The present study was carried out at Shandaweel Research Station, Cotton Research Institute, Agriculture Research Center, Sohag, Egypt during the three summer seasons of 2014-2016. The basic materials consisted of F2- population stemmed from the cross between (Giza 90 X Giza 80). The population was subjected to pedigree selection for two cycles. The selection procedures were single trait selection for lint yield/plant, in addition to eight selection indices. The index selection proposed by Pesek and Baker (1970) was used with different combinations of characters. The main objective of this study was to determine efficiency of selection index procedure in isolate elite high yielding genotypes. Average observed genetic gain of the ten selected families after tow cycles of selection indicated that lint yield/ plant ranged from 9.08% (P≤0.01) for index 7 to 22.73% (P≤0.01) for index 2. Index 5 and index 8 ranked the first and showed significant genetic gain of (12.30 and 15.00%), (17.88 and 15.43%), (13.60and 13.97%), (5.38 and 7.20%) and (17.76 and 10.36%) for seed cotton yield, lint yield, boll weight, seed index and lint index; respectively. Index 2 ranked the second who gave significant genetic gain of 20.75, 22.73, 15.81, 7.53 and 15.43% for the same previous traits respectively. While, the single trait selection for lint yield/plant ranked the last and showed significant genetic gain of 11.08, 9.69, 16.18, 7.96 and 10.78% for seed cotton yield, lint yield, boll weight, seed index and lint index; respectively. These results indicated that selection index was better than single trait selection in isolated superior families in yield and yield components.