The scale and nature of pre-Columbian human impacts in Amazonia are currently hotly debated. Whereas pre-Columbian people dramatically changed the distribution and abundance of species and habitats in some parts of Amazonia, their impact in other parts is less clear. Pioneer research asked whether their effects reached even further, changing how ecosystems function, but few in-depth studies have examined mechanisms underpinning the resilience of these modifications. Combining archeology, archeobotany, paleoecology, soil science, ecology, and aerial imagery, we show that pre-Columbian farmers of the Guianas coast constructed large raised-field complexes, growing on them crops including maize, manioc, and squash. Farmers created physical and biogeochemical heterogeneity in flat, marshy environments by constructing raised fields. When these fields were later abandoned, the mosaic of welldrained islands in the flooded matrix set in motion self-organizing processes driven by ecosystem engineers (ants, termites, earthworms, and woody plants) that occur preferentially on abandoned raised fields. Today, feedbacks generated by these ecosystem engineers maintain the human-initiated concentration of resources in these structures. Engineer organisms transport materials to abandoned raised fields and modify the structure and composition of their soils, reducing erodibility. The profound alteration of ecosystem functioning in these landscapes coconstructed by humans and nature has important implications for understanding Amazonian history and biodiversity. Furthermore, these landscapes show how sustainability of food-production systems can be enhanced by engineering into them fallows that maintain ecosystem services and biodiversity. Like anthropogenic dark earths in forested Amazonia, these self-organizing ecosystems illustrate the ecological complexity of the legacy of pre-Columbian land use.French Guiana | historical ecology | land-use legacy | raised-field agriculture | coupled human and natural systems S ome tropical landscapes long considered "pristine" are now known to have been densely occupied by humans in the past (1, 2), inspiring optimism that these environments and their biodiversity may be more resilient to intensive human use than often feared (3, 4). Among the most extensive apparently pristine landscapes are the forests and, by extension, the savannas of Amazonia. The extent to which Amazonian ecosystems were affected by preColumbian human activities is currently hotly debated. This controversy has bearings on much broader debates about how to conceptualize the history of forested and other tropical systems and the ecological footprint of past agriculture upon them (2, 5) and how to plan for the transition of these systems toward future global food, energy, and carbon needs (6-10). It still is argued whether the influence of pre-Columbian humans was minor and localized (11) or important and widespread (12, 13). The nature of this impact is also unclear. In addition to altering the species composition of commu...