A prospective follow-up study lasting 18 months in 1983-1984 was conducted in 50 villages in order to assess the risk related to the consumption of drinking water which did not meet the bacteriology standards. 119 physicians, 52 pharmacists and 118 primary-school teachers enumerated the cases of acute gastro-intestinal diseases observed among a population of 29,272 inhabitants. A weekly water sample was analysed in each village as to the presence of four indicator germs of fecal contamination: total plate count, total coliforms, fecal coliforms (thermotolerant) and fecal streptococci. 1,950 cases of acute gastro-intestinal diseases were registered by the physicians and pharmacists during a population experience of 1,873,303 persons-weeks, i.e. an incidence density of 1.06 X 10(-3) cases per person-week. The risk notified by the school teachers among children aged 7 to 11 was higher: 16.3 X 10(-3) cases per child-week. The villages were classified in 3 groups, according to the proportion of substandard samples: "good", "intermediate", and "bad" categories. The relative risk contrasting the "bad" and "good" villages was 3.5 for the cases notified by the physicians and the pharmacists among the general population (95% confidence interval = 2.5-4.8); it was 1.7 for the cases reported by the teachers among the children (95% confidence interval = 1.5-1.9). This concordant difference was very significant but not constant across the 18 months of the survey. The authors conclude that the bacteriology standards are a relevant indicator of a health hazard.