2002, Brazilian Society of Plant Breeding 369 Crop Breeding and Applied Biotechnology, v. 2, n. 3, p. 369-374, 2002 INTRODUCTION The International Rice Research Institute (IRRI) released IR8 for cultivation in 1966, the first modern irrigated rice cultivar and the precursor of the ‘Green Revolution´ that had a strong impact on world rice cultivation. Since then, breeders have used this cultivar, or lines derived from it, intensively as parents in crosses, a strategy that has limited the genetic variability of populations used in breeding programs, contributing to the establishment of a yield plateau. Studies by Peng et al. (1999) showed that the yield potential of new bred rice cultivars, indica subspecies, have remained the same over 30 years, since the release of IR8. Brazil re-organized its agricultural research structure in the mid 1970s, and irrigated rice researchers were able to take advantage of all the advances achieved by international research groups, which helped them meet their goals in a shorter period of time (Morais and Rangel, 1997). All the effort was rewarded in the beginning of the 1980s when the traditional high stand varieties were substituted with modern low stand varieties, practically doubling the irrigated rice yield in several states in Brazil. In Rio Grande do Sul, yield in the field increased by 30% (Carmona et al., 1994) due to modern cultivars and better crop management. After this great advance, irrigated rice yield has remained the same, and efforts to increase yield potential in cultivar populations have not resulted in expressive gains (Rangel et al., 2000a). Two factors may account for the stagnation of the yield potential of the irrigated rice cultivars in Brazil: a) prioritization of grain quality and disease resistance in detriment to yield; and b) the use of a few repeated parents in the formation of breeding populations. Recently researchers involved in the development of irrigated rice cultivars have tried to increase the yield potential through various strategies. Two lines of research are used in the IRRI: a) development of a new type of plant using the japonica tropical germplasm, whose results have not been encouraging up to now (Peng et al., 1999) and b) development of indica rice hybrids for tropical environments using Chinese technology. It is unlikely that these hybrids will be cultivated in Brazil because the seeds are very expensive and a highly elaborate and therefore costly cultivation system is required. In Brazil, one of the research lines used by the irrigated rice genetic breeding program at the Brazilian Agricultural Research Corporation - EMBRAPA to increase the yield potential consists of synthesizing populations of wide genetic base, followed by recurrent selection breeding. This strategy ensures the systematic obtainment of continuous gains, especially for grain yield, besides Grain yield gains in three recurrent selection cycles in the CNA- IRAT 4 irrigated rice population Paulo Hideo Nakano Rangel*; Orlando Peixoto de Morais and Francisco José Pfeilsticker Zimmermann Embrapa Arroz e Feijão, Caixa Postal 179, CEP 75375-000, Santo Antônio de Goiás, GO, Brazil. (* Corresponding Author. E-mail: phrangel@cnpaf.embrapa.br) ABSTRACT One line of research on irrigated rice genetic breeding programs at Embrapa consists of synthesizing wide genetic base populations followed by recurrent selection. This strategy ensures systematic continuous gains for grain yield and other traits of interest. To estimate the gains observed in three recurrent selection cycles in the CNA-IRAT 4 population, the grain yield data from 924 S 0:2 families in 14 experiments carried out in various Brazilian states in the 1992/93, 1994/95 and 1997/98 growing seasons were analyzed. A triple lattice (two 10 x 10 and two 8 x 8 lattices) experimental design was used in the first cycle and Federer augmented blocks in the two subsequent cycles. A non -significant gain was observed in the first cycle (only 15.7 kg/ha (0.28%). The gain observed in the second cycle and the mean gain observed due the selection in the first and second cycles were significante and of high magnitude (369.9 kg/ha (6.65%) and 259.9 kg/ha (4.67%), respectively). Results showed that recurrent selection applied to genetically divergent populations can result in considerable gains for grain yield. KEY WORDS: Oryza sativa, genetic gain, genetic breeding.