Two cycles of recurrent selection for the improvement lint percentage and other associated characters in two cotton crosses were applied at Sakha Agricultural Research Station Kafr El-Sheikh Governorate during six successive seasons (2008 – 2013).The results were as following. The mean values were increased with first and second cycles of recurrent selection as compared with the basic populations. Genotypic and phenotypic coefficients of variability were larger in F2 generation than F3 generations and GCV was reduced in later generation (second cycle) indicating that the advanced cycle was more homogenous and uniform than the basic population. In addition, moderate to high estimates of heritability in broad sense were observed for all studied traits in the two populations and increased from F2 to F3 generations. There were generally close agreement between predicted and observed responses to selection and predicted over estimated the actual ones for all traits. This suggests that dominance effects were lacking or of relatively with minor importance. Additive genetic effects would appear to predominate in the two populations. Hence, the breeder can depend on the predicted values. A comparison of direction and magnitude of associations among traits over the two recurrent selection cycles indicated that several new associations in terms of direction and magnitude were observed, these results in newer recombinants which would de due to changes in coupling to repulsion phase linkage. Lint percentage of the second cycle selection exceeded over the base population by 13.3% and17.73% in population I and II, respectively. Response to selection was linear and is predicted to continue at approximately the same rate of gain for an additional cycle. The correlation responses to selection for lint percentage were observed between some unselected traits measured. As selection increased lint percentage, simultaneous increases were observed for boll weight, fiber length and strength. While, micronair reading showed slight negative desirable change. Generally, recurrent selection has been very successful in increasing the population mean in the desired direction simultaneously for all studied traits without losing variability.