of each of our languages (Dataset S1), using data provided byBastin et al. (38) and fieldwork studies.

Ancestral Reconstructions. We inferred ancestral latitude and longitude foreach node of our tree using a Brownian motion model applied to the con-temporary data that allowed for rates of geographical movement to varythroughout the tree, following methodology we have reported elsewhere(63) and as implemented in our BayesTraits software.

Simulated Migrations of Savannah Corridor Route. We generated randomdispersal routes from the Bantu homeland for the nine nodes (nodes 0–8 ofFig. 1) corresponding to the southeasterly movement through the savannahcorridor and out into the savannah south of the Congo rainforest. We heldconstant the consensus phylogenetic tree and the timings at its nodes, so asnot to introduce a large and unknown additional source of possible geo-graphical movements. Simulated routes were allowed to go to places thatBantu have actually inhabited historically or at present. Moves into the seaor other bodies of water were prohibited, and a newly simulated positionwas not allowed to occupy a space already occupied (defined as within10 km of any previously simulated point, unless the distance to be traveledwas less than this).

These constraints narrowed the space of possiblemigration routes, makingit more likely the simulated routes would coincide with the savannah cor-ridor. We then simulated two dispersal scenarios. In the first, the distancesmoved along the backbone on the tree followed those actually observedalong the same branches but in a randomorder; in the second, these distanceswere drawn from a random distribution but normalized to have the sametotal distance moved as observed in the real data. The first yielded 9.7% ofroutes with seven or more nodes falling in the savannah corridor, and thesecond returned 6.3% using the same criterion. Only when we constrain thesimulations to move exclusively in a southeasterly direction do our simulatedroutes coincide with the savannah corridor beyond a negligible level—∼47%of route falls in savannah corridor.

ACKNOWLEDGMENTS. We thank Gérard Philippson for the data on EasternBantu languages, Jean-Marie Hombert for the Grassfields languages, theKongoKing research group (leader KB) for the data on the Kikongo languages(H), and Jean-Pierre Donzo and Guy Kouarata for data on C languages spokenin Congo and DRC. This work was supported by European Research CouncilAdvanced Investigator Award 268744 (Mother Tongue; to M.P.). K.B. was sup-ported by European Research Council Starting Grant No. 284126 (KongoKing)and by the Special Research Fund of Ghent University.

1. Pagel M (2012) Wired for Culture: Origins of the Human Social Mind (Penguin, NewYork), p 432.

2. Mellars P (2006) Going east: New genetic and archaeological perspectives on themodern human colonization of Eurasia. Science 313(5788):796–800.

3. Hung H-c, et al. (2011) The first settlement of Remote Oceania: The Philippines to theMarianas. Antiquity 85(329):909–926.

4. Gray RD, Drummond AJ, Greenhill SJ (2009) Language phylogenies reveal expansionpulses and pauses in Pacific settlement. Science 323(5913):479–483.

5. Rogers DS, Ehrlich PR (2008) Natural selection and cultural rates of change. Proc NatlAcad Sci USA 105(9):3416–3420.

6. Diamond J (1997) Guns, Germs, and Steel: The Fates of Human Societies (Norton, NewYork), p 480.

7. Diamond J, Bellwood P (2003) Farmers and their languages: The first expansions.Science 300(5619):597–603.

8. Russell T, Silva F, Steele J (2014) Modelling the spread of farming in the Bantu-speakingregions of Africa: An archaeology-based phylogeography. PLoS One 9(1):e87854.

9. Neumann K, et al. (2012) First farmers in the Central African rainforest: A view fromsouthern Cameroon. Quat Int 249(6):53–62.

10. Schwartz D (1992) Assèchement climatique vers 3000 BP et expansion bantu enAfrique Centrale atlantique: Quelques réflexions. Bull Soc Geol Fr 163(3):353–361.

11. Vincens A, et al. (1994) Changement majeur de la végétation du lac Sinnda (vallée duNiari, Sud-Congo) consécutif à l’assèchement climatique holocène supérieur: Apportde la palynologie. Comptes Rendus à l’Academie des Sciences de Paris 318:1521–1526.

