How People Domesticated Amazonian Forests

A botanist follows a village elder through the Amazon undergrowth and stops at a stand of trees — peach palms, Brazil nuts, cacao, all clustered together, none of them wild. The forest looks ancient. It is. But it wasn't made by nature alone. It was grown by people. Not planted like a farm, not wild like a jungle — something in between, built over millennia, and until recently almost entirely invisible to science. That is the argument at the center of a paper by Carolina Levis and colleagues, and it starts with a definitional move that turns out to matter enormously. When most of us picture Amazonia, we imagine pristine wilderness. However, Levis and her team draw a sharp distinction between three things that are usually lumped together. A pristine forest is one where humans haven't altered the ecological processes — the baseline state before people arrived. An old-growth forest is mature and recently undisturbed, but not necessarily untouched by human history. A domesticated forest is something else entirely: a patch where long-term human intervention has reshaped the composition of species, often so gradually and persistently that the intervention becomes invisible. The key marker is a dominance of useful species — plants that people eat, trade, or use — concentrated in ways that nature alone wouldn't produce. This matters because it overturns the story Amazonia has been told about itself. With at least thirteen thousand years of growing human populations across the basin, truly pristine forest has been rare for a very long time. Palms dominate modern forests growing on abandoned pre-Columbian mounds and anthropogenic soils left behind more than four hundred years ago. The forest didn't reset when the people left. It remembered them. So how, exactly, did human management turn wild forest into something domesticated? Levis and colleagues identified eight categories of practice from the ethnobotanical literature, and the key insight is that these practices don't work alone — they compound. The first two reshape neighborhood dynamics. Removal of non-useful plants — clearing understory, cutting lianas, weeding around favored trees — opens light and space for species like Brazil nuts and açaí. Protection practices do the complementary work: sparing seedlings during clearing, pruning carefully, even biological pest control. The Kayapó deploy Azteca ants to repel the leaf-cutting ants that would otherwise shred their useful trees. In the field interviews Levis and colleagues conducted, removal and protection were the most commonly cited practices across communities. The second cluster of practices moves genes and chosen individuals through space. People attract non-human dispersers by leaving fruits out or cultivating large-seeded trees that draw game animals — stands of Attalea maripa palms, for instance, are associated with tapir latrines. People also transport plants themselves, accidentally or deliberately, along trails and between camps. The Hotï people's habit of dropping Humiria balsamifera seeds from their baskets explains an otherwise puzzling local abundance of that species. Then, there is intentional planting — concentrating desired species near settlements and swiddens. Transportation, planting, and phenotypic selection turned out to be the most frequently documented practices across the twenty-two focal species in the literature. That selection piece is worth pausing on. Levis and colleagues document human selection for fruit size, taste, and processing qualities across many species — peach palm selected first for small oily fruits or wood, then later for large starchy ones; annatto selected for higher pigment yield and altered fruit dehiscence; multiple pequí varieties chosen by the Kuikuro for different properties. This isn't the controlled breeding of a seed company. It's millennia of preferential harvest, protection, and replanting of individuals that happen to be more useful. Unconscious and intentional selection, working in parallel, generates what the paper calls domestication syndromes — recognizable patterns of morphological and genetic divergence from wild relatives. The third engine modifies the ground itself. Fire management can favor light-demanding or fire-adapted useful species; burití palm patches are linked to fire histories, and there is abundant charcoal around old settlements. Soil improvement ranges from ash and mulch additions to the construction of mounds in floodplains, and culminates in Terra Preta de Índio — the extraordinarily fertile dark earths formed from centuries of accumulated ash, charcoal, and organic refuse at sedentary settlements. Terra Preta de Índio is both a marker of past occupation and a driver of future plant communities. Useful species thrive in these soils and keep recruiting there long after the people who made those soils are gone. The conceptual model Levis and colleagues built from all of this is both temporal and spatial. Temporally, it follows a continuum from late Pleistocene hunter-gatherers through the increasing sedentism of the Holocene. As human populations grew, management practices became more frequent and more intensive, layering interventions on top of interventions across generations. Spatially, the model describes gradients radiating out from settlements: patches directly on these anthropogenic soil sites were associated with pre-Columbian sedentary occupation; patches in nearby fallows correspond to past swiddens; more distant patches reflect old-growth contexts that were periodically managed. The intensity of useful species dominance increases as you approach ancient settlement sites — and that gradient is still readable in the forest today. Then the team went to test it. Between March 2013 and March 2015, Levis and colleagues visited thirty contemporary villages along four major Amazonian rivers — the Madeira, Solimões, Negro, and Tapajós — all of them on or near archaeological dark-earth sites. They interviewed fifty-six local informants, ran participatory mapping workshops using georeferenced satellite imagery, and conducted guided forest tours in each village to document patches of useful perennial species. The method mattered: they weren't imposing an external classification on the landscape — they were working alongside people who named these patches in their own languages, who used suffixes like "zal" in Portuguese or "tíwa" in Nheengatu to describe aggregated stands. What they found matched the model's predictions. People cited thirty-five named patch types corresponding to thirty-eight useful species. Across all patches visited, the team recorded eighty-seven useful species in total. Of the twenty-one patches they studied in detail, palm species dominated in three-quarters of them — Oenocarpus, Euterpe, and Attalea appearing again and again, with regional patterns emerging: Attalea maripa dominated Madeira patches, Euterpe precatoria the Solimões, Oenocarpus bataua along the Negro, and Oenocarpus distichus along the Tapajós. Each patch contained between four and twenty-one useful species, with a median of seven. Most patches were small, under one square kilometer. Half of all patches sat within one kilometer of an archaeological site, though some extended up to forty kilometers out. Four species — Attalea maripa, Caryocar villosum, Mauritia flexuosa, and Theobroma cacao — were managed with seven of the eight practices documented in the literature. In the fieldwork, Caryocar villosum and Endopleura uchi also reached seven practices. More than half of the useful species studied were managed with at least five of the eight practices. This isn't a picture of simple monoculture gardens. It's a picture of complex, overlapping, reinforcing interventions that accumulate over generations into something that looks, from a distance, like natural forest. Levis and colleagues report that at least eighty-five tree and palm species were domesticated to some degree during pre-Columbian times, and that domesticated palms now dominate modern forests in southwestern Amazonia. Those plants have persisted on anthropogenic soils and mounds abandoned more than four centuries ago. The useful species occurring in forest patches appear across multiple sub-basins, which suggests their spread was not local or accidental — it was the outcome of widespread, sustained human management across the entire basin. This changes what conservation means in Amazonia. Forest that looks pristine often bears the signature of long-term human modification. The biodiversity that ecologists measure and policymakers try to protect is partly a cultural product — shaped by people who knew exactly what they were doing, across timescales too long for any individual to witness. Levis and colleagues argue for integrating indigenous ecological knowledge into conservation and land management plans, not as a courtesy, but because that knowledge is the mechanism by which these forests are maintained. The central finding, stated plainly: millennia of accumulated ecological decisions are encoded in forest composition. They are readable — in the clustering of useful species, in the fertility of dark earths, and in the names communities give to forest patches along rivers they have managed for generations. The forest remembers. The question is whether we are paying attention. 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.