Previous epidemiological studies have shown a significant statistical correlation (P < 0.005 to P < 0.0005) between goiter prevalence and rock types in the watersheds that supply 37 localities in Western Colombia. Bacterial contamination has also been implicated as a cause of endemic goiter. We, therefore, did bacteriological studies to incorporate this variable into the statistical model in 34 of the 37 localities previously surveyed. Samples of the water in the 34 localities were taken at the water source or intake of water supply and at the pipeline system in households and schools. Samples were collected in sterile bottles and cultured on several media for 48 h. The number of bacteria per ml was determined by Quebec's camera and the bacteria (E. coli, Bacillus sp., K. pneumoniae, Proteus sp., and Neisseria sp.) were identified according to conventional methods. Multiple regression analysis and the logistic regression model were used to analyze the data. Only two variables had significant relationships with goiter prevalence: K. pneumoniae in the water source (P < 0.01) is related to a lower prevalence, and the overall concentration of bacteria in the pipeline system (P < 0.05) is related to a higher prevalence. Multiple regression analysis to compare the effects of the bacteriological variables to those of geological variables indicating rock types showed sedimentary rocks in the watershed associated (P < 0.005) with higher goiter prevalence and an increased concentration of K. pneumoniae in the water source again associated (P < 0.025) with lower goiter prevalence. Bacterial concentration in the pipeline system was of marginal importance (0.05 < P < 0.10). Thus, the presence of sedimentary rocks was the best indicator of disease. A second best indicator was the concentration of K. pneumoniae in the water source. We hypothesize that the lower goiter prevalence associated with K. pneumoniae may be a natural example of biodegradation of the organic contaminants that produce goiter.