Two strategies have been used in the breeding of Brachiaria (syn. Urochloa) to produce new cultivars. The first involves exploring the natural variability existing in nature, which is the selection of ecotypes, mostly apomicts, from the diversity in germplasm banks. This strategy proved efficient originally and the cultivars in use in Brazil were derived in this way, but progress with this strategy is limited in the medium to long term. The generation of novel variability through crossings is the alternative strategy to continue producing new improved varieties. This has been possible only through artificial chromosomal duplication of diploid and sexual B. ruziziensis and the fact that this species is crosscompatible with B. brizantha and B. decumbens, in Brazil the 2 most important species for tropical pastures. Crossings between species have been carried out at Embrapa Beef Cattle, Campo Grande, MS, Brazil, since 1988 (Valle et al. 1993, 2001; Valle 1997; Resende et al. 2007). The initial success in obtaining interspecific hybrids in Brachiaria is indisputable and helped establish a practical methodology as well as allowed for the identification of apomictic parents with good combining ability. However, a comprehensive genetic improvement strategy to benefit from gains in quantitative characteristics associated with the majority of agronomic traits of economic interest in these species had not been started. Therefore, 3 methodologies are being used for this purpose: reciprocal recurrent selection; intra-population recurrent selection; and directed crosses between sexual