slowly growing natural populations. Various approaches have been adopted in order to improve the sensitivity. These have included the use of multiple probes labelled with a single fluor (Lee et al. 1993); or labelled with multiple fluors (Trebesius et al. 1994) and enzyme-linked probes or detection systems that allow signal amplification (Lebaron et al. 1997, Schonhuber et al. 1999). The latter indirect approach not only has the potential for signal amplification, but may also be used in natural samples showing high levels of autofluorescence. Any thorough identification method has to include positive and negative controls. False-positive results may either be caused by cells emitting autofluorescence upon excitation or by nonspecific binding of the probe to nontarget cells. Samples should therefore be checked for autofluorescence before hybridization and a negative control with a fluorescent oligonucleotide not complementary to rRNA has to be applied to check for sequence-independent nonspecific binding. Such non-specific binding may be due to interaction of the dye compound of the probe with hydrophobic cell components. Failures to detect cells containing target sequences (false-negatives) may originate from cells with either low cellular ribosome content or limited permeability of the cell periphery for the fluorescent probe (Manz et al. 1992). With the rapidly expanding database of 16S rRNA sequences, the problem of probe specificity has become more apparent and the design of probes is becoming increasingly difficult. These problems are also applicable to PCR and other oligonucleotide-dependent techniques. The problem of probe specificity may be overcome by using multiple specific oligonucleotide probes targeting different sites on the rRNA molecule and labelled with different fluorochromes. While a single oligonucleotide target sequence may be found in a number of related taxa, the probability that target sites for three designed oligonucleotides are found in a nontarget organism is, however, much reduced.