Cultivation of the woody legume Sesbania sesban in agroforestry systems has been widely promoted for soil fertility improvement in Africa. Despite the existence of natural populations in eastern Zambia, considerable seed originating from Kenya has been introduced to the area, while further introductions from Malawi are planned. Large-scale anthropogenic seed dispersal raises concerns regarding genetic management, which were addressed here by random amplified polymorphic DNA (RAPD) analysis. Eleven populations of Sesbania from southern and East Africa were assessed, including S. sesban populations already or likely to be introduced into eastern Zambia and reference taxa. Seventeen primers revealed a very large number of polymorphisms (538) in surveyed material. Differentiation among S. sesban populations was unusually high for a woody perennial (70% of variation among populations), with all individuals stringently assigned to specific populations in cluster analysis, even for geographically proximate sites. High differentiation, combined with the unusually large number of polymorphisms determined, resulted in the identification of a number of population-diagnostic markers. High differentiation may reflect the restricted ecology of S. sesban, human disturbance and frequent natural fluctuations in niche availability. Hydrochory, posited as a mechanism for gene flow between populations located along the same drainage system, appears to be discounted. Geographic proximity was not always a good guide for genetic similarity among S. sesban populations. In particular, rather proximate populations from Zambia and Malawi appeared highly differentiated genetically, while Zambian populations were relatively similar to a geographically distant population from Namibia. Differentiation was consistent with the presence of two botanic varieties in the southern Africa region, vars. nubica and zambesiaca. Estimates of Nei's genetic diversity suggested that S. sesban populations from East Africa were most diverse. Overall, population diversity estimates were low, possibly due in part to frequent bottlenecks, short longevity and a short juvenile phase. The unusually large number of polymorphisms detected here provided an opportunity to test for data redundancy and primer selection methodologies during RAPD analysis. Combined data from a subset of eight primers selected according to three different approaches provided a very good overall fit to the complete data set (rM > 0.95). Different primer selection approaches influenced the reliability of clustering and diversity analysis, with a minimum of eight primers required for reliable ranking of diversity estimates. The implications of our data for conservation, cultivation and monitoring of S. sesban in the southern Africa region are discussed. Caution is advised during the introduction of germplasm into eastern Zambia from both distant (Kenya) and relatively proximate (Malawi) sources. Keywords: Genetic management, Germplasm introduction, Random amplified polymorphic DNA (RAPD), Sesbania sesban.