Wallace's riverine barrier hypothesis postulates that large rivers, such as the Amazon and its tributaries, reduce or prevent gene flow between populations on opposite banks, leading to allopatry and areas of species endemism occupying interfluvial regions. Several studies have shown that two major tributaries, Rio Branco and Rio Negro, are important barriers to gene flow for birds, amphibians and primates. No botanical studies have considered the potential role of the Rio Branco as a barrier, while a single botanical study has evaluated the Rio Negro as a barrier. We studied an Amazon shrub, Amphirrhox longifolia (A. St.-Hil.) Spreng (Violaceae), as a model to test the riverine barrier hypothesis. Twenty-six populations of A. longifolia were sampled on both banks of the Rio Branco and Rio Negro in the core Amazon basin. Double digest RADseq was used to identify 8,010 unlinked SNP markers from the nuclear genome of 156 individuals. Data relating to population structure support the hypothesis that the Rio Negro acted as a significant genetic barrier for A. longifolia. On the other hand, no genetic differentiation was detected among populations spanning the narrower Rio Branco, which is a tributary of the Rio Negro. This study shows that the strength of riverine barriers for Amazon plants is dependent on the width of the river separating populations and species-specific dispersal traits. Future studies of plants with contrasting life history traits will further improve our understanding of the landscape genetics and allopatric speciation history of Amazon plant diversity.