400 mV), the biofilm on the steel surface was characterized using confocal laser scanning microscopy (CLSM) in combination with functional and phylogenetic stains. The biofilm consisted of microbial cell clusters covering 10–20% of the surface. The clusters were loaf-formed, with a basal diameter of 20–150 μm, 5–20 per mm−2, each holding >104 cells in a density of 1–5 × 107 cells mm−3. The typical cluster contained mainly small Gram-negative bacteria (binding the EUB338 probe when hybridized in situ on the steel surface), and often carried one to three spherical colonies, either homogeneously composed of large Gram-negative cocci or more often small bacterial rods in high density, 108–109 cells mm−3. The clusters in live biofilms contained no pores, and clusters over 25 μm in diameter had a core nonpenetrable to fluorescent nucleic acid stains and ConA lectin stain. Fluorescently-tagged ConA stained cells at a depth of 10 cells in a stack) into the cluster than did the less polar dyes SYTO 16 (log Kow 1.48) and acridine orange (log Kow 1.24), which stained five cells in a stack. Fluorescent hydrophobic and hydrophilic latex beads (diameter 0.02, 0.1 or 1.0 μm) coated patchwise the cluster surface facing the water, but penetrated only to depths of ⩽2 μm indicating a permeability barrier. About 1/3 of the stainable cells hybridized in situ with Alf1b, while fewer than 1/7 hybridized to GAM42, probes targeted towards α- and γ-Proteobacteria, respectively. Our results represent a microscopic description of an ennobling biofilm, where the ennoblement could follow the sequence of redox events as suggested by the model of Dickinson and Lewandowski (1996) for the structure of corrosive biofilms on a steel surface.