Studies in western Amazonian forests have found that similarities in soil cation concentration and texture explain floristic similarities between sites, when these are measured using trees, pteridophytes or Melastomataceae. However, it is not known to what extent the three plant groups react to the same soil characteristics, because tree studies have almost always been conducted in different areas than studies on the understorey plant groups. We made inventories in 23 sites representing non-inundated rain forests on clayey to loamy soil in three regions of western Amazonia. Significant Mantel correlations between the floristic patterns of trees and pteridophytes were found in all three regions when floristic differences were measured with species presence–absence data. When species abundance data were used, and when the floristic patterns of trees and Melastomataceae were compared, significant correlations were found in one or two regions. Mantel correlations between plant groups were highest in the two regions where the observed variation in soil characteristics was largest. In all regions, the same soil variables emerged with significant Mantel correlations with trees, pteridophytes and Melastomataceae. Soil calcium and magnesium were most frequently retained in the models of multiple regression on distance matrices. On average, soil differences explained 50% of the variation in floristic differences (range = 14–84%), and geographical distances explained 16% (range = 0–64%). Our results demonstrate that beta diversities of the three plant groups are highly correlated, and that much of this congruence is explained by similar reactions to soil variation. These results support the idea that pteridophytes, and to a lesser degree Melastomataceae, can be used as indicators of general floristic and edaphic patterns in Amazonian rain forests. Since understorey plants are much quicker to inventory than trees, this would make it possible to recognize and map floristic patterns over huge areas of lowland Amazonia within a reasonable time.