12. Vincens A, Buchet G, Servant M (2010) Vegetation response to the” African HumidPeriod” termination in Central Cameroon (7° N)–new pollen insight from Lake Mbalang.Clim Past 6(3):281–294.

13. Maley J, Brenac P (1998) Vegetation dynamics, palaeoenvironments and climaticchanges in the forests of western Cameroon during the last 28,000 years BP. RevPalaeobot Palynol 99(2):157–187.

14. Maley J (2001) La destruction catastrophique des forêts d’Afrique centrale survenue ilya environ 2500 ans exerce encore une influence majeure sur la répartition actuelledes formations végétales. Syst Geogr Plants 71:777–796.

15. Maley J, Willis K (2010) Did a savanna corridor open up across the Central Africanforests 2500 years ago? CoForChange. Available at www.coforchange.eu/products/notes/did_a_savanna_corridor_open_up_across_the_central_african_forests_2500_years_ago.Accessed 18 August 2015.

16. Ngomanda A, et al. (2009) Western equatorial African forest-savanna mosaics: Alegacy of late Holocene climatic change. Clim Past 5(4):647–659.

17. Bostoen K, et al. (2015) Middle to Late Holocene Palaeoclimatic change and the earlyBantu expansion in the rain forests of West Central-Africa. Curr Anthropol 56(3):354–384.

18. Klieman KA (2003) The Pygmies were our compass: Bantu and Batwa in the History ofWest Central Africa, Early Times to c. 1900 C.E. (Heinemann, Portsmouth, NH), p 253.

19. Vincens A, Schwartz D, Bertaux J, Elenga H, de Namur C (1998) Late Holocene climaticchanges in western equatorial Africa inferred from pollen from Lake Sinnda, south-ern Congo. Quat Res 50(1):34–45.

20. White F (1979) The Guineo-Congolian Region and its relationships to other phy-tochoria. Bulletin Jardin Botanique National Belgique/Bulletin Nationale PlantentuinBelgie 49(1/2):11–55.

21. Gond V, et al. (2013) Vegetation structure and greenness in Central Africa fromModismulti-temporal data. Philos Trans R Soc Lond B Biol Sci 368(1625):20120309.

22. Doumenge C, Gonmadje C, Doucet J-L, Maley J (2014) Intervalle de la Sangha: Mytheou réalité biogéographique? (Résumés, Stellenbosch, Afrique du Sud).

23. Currie TE, Meade A, Guillon M, Mace R (2013) Cultural phylogeography of the BantuLanguages of sub-Saharan Africa. Proc R Soc 280(1762):20130695.

24. Gignoux CR, Henn BM, Mountain JL (2011) Rapid, global demographic expansionsafter the origins of agriculture. Proc Natl Acad Sci USA 108(15):6044–6049.

25. Li S, Schlebusch C, Jakobsson M (2014) Genetic variation reveals large-scale pop-ulation expansion and migration during the expansion of Bantu-speaking peoples.Proc R Soc 281(1793):20141448.

26. Adams J, Faure H, eds (1997) Review and Atlas of Palaeovegetation: Preliminary LandEcosystem Maps of the World Since the Last Glacial Maximum (Oak Ridge NationalLaboratory, Oak Ridge, TN).

27. Phillipson DW (1977) The Later Prehistory of Eastern and Southern Africa (Africana,New York), p 323.

28. Blench R (2012) Two vanished African maritime traditions and a parallel from SouthAmerica. Afr Archaeol Rev 29(2/3):273–292.

29. Heine B, Hoff H, Vossen R (1977) Neuere Ergebnisse zur Territorialgeschichte derBantu. Zur Sprachgeschichte und Ethnohistorie in Afrika, eds Möhlig WJG, Rottland F,Heine B (Dietrich Reimer, Berlin), pp 57–72.

30. Henrici A (1973) Numerical classification of Bantu languages. African Language Studies14:82–104.

13300 | www.pnas.org/cgi/doi/10.1073/pnas.1503793112 Grollemund et al.

