Abstract
Three replications of twenty-two grain sorghum genotypes were planted in a randomized block design. Plots consisted of three rows. Yield records were taken from eight single plants as well as from 2-m samples of each row. The components of variation and estimated gain from selection were used to indicate optimum allocation of experimental material. Both methods yielded similar results. At least six plots were indicated to be necessary to obtain a standard error of 10% of the experimental mean yield. The number of rows per plot played a negligible role in increasing the accuracy of the experiment. The number of plants per row could be as few as ten. The trial occupying the smallest area would be one that included single rows of ten plants each replicated eleven times. The results were discussed with respect to the variability observed in and the limitations of the present experiment.