A cultivation-independent approach based on polymerase chain reaction (PCR)-amplified partial small subunit rRNA genes and genetic profiling by single-strand conformation polymorphism (SSCP) was used to characterize the bacterial diversity inhabiting the rhizosphere of maize plants grown on an agricultural field. The community structures of two cultivars, a genetically engineered and a nonengineered variety, different herbicide regimes and soil tillage were compared with each other at two sampling dates. SSCP-profiles were generated with DNA from bacterial cell consortia with primers hybridizing to evolutionarily highly conserved rRNA gene regions. On silver-stained gels, each profile consisted of approx. 50 distinguishable bands. Similarity analyses of patterns recorded by digital image analyses could not detect any difference between cultivars or treatments that was greater than the variability between replicates. A total of 54 sequences recovered from different bands were identified and grouped into operational taxonomical units (OTUs). Surprisingly, only five of 40 OTUs contained sequences of both samplings. Three different bands from a profile were selected to test whether this small overlap was due to an incomplete recovery of sequences. From a faint band, two different OTUs were found when 12 clones were analysed, and from two strong bands 24 and 22 OTUs were detected from a total of 26 and 36 clones, respectively. The OTUs belonged to phylogenetically different groups of bacteria. Gene probes that were developed to target different bands of the profiles, however, indicated in Southern blot analyses that patterns between treatments, replicates and samplings, and even from two different growing seasons were highly conserved. Our study demonstrates that community profiles can consist of more sequences than detectable by staining and that gene probes in Southern blot can be a useful control to investigate the composition of microbial communities by genetic profiles.