Mass selection for grain protein concentration based on seed density in common wheat

Mass selections for high and low imbibed seed density among F$\sb3$-F$\sb6$ families were evaluated for their effect on grain protein concentration of 12 winter wheat (Triticum aestivum L.) populations. In cycles I and II, low (LCI, LCII) and high (HCI, HCII) seed density fractions were derived using 10 and 15% selection intensities, respectively. Mean grain protein concentrations over locations in 1986 of 180, 178, 176, and 173 g kg$\sp{-1}$ for LCI, LCII, unselected, and HCII families, respectively, were significantly different. In cycle I, eight and four populations, and ten and seven populations in cycle II had significant increases and decreases in protein with selection for low and high seed density, respectively. Generally, selection for low seed density resulted in significant reductions in grain yields, while grain yields varied between cycles with selection for high seed density. Kernel weights among fractions varied between cycles. Genetic variation for grain protein concentration in the 12 populations was determined via evaluation of F$\sb4$-F$\sb7$ headrows derived from cycle I density selected and unselected bulks. Headrows derived from unselected populations had variances for grain protein concentration ranging from 1.12 to 1.64. Apparently, this was not sufficient to distinguish between populations, and there was little relationship between variance in grain protein concentration and change in protein level with selection. Headrows derived from low density selected populations had seed protein levels that were 2 g kg$\sp{-1}$ higher (p $\u3c$ 0.05) than headrows derived from unselected bulks. Headrows derived from high density selected and unselected populations were similar in grain protein concentration. Precision in estimation of variances and differentiation of means was limited due to the large amount of nongenetic variation governing grain protein concentration. Selection for seed density may allow plant breeders to increase the frequency of high protein genotypes in bulk populations and to increase the rate of gain from cyclic selection programs; however, the large nongenetic variation in protein of individual seeds may limit progress