[[abstract]]本試驗係將台中在來1號 × CP231、嘉農242號 × IRP-60與C223 × Piro4等三種雜交組合之F2集團分別以世代促進法及一般慣行法繁殖，所得F3集團則一方面再以世代促進法及慣行法分別在室內及田間栽培，另一方面則將此兩種來源不同之F3集團同時按慣行法在田間栽培，以觀察其主要農藝性狀如生育日數、株高、穗長、一穗粒數，每株穗數及每株谷重之變異性。 以世代促進法栽培之F3集團，因受空間及養料之限制，致其主要農藝性狀之平均值均較以慣行法栽培者為小，而生育日數受世代促進法栽培條件之影響顯然較其他性狀為輕。兩種來源不同之F3集團同時以一般慣行法在田間栽培時，兩者之生有日數頗為相似，惟來自世代促進法之F3集團，其平均株高似較來自慣行法者為小，而前者之平均每株穗數及每株谷重則較後者為大。 主要農藝性狀之變方及變異係數，以世代促進法栽培之F3集團與以慣行法栽培者，雖因雜交組合之不同而有大小之別，但兩者之間差異不大，兩種來源不同之F3集團以慣行法栽培時，來自世代兄進法者其有關性狀之變方及變異係數在台中在來1號 × CP231什交組合均較來自慣行法者為小，而在嘉農242號 × IRP-60則均較來自慣行法者為大，C223 × Piro4則因性狀而不同，惟來源不同之F3集團間之差異不大，由此可知水稻雜種集團若以世代促進法繁殖時，亦可保持該集團內有關性狀之正常變異性。 F2 generation of the crosses, Taichung native 1 × CP 231, Chiangung 424 × IR 9-60, and C 223 × Pi No.4 were raised by Both “Generation Acceleration Method” and conventional one, and F3 populations were continuously grown by the sane method as their respective F3 populations. At the same tine, F3 populations derived from different methods were grown together in the field by the conventional method. The general performance as well as variability of some agronomic characters in the above mentioned F3 populations were investigated. The growing conditions of the “Generation Acceleration Method” greatly reduced the means of all characters except that of heading date when compared with those grown by the conventional method. When two F3 populations of different origin were grown together in the field, they appeared to have similar heading date but the plants from the “Generation Acceleration Method” were slightly shorter with more panicles per hill and heavier grain weight per hill in comparison with those of from conventional method. Variance and coefficient of variation for agronomic traits differed slightly with growing conditions. For example, variance associated with heading date in the F3 populations grown by “Generation Acceleration Method” was considerably greater than those of conventional method. When two F3 populations of different origin were grown together in the field, it was observed that in the cross of Taichung native 1 × CP 231, variance and coefficient of variation of agronomic traits were larger in the F3 population derived from “Generation Acceleration Method” than those from the conventional method, whereas in the cross of Chianung 242 × IR 9-60, the opposite situation prevailed. In the cross of C 223 × Pi No. 4, however, F3 population of “Generation Acceleration Method” had smaller variance and coefficient of variance and coefficient of variation in heading date and plant height but had larger variance and coefficient of variation in number of panicles per hill and weight of grains per hill than F3 population from conventional method. The difference in variance and coefficient of variance between two F3 populations of differ rent origin was generally small, indicating that hybrid population treated with “Generation Acceleration Method” may have the same variability for the agronomic traits as that handled by the conventional method.