31. Heine B (1973) Zur genetischen Gliederung der Bantu-Sprachen. Afr Uebersee 56(3):164–185.

32. Ehret C (1999) Subclassifying Bantu: The evidence of stem morpheme innovation.Bantu Historical Linguistics. Theoretical and Empirical Perspectives, eds Hombert JM,Hyman LM (CSLI, Stanford, CA), pp 43–147.

33. Ehret C (2001) Bantu expansions: Re-envisioning a central problem of early Africanhistory. Int J Afr Hist Stud 34(1):5–41.

34. Holden CJ (2002) Bantu language trees reflect the spread of farming across sub-Saharan Africa: A maximum-parsimony analysis. Proc Biol Sci 269(1493):793–799.

35. Holden CJ, Gray RD (2006) Rapid radiation, borrowing and dialect continua in theBantu languages. Phylogenetic Methods and the Prehistory of Languages, edsForster P, Renfrew C (MacDonald Institute for Archaeological Research, Cambridge,U.K.), pp 19–31.

36. Möhlig WJG (1981) Stratification in the history of the Bantu languages. Sprache undGeschichte in Afrika 3:251–317.

37. Bastin Y, Coupez A, de Halleux B (1983) Classification lexicostatistique des languesbantoues (214 relevés). Bull Seances Acad R Sci Outre Mer 27(2):173–199.

38. Bastin Y, Coupez A, Mann M (1999) Continuity and Divergence in the BantuLanguages: Perspectives from a Lexicostatistic Study (Musée Royal de l’Afrique Cen-trale, Tervuren, Belgium), p 225.

39. Vansina J (1995) New linguistic evidence and ‘the Bantu expansion’. J Afr Hist 36(2):173–195.

40. de Filippo C, Bostoen K, Stoneking M, Pakendorf B (2012) Bringing together linguisticand genetic evidence to test the Bantu expansion. Proc R Soc B Biol Sci 279(1741):3256–3263.

41. Desjardins T, et al. (2013) d13C variation of soil organic matter as an indicator ofvegetation change during the Holocene in central Cameroon. C R Geosci 345:266–271.

42. Bayon G, et al. (2012) Intensifying weathering and land use in Iron Age Central Africa.Science 335(6073):1219–1222.

43. Maley J, Giresse P, Doumenge C, Favier C (2012) Comment on “Intensifying weath-ering and land use in Iron Age Central Africa”. Science 337(6098):1040, author reply1040.

44. Neumann K, et al. (2012) Comment on “Intensifying weathering and land use in IronAge Central Africa”. Science 337(6098):1040, author reply 1040.

45. Eggert MKH (1992) The Central African rain forest: Historical speculation and ar-chaeological facts. World Archaeol 24(1):1–24.

46. Wotzka H-P (1995) Studien zur Archäologie des Zentralafrikanischen Regenwaldes:Die Keramik des Inneren Zaïre-Beckens und Ihre Stellung im Kontext der Bantu-Expansion (Heinrich-Barth-Institut, Köln, Germany).

47. Vansina J (1990) Paths in the Rainforests (Univ of Wisconsin Press, Madison, WI), p 428.48. Basalla G (1988) The Evolution of Technology (Cambridge Univ Press, Cambridge, U.K.).49. Arthur WB (2009) The Nature of Technology: What It Is and How It Evolves (Penguin,

London).50. Powell A, Shennan S, Thomas MG (2009) Late Pleistocene demography and the ap-

pearance of modern human behavior. Science 324(5932):1298–1301.51. Pagel M (2009) Human language as a culturally transmitted replicator. Nat Rev Genet

10(6):405–415.52. Hombert J-M, Van der Veen L, Medjo Mve P (2011) ALGAB, Atlas Linguistique du

GABon (Laboratoire Dynamique du Langage, Lyon). Available at www.ddl.ish-lyon.cnrs.fr/equipes/index.asp?Langue=FR&Equipe=8&Page=Action&ActionNum=48. Ac-cessed November 10, 2014.

