Various sample pre-treatment techniques and different growth media for the isolation of Legionellae from hot water supplies in public buildings were compared. A total of 102 hot water samples from taps and showers was examined. The highest recovery frequency was obtained with the heat pre-treatment method and using the selective medium GVPC. However, the results differed according to the concentration of legionellas. In the case of low plate counts (< or =5000 cfu l(-1)), the heat pre-treatment technique gave a significantly higher percentage of positive samples compared with other techniques (P < 0.05). With increasing concentration, the differences between the procedures decreased until they became statistically not significant for concentrations above 50 000 cfu l(-1). The direct inoculum method allowed a significantly higher detection of concentrations (P < 0.001) compared with heat and acid decontamination methods, which brought about a 67-68% reduction in detectable Legionellae. Heat decontamination techniques show greater sensitivity and specificity. However, they underestimate the number of legionellas. In environmental surveillance programmes, this underestimate must be taken into consideration when assessing the health risk.