Counting bacteria in drinking water samples by the epifluorescence technique after 4',6-diamidino-2-phenylindole (DAPI) staining is complicated by the fact that bacterial fluorescence varies with exposure of the cells to sodium hypochlorite. An Escherichia coli laboratory-grown suspension treated with sodium hypochlorite (5 to 15 mg of chlorine liter-1) for 90 min was highly fluorescent after DAPI staining probably due to cell membrane permeation and better and DAPI diffusion. At chlorine concentrations greater than 25 mg liter-1, DAPI-stained bacteria had only a low fluorescence. Stronger chlorine doses altered the DNA structure, preventing the DAPI from complexing with the DNA. When calf thymus DNA was exposed to sodium hypochlorite (from 15 to 50 mg of chlorine liter-1 for 90 min), the DNA lost the ability to complex with DAPI. Exposure to monochloramine did not have a similar effect. Treatment of drinking water with sodium hypochlorite (about 0.5 mg of chlorine liter-1) caused a significant increase in the percentage of poorly fluorescent bacteria, from 5% in unchlorinated waters (40 samples), to 35 to 39% in chlorinated waters (40 samples). The presence of the poorly fluorescent bacteria could explain the underestimation of the real number of bacteria after DAPI staining. Microscopic counting of both poorly and highly fluorescent bacteria is essential under these conditions to obtain the total number of bacteria. A similar effect of chlorination on acridine orange-stained bacteria was observed in treated drinking waters. The presence of the poorly fluorescent bacteria after DAPI staining could be interpreted as a sign of dead cells.