Soil variability affects experimental results and we can minimize this effect by choosing a proper design and the use of optimum number of plots, blocks, samples as well as their sizes. This study aimed to investigate effect of these interrelationships on experimental and sampling errors, relative efficiency and coefficient of variation (CV) as measures of accuracy for randomized complete block design in sorghum experiments. Two uniformity trail were conducted at Shandwee' Agricultural Research Station, Sohag Governorate during 2004 and 2005 seasons. Dorado variety was used in both Seasons. Total field area was 2287.5 m2 (0.545 feddan ) for each trial that divided into 6 strips consisted of 120 ridges each considering ridge as the basic unit. Samples consisting of 5, 10, 15, 20, and 25 plants and data were assigned in randomized complete block design with different number of plots (it), blocks (b), sample (5) as well as their sizes in the two seasons. Results cleared that increasing sample size and number of blocks led to Increase relative efficiency. Increasing number of plots per block, with number of blocks and size of sample also led to increase relative efficiency and precision of the analysis. Results also cleared that the optimum combinations were 2,3 and 4 blocks of area 97.2, 129.6 and 144 m2 with number of plots of 9, 12 and 16 plot per block of area 10.8, 10.8 and 9 m2 as plot sizes and 25, 25 and 20 plants as size of sample, respectively. These combinations scored RE equal to 190, 197and 230% in first season and 149, 173 and 217% in second season, respectively. The coefficient of variation (CV) estimates scored 13.62. 8.64 and 7.22 % in first season and 11.45, 10.34 and 6.04 "/u in the second season, respectively. These results also showed that number of blocks is inversely proportional with sample size. Experimental (E) and sampling (8) errors values also compared with each other to estimate the optimum sample size. Results cleared that sampling error scored 0 075, 0.127 and 0.234 in first season and 0.254, 0-124 and 0.208 in second season while experimental error scored 0.326. 0.256 and 0.738 in the first season and 0.424, 0.333 and 0.959 in the second season for 2,3 and 4 blocks, respectively, these indicating homogeneity of sample size .