The banana genome (Musa sp.) is invaded by numerous badnavirus sequences including those of banana streak viruses (BSVs). Badnaviruses are double stranded DNA pararetroviruses belonging to the family Caulimoviridae with any integration step into the host genome during their life cycle. The majority of viral integrants is mostly defective as a result of pseudogenisation driven by the banana host genome evolution. Conversely, some from BSVs only and named endogenous BSVs (eBSVs), can release a functional viral genome following stresses. All the badnavirus sequences described so far in banana [1] are spread among the three main groups structuring the badnavirus genus diversity [2]. We established that eBSVs are restricted to group 1, have an episomal counterpart, and are present in Musa balbisiana (B) genomes only [1;3;4] whereas the others are restricted to group 2 with any episomal counterpart described so far. We elucidated the sequence and the allelic organization for three BSV species (BSOLV, BSGFV and BSImV) present in the B genome of the seedy diploid PKW (BB), and developed several PCR and deCAPS markers for a specific eBSV genotyping [3-5]. Then, we investigated the early evolutionary stages of infectious eBSV through the study of their distribution, their polymorphism of insertion and their structure evolution when possible, among the banana B genomes diversity available through a sampling of BB seedy diploids extended to interspecific hybrids exhibiting different levels of ploidy for the B genome. We selected our sampling according to the two areas of sympatry between M. acuminata and M. balbisiana representing the centers of origin for the most largely cultivated AAB cultivars. One was in India and the other one in East Asia going from Philippines to New Guinea (Perrier et al, 2009). We used the PCR markers and Southern blots analysis to characterize PKW-related eBSV allelic polymorphism and we coded these results in order to create a common dissimilarity m