Amazonia is the largest continuous forest in the world, and is a terrestrial ecosystem with one of the highest estimated number of plant species globally. Even with the accumulated knowledge on species distributions in lowland Amazonia, long-standing questions as to species-habitat associations remain unanswered. This is because for most parts of Amazonia the associations between species and environmental conditions are still unknown. In this thesis, I investigated species-habitat associations from multiple perspectives. In the first paper I evaluated evidence of community assembly arising from soil vs. hydrological niche partitioning. I found that when soil cation concentrations were relatively uniform, communities of understorey plant species were strongly associated with local hydrological conditions as determined by topography. In the second paper I found that when soil cation concentrations ranged more widely, floristic patterns were strongly associated with soils. Single-species models showed that species’ realized niches are to a large degree defined by soil nutrients and hydrological conditions. In the third paper, species-habitat associations were quantified across potential zones of species turnover. Turnover zones have been hypothesized to be associated with the main rivers of the Amazon basin, if these represent dispersal barriers that trigger vicariance. I showed that a previously mapped, ca. 1000-km-long geological limit in western Amazonia is a biotic boundary associated with high species turnover. Species composition on the western side of the boundary was very different to that on the eastern side. This barrier runs in a northsouth direction perpendicularly to the Juruá River. No barrier effect across the Juruá River was found. It has been assumed that the current positions of rivers in central Amazonia became established in the Pliocene, promoting speciation on both sides of the rivers by vicariance. In the fourth paper, however, I present evidence that central Amazonian river channels and floodplains have continued to be highly dynamic during the Quaternary. This means that current biogeographical models will need to be revised. Lastly, knowledge of species-habitat associations makes spatial predictions of species distributions possible if appropriate environmental data layers are available. In final paper, I used species associations with soil cation concentrations, as measured from soil samples taken in the field, to test whether digital soil maps allow reconstruction of these associations. The correspondence between field and digital data was low, indicating that the mapping of Amazonian diversity needs better digital environmental data. In sum, the results of this thesis suggest that species-habitat associations in Amazonia are strongly related to habitat heterogeneity at different scales. Floristic patterns therefore need to be investigated from different perspectives, to provide a broad picture of how species are currently distributed. This is crucial in an ecosystem that is highly affected by diversity loss.