The maize population Csyn 4 was improved for three cycles with three recurrent selection methods MSI, MS1-HS, and MHSRRS in northern China. A total of 40 NC II testcrosses were made by four testers with Csyn 4 and 10 improved populations, which were evaluated in four environments in the cropping season of 2005. Analysis of variance indicated a significant progress (P＜0.05) in yield and other chief agronomic traits in the improved populations and testcrosses, demonstrating that the three recurrent selection methods were effective for increasing grain yield of testcrosses and improvement of general combining ability in maize population. The average grain yield increase of population Csyn 4 in MSI, MS1-HS, and MHSRRS recurrent selections was 266.7 kg ha-1 (5.3%), 288.0 kg ha-1(5.7%), and 231.3 kg ha-1(4.6%) per cycle, while the grain yield of SI progeny of population for Csyn 4 was increased by 420.0 kg ha-1 (10.9%), 376.0 kg ha-1(9.8%), and 414.7 kg ha-1(10.8%) per cycle in MSI, MS1-HS, and MHSRRS recurrent selections, respectively. Linear responses (b) in the MSI, MS1-HS, and MHSRRS recurrent selections were the same as the realized responses of the selection (G) for grain yield increase of Csyn 4 and SI progeny of the populations. The grain yield of testcrosses increased by 3.3-5.2% on average per cycle. The value of heterosis did not display any differences amongst the different cycles of selections. The GCA for yield component traits of population Csyn 4 was improved significantly by means of three selection methods, whereas the GCA for plant height, ear height, days to silking, and days to pollen did not show significant increase in the recurrent selections. GCA for the number of tassel branches were decreased in the selections, while GCA for the number of plant leaves were increased, but the difference was not significant.