Ochre communities of practice in Stone Age Eswatini

The study investigates how Stone Age communities in Eswatini selected, procured, processed, and transported ochre and related mineral pigments, and whether these behaviors reflect regional-scale communities of practice with localized, site-specific traditions. Contextualized within Middle Stone Age/Late Stone Age (MSA/LSA) research in sub-Saharan Africa, the paper emphasizes ochre’s roles in symbolic behavior, hafting adhesives, bedding, and rock art. The authors aim to move beyond isolated case studies by building a framework that links geochemical provenance, mineralogical properties, and diachronic site records to reconstruct ochre provisioning networks, social learning, mobility, and potential exchange across western highveld, central lowveld, and eastern Lebombo regions of Eswatini.

Prior research has established the antiquity and ubiquity of ochre use in Africa, connecting it to cognitive complexity, symbolic practices, adhesives, and bedding. Numerous provenance and compositional studies have been conducted at sites such as Blombos, Diepkloof, Pinnacle Point, Sibudu, and Porc-Epic. However, the concept of communities of practice (after Lave and Wenger) has rarely been applied to ochre. The paper builds on ethnographic, archaeological, and materials science work to conceptualize ochre behaviors as a chaîne opératoire embedded in social learning, mobility, and intergroup relations. It also engages the broader debate on the timing and nature of the MSA/LSA transition, noting regionally variable trajectories and the possibility of gradual, non-linear change. The long-held assumption that Ngwenya supplied most regional ochre (including to nearby Border Cave) is reassessed with new geochemical evidence.

• Study area and sampling: Ochre artifacts from ten archaeological sites and raw materials from nine geological source localities across Eswatini were analyzed, covering the final MSA, LSA, and into the Iron Age. In total, 361 artifacts (of 931 catalogued) and 173 source samples were analyzed for provenance and material properties.
• Field surveys and sources: Systematic surveys were conducted across ~22 km of the Ngwenya Massif (including Lion Cavern, Banda Cavern, SW Spur and the Bulembu outcrop), and six additional Fe/Mn sources elsewhere: Maloma, Kubuta manganese ore field, Gege goethite–magnetite, Mahamba Mountain (secondary banded ironstone nodules), Mnyogane and Lupholo Dam lateritic–saprolitic clays, and a locality around Siphiso in the Lebombo Mountains.
• Analytical techniques: All samples were screened by pXRF (Bruker Tracer 5i). Elemental characterization employed NAA (388 specimens: sample prep with washing, grinding, controlled irradiations; detection of short-, medium-, and long-lived radioisotopes; concentration calculation via comparator method using NIST SRMs). LA-ICP-MS (PerkinElmer SCIEX NexION 300 with Teledyne 193 nm laser) measured a suite of 60 isotopes with controlled ablation parameters, bracketing standards/QC. Mineralogical identification used XRD (Rigaku Ultima IV, Cu-Kα1) and handheld Raman spectroscopy (Bruker Bravo, dual lasers).
• Statistics and grouping: Multivariate analyses included PCA, hierarchical clustering, element pair plots, and Mahalanobis distance for group assignment and source membership probabilities. Iron oxide (FeOx) and manganese oxide (MnOx) datasets were analyzed independently. Nine FeOx compositional groups (G1/G1b to G5) and six MnOx groups (Mn-G1 to Mn-G6) were defined; unassigned samples were noted.
• Chronology: Optically Stimulated Luminescence (OSL) dating of Lion Cavern deposits was undertaken at the University of Cologne (fine-grain quartz, SAR protocol on Risø DA20). Six samples from Lion Cavern units yielded ages reaffirming intensive ochre mining at ~48 ka.
• Stratigraphic analysis: Diachronic distributions of compositional groups were examined at two deeply stratified sites, Sibebe and Siphiso, across multiple temporal phases to infer changes in procurement and transport through time.

