Fluorescence in situ hybridization (FISH) is widely used to describe bacterial community composition and, to a lesser extent, to describe the physiological state of cells. One of the limitations of the technique is that the effectiveness of the detection of target cells appears to vary widely. Here, we present a quantitative review of published reports on the percentage of cells detected using the common EUB338 probe (%Eub) in aquatic ecosystems. The %Eub varies from 1 to 100% in the different published reports, with an average of 56%. There is a methodological component in this variation, with a significant effect of the fluorochrome type and the stringency conditions of the reaction. But there is also a strong environmental component, and the type of ecosystem and dominant phylogenetic group significantly influence %Eub. We argue that the optimization of the FISH protocol to describe the phylogenetic composition of bacterial assemblages will probably lead to techniques that are not effective to describe the physiological state of cells.