The present study was carried out in three successive seasons from 2007 to 2010.The first and second seasons were at El-Giza Research Station while the third season was at Kafr El- Hamam Research Station. The main objective of this study was to estimate combining ability variances and effects of eighteen wheat genotypes which were crossed to three local wheat cultivars, i.e., Gemmiza 9, Sakha 93 and Giza 169 as testers using Line x Tester analysis in F1 and F2 populations for grain yield and its components. The differences among genotypes, parents and crosses were highly significant for all studied characters. Both general and specific combining ability variances were highly significant for all characters. The results revealed that non additive gene effects were larger than those of the additive ones and played the major role in the inheritance of all studied characters. Also, the maximum contribution to total variance of grain yield per plant and other traits was made by female lines. The parental lines 3, 4, 14 and 15 proved to have better general combiners for grain yield per plant and most of its components in grain yield improvement. Hence, four female parents could be used in future programs to improve grain yield. Tester number 1 appeared to have better general combiner for grain yield per plant and some other traits. These parents presumed to have a relatively large number of favorable alleles for these traits. However, the hybrids (L3 X T3), (L4 X T1), (L14 X T2) and (L15 X T2) were excellent harmonious combinars for specific combining ability effects for all characters including grain yield per plant. Hence, it would result in transgressive segregants for grain yield per plant in late segregating generations due to the predominant non – additive gene action.