Replicate soil samples of 0.01, 0.1, 0.25, 1.0 and 10.0 g were taken from a single, large, homogenized sample from a field maintained as continuous meadow. The samples were processed for direct enumeration of bacterial cells and community structure assays by DGGE analysis of PCR-amplified 16S-rDNA fragments from whole community extracts. The goal was to determine the sample size or size range that produced the most consistent results (i.e., mean values) and the lowest variance. Enumeration data were analyzed by ANOVA, and the community composition fingerprints were analyzed by discriminant analysis (DA). Acceptable results were obtained for sample sizes from 0.1 to 1.0 g for both enumeration and community fingerprinting, but the size that yielded the best results for both measures was 0.25 g. The results suggest that for well homogenized silt loam soils with moderate organic matter concentrations, this sample size should produce high quality consistent results. For soils that differ in organic concentrations or clay content, a reconnaissance survey similar to the present examination is recommended.