The genetic materials were F₂, F₃, F₄ and F₅ generations of the cross “Giza164/Sids4" of bread wheat. The experiments were conducted under drought stress and normal irrigation. Three cycles of single trait selection for grain yield/plant (GY/P) and five models of desired gain selection index (combinations from seven traits) were achieved in both environments separately. After the third cycle of selection, the selected families for the six types of selection under drought stress, and the selected families under normal irrigation were evaluated under both environments. The GCV in GY/P decreased from the F₂ to the F_5. The remained variability in GY/P for the five selection indices was higher after the three cycles of selection than the single trait selection for GY/P. Heritability estimates as calculated from regression of F₅/F₄ generations for the traits under selection pressure were higher under drought stress than under normal irrigation except few cases. The observed gain increased from index₁ to index_5, and all indices were better in improving GY/ P than direct selection for GY/P. It could be concluded that selection index was better in improving grain yield than single trait selection for GY/P. Furthermore, selection under drought stress (antagonistic selection) in these materials was better than selection under normal irrigation (synergistic selection) in improving GY/P either selection evaluated under drought stress or under normal irrigation. Selection index₅ which involved GY/P, 100GW and NG/S was the best index in improving GY/P.