[Objective] Sampling site spacing and soil classification granularity are two important indices for measuring representativeness of soil samples and accuracy of a survey. The objectives of this study were to characterize variability of soil bacterial diversity on different scales and explore effects of soil sampling site spacing and soil classification granularity on analysis of soil bacterial diversity.[Method] Considering differences in soil classification granularity and spatial scale (village, town and county), this study had 8 sample farmland areas set up in a region under the rice-wheat rotation system in Changshu City, Jiangsu Province, and four adjoining sampling sites (6 m by 10 m) in each sample area. Surface soil samples were collected from five sampling points, arranged in a plum blossom pattern, in each of the sampling site and then blended into one mixture sample, thus making up a total of 32 soil samples for analysis of soil bacterial diversity by means of the High-throughput sequencing analysis technique and eventually effects of spacing of sampling sites and classification granularity on variability of the diversity. Spacing of the sampling sites was designed to have four scales, i.e. <50 m, <10 km, 10-20 km and >20 km, and soil classification granularity, four levels, i.e. species, family, subgroup, and great group, which were formed into various spacing-classification combinations. Then comparisons were made between the combinations in coefficient of variation (CV) of α diversity index and bacterial community dissimilarity (β diversity).[Result] Results show that CV of the α diversity index of soil bacteria decreased with declining soil classification granularity and with declining sampling site spacing, as well, while bacterial community dissimilarity tended to decline slightly with lowering soil classification granularity, but to increase significantly with increasing sampling site spacing from a field scale to a town scale, but not so significantly when the spacing went from a town scale to a county scale. CVs of the α diversity index of soil bacteria and bacterial community dissimilarity were lower in the group of two factor combinations than in the one factor group. The decline of the CV of α diversity and bacterial community dissimilarity with declining soil classification granularity on the town scale(10-20 km)was more obvious than that on the village scale (<10 km). Dummy correlation analysis shows that bacterial community dissimilarity was more closely related to sampling site spacing (Pearson correlation coefficient:0.585, R²=0.34; Spearman correlation coefficient:0.503, P<0.01)than to soil classification granularity(Pearson correlation coefficient:0.340, R²=0.11; Spearman correlation coefficient:0.318, P<0.01). Variation of soil physico-chemical properties and farmland management measures with elevating spatial scale was an important factor affecting effect of special spacing of sampling sites on soil bacterial community structure, which indicates that sampling area and sampling site spacing should be the prime factors in the surveys of soil microbial community structure in farmland, and soil classification granularity was another one to be taken into account in surveys of areas above a town scale. The lower the soil classification granularity, the more representative the sampling points.[Conclusion] All the findings in this study may serve as scientific reference for future studies on soil microbial diversity in farmland on a county scale and arrangement of sampling points for the study.