Genetic effects from long-term selection in populations of maize (Zea mays L.)

Generation means analysis was adopted in the studies reported herein to determine estimates of components of genetic effects governing yield, index, and other agronomic traits in three maize (Zea mays L.) breeding stocks, namely Nebraska Krug Full-sib (NKF), Kitale Synthetic II (KSII) and Ecuador 573 (Ec573), the latter two being used in Kenya. Three and six generations were available for reduced- and full-model NKF, respectively. One and two generations were available for estimation of genetic effects in Kitale Experiments 1 and 2, respectively, involving KSII and Ec573. More than three generations, in an additive-dominance model, were required to estimate genetic effects in NKF. Response to selection in NKF for yield, index, ear height, plant height, and days to mid-anthesis was determined more by dominance than additive effects. However, non-barren plants and percent prolificacy were apparently controlled by additive effects. Random drift influenced response in a direction opposite to that due to selection for yield, index and days to mid-anthesis, but not for plant height and ear height. Index increased linearly at a rate of 0.14 Mg ha$\sp{-1}$ cycle$\sp{-1}$. However, response in yield and standing plants were quadratic, albeit with opposite signs. Non-barren plants and ears per 100 plants followed yield response trend. Plant and ear heights decreased linearly, the latter at a rate of 0.53 cm cycle$\sp{-1}$. Days to mid-anthesis followed a quadratic trend over cycles, reaching a peak at cycle seven. Selfing in the NKF population per se (S$\sb0$) decreased yield and index in the S$\sb1$ and S$\sb2$ generations, but percent standing plants increased. Parsimonious generation means models were applied to data from Kitale Experiments 1 and 2, due to paucity in number of generations available for estimation of genetic effects. No information was gained on the relative magnitudes of additive and dominance effects at play in KSII and Ec573 populations because the two effects could not be estimated separately. Nonetheless, useful comparisons of the two breeding stocks and two selection methods were made from contrast analysis and regression of trait means on cycles of selection