53. Drummond AJ, Ho SY, Phillips MJ, Rambaut A (2006) Relaxed phylogenetics anddating with confidence. PLoS Biol 4(5):e88.

54. Yule GU (1925) A mathematical theory of evolution, based on the conclusions ofDr. JC Willis, FRS. Philos Trans R Soc Lond, B Contain Pap Biol Character 213:21–87.

55. Larget B, Simon DL (1999) Markov chain Monte Carlo algorithms for the Bayesiananalysis of phylogenetic trees. Mol Biol Evol 16:750–759.

56. Tuffley C, Steel M (1998) Modeling the covarion hypothesis of nucleotide substitution.Math Biosci 147(1):63–91.

57. Pagel M, Meade A, Barker D (2004) Bayesian estimation of ancestral character stateson phylogenies. Syst Biol 53(5):673–684.

58. Lavachery P (2001) The Holocene archaeological sequence of Shum Laka Rock Shelter(Grassfields, Western Cameroon). Afr Archaeol Rev 18(4):213–247.

59. Oslisly R (2001) The history of human settlement in the middle Ogooué valley (Gabon).African Rain Forest Ecology and Conservation, eds Weber W, White LJT, Vedder A,Naughton-Treves L (Yale Univ Press, New Haven), pp 101–118.

60. Ngomanda A, et al. (2005) Vegetation changes during the past 1300 years in westernequatorial Africa: A high-resolution pollen record from Lake Kamalété, Lopé Reserve,Central Gabon. Holocene 15(7):1021–1031.

61. Lavachery P (2003) A la lisière de la forêt. Peuplements Anciens et Actuels des ForêtsTropicales, eds Froment A, Guffroy J (IRD Editions, Paris), pp 89–102.

62. Ashley CZ (2010) Towards a socialised archaeology of ceramics in Great Lakes Africa.Afr Archaeol Rev 27(2):135–163.

63. Venditti C, Pagel M (2010) Speciation as an active force in promoting genetic evo-lution. Trends Ecol Evol 25(1):14–20.

64. Guthrie M (1948) The Classification of the Bantu Languages (Oxford Univ Press for theInternational African Institute, London), p 91.

65. Guthrie M (1967–1971) Comparative Bantu: An Introduction to the Comparative Lin-guistics and Prehistory of the Bantu Languages (Gregg International, London), 4 vols.

66. Grollemund R (2012) Nouvelles approches en classification: Application aux languesbantu du Nord-Ouest. Thesis (Université Lumière Lyon 2, Lyon).

67. Grollemund R, Branford S, Hombert J-M, Watters JR (2015) Towards a new phylo-genetic classification of Bantoid languages. Towards Proto-Niger-Congo: Comparisonand Reconstruction, ed Watters JR (Language Science Press, Berlin).

68. Yang Z (1996) Among-site rate variation and its impact on phylogenetic analyses.Trends Ecol Evol 11(9):367–372.

69. Pagel M, Meade A (2006) Estimating rates of lexical replacement on phylogenetictrees of languages. Phylogenetic Methods and the Prehistory of Languages, edsForster P, Renfrew C (McDonald Institute Monographs, Cambridge, U.K.), pp 173–182.

70. Cornelissen E (2003) On microlithic quartz industries at the end of the Pleistocene inCentral Africa: The evidence from Shum Laka (NW Cameroon). Afr Archaeol Rev20(1):1–24.

71. Lavachery P (2001) The Holocene archaeological sequence of Shum Laka rock shelter(Grassfields, Cameroon). Afr Archaeol Rev 18:213–247.

72. Asombang R, De Maret P (1992) Re-investigating Shum Laka (Cameroun). Nsi 10/11:13–16.

73. de Maret P, et al. (1993) Preliminary results of the 1991-1992 field season at ShumLake, Northwestern province, Cameroon. Nyame Akuma 39:13–15.

