Dozens of chemicals, both natural and manmade, are often found in drinking water. Some, such as the natural contaminants uranium and arsenic, are well-known toxicants with a large toxicology database. Other chemicals, such as methyl tertiary-butyl ether (MTBE) from leaking fuel tanks, we learn about as we go along. For still others, such as the alkyl benzenes, there are very little available data, and few prospects of obtaining more. In some cases, chemicals are purposely added to drinking water for beneficial purposes (e.g., chlorine, fluoride, alum), which may cause a countervailing hazard. Removing all potentially toxic chemicals from the water is virtually impossible and is precluded for beneficial uses and for economic reasons. Determination of safe levels of chemicals in drinking water merges the available toxicity data with exposure and human effect assumptions into detailed hazard assessments. This process should incorporate as much conservatism as is needed to allow for uncertainty in the toxicity and exposure estimates. Possible sensitive subpopulations such as unborn children, infants, the elderly, and those with common diseases such as impaired kidney function must also be considered. However, the range of sensitivity and the variability of toxicity and exposure parameters can never be fully documented. In addition, the validity of the low-dose extrapolations, and whether the toxic effect found in animals occurs at all in humans, is never clear. This publication discusses how these competing needs and uncertainties intersect in the development of Public Health Goals for uranium, fluoride, arsenic, perchlorate, and other highly debated chemicals.