Adequate characterization of soil heterogeneicy' of an experimental site is a goodrngu,id e and at times even a prerequisite to choose a good experimentation technique.rnBased on data from a uniformity trial on wheat, soil fertility pattern of an experimentalrnfarm was studied using the techniques of uniformity trials and spatial autocorrelationrnanalysis.rnFrom the soil productivity contour map, a uni-directional fertility gradient (from left tornright) was observed on fields using broadcast sowing method than using rowrnplantation method. The same result was a/so confirmed using the mean squarernbetween stripes method. The distribution of fertile plots was found to be relativelyrnheterogeneous for row planted fields than for broadcast sown fields. The degree ofrnadjacent plot correlation between yield of plots using broadcast sowing method wasrnrelatively higher than for row plantation method as measured by Smith 's coefficient ofrnheterogeneity indicating that the variability in yield due to any given plot size doesrnalso depend on crop geometry (type of plantation method). Both square andrnrectangular plots were equally efficient in reducing the error variance. A plot size of 8rnto 12 m2 was found to be optimum on both plantation methods.rnThe Moran's, I, and Geary's, c, coefficients were used to test against spatialrnautocorrelation between yields of neighbOring experimental plots. For all the threerncases considered to represent 'neighborhood' between plots, namely, Bishop's case,rnRook's case and Queen's case, both tests confirmed that there is a significant spatialrnautocorrelation between neighboring plots using broadcast sowing method. For rowrnplanted fields, the observed spatial autocorrelation became insignificant for a plotrnsize 30 m2 with bishop's case, 4 m2 with rook's case (using Moran's I), and 12 m2rnwith queens case (using Moran 's I).