Research ArticleANTHROPOLOGY

Population collapse in Congo rainforest from 400 CE urges reassessment of the Bantu Expansion

See allHide authors and affiliations

Science Advances  12 Feb 2021:
Vol. 7, no. 7, eabd8352
DOI: 10.1126/sciadv.abd8352
  • Fig. 1 Geographic distribution of pottery finds in the Congo rainforest and adjacent woodland areas of Central Africa.

    Each circle represents a site with an assemblage of well-described pottery groups (n = 472), with circle opacity proportional to the number of pottery groups per site. Each small solid dot represents an additional dated assemblage of unclassified pottery, remains of metallurgy, or charred fruit remains with a reliable archeological contextualization (class I; n = 254). Colors and associated capital letters (A to K) demarcate the different study regions, as defined in table S2 and Materials and Methods. Regions A to H are located in the Congo rainforest (light gray background), regions I to K in adjacent woodland areas within the Congo basin, and on Bioko Island (white background). Dark gray surfaces and lines represent the hydrographic network (lakes and rivers); thin black lines are country borders.

  • Fig. 2 Temporal variation in the activity of ancient pottery-producing communities in the Congo basin over the past 4000 years.

    Patterns are shown separately for the Congo rainforest (A) and the adjacent woodland areas including Bioko Island (B). Activity is based on the SPD (full black lines) of all class I archeological 14C dates (table S1), smoothed using a 60-year moving average [(A) n = 1075 dates, 675 bins; (B) n = 74 dates, 45 bins]. Gray background shading represents the 95% uncertainty envelope of summed probability in a logistic model of hypothetical population growth drawn from the same 14C datasets; see fig. S3 for uncertainty envelopes of alternative (uniform, linear, and exponential) growth models. Color shading demarcates periods of more or less intense human activity, defined as time windows during which the observed SPD surpasses (blue) or falls below (red) one (light shading) or multiple (dark shading) growth models. Phases of inferred population expansion and collapse are defined as time intervals during which the rate of SPD change exceeds the first quartile of all rates of change. The thin black lines represent a composite kernel density estimate (CKDE) of 500 sets of 675 randomly sampled calibrated 14C dates, smoothed using a 30-year kernel bandwidth.

  • Fig. 3 Evolution of the numerical abundance and geographical distribution of pottery styles in the Congo rainforest over the past 3000 years.

    Abundance (A) is quantified as the number of pottery groups recorded within each century bin; spatial distribution is quantified as the number of sites where each pottery group is found (B) and by its total area of distribution [(C) in units of 1000 km2]. Background color shading demarcates periods of high (blue) or low (red) human activity, as in Fig. 2. The evolution of pottery styles confirms the temporal fluctuation in human activity as reflected in the cross-regional SPD of archeological 14C dates (Fig. 2) and unveils a two-phase settlement pattern during both the Early and Late Iron Age periods. Each starts with a phase of population expansion (A) during which stylistically homogeneous pottery groups became widely distributed (B and C) and end with a phase of high activity (A) characterized by increasing abundance of local pottery styles reflecting a process of regionalization (B and C). Maps in (D) illustrate these phases of homogeneity and regionalization for major pottery groups in the well-studied inner Congo basin (region F). Detailed temporal occurrences of each pottery group are shown in fig. S1.

  • Fig. 4 Variation in effective population size (Ne) of Bantu-speaking agriculturalist communities in Gabon (region B) over the last 130 generations.

    Ne trajectories [colored lines in (A)] are based on identity-by-descent (IBDNe) analysis of genome-wide single-nucleotide polymorphism (SNP) data from a total of 816 individuals (5) in 16 present-day communities throughout Gabon (B) belonging to North-Western (green) and West-Western (brown) language groups (table S3). Background shading shows the periods of more (blue) and less (red) intense human activity across the Congo rainforest, as in Fig. 2. The lower x axis shows the corresponding timeline in calendar years BCE/CE, based on a generation time of 30 years (78).

Supplementary Materials

Stay Connected to Science Advances

Navigate This Article