Assessing soil microbial community structure by the use of molecular techniques requires a satisfactory sampling strategy that takes into account the high microbial diversity and the heterogeneous distribution of microorganisms in the soil matrix. The influence of the sample size of three different soil types (sand, silt and clay soils) on the DNA yield and analysis of bacterial and fungal community structure were investigated. Six sample sizes from 0.125 g to 4 g were evaluated. The genetic community structure was assessed by automated ribosomal intergenic spacer analysis (A-RISA fingerprint). Variations between bacterial (B-ARISA) and fungal (F-ARISA) community structure were quantified by using principal component analysis (PCA). DNA yields were positively correlated with the sample size for the sandy and silty soils, suggesting an influence of the sample size on DNA recovery, whereas no correlation was observed in the clay soil. B-ARISA was shown to be consistent between the different sample sizes for each soil type indicating that the sampling procedure has no influence on the assessment of bacterial community structure. On the contrary for F-ARISA profiles, strong variations were observed between replicates of the smaller samples (<1 g). Principal component analysis analysis revealed that sampling aliquots of soil > or =1 g are required to obtain robust and reproducible fingerprinting analysis of the genetic structure of fungal communities. However, the smallest samples could be adequate for the detection of minor populations masked by dominant ones in larger samples. The sampling strategy should therefore be different according to the objectives: rather large soil samples (> or =1 g) for a global description of the genetic community structure, or a large number of small soil samples for a more complete inventory of microbial diversity.