Questions Secondary succession in the tropics can follow alternative pathways. Land-use history is known to engender alternative successional communities, but the underlying mechanisms driving and sustaining divergence remain unclear. In this study we aim to answer the following questions: (1) does previous land use act as a filter for species composition in secondary forests; and (2) what are the relative roles of management practices, soil properties and landscape composition in determining species composition? Location Central Amazon, Brazil. Methods We sampled trees, shrubs and palms (≥1cm diameter) in 38 early secondary forests (5 yr after abandonment) located along gradients of land-use intensity in five shifting cultivation landscapes. We measured the diameter and height of each sampled plant, identified it to species or morpho-species level and checked if it was resprouting or not. At each secondary forest we also collected soil samples for chemical and physical analyses and estimated the amount of old-growth forest surrounding it (landscape composition). Results We found that previous land-use intensity determined species composition. With increasing land-use intensity, management practices of cut-and-burn and associated reduction in soil quality filtered out seed-dependent species and favoured strong sprouters and species that can cope with low nutrient availability. Landscape composition had a weak effect on species assemblages. We found specific species assemblages and indicator species associated with different levels of previous land-use intensity. As a consequence of these local filters, species α- and β-diversity decreased and therefore early successional communities became more similar to each other. Conclusion Species composition of successional forests is strongly determined by different land-use intensities. Dispersal limitation has a limited effect on determining the composition of the dominant species. Filtering effects of management practices and soil quality determine the species dominating the canopy at early stages of succession and narrow down the range of species able to colonize and establish. This study highlights how land use shapes successional communities and suggests that alternative successional pathways are determined at early stages of succession. Therefore, accounting for land-use history is crucial to improve the understanding of tropical secondary succession. We present a list of indicator species for different levels of previous land-use intensity that can be used to support conservation and restoration decisions in the Amazon. Questions: Secondary succession in the tropics can follow alternative pathways. Land-use history is known to engender alternative successional communities, but the underlying mechanisms driving and sustaining divergence remain unclear. In this study we aim to answer the following questions: (1) does previous land use act as a filter for species composition in secondary forests; and (2) what are the relative roles of management practices, soil properties and landscape composition in determining species composition?. Location: Central Amazon, Brazil. Methods: We sampled trees, shrubs and palms (≥1cm diameter) in 38 early secondary forests (5 yr after abandonment) located along gradients of land-use intensity in five shifting cultivation landscapes. We measured the diameter and height of each sampled plant, identified it to species or morpho-species level and checked if it was resprouting or not. At each secondary forest we also collected soil samples for chemical and physical analyses and estimated the amount of old-growth forest surrounding it (landscape composition). Results: We found that previous land-use intensity determined species composition. With increasing land-use intensity, management practices of cut-and-burn and associated reduction in soil quality filtered out seed-dependent species and favoured strong sprouters and species that can cope with low nutrient availability. Landscape composition had a weak effect on species assemblages. We found specific species assemblages and indicator species associated with different levels of previous land-use intensity. As a consequence of these local filters, species α- and β-diversity decreased and therefore early successional communities became more similar to each other. Conclusion: Species composition of successional forests is strongly determined by different land-use intensities. Dispersal limitation has a limited effect on determining the composition of the dominant species. Filtering effects of management practices and soil quality determine the species dominating the canopy at early stages of succession and narrow down the range of species able to colonize and establish. This study highlights how land use shapes successional communities and suggests that alternative successional pathways are determined at early stages of succession. Therefore, accounting for land-use history is crucial to improve the understanding of tropical secondary succession. We present a list of indicator species for different levels of previous land-use intensity that can be used to support conservation and restoration decisions in the Amazon.