• Provenance structure: Source datasets meet the provenance postulate with variability between sources exceeding within-source variability. Ngwenya Massif shows internal distinctions (high-grade Fig Tree ironstones vs ferruginous shale exposed by modern mining) and localized MnOx mineralization at Banda Cavern.
• FeOx artifact groups and sources (n=337 assigned; 37 unassigned): • Fe-G1 (Ngwenya high-grade): 82 artifacts, 50–66% Fe with low impurities; >80% probability matches Lion Cavern, Banda Cavern, or SW Spur. Dominant at Sibebe (n=66); also at Nyonyane, Hlalakahle, Nsangwini, and in small numbers at Siphiso and Mlawula. • Fe-G1b (Bulembu): 25 artifacts; higher Si (~12%) and Mn (~0.64%) than Fe-G1; matches the Bulembu outcrop ~22 km north of Lion Cavern. Mostly at Sibebe, rare at Nsangwini, Nyonyane, and Siphiso. • Fe-G2a (Siphiso area): 20 artifacts (7 from Siphiso excavations; 5 local surface finds); volcanogenic signature typical of Lebombo rhyolites; also at Muti Muti and a single long-distance occurrence at Ndhlozane (>80 km transport). • Fe-G2b (Eastern region): 24 artifacts, mainly Siphiso; higher K, Si, Ba, Ca, Eu, Mn, Sr; inferred local to lowveld/Lebombo, possibly tied to oxidized Sabie River basalts (<5 km from Siphiso/Muti Muti). • Fe-G2c (Eastern region): 25 artifacts, all from Siphiso; Fe ~30–40%, elevated Mg and K; mineral phases include muscovite-like mica, pyrophyllite, almandine; inferred local. • Fe-G2d (Eastern region): 66 artifacts; high-purity magnetite with trace elements suggesting a volcanogenic Lebombo origin; dominant at Siphiso throughout the LSA sequence, with rare pieces at Mlawula, Nyonyane, and Sibebe. • Fe-G3 (Eastern region): 18 artifacts; titaniferous magnetite (to ~19.9% Ti); present at Mlawula, Siphiso, Muti Muti, and one at Sibebe; source likely near Mbuluzi River (20–25 km from these sites). • Fe-G4 (Unknown locality): 17 artifacts (15 Sibebe, 2 Siphiso); high Fe and Al; poor statistical match to western highveld sources; origin unresolved. • Fe-G5 (Mahalamanti Valley): 14 artifacts across Sibebe, Siphiso, Nyonyane, Hlalakahle, Muti Muti, plus two nodules from the Mphaphati I iron-working kiln ~4 km from the Mahalamanti ore source; indicates a southern-central lowveld source with wide distribution.
• MnOx groups and sources (n=56 total Mn samples): • Mn-G1 (Bulembu): 10 samples (6 source, 4 Sibebe artifacts); minerals include pyrolusite, hematite, alabandite. • Mn-G2 (Eastern region): 2 Siphiso artifacts; high-purity pyrolusite (>70%) with BaO (3–4%) and K2O (0.8–1.0%), suggesting hollandite-type mineralization. • Mn-G3 (Western region): 3 Sibebe artifacts; 50–62% Mn with 10–20% Fe and 0.8–2.1% Ba; barium iron oxide, Fe–Mn sulfides, alabandite, hausmannite identified. • Mn-G4 (Banda Cavern): 14 samples (2 source, 12 artifacts from Sibebe and Siphiso); elevated trace/REEs; mineral phases include pyrolusite, braunite, BaMnO3, bixbyite, jacobsite, magnetite. • Mn-G5 (Kubuta): 14 samples (mostly source; 1 Sibebe artifact); composition similar to umber pigment with magnetite/hematite/pyrolusite. • Mn-G6 (Eastern region): 10 Siphiso artifacts; high Mn (to 64.8%), low Fe; high K, quartz, Al impurities; pyrolusite and manganite present. One Sibebe artifact unassigned.
• Diachronic patterns: • Sibebe: Final LSA with pottery (Strata I/II) dominated by Ngwenya (Fe-G1a) and Bulembu (Fe-G1b); MnOx mainly Mn-G1 and site-specific Mn-G3. Pre-ceramic LSA (Stratum III) similar dominance with some Fe-G4 and weak matches to Maloma; MnOx shows Mn-G1 and Mn-G3. Final MSA (Strata V/VI) exhibits greater diversity, including small amounts of Fe-G2d, Fe-G3, Fe-G4, and Fe-G5, and Mn-G5 (Kubuta), indicating increased long-distance transport during these phases. • Siphiso: Across at least 11 strata (8 with ochre), assemblages consistently favor local eastern groups (Fe-G2a–d, Fe-G3; Mn-G2, Mn-G6). Ngwenya/Bulembu (Fe-G1/G1b) appear only in Strata III–V (Oakhurst phase ~12.2–9.7 cal BP) and later, not earlier. Fe-G4 appears only in VI and VIII. Fe-G5 (Mahalamanti) occurs in low quantities from IV–VII, indicating selective long-distance transport during Oakhurst–Wilton, largely absent by ceramic LSA.
• Transport networks: Western sites (Nyonyane, Hlalakahle, Nsangwini) predominantly used Ngwenya/Bulembu ochre, with occasional Fe-G2d and Fe-G5 indicating east–west and south–north links. In the eastern Lebombos, Mlawula sites used local eastern Fe groups; Muti Muti surface finds align with Fe-G2a, G2b, G3, and one Fe-G5. Ndhlozane contains one Fe-G2a nodule (eastern origin) among otherwise unmatched samples, evidencing long-distance movement.
• OSL chronology: New OSL dates from Lion Cavern reaffirm intensive ochre mining at ~48 ka, the oldest known worldwide.
• Usewear and processing: Only 9/931 (<1%) artifacts show clear usewear (grinding/abrasion/striations), suggesting powder production often involved pulverizing on anvils or mortar-and-pestle rather than repeated abrasion. Rock art associations are abundant (52 documented sites in Eswatini); Nsangwini yielded lithics stained with ochre and paint spatters.

