Early and Middle Holocene Hunter-Gatherer Occupations in Western AmazoniaThe Hidden Shell Middens

For decades, archaeologists working in Amazonia operated under a quiet assumption: the western lowlands were essentially empty for most of human prehistory. The Llanos de Moxos, a vast seasonal wetland in the Bolivian lowlands, was famous for its late Holocene earthworks, the canals and raised fields built by so-called Earthmovers. But when researchers pushed the question back further, before four thousand years ago, the map went blank. The environment was too wet, the landscape too dynamic, and the organic remains too degraded. No stone tools and no obvious sites. The absence of evidence was interpreted as evidence of absence. Then Lombardo and colleagues went looking in the right places, with the right questions, and found people who had been there all along—ten thousand years ago, sitting on mounds of freshwater snail shells, cracking open apple snails, lighting fires, and living seasonally in a landscape that science said couldn't accommodate them. The key to finding those people was understanding why they'd been invisible. The Llanos de Moxos has three characteristics that conspire against archaeological visibility. First, the region has no local stone resources, so there are no lithic artifacts—the durable traces archaeologists most reliably find. Second, it’s a highly seasonal and inundated system: rivers flood, deposit sediment, and bury older surfaces. Later Holocene fluvial processes have covered large swaths of the lowlands. Third, early groups here were almost certainly mobile, exploiting scattered wetland resources rather than settling permanently, which produces an inherently ephemeral record. Put those three factors together and you get what the authors call "poor visibility and the absence of adequate archaeological correlates." This is a methodological trap, not a historical truth. What broke the trap open was a different kind of site. Scattered across the Llanos de Moxos are small forest islands—earthen mounds rising above the seasonally flooded grassland that support patches of trees. They're refugia during flood season, and they stand out as places where stratified deposits might survive. Lombardo and colleagues selected three islands for investigation, labeling them SM1, SM2, and SM3, with SM1 being locally known as Isla del Tesoro. A shell midden, to clarify terms, is an anthropogenic accumulation of food refuse—here dominated by Pomacea, the freshwater apple snail, interbedded with charcoal, burnt earth, and animal bone. These deposits record seasonally repeated human activity: food processing, burning, and refuse disposal. In a landscape without stone tools, they're the archaeological signal that had simply never been looked for. Field methods combined coring across all three islands with a test excavation at SM1 that exposed a 1.7 meter stratigraphic profile before hitting the water table. Cores showed that the shell accumulation continued below. What the team recovered was consistent and striking: dense layers of Pomacea shell fragments throughout the sequence, with animal bones and charcoal present at every level, discrete burnt lenses marking episodic occupation, and in the uppermost thirty centimeters, late Holocene material—pottery, bone tools, and a Caiman cranium with butchery marks dated to around 400 years ago. Deeper down lies the early and middle Holocene midden. The chronological spine of the argument is a radiocarbon program of twenty-eight dates at SM1 alone. The results follow a coherent vertical trajectory. At the base of the midden, around 150 to 165 centimeters in depth, charcoal calibrates to 10,604 plus or minus 126 calendar years before present. Bulk organics from the same level calibrate to 10,382 plus or minus 141 years. The midden accumulated through the early and middle Holocene before a notable hiatus after about 4,300 years ago, with late Holocene reoccupation above. The neighboring mounds tell a similar story: SM3 yields shell and charcoal dates around 8,550 years ago, SM2 produces middle Holocene dates around 6,200 years ago. Paired dates from shells and charcoal agree closely, ruling out significant freshwater reservoir effects. But radiocarbon ages alone don't prove that humans made these deposits. That's where the convergence of independent evidence becomes decisive. Steroid analysis shows coprostanol, a biomarker produced specifically by human gut bacteria, dominating the steroid fraction in SM1, with particularly strong enrichment at depth. This indicates that human feces concentrated in the midden across thousands of years. Black carbon measurements, via benzene-polycarboxylic acid analysis, show combustion residue in the midden running roughly two point three times higher on average than in the surrounding paleosol, with values in individual midden samples reaching up to nine grams of carbon per kilogram of sediment compared to a paleosol maximum of about two point nine. Fire, repeatedly, in the same spot. Multi-element analysis by X-ray fluorescence adds another line. The midden and surrounding paleosol have sharply different elemental signatures—driven by phosphorus, calcium, aluminum, and silicon—with the highest phosphorus values concentrated at the midden's base, consistent with dense biological and anthropic input. Micromorphology confirms in-situ burning: thin sections show mixed burnt and unburnt shell fragments, many aragonitic shells transformed to microcrystalline calcite, a conversion that begins around 200 degrees Celsius and completes after roughly an hour at 330 degrees Celsius. Burnt earth fragments and charred wasp nest chambers appear throughout the profile. The surrounding paleosols, sampled and dated contemporaneously, serve as controls: lower black carbon, distinct elemental signatures, and no shell accumulation. The midden is anomalous in every dimension the team measured. The faunal evidence speaks to behavior as much as antiquity. The assemblage is overwhelmingly Pomacea, with vertebrate remains—brocket deer, marsh deer, fish, reptiles, and birds—most diverse and best preserved in the deepest unit, at 140 to 170 centimeters. Two refitted pieces of a marsh deer long bone at 160 centimeters show a percussive impact fracture consistent with butchery. That’s a person, ten thousand years ago, breaking open a bone for marrow. So how did these people actually live here? The evidence points to mobile wetland foragers making seasonal return visits rather than permanent occupants. Apple snails are most easily harvested in the dry season. Yet only two opercula—the trapdoor-like structures snails use to seal themselves in—were found among several hundred shells in one sampled unit. This suggests the meat was extracted and cooked or transported rather than consumed on the spot. The burnt wasp nest chambers are particularly telling: wasps build nests during hiatuses in human occupation, and those chambers are charred when people return and light fires. This indicates intermittent, probably seasonal reuse of the same mound, over thousands of years. The burning patterns—discrete, concentrated lenses of burnt shell in two of the deeper units—and the coprostanol enrichment across the profile together paint a picture of a logistical campsite. People came back to the same island because Pomacea was reliably abundant there during the dry season. They combined this low-effort resource with hunting large game and then moved on. The midden thickness, roughly ten times deeper than the contemporaneous paleosol accumulation, reflects not dense permanent occupation but six thousand years of seasonal return. What does this reveal? Two things, and they matter at different scales. At the regional scale, the late Holocene Earthmovers of the Llanos de Moxos—those engineers of canals and raised fields—now have plausible local antecedents. The forest islands they built on, or near, in some cases turn out to be early Holocene shell mounds. The cultural sequence doesn't begin with monument-building. It begins with mobile foragers reading the seasonal wetland pulse, returning to the same elevated refugia year after year, and leaving behind ten-meter mounds of snail shells that persisted through the millennia. Whether that represents direct cultural continuity is a question the data don't yet answer, but the geographic and temporal connection is no longer a gap—it's a research frontier. At the broader scale, this study demonstrates method. The absence of early sites across Amazonia has long been treated as a historical fact. Lombardo and colleagues show it has been a methodological failure—a consequence of looking for stone tools in a landscape with no stone, and not understanding what the actual archaeological correlate of a mobile tropical forager looks like. Shell middens, invisible without geoarchaeological analysis, buried under alluvial sediment, have been dismissed as natural features. The oldest archaeological sites in western Amazonia were there all along. We just didn't know what we were looking at. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.