Selection for Agronomic Characters in Hill and Row Plots of Soybeans

The objectives of the study were to compare the effectiveness of selection for yield, maturity, height, lodging, and phenotypic score among soybean [Glycine max (L.) Merr.] lines in hill and row plots and in random and nonrandom complete-block designs. Progeny from 50 random S₃ plants of Maturity Group II were grown in two-row, unbordered plots with 69 cm between rows within the plot and 102 cm between plots and in single, unbordered hill plots spaced 102 ✕ 102 cm. A randomized complete block design was used for the row plots, and both randomized and nonrandomized complete-block designs were used for the hill plots. There was no difference in the effectiveness of selection for yield between random and nonrandom hills. The hill and row plots were equally effective in selecting the top two lines. Actual genetic gain was 2.5 q/ha for random hills, 2.8 q/ha for nonrandom hills, and 3.0 q/ha for row plots with a 10% selection intensity. Considering possible differences in the selection differential and number of years per cycle between hill and row plots, the predicted genetic gain per year was 0.6 q/ha for random hills compared with 0.4 q/ha for rows. Selection for maturity, height, lodging, and phenotypic score was similar in effectiveness for the three plot arrangements. Genotypic correlations of line performance between random and nonrandom hills had coefficients of 0.99 to 1.02 for the four characters. The genotypic correlations between hills and rows ranged from 0.89 to 0.99. Hill plots were as effective as rows for phenotypic selection of agronomically desirable genotypes. Evaluation of yield, maturity, height, lodging, or phenotypic score can be done effectively in single, unbordered hill plots. The primary advantages of hills are the reduction in number of years required for cultivar development or recurrent selection by eliminating the need for a generation of seed increase and the larger number of lines that can be tested. Nonrandom hill plots may be useful in cases where randomization prevents efficient discarding of inferior lines before harvest.