Two cycles of pedigree selection method for grain yield/plant were conducted on a segregating population (TP24D) of grain sorghum in the F₃ generation . Selection was practiced at Shandaweel, Agric. Res. Center in 2002 and 2003 summer seasons . After the second cycle, the selected  families were evaluated at  Shandaweel (Sohag), Agric. Res., Station, Sohag (optimal environment) and Qena  (sub-optimal environment), South Valley University Experiment    Farm, Faculty of Agric. After two cycles of pedigree selection showed 5.87 and 3.72% increase in  grain yield/plant over the bulk sample at Sohag and Qena,  respectively, with an average of 4.98% over the two locations .  Selection for grain yield/plant was accompanied by decrease in  days to 50%  flowering of -7.43, -6.66% and -7.03 %, increase in  1000-grain weight of 4.27, 5.70 and 4.89% over the bulk sample at  Sohag, Qena and across locations,  respectively .                                                        Synergistic selection for grain yield/plant under optimal environment (Sohag) was better than the antagonistic one under sub-optimal environment (Qena). Sufficient genotypic coefficient of variability for grain yield/plant and other studied traits was observed .Phenotypic and genotypic variances expressed as pcv% and gcv% slightly decreased after the second cycle of selection in 1000-grain weight and  days to 50%  flowering compared to the first cycle . However, the variability in plant height was increased by selection for grain yield/plant . Family no. 6 could be considered the best selected family at  Sohag,  Qena and across locations which showed 44.26, 27.20 and 36.56% increase in yield, and was earlier than the bulk sample 4.44, 4.01 and 4.22% and gave heavier grains than it by 21.25, 11.30 and 16.57%, respectively . These results reflect that the direct selection for grain yield/plant was effective in improving these materials.