An understanding of the patterns of variation within and among populations of tropical trees is essential for devising optimum genetic management strategies for their conservation and sustainable utilization. Here, random amplified polymorphic DNA (RAPD) analysis was used to partition variation within and among 10 populations of the endangered Afromontane medicinal tree, Prunus africana, sampled from five countries across the geographical range of the species (Cameroon, Ethiopia, Kenya, Madagascar and Uganda). Analysis of molecular variance ( AMOVA) employed 48 RAPD markers and revealed most variation among countries (66%, P < 0.001). However, variation among individuals within populations and among populations within Cameroon and Madagascar was also highly significant. Analysis of population product frequency data indicated Ugandan material to be more similar to populations from Cameroon than populations from Kenya and Ethiopia, while Malagash populations were most distinct. The implications of these findings for determining appropriate approaches for conservation of the species, particularly in Cameroon and Madagascar, are discussed.