The objective of this work was to evaluate the perspectives of using early generation testing for grain yield in soybeans. The base material comprises a population derived from a two-way cross between lines 14 and 38, divergent for grain yield. One hundred progenies F2:4 and one hundred progenies F4:6 were derived from this population and evaluated in the 2006/07 growing season, using a lattice design with 4 replications, in two locations, at Piracicaba, SP. Plots consisted of 2 meter long rows spaced 0,5 meter apart, with 35 plants after thinning. The components of variance, coefficients of heritability on a plot mean basis and expected response to selection were estimated for grain yield from the joint analysis of variance, for F2:4 and F4:6 generations. Estimates of expected response to selection were about 60% higher in the F4:6 generation; however this superiority would probably be offset by the larger time to complete a cycle. General results indicate that early generation testing for grain yield in soybeans can be effective for moderate selection intensities.
early generation testing; early selection; selection response