The effectivinese of early generation testing for combinig ability for grain yield in grain sorghum restorer lines was studied at College Station, Texas, Two well-Known restorer lines, TX 7078 and TX 09, were crossed and resulting F@, F$, and F^ segregates were evaluated for combining ability. In the F2 population, 70 F2 plants were crossed to each of four male-sterile lines, A Tx 3197 (Combine Kafir-60), A Tx 398 (Martin), A Tx 378 (Redlan), and A Tx 3048 (Redbine sel.). These "F2 hybrids" were grown in a replicated test in 1962. Based on the mean yield of the four hybrids with each F2, 12 high- and 12 low-combining F2's were selected for advancement to the F4 generation. The F4 lines from each of these 24 selected F2's were crossed to each of the same four male-steriles and the hybrids tested in 195=63. Three high-and three low-combining F2 families were closen for further study in the F6. The F6 lines were crossed of the same four male-steriles and the hybrids tested in 1964. By compring, within families, the yield of "F2 hybrids" with the yield of "F4 hybrids" and "F6 hybrids, the effectiveb=ness of early generation testing for combining ability was estimated. High-combining F2's produced F4 lines and F6 lines that were significantly higher in combining ability than the F4 and F6 lines derived from low-combining F2's. Correlation coefficients for yield between "F2 hybrids" and "F4 hybrids', F2 hybrids" and "F6 hybrids"' and "F4 hybrids" and F6 hybrids" were generally large, positive, and significant (...).