Direct archaeological evidence for Southwestern Amazonia as an early plant domestication and food production centre

A stone tool pulled from a riverbank bluff, forty meters above the Madeira River in Brazil. A researcher scrapes residue from its surface, runs it under a microscope, and finds starch grains that are nine thousand years old. That single object raises a question: who was grinding this, and what were they growing? According to Jennifer Watling and colleagues, the people living at this site — called Teotonio — were cultivating manioc, squash, and beans at a time when agriculture in the Americas was barely getting started anywhere. Their paper provides the first direct archaeobotanical evidence for what geneticists had long suspected: that southwestern Amazonia was an independent center of plant domestication in the New World. The genetic case had been building for years. Manioc, peanut, peach palm, annatto, and at least one lineage each of squash and chili all trace their origins to the seasonal forests of the upper Madeira and Guaporé watersheds. But genetics is an indirect argument. It tells you where a crop came from, not when people started growing it or what their lives looked like when they did. The archaeological evidence had been thin — partly because systematic fieldwork in this remote corner of Brazil only began recently, and partly because the taphonomic obstacles are real. Organic material degrades quickly in tropical soils. Pre-ceramic deposits can be shallow. Post-depositional mixing can blur the timeline. Watling and colleagues honestly lay out these difficulties, noting that manioc phytoliths — the microscopic silica bodies formed inside plant cells — are produced in low quantities and are hard to detect even when the plant was present. The absence of evidence, in this case, was not evidence of absence. Teotonio begins to close that gap. The site sits on a forty-meter bluff on the south bank of the Madeira River, just above a waterfall that made the location both a premier fishing spot and, before modern damming, a crucial trade and communication node — the kind of place people returned to for millennia. And return they did. The stratigraphic sequence at Teotonio captures almost the entire regional cultural history from the early Holocene onward, organized in the study around two key pre-ceramic phases. The older Girau phase sits in orange-brown ferralsol deposits; charcoal from the base of those levels in one excavation unit returned a date of nine thousand five hundred twenty-four to nine thousand four hundred calibrated years before present. Overlying those levels are deposits from the Massangana phase, associated with Anthropogenic Dark Earths — terra preta — the nutrient-rich, human-made soils that signal sustained, cumulative occupation and land management. In Unit 5, the Massangana horizon runs one hundred ten centimeters deep and contains hearths dated between six thousand four hundred ninety-five and six thousand four hundred calibrated years before present. Over roughly six thousand years of occupation, the paper reports that the inhabitants of Teotonio created more than fifty hectares of these dark earths. To extract the plant record, the team combined two analytical methods on twenty-nine retouched stone tools — twenty-three from the Massangana horizon and six from Girau. The first method targets starch grains, microscopic carbohydrate structures whose shape is distinctive enough to identify by genus. The second targets phytoliths, those silica bodies that form inside plant cells and survive long after the rest of the plant has rotted away. Both methods demand rigorous contamination controls. All glassware, brushes, and slides were autoclaved between uses; latex gloves were avoided because prior research had found maize starch in them. Each artifact went through two extraction stages — first a wet brush to collect surface contamination, then an ultrasonic bath to dislodge tougher, more deeply adhering residues. The logic is that use-related residues should survive in the ultrasonic extract even after the surface wash removes modern contamination. That logic paid off. Across fourteen artifacts, the analysis recovered eleven starch-grain morphotypes. Three of those grains were confidently identified as Phaseolus — the bean genus — oval in front view, roughly thirty micrometers long, recovered from the ultrasonic extracts of three separate Massangana tools at depths of fifty to seventy centimeters. Crucially, none of those Phaseolus grains appeared in the wet brush samples from the same artifacts. That is exactly the pattern you would expect from original tool use rather than modern contamination. An abundant unidentified morphotype — Type 1 — appeared on residues from eleven different artifacts and showed damage consistent with roasting. The phytolith assemblages were more complex: many were dominated by arboreal and palm types that also dominate the surrounding soils, making them hard to interpret as use-related. But two phytolith morphotypes found on tools were absent from surrounding soils, suggesting they came from plants actually processed on those tools, even if their botanical identity remains unresolved. Now step back from the methodology and look at the full plant picture that Teotonio reveals. In the earlier Girau-phase contexts, the record points to managed wild resources and root and tuber use. The team found Calathea-type rhizome phytoliths — tentatively identified as leren, an edible starchy root — alongside carbonized tuber and root fragments. Macrobotanical remains from those same levels include pequiá, guava, and Brazil nut. Watling and colleagues interpret the phytolith assemblage from these contexts as already showing disturbance indicators and palm concentrations inconsistent with unmodified forest, which leads them to argue that landscape domestication — the purposeful reshaping of the surrounding forest to favor useful species — had begun at least by nine thousand five hundred calibrated years before present. By the Massangana phase, the evidence for cultivation sharpens considerably. Manioc phytoliths — distinctive heart-shaped forms, around eleven to twelve micrometers long — appear in Unit 5 within Anthropogenic Dark Earth contexts. Squash rind phytoliths show up in both Unit 1 and Unit 5; two of the specimens from Unit 5 fall in the domesticated size range. And the Phaseolus bean starch from the lithic tools places an exotic legume — not native to this part of Amazonia — in the hands of Massangana-phase people. Watling and colleagues point out that the formation of nutrient-rich dark earths was itself likely what made these exotic, phosphorus-demanding crops viable in the lowland forest environment. The dark earths were not just a byproduct of occupation. They were, in a real sense, agricultural infrastructure. What this all adds up to is not a sudden shift to full-scale farming. Watling and colleagues characterize the Teotonio economy as something closer to what they call a hunter-fisher-gardener system — cultivation embedded in a broader strategy of wild resource use, gathered fruits, fishing, and hunting. The charcoal record backs this up: burnt wood accounts for sixty to eighty percent of macrobotanical fragments in some units, pointing to domestic activities and hearth use rather than field-clearing. Palm macroremains make up less than two percent of macrobotanical samples even though palm phytoliths dominate the microbotanical record in some contexts — a discrepancy that suggests complex inputs from habitation debris rather than simple on-site palm cultivation. And then there is the larger view. Manioc alone currently feeds at least five hundred million people. Squash and beans are dietary staples across two hemispheres. The crops that changed global food systems trace their origins, in significant part, to the people who lived on this bluff above the Madeira River and left their grinding stones behind. Watling and colleagues connect their findings to a growing consensus — supported by work from the Colombian Amazon dating anthropogenic forest patches to ten thousand to eight thousand years ago — that the biological richness of Amazonia today is not a pristine accident of nature but partly a product of sustained human management stretching back through the entire Holocene. What remains unresolved is the full regional picture. Systematic fieldwork in southwest Amazonia is recent, and Teotonio is one site. The relative contribution of cultivated versus gathered resources, and how connected different early Amazonian communities were to one another, are questions this study opens rather than closes. But for the first time, the starch grains are there, on the tools, datable and diagnostic. The genetic hypothesis now has physical evidence. Nine thousand years ago, someone on a bluff above the Madeira was grinding beans. 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.