For cotton like for most important crops, future improvements of adaptation to adverse environment, of agronomic fitness and of quality of agricultural products, are expected to rely on the better utilization of genetic resources. Opportunities for such utilization are offered by the techniques of molecular biology through molecular breeding and genetic transformation. Within the major tetraploid cultivated species, G. hirsutum, two pools can be distinguished as sources of genetic variability for cotton breeders: the cultivated pool and the primitive, or exotic, pool. It is recognized that a severe bottleneck has accompanied the domestication process leading to modern commercially important cotton cultivars, resulting in a reduced genetic base. On another hand, the exotic tetraploid germplasm pool is constituted of landraces (7 geographical races belong to the G. hirsutum sp.), feral accessions and wild tetraploids, most being perennial and photoperiodic shrubs. This important gene reservoir, though constituted of morphologically and ecologically diverse germplasm, has received minor attention. Studies using isozymes and RFLP markers have early shown the limited diversity in the cultivated pool of G. hirsutum sp, thus explaining the choice made by many teams to initiate in the early 90s essentially interspecific marker-assisted breeding programs. The development of more efficient marker-systems, like AFLPs, and the availability of an increasing number of sources of microsatellites (SSRs) in the public (ca 1000) and private (min 2000) has now changed the vision. We used different sources of SSRs to assess the molecular polymorphism of a set of G. hirsutum cultivars and races, as well as of exotic tetraploid accessions. The 2 primary sources of SSRs belong to the "BNL" and "CIR" series, all chosen after placement of corresponding loci on our genetic map (Nguyen et al. TAG, 2004, 109:167-175). The plant material comprised a collection of 48 accessions, chosen in our ge