Isotopic and microbotanical insights into Iron Age agricultural reliance in the Central African rainforest

The study investigates how and to what extent incoming cereal crops (C4 plants such as pearl millet and sorghum) were adopted by populations living within the Central African rainforest during the Iron Age. Prevailing models link agriculture’s emergence to Bantu-speaking population movements and to a hypothesized ‘rainforest crisis’ that purportedly opened forest canopies. Yet, direct dietary evidence from the Congo Basin has been scarce. The authors aim to test assumptions about agricultural adoption by directly reconstructing human diets across multiple sites and time periods, evaluating the relative importance of C4 crops versus C3 forest and freshwater resources. This work is important because it refines our understanding of agricultural adaptation in tropical rainforests, challenges broad-brush narratives of the Bantu expansion, and informs on long-term subsistence mosaics in a region critical for biodiversity and the global carbon cycle.

Background research highlights: (1) decades of debate on the timing and processes of Bantu-speaking community dispersals and their links to farming; (2) experimental and archaeobotanical evidence that pearl millet could be (and was) cultivated within areas now under rainforest, questioning the necessity of large-scale deforestation; (3) pollen and sediment records that suggest environmental change around 2.5 ka, but with uncertainty over the extent and drivers; (4) limited direct dietary studies in Central Africa compared to other African regions; (5) evidence for complex, time-transgressive settlement histories across the Congo Basin with interactions among diverse populations. Together, prior work motivates direct, context-specific dietary analyses to evaluate the degree of reliance on C4 crops versus persistent use of forest and aquatic resources.

Study area and materials: Four sites in the Democratic Republic of Congo representing early to late Iron Age contexts and diverse ecologies were analyzed: Imbonga (IMB), Longa (LON), Bolondo (BLD), and Matangai Turu Northwest (MTNW). IMB and LON lie on tributaries of the Congo River in western DRC; BLD is on the Tshuapa floodplain; MTNW is a rockshelter in the Ituri rainforest (northeast DRC). Chronology spans ~2050 BP to ~130 BP; MTNW individual dates to 813 ± 35 BP.

• Human remains: collagen δ13C and δ15N from BLD (n=11 carried forward; 2 excluded for poor preservation) and LON (n=1). Human enamel δ13C and δ18O from IMB (n=1), LON (n=1), BLD (n=11), and MTNW (n=1; multiple molars to track childhood diet).
• Faunal remains: BLD bone collagen (n=10) including goats, duiker/antelope, dog, fox-sized carnivore, fish, crocodile; BLD enamel (n=6) to establish ecological baselines.
• Charred food residues: BLD charred fragments (n=8) analyzed for δ13C and δ15N.
• Dental calculus microparticles: MTNW individual—three mandibular molars processed, retrieving starch granules (n=38) and phytoliths (n=9).

Laboratory protocols and quality control:

• Collagen extraction followed a modified Longin protocol; preservation criteria included atomic C/N 2.9–3.6, %C ~15–48%, %N ~5–17%. δ13C (VPDB) and δ15N (AIR) measured via EA-IRMS with long-term precision ±0.2‰ for both.
• Enamel pretreatment with NaClO and acetic acid; carbonate δ13C and δ18O measured using GasBench-IRMS, precision ±0.1‰ (δ13C) and ±0.2‰ (δ18O). Enamel integrates total diet (carbohydrates, fats, proteins) and records specific childhood windows (M1: −0.2 to 4 years; M2: 4–7 years; M3: 9–16 years).
• Charred food fragments analyzed on CF-IRMS; δ13C uncertainty ±0.13‰; δ15N total analytical uncertainty ±0.61%.
• Dental calculus decontaminated (2% NaOH), decalcified in cleanroom; starch granules identified morphometrically against sub-Saharan reference collections; phytoliths characterized.
• AMS radiocarbon dating performed at SUERC; calibration with OxCal and IntCal13.

Data interpretation framework:

• δ13C distinguishes C3 (forest plants and their consumers) versus C4 (millet, sorghum, later maize) contributions, with canopy effect considered for C3 values in closed forests.
• δ15N informs trophic position and aquatic (freshwater) resource consumption.
• δ18O reflects water sources and environmental context. Comparison across tissues (collagen vs enamel) allows assessment of protein bias and potential life-stage dietary shifts.

