An understanding of the level, structure and origin of genetic variation within and among populations of tropical trees is essential for devising optimum management strategies for their sustainable utilization and conservation. Here, amplified fragment length polymorphism (AFLP) analysis was used to partition genetic variation within and among nine populations of the predominantly riverine tree, Calycophyllum spruceanum, sampled across a wide geographical range along river tributaries of the Peruvian Amazon Basin. Analysis of molecular variance (AMOVA) employed 65 AFLP markers and revealed most variation among individuals within populations (91%), although variation among populations was highly significant (P < 0.001). Calculation of genetic distances and nested AMOVA indicated a degree of structuring among populations based on geographical proximity, although clustering did not depend on geographical distance alone. No firm evidence was obtained for unidirectional seed dispersal by water playing an important role in determining genetic structure over the geographical range sampled. Implications of data for optimising genetic management of the species are discussed and areas for further study identified.