This study compared the gains obtained using different indexes of selection in segregating populations of soybean. A total of 381 progenies were evaluated in 28 augmented Federer blocks. Parameters evaluated: number of days to flowering, number of days to maturity, first pod height, plant height at maturity, lodging, agronomic value, number of branches, number of nodes, number of pods and grain yield per plant. Selection pressure of 30% was considered for direct and indirect selection index, classical index of Smith and Hazel, Williams index, weights and parameters free index of Elston, multiplicative index of Subandi, Compton and Empig, index based on the desired gain of Pešek and Baker, and the sum of ranks index of Mulamba and Mock. The selection index with good response for many traits was the multiplicative index of Subandi, Compton and Empig; the Williams index; the sum of ranks index of Mulamba and Mock; and, the index based on the desired gains of Pešek and Baker. Different economic weights did not influence the sum of the positive genetic gain significantly. The Restricted maximum likelihood methodology provided better quality variance components that were used in the indices to compose the matrix G or as economic weights.

Keywords: Economic weights, Genetic covariance, Glycine max, mixed models, REML.