74. de Maret P, Asombang R, Cornelissen E, Lavachery P, Moeyersons J (1995) Continuingresearch at Shum Laka rock shelter, Cameroon (1993-1994 field season). Nyame Akuma43:2–3.

75. de Maret P, Clist B, Van Neer W (1987) Résultats des premières fouilles dans les abris-sous-roche de Shum Laka et Abeke au Nord-Ouest du Cameroun. Anthropologie91(2):559–584.

76. Lavachery P (1998) De la pierre au métal: Archéologie des dépôts holocènes de l’abrisous roche de Shum Laka (Cameroun). PhD thesis (Université Libre de Bruxelles,Bruxelles).

77. Lavachery P (2003) A la lisière de la forêt: 10.000 ans d’interactions entre l’homme etl’environnement dans les Grassfields (Cameroun). Peuplements Anciens et Actuels de desForêts Tropicales, Actes du Séminaire-Atelier Orléans 15-16 Octobre 1998, CollectionColloques et Séminaires, eds Froment A, Guffroy J (IRD Editions, Paris), pp 89–102.

78. Lavachery P, Cornelissen E (2000) Natural and cultural spatial patterning in the LateHolocene deposits of Shum Laka Rock Shelter, Cameroon. J Field Archaeol 27(2):153–158.

79. Ribot I, Orban R, de Maret P (2001) The Prehistoric Burials of Shum Laka Rockshelter(NorthWest Cameroon) (Royal Museum for Central Africa, Tervuren, Belgium).

80. de Maret P (1994-1995) Pits, pots and the Far West streams. Azania 29–30:318–323.81. Hassan FA (1996) Abrupt Holocene climatic events in Africa. Aspects of African

Archaeology, eds Pwiti G, Soper R (Univ of Zimbabwe, Harare, Zimbabwe), pp 83–89.82. de Maret P (1992) Sédentarisation, agriculture et métallurgie du Sud-Cameroun:

Synthèse des recherches depuis 1978. L’archéologie au Cameroun. Actes du premierColloque international de Yaoundé, 6-9 janvier 1986, ed Essomba JM (Karthala, Paris),pp 247–262.

83. de Maret P (1982) New survey of archaeological research and dates for West-Centraland North-Central Africa. J Afr Hist 23:1–15.

84. de Maret P, Clist B, Mbida C (1983) Belgian archaeologial mission in Cameroon, 1983field season. Nyame Akuma 23(5–6):5–6.

85. Clist B (1987) A critical reappraisal of the chronological framework of the early UreweIron Age industry. Muntu 6:35–62.

86. Phillipson DW (1993) African archaeology (Cambridge Univ Press, Cambridge, U.K.), p 389.87. Organ CL, Shedlock AM, Meade A, Pagel M, Edwards SV (2007) Origin of avian ge-

nome size and structure in non-avian dinosaurs. Nature 446(7132):180–184.88. Lemey P, Rambaut A, Welch JJ, Suchard MA (2010) Phylogeography takes a relaxed

random walk in continuous space and time. Mol Biol Evol 27(8):1877–1885.89. Bouckaert R, et al. (2012) Mapping the origins and expansion of the Indo-European

language family. Science 337(6097):957–960.90. Anhuf D, Frankenberg P (2000) Die mittelholozäne Feuchtphase 5000 BP–Eine Veg-

etationsrekonstruktion für Afrika. Regensburger Geographische Arbeiten 33:99–126.91. Maley J (2002) A catastrophic destruction of African forests about 2,500 years ago still

exerts a major influence on present vegetation formations. IDS Bull 33(1):13–30.92. White F (1983) The Guineo-Congolian Regional Centre of Endemism. The Vegetation of

African: A descriptive Memoir to Accompany the UNESCO/AETFAT/UNSO VegetationMap of Africa (UNESCO, Paris), p 356.

93. White F (1986) La Vegetation de L’Afrique (The Vegetation of Africa) (ORSTOM –UNESCO,Paris), p 384.

Grollemund et al. PNAS | October 27, 2015 | vol. 112 | no. 43 | 13301

EVOLU

TION