• Faunal baselines (BLD): Mammals show largely C3-based diets (collagen δ13C −23.7 to −18.2‰). Herbivores (duiker, small antelope) show δ15N 5.9–7.3‰; carnivores (dog 12.3‰; fox-sized carnivore 13.7‰) higher δ15N; aquatic species elevated δ15N (crocodiles 11.1–11.4‰; bichir 12.7‰; catfish 9.4‰), highlighting the need to consider both δ13C and δ15N to differentiate freshwater vs forest resources.
• Human bone collagen (BLD n=11; LON n=1): δ13C −21.0 to −16.3‰, consistent overall with C3 and freshwater reliance. Mean human δ15N at BLD ~14.5‰ (range 13.4–16.9‰), higher than goats (mean 9.9‰), indicating substantial freshwater fish consumption and/or higher trophic protein. All humans/domestic animals have higher δ13C than wild C3/fauna, implying an additional δ13C-enriched resource; a negative correlation between human δ15N and δ13C suggests this resource was plant-based (i.e., some C4 input), though correlation significance is marginal (r ≈ −0.60; p=0.05; small n).
• Human enamel δ13C (IMB, LON, BLD): IMB −14.1‰ and LON −15.1‰ indicate heavy reliance on tropical rainforest and/or freshwater resources during late childhood-adolescence. BLD enamel values split: two individuals −14.7 and −13.6‰ (rainforest/freshwater), others −12.0 to −10.8‰ consistent with dominant C3 foods grown in more open conditions (e.g., yams, plantain, oil palm) or mixed closed canopy and freshwater diets. No clear enamel evidence for substantial C4 reliance at IMB, LON, or BLD.
• Tissue contrasts (BLD paired enamel-collagen): Several individuals show enamel consistent with C3/freshwater but collagen indicating more C3/C4-based protein (potentially adults consuming C4-fed animal proteins or more C4 in adulthood), suggesting life-stage dietary change and protein bias in collagen.
• MTNW enamel: M3 δ13C −3.2‰ (and multiple teeth) demonstrates dominant C4 resource reliance throughout childhood/adolescence.
• Dental calculus (MTNW): 38 starch granules and 9 phytoliths. Starch types indicate grass seeds (Poaceae) including morphologies consistent with sorghum (Sorghum bicolor) and other millets (pearl/finger millet not specifically assignable), plus elongated parabolic/oblong granules matching Dioscoreaceae (wild yams) and an ovate type potentially from Asphodelaceae underground storage organs. Phytoliths (globular tuberculate–echinate) match Elaeis guineensis (oil palm) nutshell.
• Charred food residues (BLD): Three δ13C groupings—<−27‰ and an intermediate at −24‰ (C3/aquatic-based foods) and two around −9‰ indicating C4-based foods. One high-δ13C sample (−9.3‰) also yielded δ15N 7.8‰. Together with identified charred pearl millet grains, this confirms processing of C4 cereals (e.g., millet/sorghum) at BLD. Overall: Western DRC sites (IMB, LON, BLD) show persistent reliance on forest and freshwater resources with modest or context-limited C4 inputs; northeastern MTNW shows strong C4 dependence (likely sorghum) alongside forest plant use.

Findings demonstrate heterogeneous agricultural adoption across Central Africa during the Iron Age. Despite archaeobotanical presence of C4 cereals (e.g., pearl millet at BLD), enamel data from IMB, LON, and most BLD individuals indicate diets centered on rainforest and freshwater resources, with collagen revealing some C4-derived protein inputs—possibly through domesticates (e.g., goats) or episodic cereal consumption more visible in adult protein intake. This suggests that C4 cereals did not uniformly become staples in western DRC; instead they may have been incorporated in specific social or seasonal contexts (e.g., feasting, brewing, prestige) while mixed subsistence persisted. In contrast, the MTNW individual shows sustained, dominant C4 consumption throughout childhood, with dental calculus evidence for sorghum and continued forest plant use (yams, oil palm), pointing to forager–farmer interactions and exchange with eastern African agricultural groups. These results argue against monolithic models of the ‘Bantu expansion’ and challenge the assumption that a widespread ‘rainforest crisis’ was necessary for cereal cultivation. Rather, agricultural practices were flexible, context-specific, and integrated within longstanding mosaic subsistence strategies that combined C4 crops, forest plants, and aquatic resources.

This study provides the first multi-tissue (collagen and enamel) stable isotope and microbotanical evidence tracking Iron Age diets across multiple regions of the Congo Basin. It shows that adoption of C4 cereals varied markedly: in the western DRC, cereal use occurred alongside a durable focus on forest and freshwater resources, while in the northeast (Ituri) a strong, sustained C4 signal likely reflects sorghum consumption within forager–farmer interaction spheres. The work underscores the longevity of heterogeneous, mixed subsistence strategies in Central African rainforests and cautions against singular narratives linking language dispersal, farming, and deforestation. Future research should expand regionally comparative, context-specific sampling (including more individuals, fauna baselines, and direct AMS dating), integrate additional biomolecular proxies, and examine social uses of cereals (e.g., feasting) to resolve the role of C4 crops across life stages and communities.

• Sample sizes are small and constrained by preservation and availability; only one individual each from IMB and LON, and a single individual from MTNW.
• Two BLD human collagen samples were excluded due to poor preservation; collagen could not be obtained for MTNW.
• Faunal baselines are site-biased (robust at BLD, absent at MTNW), limiting inter-site comparability.
• Collagen δ13C is protein-biased, potentially underrepresenting low-protein plant contributions; enamel integrates total diet but only for tooth-formation periods (childhood/juvenile windows), complicating adult diet inference.
• Charred food fragments often had low nitrogen content, limiting δ15N determinations.
• Statistical power is limited (e.g., marginal significance for δ13C–δ15N correlation) due to small n.
• Temporal resolution varies; some individuals have indirect dates; life-history dietary changes are inferred from limited paired tissues.