Natural fluctuations in soil microbial communities are poorly documented because of the inherent difficulty to perform a simultaneous analysis of the relative abundances of multiple populations over a long time period. Yet, it is important to understand the magnitudes of community composition variability as a function of natural influences (e.g., temperature, plant growth, or rainfall) because this forms the reference or baseline against which external disturbances (e.g., anthropogenic emissions) can be judged. Second, definition of baseline fluctuations in complex microbial communities may help to understand at which point the systems become unbalanced and cannot return to their original composition. In this paper, we examined the seasonal fluctuations in the bacterial community of an agricultural soil used for regular plant crop production by using terminal restriction fragment length polymorphism profiling (T-RFLP) of the amplified 16S ribosomal ribonucleic acid (rRNA) gene diversity. Cluster and statistical analysis of T-RFLP data showed that soil bacterial communities fluctuated very little during the seasons (similarity indices between 0.835 and 0.997) with insignificant variations in 16S rRNA gene richness and diversity indices. Despite overall insignificant fluctuations, between 8 and 30% of all terminal restriction fragments changed their relative intensity in a significant manner among consecutive time samples. To determine the magnitude of community variations induced by external factors, soil samples were subjected to either inoculation with a pure bacterial culture, addition of the herbicide mecoprop, or addition of nutrients. All treatments resulted in statistically measurable changes of T-RFLP profiles of the communities. Addition of nutrients or bacteria plus mecoprop resulted in bacteria composition, which did not return to the original profile within 14 days. We propose that at less than 70% similarity in T-RFLP, the bacterial communities risk to drift apart to inherently different states.