Three crosses were made among four bread wheat cultivars, Sahka 93 x Gemmeiza 7, Sakha 94 x Gemmeiza 7 and Sakha 94 x Giza 168. The five populations i.e P1, P2, Fl, F2 and F3 of each cross were used to estimate genetic variance, gene action and heritability for grain yield and its components. The experiment was conducted at Sakha Agricultural Research Station from 2003/2004 through 2006 / 2007 season. Significant positive heterotic effects were obtained relative to the better parents for plant height, spike length and 100-kernel weight in some crosses. Meanwhile, for maturity date, significant negative (favorable) heterosis relative to the earlier parent was detected in the first cross. Overdominance towards the higher parent was detected for days to heading and to maturity in the first cross, and plant height and spike length in the second and third crosses, respectively. While, overdominance towards the lower parent was detected for spike length only in the first cross. Partial dominance towards the higher parent was obtained for days to maturity in the second and third crosses , spike length in the second cross, and grain yield plant-1 in the first and second crosses. On the other hand, partial dominance towards the lower parent was observed for plant height in the first and third crosses, days to heading in the second and third crosses, number of kernel spike-1, 100 - kernel weight In the three crosses, and grain yield plant-1 in the third one. Significant positive estimates of inbreeding depression were detected for all studied characters except for days to heading in the first and second cross and days to maturity in the third one. The additive gene effect was significant for all characters in all crosses. In addition, dominance and epistasis effects were significant for most of the studied attributes. Estimates of heritability in broad sense were high to moderate for all studied characters In the three crosses. Moreover, moderate estimates of narrow sense heritability were detected for most studied characters. These results indicated that selection for the studied characters could be used in the early generations but would be more effective if postponed to late generations.