The findings support the existence of a regionally coherent, long-lived community of practice surrounding ochre in Eswatini, manifested in stable procurement preferences and intergenerational knowledge transmission, alongside localized technological choices. Western highveld/middleveld communities consistently preferred high-quality Ngwenya/Bulembu ochres, implying shared knowledge of geology, mining techniques, and performance characteristics (colorfastness, adhesion, weathering resistance). In contrast, eastern Lebombo communities used a more diverse suite of locally available, softer, impurity-rich ochres, necessitating different extraction and processing knowledge (e.g., levigation), reflecting distinct but connected communities of practice. Long-distance transport events are present but infrequent, linking western, central, and eastern regions, with Fe-G5 (Mahalamanti Valley) serving as a key south-central lowveld connector. Comparative site chronologies indicate that direct, localized procurement dominated, with limited evidence that ochre formed part of intensive reciprocal exchange systems such as hxaro. The temporal overlap (or lack thereof) between occupations at Sibebe and Siphiso suggests many transfers were not mediated by intergroup exchange at the times of highest transport. Overall, the study demonstrates how geochemical provenance, mineralogy, and stratigraphic distributions can diagnose social learning networks, mobility strategies, and context-dependent chaînes opératoires of pigment use across deep time.

This study develops and applies a framework for identifying ochre communities of practice across Stone Age Eswatini by integrating compositional sourcing, mineralogical characterization, and stratigraphic analyses. It refines the chronology of Lion Cavern as the oldest known intensive ochre mine (~48 ka), documents strong and persistent western preferences for Ngwenya/Bulembu pigments, reveals diverse local procurement strategies in the eastern Lebombo region, and maps episodic long-distance movements that linked regions—especially via Mahalamanti Valley ochre. The results underscore the social embeddedness of ochre provisioning and processing, and their value for interpreting mobility, knowledge transmission, and symbolic practices. Implications for the MSA/LSA debate favor regionally variable, non-linear cultural trajectories rather than a singular transition. Future work should locate unresolved sources (e.g., Fe-G4 and eastern Mn groups), expand surveys in the Lebombo Mountains, further analyze ochre-stained lithics and paint residues, and apply additional chronological controls to refine diachronic interpretations.

• Several compositional groups lack identified source localities (e.g., Fe-G4; Mn-G2, Mn-G3, Mn-G6), limiting spatial reconstructions.
• Some sites/strata have small sample sizes, reducing statistical power for diachronic trends.
• Industrial mining destroyed parts of Ngwenya (Castle Cavern/Quarry), constraining direct sampling of all prehistoric sources.
• The very low frequency of artifacts with usewear complicates inferences about processing techniques and end-use contexts.
• Distinguishing direct procurement from exchange remains challenging without complementary lines of evidence (e.g., habitation overlap, social network proxies).
• Potential biases from preservation, collection histories (1980s excavations), and surface finds may affect assemblage representativeness.