Abstract: 【Objective】 Two synthetic narrow-base maize populations, P3C0 and P4C0, were improved by 5 cycles of biparental mass selection, the effects on the population per se and combining ability were analyzed with the objective to make suggestions on how to raise the efficiency of improvement and utilization of synthetic populations. 【Method】 In different ecological environment, the phenotype of the populations per se were analyzed, and the combining ability were tested according to an incomplete diallel model.【Result】 Most traits of the populations per se and the GCA were improved significantly, whereas the SCA of most traits showed no significant increase over biparental mass selection. For P3C0 and its improved descendants, the mean of kernel depth and rows per ear were smallest at C0 and largest at C5. The GCA effects of them were smallest at C0, and largest at C2 and C5, respectively. The yield per plant of populations per se and its GCA effects were both smallest at C0, and largest at C4 and C2, respectively. For P4C0 and its improved descendants, the mean of kernel depth and rows per ear were smallest at C0, and largest at C5 and C4, respectively. The GCA effects of them were both smallest at C0, and largest at C4. The yield per plant of the populations per se and its GCA effects were both smallest at C0, and largest at C2 and C4, respectively. For most of the other traits per se and their GCA effects, the improved descendants were larger than their respective C0. There were different effects of improvement on different populations and traits at different cycles. The general trend showed that after cycles of biparental mass selection, some traits of populations per se and their GCA were improved simultaneously, while the improvement of population per se and the GCA did not display a synchronicity for some others. After the largest selection response to biparental mass selection was obtained, a sustained biparental mass selection would lead to a decrease in genetic gain or even a negative genetic gain. In addition, the populations retained their genetic diversity over 5 cycles of biparental mass selection with a population size of 600 plants and a selective ratio of 10%. 【Conclusion】 The improvement effect of the populations should be analyzed periodically, then adjust the improvement program according to the results. It is necessary to coordinate the relationship of the directly selected traits and the remaining traits in a narrow-base population improvement program via recurrent selection. When the major traits of population per se and GCA are improved effectively, a reciprocal recurrent selection method should be employed to improve the GCA and SCA synchronously, and then the germplasm improvement could combine with hybrid breeding.

Key words: maize, population improvement, biparental mass selection, trait, combining ability