The main objective of this study was to estimate and compare the genetic gain obtained from Simith-Hazel index model of selection index with direct and indirect selection to enhance selection efficiency of superior promising families in early segregating generations of the cotton cross Giza 86 x Karshenky. The data revealed increase in mean values for all characters with advanced generations' from F2 to F4 except micronaire reading which showed lower values (desirable values). This shifting in mean values in desirable direction could largely be attributed to the possible accumulation of favorable alleles as a result of selection procedures adapted in this study. The range was comparatively wider in F2 generation as compared with the later generationsF3 and F4 for all studied characters. The advanced generations (F3 and F4 generations) showed reduction in PCV and GCV values as compared with F2, this may due to reduction in genetic variability and heterozygosity as a result of using different selection procedures which exhausted a major part of variability. Most characters showed high heritability values over 60% over generations indicating high magnitude of genetic variability. Significant desirable correlations between boll weight and each of seed/boll and seed index were existed over the three generations. The significant undesirable association existed between seed cotton yield with most traits in F2 generation were broken up and converted to desirable relation in later generations. Principal component analysis grouped estimates variables into six main components. In the first PC1 yield characters (lint yield followed by (x1) and seed cotton yield/plant showed negative loading and more contributed than the other fiber characters which gave negative loadings. The PC2 was great influenced by seed index followed by lint/seed and boll weight which had positive loadings, in same time micronaire reading and fiber strength showed negative loadings. Ten out eleven selection indices were more efficient than direct selection for improvement of lint yield in F2 population. The highest predicted genetic gain from F2 generation for lint yield/plant was observed when selecting for lint yield/plant with bolls/plant (IW1) followed by (I12) selecting for boll/plant with seed/boll and selection index in involving selection index involving lint yield/plant, bolls/plant, seeds/boll and lint/seed. The highest actual genetic gains from F3 generation for lint yield/plant occurred when selecting directly for lint yield/plant followed by selecting for boll/plant. However the indices IW12 (selection index involving lint yield/plant, bolls/plant and seeds/boll) followed by IW13 (selection index involving lint yield/plant, bolls/plant and lint/seed) and IW3 (selection index involving lint yield/plant and lint/seed) were superior to all selection procedures in amount of actual gain. Most indices showed high discrepancy between predicted and actual genetic gain as lint yield/plant, this was due to non-additive gene effect and large effect of environmental factor. However maximum actual genetic advance from F4 generation for lint yield/plant were achieved when selecting for lint yield/plant, seeds/boll and lint/seed followed by selection indices containing lint yield/plant, boll/plant, seed/boll. The direct selection for lint yield and pedigree selection for boll/plant followed by selection for lint/ seed gave desirable actual values and surpassed most indices. selection index involving lint yield/plant and boll/plant surpassed all selection procedures for predicted gain followed by Selection index involving lint yield/plant, bolls/plant, seeds/boll and lint/seed. However direct selection for lint yield followed by pedigree selection for boll/plant appeared to be most effective for the improvement lint yield and gives reasonable actual gains. The predicted and actual advances determine from F3 generation were higher than F4 generation for most selection procedures. On the basis of various selection procedures, six selected families were isolated from F4 generation by superiority of these families from better parents, F3 families and point start of F2 plants mean. The breeder may utilize such selected families in breeding programs aiming to improve yield and quality.