Since one purpose of reactor site selection is to minimize risk to the public, the possible range of solutions to this siting problem can be explored using probabilistic concepts about reactor fission-product releases and risks from radiation. A mathematical definition based on consideration of radiobiological hazards from releases of gaseous and volatile fission products and their associated risks would be preferred. Using an assumption that the safety of the reactor can be made to conform to a release frequency limit line proposed in 1967 by F. R. Farmer, the highest risk to an individual member of the population is calculated, and it is suggested that this will be negligible compared to the risks of everyday life, and negligible compared to the collective risk to the population from reactor accident releases. These collective risks are assessed and analyzed, with two aspects—the risk of deaths occurring from radiation-induced cancer and the risk of loss of productive capacity by the community as a result of radioactive contamination of property—being selected as the most important risks to study. A factor of < 10 in probability of a given degree of injury or damage distinguishes remote sites from present-day semiurban sites in the United Kingdom. If the suggested release frequency limits can be achieved by existing reactor safety procedures and technology, population safety would appear to be adequately safeguarded. Therefore one may argue that the resources of society might be better employed in tackling the safety problems of non-nuclear industries and activities with the same diligence that has characterized the safe development and deployment of nuclear power.