In a study of the variation among isolates of Banana streak virus (BSV) in Uganda, 140 sequences were obtained from 49 samples by PCR across the conserved reverse transcriptase/RNaseH region of the genome. Pairwise comparison of these sequences suggested that they represented 15 different species and phylogenetic analyses showed that all species fell into three major clades based on 28% sequence difference. In addition to the Ugandan sequences, clade I also contained BSV species that are known as both integrated sequences and episomal viruses; clade II also contained integrated BSV sequences but which have not previously been identified as episomal viruses. Clade III comprised of Sugarcane bacilliform virus isolates and Ugandan BSV sequences and for which there is no evidence of integration. The possible reasons for the extraordinary levels of virus sequence variation and the potential origins and epidemiology of these viruses causing banana streak disease are discussed.