Response to selection for grain yield and harvest index in F2, F3 and F4 derived lines of two wheat crosses

Summary Aspects of selection for yield and harvest index were investigated by simulating selection using data from random pedigreed F₂, F₃, F₄ and F₅ derived lines from two crosses grown in plots at two sites over two years.Improvement in yield through selection was obtained when the response was measured at the same site and in the same year as the selection. Selecting the best 10 per cent of F₂ to F₄ derived lines gave F₅ derived lines that outyielded random selections by 19 to 53 per cent for one cross and 5 to 23 per cent for the second cross. These lines were 41 to 50 per cent better than the mid-parent in one cross, but were less than the mid-parent in the other cross.However, the response to selection when measured in a different year was little better than random selection. The effect of different sites also reduced the effectiveness of selection.Selection of harvest index in early generations for improvement of yield was ineffective when response was measured at the same site in the same year, or in different years.Contrary to some theoretical proposals, the same improvement in yield was obtained by selecting in early or late generations. While high yielding genotypes may be lost by delaying selection, this is counteracted by the better predictive value of late generations due to their greater homozygosity and homogeneity.