Selection for Yield and Yield Components in Wheat1

Abstract

From an F2 population of about 700 plants from across of two spring wheats (Triticum aestivum L.) CI 13242 and ‘Thatcher,’ 10 plants were selected for maximum expression of grain yield per plant, spike number/plant, kernel number/spike, spikelet number/spike, and kernel weight. In each subsequent generation through the F8, selection lines for each of the five characters were advanced by selecting the 10 plants with the highest value from a population of about 400 plants. A performance trial was conducted at three locations in 1 year with five generations (F4 to F8) of each of the five selection lines and the two parents. Yield and yield components were evaluated and compared with midparent values and as linear regressions on generation number. Selection for kernel weight and kernels/spike at the F8 generation gave 11 and 16% increases in yield over the midparent, whereas direct effects of selection for these two characteristics were 13 and 10%, respectively. Direct effects of selection for grain yield and spikes/m2 at the F8 generation were significantly lower than the midparent (—13 and —7%). Nonsymmetrical correlated responses for increased kernel weight with selection for high kernel number and for decreased kernel number with selection for high kernel weight were large and unexpected. Spike number per plant was not an effective selection criterion for increasing spike number per m2 or grain yield. Only 3 of 20 regression coefficients for response in the F4 through F8 generations were significant, compared with 15 of 20 significant differences from the midparent at F8. Apparently selection was effective in the F2 and F3 generations (13 of 20 midparent differences were significant at F4), with little subsequent response. In this population, kernel weight and kernel number per spike were good characters for indirect selection for yield improvement.