Expected genetic responses were used to compare the simultaneous improvement of several agronomic traits for three selection index methods in recurrent selection programs. Observations for various agronomic traits in maize (Zea mays L.) were obtained from 144 S1 lines of the BSSS2 maize population grown in several locations during the 1970‐72 growing seasons. These traits were European corn borer (Ostrinia nubilalis (Hubner)) resistance, cold tolerance (percentage of emergence, emergence index, and seedling dry weight), grain yield, grain moisture, root and stalk lodgings, and corn root worm (Diabrotica spp.) resistance (root damage, root size, and secondary root growth). Seven of them, grain yield, corn borer resistance, percentage emergence, emergence index, seedling dry weight, root size, and secondary root growth, were used to construct three types of index selection for S1 selection.The conventional indices constructed from two sets of arbitrarily assigned relative economic weights did not give satisfactory predicted improvement for all traits included in individual indices. Most of the predicted improvement went to grain yield and percentage of emergence. Base indices, in which relative economic weights were used as index coefficients, were 95 and 97% as efficient as conventional selection indices. Hence, the application of the base index would be preferable with S1 selection when precise relative economic values of the traits included in the indices are known. When modified selection indices (based on desired gains) were used, final goals from recurrent selection of the traits considered (S1 testing and 10% selection intensity) were predicted to be attained after 14 cycles or 28 years of selection. In another scheme of recurrent selection, corn borer resistance and cold tolerance traits would be evaluated in the S1 generation, and yield, root size, and secondary root growth would be tested hi the S2 generation. With 44 and 8% selection intensities in the S1 and S2 generations, respectively, improvement of all traits was expected to reach final goals after 10 cycles or 30 years of selection. Assuming that the two sets of the arbitrarily assigned relative economic weights were realistic, modified selection indices (based on their contributions to aggregate genetic advance) were shown to be only 46 and 61% as efficient as conventional indices for the specified desired gains and relative economic weights considered. Modified selection index with the S1 testing scheme was expected to give 53% of improvement for yield, compared with single trait selection based on yield alone. Use of the modified index selection is recommended for the simultaneous improvement of several agronomic traits in recurrent selection programs when the relative economic values of the traits are difficult to determine.Results from comparisons of predicted gain from recurrent selection with S1 testing and a combination of S1 and S2 testings suggested that S1 testing would be preferable, because fewer resources would be required and the construction of selection indices would be less complicated.