The breeding materials used in this study were 48 families of flax derived from four crosses along with their four parents (P₁=Giza7, P₂=Giza8, P₃=S.2419/1, and P₄=S.148/6/1) as well as a bulk of each cross in F₃ and F₄ generations. These genotypes were grown in RCBD with three replicates at Etay El-Baroud exp. Sta., El-Beheira Governorate through the two successive seasons (2000/2001 and 2001/2002). The present study aimed to compare the improvement resulting from application of independent culling levels selection method with the bulk hybrid for straw and seed yields in early segregating generations of flax hybrids. The results obtained could be summarized as follows: 1- The results indicated that the possibility of using number of basal branches, technical length and plant height as selection criteria for improving straw yield per plant. In the meantime these traits gave low discrepancy between PCV and GCV values with high heritability as well as high genetic advance (GA%).  2- The cross, P₁xP₂ showed high heritability (h²) associated with high genetic advance (GA%) as well as narrow range between PCV and GCV for the three components of straw yield (plant height, technical length and number of basal branches), whereas the cross, P₃xP₄ gave similar results for only two components of straw yield (plant height and technical length) in both F₃ and F₄ generations. Therefore, these two crosses may be recommended for isolating superior pure lines for these traits in later generations. 3-The crosses P₁xP₃ and P₂xP₃ showed for all seed components high heritability and high genetic advance which indicated that it is due to additive gene effects. Particularly, the cross P₂xP₃ may be recommended for isolating superior pure lines for seed yield and its components in later generations. 4- The use of independent culling level selection (ICL) method in most cases, was more efficient in improving straw yield through selection for the two important components of straw yield viz., plant height and technical length than bulk hybrid and so, number of capsules per plant and1000-seed weight could be used as selection criteria to improve seed yield per plant. 5- Straw yield showed significant positive correlation with each of plant height, technical length, number of capsules per plant and seed yield per plant of the cross P₁xP₂. Seed yield exhibited significant positive correlation with number of capsules per plant in all crosses in both generations and 1000-seed weight in F₄ only. However, number of capsules per plant showed significant and negative correlation with both of 1000-seed weight and number of seeds per capsule in both generation of the cross P₃xP₄. Also, 1000-seed yield and number of seeds per capsule exhibited significant and negative correlation in two crosses, P₁xP₂ and P₁xP₃.