New genetic variance and covariance components related to intra- and interpopulational recurrent selection methods have been theoretically developed by Souza Jr. (Rev. Bras. Genet. 16: 91-105, 1993) to explain the failure of these methods to concomitantly develop hybrid and per se populations. Intra- and interpopulation half-sib progenies of 100 genotypes were sampled from maize (Zea mays L.) populations BR-106 and BR-105 to estimate variance and covariance components and to compare the expected responses to reciprocal (RRS), intrapopulational (HSS), and modified (MRS) recurrent selection in interpopulation hybrid, populations per se, and to determine heterosis. Four sets of 100 progenies, two intra- and two interpopulational, were evaluated in partially balanced 10 x 10 lattices arranged in split-blocks with two replications in two years (1991/92 and 1992/93) and two locations in Piracicaba, SP. Data for ear weight, plant and ear height, and ear by plant height ratio were recorded. Populations and interpopulation crosses were high yielding and showed high breeding potential for production of hybrids from inbred lines. Mid parent and the highest parent heterosis were relatively high, but lower than values reported for these populations under other environmental conditions. Additive variance estimates of populations per se and interpopulation crosses confirmed the high potential of these materials. The magnitude of the variance estimates for the deviations from intra- and interpopulation additive effects (<img src="http:/img/fbpe/gmb/v21n4/1918f1.jpg" alt="1918f1.jpg (1262 bytes)" align="middle"> for BR-106 and <img src="http:/img/fbpe/gmb/v21n4/1918f3.jpg" alt="1918f3.jpg (1259 bytes)" align="middle">for BR-105) and covariance between additive effects with these deviations (<img src="http:/img/fbpe/gmb/v21n4/1918f2.jpg" alt="1918f2.jpg (2022 bytes)" align="middle"> for BR-106 and <img src="http:/img/fbpe/gmb/v21n4/1918f4.jpg" alt="1918f4.jpg (1995 bytes)" align="middle"> for BR-105) indicated that these new components can significantly influence the effectiveness of breeding methods. Genetic component estimates for BR-105 had relatively small errors, with <img src="http:/img/fbpe/gmb/v21n4/1918f4.jpg" alt="1918f4.jpg (1995 bytes)" align="middle"> negative for all traits. Estimates of <img src="http:/img/fbpe/gmb/v21n4/1918f1.jpg" alt="1918f1.jpg (1262 bytes)" align="middle">and <img src="http:/img/fbpe/gmb/v21n4/1918f2.jpg" alt="1918f2.jpg (2022 bytes)" align="middle"> had relatively larger errors for BR-106. The MRS method was more effective than the RRS and HSS methods in producing hybrids from inbred lines. The choice of a population tester for the MRS method based on population means per se may be incorrect. The additional use, when possible, of intra- and interpopulation additive genetic variances from each population would be more appropriate.
