This investigation was undertaken to study the optimum plot size and number of replications in fertilization experiments for wheat yield trials at Sakha Experimental Station during 1994/95 and 1995/96 sea‌sons. The experiment included 24 treatments which were the combina‌tions of two wheat cultivars, three organic manure level and four nitro‌gen levels. A split-split plot design with five replications was selected. The sub-sub plot unit area was 16 m2. The harvesting area was 9.0 m2. The grain yield data were recorded for each plot (kg/plot). They were subjeCted to two procedures of statistical analysis to estimate the opti‌mum plot size. The first was developed by Smith (1938), the second was the maximum curvature devloped by Lessman and Atkins (1963). The results obtained could be summarized as follows : 1. increasing the plot size decreased the variance per basic unit and the coefficient of variability. However, the reduction was not in propor‌tion with the increase in plot size. 2_ The index of soil variability ranged from 0.456 to 0.363 with an aver‌age of 0.409. 3.The exponential relationships between the coefficient of variability (C.V.) and plot size (X) were : C.V. = 20.57 X -0.3214for the first season. C.V. = 17.55 X -0.3129for the second season. 4.The optimum size of plot ranged from 1/229 to 1/140 faddan. 5.plot shape was important for small sized plot only, square plots were recommended in fertilization experiments. 6.Increasing plot size and/or number of replications reduced the magni‌tude of the difference detected at specified level of significance. The reduction of difference with increasing plot size was less than that obtained by equivalent increase in the number of replications.